International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 745-750
ISSN 2078-2365
Power Quality Enhancement using Custom Power
Devices
Pudi Sekhar#1, K. Venkateswara rao#2, T.Devaraju#3
#1
M.Tech Student, EEE Department, GITAM University
#2
Lecturer, EEE Department, GITAM University
#3
HOD, EEE Department, Sri Vidyanikethan Engineering college
Abstract— A Power quality problem is an occurrence manifested
as a nonstandard voltage, current or frequency that results in a
failure or mis-operation of end use Equipments. Utility
distribution networks, sensitive industrial loads, and critical
commercial operations all suffer from various types of outages
and service interruptions which can cost significant financial 1oss
per incident based on process down-time, lost production, idle
work forces, and other factors. With the restructuring of Power
Systems and with shifting trend towards distributed and dispersed
Generation, the issue of Power Quality is going to take newer
dimensions. The aim therefore, in this work, is to identify the
prominent concerns in the area and thereby to recommend
measures that can enhance the quality of the power, keeping in
mind their economic viability and technical repercussions. In this
paper three custom power controllers: DVR, DSTATCOM
(modelled
in
MATLAB/Simulink),
PWM
Switched
Autotransformer
(modelled
in
PSCAD/EMTDC
&
MATLAB/SIMULINK) are presented. Comprehensive results are
presented to assess the performance of each device to mitigate the
Voltage sag and economical device is proposed.
Index terms— Power Quality Problems, Voltage sag, DVR,
DSTATCOM, PWM Switched Autotransformer
I. INTRODUCTION
Power quality is certainly a major concern in the present era; it
becomes especially important with the introduction of
sophisticated devices, whose performance is very sensitive to
the quality of power supply. Modern industrial processes are
based a large amount of electronic devices such as
programmable logic controllers and adjustable speed drives.
The electronic devices are very sensitive to disturbances [1]
and thus industrial loads become less tolerant to power quality
problems such as voltage dips, voltage swells, and harmonics.
Voltage dips are considered one of the most severe
disturbances to the industrial equipment. Swells and over
voltages can cause over heating tripping or even destruction of
industrial equipment such as motor drives. Electronic
equipments are very sensitive loads against harmonics because
their control depends on either the peak value or the zero
crossing of the supplied voltage, which are all influenced by
the harmonic distortion.
This paper analyzes the key issues in the Power Quality
problems. As one of the prominent power quality problems, the
origin, consequences and mitigation techniques of voltage sag
problem has been discussed in detail. The study describes the
techniques of mitigating voltage sag in a distribution system by
two power electronics based devices called Dynamic Voltage
Restorer (DVR) and Distribution STATCOM (D-STATCOM)
and a new mitigation technique called PWM Switched
Autotransformer.
II. SOURCES AND EFFECTS OF POWER QUALITY PROBLEMS
The distortion in the quality of supply power can be introduced
/enhanced at various stages; however, some of the primary
sources of distortion [2] can be identified as below:
A. Power Electronic Devices
B. IT and Office Equipments
C. Arcing Devices
D. Load Switching
E. Large Motor Starting
F. Embedded Generation
G. Electromagnetic Radiations and Cables
H. Storm and Environment Related Causes etc.
Some of the common power quality issues and their
prominent impact are summarized in the table below:
Problem
TABLE I
Effects
Page 745
Pudi et. al., Power Quality Enhancement using Custom Power Devices
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 745-750
ISSN 2078-2365
Voltage sags
Transients
Harmonics
Flicker
Devices /Process down time, effect on
product quality, failure/malfunction of
customer equipments and associated scrap
cost, clean up costs, maintenance and repair
costs etc.
Tripping, components failures, flashover of
instrument insulation hardware re booting,
software glitches, poor product quality etc.
Excessive losses and heating in motors,
capacitors and transformers connected to the
system
Visual irritation, introduction of many
harmonic components in the supply power
and their associated equipment.
Dynamic Voltage Restorer is a series connected device
designed to maintain a constant RMS voltage value across a
sensitive load. The DVR considered consists of:
an injection / series transformer
a harmonic filter
a Voltage Source Converter (VSC)
an energy storage and
a control system as shown in Figure 1
III. USE OF CUSTOM POWER DEVICES TO IMPROVE POWER
QUALITY
In order to overcome the problems mentioned above
conventional devices such as:
•
•
•
•
•
•
•
•
Line- voltage regulators: Tap changers, buck-boost
regulators, CVT (Constant- voltage transformer).
M-G Sets (Motor-generator Sets)
Magnetic Synthesizers
SVC (Static VAR Compensators)
UPS (Uninterruptible Power Supplies)
SMES (Superconducting magnetic energy storage)
Static Transfer Switch
Fuel Cell Based Inverter System
can be used. But Present day modern equipments are very
sensitive to voltage sags and they need the mitigating device to
be very fast in acting, which cannot possible by the above
conventional devices. So in order to overcome the above
disadvantages, a new category of devices called custom power
devices are developed. Custom power devices are the new
generation of power electronics-based equipment aimed at
enhancing the reliability and quality of power flows in lowvoltage distribution networks. There are various custom power
devices available such as DVR (dynamic voltage restorer),
Dstatcom, UPQC and PWM switching auto transformer.
IV. MODELING OF CUSTOM POWER DEVICES AND SIMULATION
RESULTS
Fig. 1 Schematic diagram of DVR
The main function of a DVR is the protection of sensitive loads
from voltage sags/swells coming from the network. Therefore
as shown in Figure 1, the DVR is located on approach of
sensitive loads. If a fault occurs on other lines, DVR inserts
series voltage VDVR and compensates load voltage to pre fault
value. The momentary amplitudes of the three injected phase
voltages are controlled such as to eliminate any detrimental
effects of a bus fault to the load voltage VL. This means that
any differential voltages caused by transient disturbances in the
ac feeder will be compensated by an equivalent voltage
generated by the converter and injected on the medium voltage
level through the booster transformer.
The series injected voltage of the DVR can be written as [3]
Vinj = Vload + Vs
(1)
Where;
Vload is the desired load voltage magnitude
Vs is the source voltage during sags/swells condition
The test system employed to carry out the simulations
concerning the DVR actuation is shown in Figure 2
A. DVR
Page 746
Pudi et. al., Power Quality Enhancement using Custom Power Devices
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 745-750
ISSN 2078-2365
Fig. 3 Simulation results showing: (a) Supply Voltage, (b) Injected Voltage
(c) Load Voltage
Fig. 2 Proposed System Configuration (DVR)
To verify the working of a DVR employed to avoid voltage
sags, simulation studies are carried out as follows:
A three–phase short-circuit fault is applied during the period of
200 to 500 ms with a fault resistance of 0.001.
The first simulation shows three phase voltage sag is simulated.
The simulation started with the supply voltage 30% sagging as
shown in Figure 3(a).Figure 3(a) also shows a 30% voltage sag
initiated at 0.2s and it is kept until 0.4s, with total voltage sag
duration of 0.2s and Figure 3(b) shows voltage injected by the
DVR and the corresponding load voltage in Figure 3(c) with
compensation. As a result of DVR, the load voltage is kept at
1pu.
When the DVR is in operation the voltage sag is mitigated
almost completely, and the rms voltage at the sensitive load
point is maintained at 98%, as shown in Figure 3.
B. D-STATCOM
The STATCOM consists mainly of a PWM inverter connected
to the network through a transformer. The dc link voltage is
provided by capacitor C which is charged with power taken
from the network [4].The control system ensures the regulation
of the bus voltage and the dc link voltage. The D-STATCOM
function is to regulate the bus voltage by absorbing or
generating reactive power to the network, like a thyristor static
compensator.
Page 747
Pudi et. al., Power Quality Enhancement using Custom Power Devices
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 745-750
ISSN 2078-2365
Fig. 6 Voltage sag generated due to sudden increase of load
Fig. 4 Block diagram of the proposed model
The block diagram of the proposed model is shown in Fig. 4.
The load 1 shown is the normal running load. At any specified
time interval, a sudden load is included. The introduction of
this sudden load introduces voltage sag. Hence if a
DSTATCOM is connected in parallel to the distribution, it will
correct the voltage sag in that interval.
The SIMULINK model for generating a three phase sag in a
power system is shown in Fig. 5. A three phase linear
transformer is provided for isolation. Two feeders originate
from this transformer. In one of the feeders, a sudden load is
included from the period of 0.2 to 0.3s. This introduction of
sudden load produces a reduction in voltage causing a voltage
dip is shown in Fig.6.
Fig. 7 Restored Voltage Waveform
The error signal is obtained by comparing the per unit value of
the voltage with a constant 1. The reference sinusoidal signals
required for the generation of PWM pulses is obtained by using
a PI controller in conjunction with the error signal obtained.
The PWM techniques have several advantages compared to
other techniques. The restored voltage waveform is shown in
Fig. 7.
C. VOLTAGE SAG MITIGATION USING PWM SWITCHED
AUTOTRANSFORMER
Fig. 5 Simulink model for Voltage Sag mitigation using D-STATCOM
Fig. 8 Block diagram of the single phase voltage sag mitigation scheme
Page 748
Pudi et. al., Power Quality Enhancement using Custom Power Devices
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 745-750
ISSN 2078-2365
(i) Principle of operation
An IGBT is used as power electronic device to inject the
error voltage into the line so as to maintain the load voltage
constant. Four power diodes (D1 to D4) connected to IGBT
switch (SW) controls the direction of power flow and
connected in ac voltage controller configuration. This
combination with a suitable control circuit maintains constant
rms load voltage. In this scheme sinusoidal PWM pulse
technique is used. RMS value of the load voltage V L is
calculated and compared with the reference rms voltage Vref
Under normal condition when there is no voltage
disturbance the power flow is through the anti parallel
thyristors used as the ac bypass switch. Output filters
containing a main capacitor filter and a notch filter are used at
the output side to filter out the switching noise and reduce
harmonics. During this normal condition VL=Vref, and the error
voltage V err is zero. The gate pulses are blocked to IGBT.
A sag or swell occurs in the system may be due to sudden
increase or decrease in the load, or due to faults. The supply
voltage Vs and hence VL decreases during disturbance. When
the sensing circuit detects an error voltage greater than of the
normal voltage the voltage controller acts immediately to
switch off the thyristors. Voltage Verr applied to the pi
controller gives the phase angle . The control voltage given
in (2) is constructed at power frequency f = 50 Hz.
Vcontrol = sin (wt+ )
(2)
Where ma is the modulation index.
The phase angle delta is dependent on the percentage of
disturbance and hence controls the magnitude of Vcontrol.This
control voltage is then compared with the triangular voltage Vtri
to generate the PWM pulses VG which are applied to the IGBT
to regulate the output voltage. Hence the IGBT switch operates
only during voltage sag condition and regulates the output
voltage according to the PWM duty-cycle.
Single phase voltage sag supporter using a PWM Switched
autotransformer with R.M.S voltage as a reference
Fig. 9 PSCAD/EMTDC model of single phase PWM switched auto
transformer with RMS voltage as reference
The Fig. 9 shows the PSCAD/EMTDC model for single phase
PWM switched auto transformer for mitigation of voltage sag
with peak voltage as a reference. It is tested on a simple a.c
system [5] [6] at which a single phase to ground fault is created
at 0.2 sec for a duration of 0.2 sec with a fault resistance of 0.1
ohm, which creates a voltage sag of 50% at load .At 0.2 sec,
the PWM switched auto transformer switches ON and it takes
one cycle to detect the sag and compensate the voltage sag. The
waveforms for source and load voltages, load currents before,
during and after the fault are as shown in Fig. 10.
RMS detection method: In this method the instantaneous load
voltage is converted into rms voltage and compared with the
reference rms voltage, which gives an error voltage and this
voltage given PI controller, which calculates the required angle
delta in proportion to error voltage.
Page 749
Pudi et. al., Power Quality Enhancement using Custom Power Devices
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 745-750
ISSN 2078-2365
Fig. 13 Waveform of load voltage showing mitigated
sag using PWM Switched autotransformer
Fig. 10 Simulation results (a) Source voltage with sag (b) Load Voltage
without sag
Fig. 11 Simulink model PWM Switched Autotransformer of a 3-phase system
used for voltage sag studies
Figure 11 shows Simulink model PWM Switched
Autotransformer [7] of a 3-phase system for voltage sag
mitigation. A sudden loading is applied at 0.1 sec to 0.2 sec and
a sag is observed as shown in Figure 12 and sag is mitigated
using PWM Switched Autotransformer and the corresponding
waveform is shown in 13.
V. CONCLUSIONS
Power quality measures can be applied both at the user end and
also at the utility level. This paper has presented models of
custom power equipment, namely DVR, D-STATCOM and
PWM switched autotransformer, applied them to mitigate
voltage dip which is very prominent as per utilities are
concerned. Though conventional techniques are available, the
proposed devices are very fast acting and efficient. A new
Voltage sag mitigation topology called PWM switched auto
transformer is modelled and simulated. This topology requires
only one PWM switch per phase as compared to DVR or
DSTATCOM requires two switches per phase. The PWM
switched auto transformer does not require energy storage
device for mitigation of voltage sag as compared to DVR and
DSTATCOM requires energy storage elements. so, PWM
switched autotransformer is efficient and economical among
the custom power devices.
REFERENCES
[1]
[2]
[3]
[4]
[5]
Fig. 12Waveform of the load voltage with sag during 0.1 sec to 0.2 sec
[6]
H. Hingorani “Introducing Custom Power” IEEE Spectrum, vol.32 no.6
June 1995 p 41-48.
N.G. Hingorani and L. Gyugyi, “Understanding FACTS: Concepts and
Technology of Flexible AC Transmission Systems”.
Mitigation Of Voltage Sags/Swells Using Dynamic Voltage Restorer
(DVR) Rosli Omar and Nasrudin Abd Rahim. 2008 Australasian
Universities Power Engineering Conference (AUPEC'08) Paper P-027.
Mitigation of voltage sag using Distribution Static Compensator (DSTATCOM) S. Elango Dr. E.Chandra Sekaran, 2011 IEEE.
Mitigation of Voltage Sags/Swells using PWM Switched
Autotransformer C. Venkatesh, Student Member, IEEE, V. Prasad
Reddy, Student Member, IEEE, and Dr. D.V.S.S. Siva Sarma, Senior
Member, IEEE.
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 22, NO.
2, MARCH 2007 A Voltage Sag Supporter Utilizing a PWM-Switched
Autotransformer Dong-Myung Lee, Member, IEEE, Thomas G.
Habetler, Fellow, IEEE, Ronald G. Harley, Fellow, IEEE, Thomas L.
Keister, Member, IEEE, and Joseph R. Rostron, Member, IEEE.
Page 750
Pudi et. al., Power Quality Enhancement using Custom Power Devices
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 745-750
ISSN 2078-2365
[7]
Modeling and simulation of PWM Switched Autotransformer for
voltage sag mitigation using MATLAB, International Journal of
electrical and power engineering 4(3), 164-168, 2010.
Page 751
Pudi et. al., Power Quality Enhancement using Custom Power Devices
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
Dynamic analysis of Single Machine Infinite Bus
system using Single input and Dual input PSS
P. PAVAN KUMAR
EEE Department, Gitam University,
Visakhapatnam, Andhra Pradesh,
India-533045
Email:
pavan.kumar270489@gmail.com
M. RAVINDRA BABU
EEE Department, Gitam University,
Visakhapatnam, Andhra Pradesh,
India-533045
Email: raviravi1983@gmail.com
Abstract – This paper deals with the design of both single input and
dual input conventional PSS which is used to damp the low
frequency rotor oscillations taking place in power systems. The
single input PSS used here are power based derivative type and
speed based lead-lag type stabilizer, the dual input stabilizer, PSS3B
has two inputs namely from change in speed and deviation of
electrical power and has two frequency bands, lower and higher
unlike the single input PSS. The PSS parameters are tuned,
considering the machine data and operating point of the system
used. The optimal parameters of the PSS are obtained using pole
placement and genetic algorithm technique and the respective
results are compared graphically .The system used is Single
Machine Infinite Bus (SMIB) system which is modelled using state
space analysis and its dynamic response is analyzed both for system
without PSS and with PSS (both single and dual input) using
Simulink/Matlab.
Keywords: temperature microchange, wireless sensor network,
global change, integral equation, WSN.
I.
INTRODUCTION
Power systems experience low-frequency oscillations due
to disturbances. These low frequency oscillations are related to
the small signal stability of a power system. The phenomenon
of stability of synchronous machine under small perturbations
is explored by examining the case of a single machine
connected to an infinite bus system (SMIB). The analysis of
SMIB gives physical insight into the problem of low
frequency oscillations. These low frequency oscillations are
classified into local mode, inter area mode and torsional mode
of oscillations. The SMIB system is predominant in local
mode low frequency oscillations [7]. These oscillations may
SARASWATHI
EEE Department, Gitam University,
Visakhapatnam, Andhra Pradesh,
India-533045,
Email: g_saraswathi@gitam.edu
sustain and grow to cause system separation if no adequate
damping is available.
Small signal disturbances observed on the power system
are caused by many factors such as heavy power transmitted
over weak tie line and the effect of fast acting, high gain
automatic voltage regulator (AVRs) [6]. The main function of
the AVR is to improve the transient stability during faults
conditions. However, its high gain and fast acting effect have
an adverse effect on the system damping which is reduced to a
negative value. The under damped system exhibits low
frequency oscillations also known as electromechanical
oscillations. These oscillations limit the power transfer over
the network and if not properly damped, they can grow in
magnitude to cause system separation. To counteract the
adverse effects of the AVRS, Power system stabilizer (PSS) is
used in the auxiliary feedback to provide supplementary
damping [6] to the system to damp these low frequency
oscillations on the rotor.
To overcome this problem, several approaches based on
modern control theory, such as Optimal control, Variable
control and intelligent control were simulated and tested with
satisfactory results. But these stabilizers have been proved to
be difficult to implement in real systems. Thus, CPSS remains
widely used by power utilities for its simple structure and
reliability. Over the past 15 years, interests have been focused
on the optimization of the PSS parameters to provide adequate
performance for all operating conditions. Hence, many
optimizations techniques based on artificial intelligence have
been used to find the optimum set of parameters to effectively
tune the PSS.
In this paper both single input (speed & power based) and
dual input stabilizers (PSS3B) are used to damp the low
frequency oscillations associated with the system. PSS3B is
used with combination of shaft speed deviation (∆ω) and
Page 632
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
change in electrical power (∆Pe) which has its own advantages
when compared to single input PSS which is described below
in section 3. The parameters of both the types of PSS are tuned
using Pole Placement technique and Genetic Algorithm and
results are thus analyzed.
II.
METHODOLOGY
A single machine-infinite bus (SMIB) system is considered for
the present investigation. A machine connected to a large
system through a transmission line may be reduced to a SMIB
system, by using Thevenin’s equivalent of the transmission
network external to the machine.
The synchronous machine is described as the fourth order
model. The two-axis synchronous machine representation with
a field circuit in the direct axis but without damper windings is
considered for the analysis. The system dynamics of the
synchronous machine can be expressed as a set of four first
order linear differential equations given in equations below
[6]. These equations represent a fourth order generator model.
The Heffron-Phillips constants are dependent on the machine
parameters and the operating condition considered for the
system. Here K1, K2, K3 and K6 are positive [6]. K4 is mostly
positive except for cases where Re is high. K5 can be either
positive or negative and K5 is positive for low to medium
external impedances (Re+ jXe) and low to medium loadings. K5
is usually negative for moderate to high external impedances
and heavy loadings [6]. The overall linearized block diagram
of the SMIB system is shown in Fig.1 below.
For the system considered four state variables are considered
and linearized differential equations can be written in the state
space form as,
(9)
Where,
(10)
(1)
(2)
(11)
(3)
(4)
(12)
The constants (K1-K6) are called Heffron-Phillips constants
and are computed using the equations given in Appendix.
The system data considered is:
xd = 0.973
= 0.19
xq = 0.55
= 7.765s
D=0 H=5
f=60Hz
(5)
Transmission line (p.u):
Re = 0 Xe = 0.4
(6)
Exciter:
KE = 200 TE = 0.05s
(7)
Operating point:
Vto = 1.0
P0 =1.0
Q0 = 0.2
δ0 = 28.26 o
(8)
Page 633
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
Since PSSs are tuned at the nominal operating point, the
damping is only adequate in the vicinity of those operating
points. But power systems are highly nonlinear systems,
therefore, the machine parameters change with loading and
time. The dynamic characteristics also vary at different points.
IV.
CONVENTIONAL POWER SYSTEM
STABILIZER
The basic function of CPSS is to damp electromechanical
oscillations. To achieve the damping, the CPSS proceeds by
controlling the AVR excitation using auxiliary stabilizing
signal. The CPSS’s structure is illustrated in Figure 2.
Fig. 1. Linearized block diagram of SMIB
In the above state space equation system state matrix A is a
function of the system parameters, which depend on operating
conditions, control matrix B depends on system parameters
only and control signal U is the PSS output. Using these state
equations and state matrices the overall transfer function of the
system is computed, since here no controller is used, it is
considered as open loop system whose transfer function is
G(s).
III.
Fig. 2. Structure of CPSS
The CPSS classically uses the following inputs [5]:
· The shaft speed deviation ∆ω
· Active power output, ∆Pa (Change in accelerating power)
· ∆Pe (change in electric power),
· Bus frequency ∆f
POWER SYSTEM STABILIZER
One problem that faces power systems nowadays is the low
frequency oscillations arising from interconnected systems.
Sometimes, these oscillations sustain for minutes and grow to
cause system separation. The separation occurs if no adequate
damping is available to compensate for the insufficiency of the
damping torque in the synchronous generator unit. This
insufficiency of damping is mainly due to the AVR exciter’s
high speed and gain and the system’s loading.
In order to overcome the problem, PSSs have been
successfully tested and implemented to damp low frequency
oscillations. The PSS provides supplementary feedback
stabilizing signal in the excitation system. The feedback is
implemented in such a way that electrical torque on the rotor is
in phase with speed variations [7]. PSS parameters are
normally fixed for certain values that are determined under
particular operating conditions. Once the system operating
conditions are changed, PSS may not produce adequate
damping into an unstable system.
1. Gain
The gain determines the amount of damping introduced by
the stabilizer. Therefore, increasing the gain can move
unstable oscillatory modes into the left – hand complex plane.
Ideally, the gain should be set to a value corresponding to a
maximum damping. However, in practice the gain Kpss is set to
a value satisfactory to damp the critical mode without
compromising the stability of other modes.
2. Washout
The washout stage is a High Pass Filter (HPF) with
purpose to respond only to oscillations in speed and block the
Page 634
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
dc offsets. The Washout filter prevents the terminal voltage of
the generator to drift away due to any steady change in speed.
3. Phase compensation
This stage consists of two lead – lag compensators as
shown in Figure 2 (lead – lag compensation stage). The lead
stage is used to compensate for the phase lag introduced by the
AVR and the field circuit of the generator. The lead – lag
parameters T1-T4 are tuned in such as way that speed
oscillations give a damping torque on the rotor. When the
terminal voltage is varied, the PSS affects the power flow from
the generator, which efficiently damps the local modes.
1)
Speed based lead-lag PSS: These stabilizers employ
the direct measurement of shaft speed (∆ω) and employ it as
input signal for it. The stabilizer, while damping the rotor
oscillations, could reduce the damping of the lower-frequency
torsional modes if adequate filtering measures were not taken
[1 & 5]. In addition to careful pickup placement at a location
along the shaft where low-frequency shaft torsionals were at a
minimum electronic filters called torsional filters should be
used for adequate damping of low frequency oscillations.
The structure of this PSS is in the form as shown below [1],
for which the parameter such as stabilizer gain Kc, lead lag
time constants T1 and T2 are to be computed such that the
overall closed loop system will be stable when the PSS is
included in the feedback loop.
4. Torsional Filter
This stage is added to reduce the impact on the torsional
dynamics of the generator while preventing the voltage errors
due to the frequency offset.
5.
Limiter
The PSS output requires limits in order to prevent conflicts
with AVR actions during load rejection. The AVR acts to
reduce the terminal voltage while it increases the rotor speed
and the bus frequency. Thus, the PSS is compelled to
counteract and produce more positive output. As described in
by P. Kundur in [8], the positive and negative limit should be
around the AVR set point to avoid any counteraction. The
positive limit of the PSS output voltage contributes to improve
the transient stability in the first swing during a fault. The
negative limit appears to be very important during the back
swing of the rotor.
V.
(13)
2)
Power based derivative PSS: Due to the simplicity of
measuring electrical power and its relationship to shaft speed,
it was considered to be a natural candidate as an input signal to
early stabilizers. The equation of motion for the rotor can be
written as follows [1 & 5]:
(14)
Where, H = inertia constant
ΔPm= change in mechanical power input
ΔPe= change in electric power output
Δω = speed deviation
As previously mentioned this type of stabilizer uses electrical
power (∆Pe) as input and is of derivative type whose structure
is as shown below [1], and the optimal stabilizer parameter K
and T are to be computed which ensure closed loop stability of
the system.
(15)
SINGLE INPUT PSS
The input signals include deviations in the rotor speed
(∆ω=ωmech – ωo), the frequency (∆f), the electrical power (∆P e)
and the accelerating power (∆Pa) [5].
As mentioned above in this paper two types of PSS are
considered to damp the low frequency oscillations they are,
1.1 Dual input CPSS (PSS3B)
In this paper a dual input PSS is used, the two inputs to dualinput PSS are Δω and ΔPe, with two frequency bands, lower
frequency and higher frequency bands, unlike the conventional
single-input (Δω) PSS [2]. The performance of IEEE type
Page 635
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
PSS3B is found to be the best one within the periphery of the
studied system model. This dual input PSS configuration is
considered for the present work and its block diagram
representation is shown in Figure 3
for the control of low frequency speed oscillations, whose
transfer function is taken as H(s) [4]. The simple block
diagram considered for pole placement technique is shown
below.
Fig. 4. Closed loop system including PSS
.Fig. 3. IEEE type PSS3B structure
In the above PSS structure used [2], the unknown
parameters are computed using pole placement and genetic
algorithm techniques, in case of pole placement technique the
transfer function of the pss is computed and is used in
feedback to form a closed loop system, for which
characteristic equation is formed to compute the unknown
parameters of PSS by placing dominant eigen values in place
of ‘s’ in the characteristic equation..
Let the linearized equations of single machine, infinite bus
system be expressed in the form,
sX(s) = AX(s)+BU(s)
(17)
Y(s) = CX(s)
(18)
The PSS with the following structure is used [4],
(19)
(16)
The transfer function of the PSS3B used is shown above, and
the pole placement technique is explained in detail in section
VI.
VI.
Where the PSS parameter are to determined such that
system dominant eigen values are equal to desired eigen
values. Using equations (17),(18) and (19), it can be readily
shown that the closed loop system characteristic equation is
given by,
POLE PLACEMENT TECHNIQUE
(20)
Pole placement is a method employed in feedback control
system theory to place the closed-loop poles of a plant in predetermined locations in the s-plane. This method is also
known as Full State Feedback (FSF) technique. Placing poles
is desirable because the location of the poles corresponds
directly to the eigen values of the system, which control the
characteristics of the response of the system.
Based on the system data considered and the operating
condition, the Heffron-Phillips constants for the system are
computed. The state equations are then considered using these
constants to compute the state matrices and then the transfer
function of the open loop system is computed in matlab using
these state matrices. The open loop system transfer function is
taken as G(s). Now in the feedback loop, the stabilizer is used
From eqn.(20) the required stabilizer parameters can be
computed by replacing ‘s’ by the desired eigen value λ and
equating the real and imaginary terms on both sides of the
equation [4].
Using the state equations and state matrices mentioned in
section 2, the open loop transfer function G(s) of the system is
obtained, and the PSS of structure shown in eqns.(13),(15) and
(16) is used as feedback H(s) for the open loop system and
thus forming the closed loop system with unknown parameters,
which are computed as mentioned above by replacing ‘s’ by
dominant eigen values.
VII.
GENETIC ALGORITHM (GA)
Page 636
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
Genetic Algorithms (GAs) are heuristic search procedures
inspired by the mechanism of evolution and natural genetic.
They combine the survival of the fittest principle with
information exchange among individuals. GA’s are simple yet
powerful tools for system optimization and other applications
[11].
This technique has been pioneered few decades ago by
Holland, basing the approach on the Darwin’s survival of the
fittest hypothesis. In GA’s candidates solutions to a problem
are similar to individuals in a population. A population of
individuals is maintained within the search space of GAs, each
representing a possible solution to a given problem. The
individuals are randomly collected to form the initial
population from which improvement is sought. The
individuals are then selected according to their level of fitness
within the problem domain and breed together. The breeding
is done by using the operators borrowed from the natural
genetic, to form future generations (offsprings) [11]. The
population is successively improved with respect to the search
objective. The least fit individuals are replaced with new and
fitter offspring from previous generation.
trapped in a local minimum. Mutation plays the role
of recovering the lost genetic materials as well as for
randomly disturbing genetic information. Mutation
has traditionally considered as a simple search
operator [11]. If crossover is supposed to exploit the
current solution to find better ones, mutation is
supposed to help for the exploration of the whole
search space.
4) Replacement: Replacement is the last stage of any
breeding cycle. It is in this process that children
populate the next generation by replacing parents, if
fitter. Reinsertion can be made partially or
completely, uniformly (offspring replace parents
uniformly at random) or fitness-based.
The most common operators handled in genetic algorithm
are described in detail below, which in whole called as
breeding cycle.
1) Selection (Reproduction): In this stage, individuals
are selected from the current population according to
their fitness value, obtained from the objective
function previously described. The purpose of the
selection is to choose individuals to be mated. The
selection can be performed in several ways. But many
selection techniques employ a “roulette wheel” [11].
It is a mechanism to probabilistically select
individuals based on some measure of their
performances.
2) Crossover (Recombination):
In this stage, the
individuals retained (in pairs), from the above stage,
exchange genetic information to form new
individuals (offsprings). This process helps the
optimization search to escape from possible local
optima and search different zones of the search space
[11]. The combination or crossover is done by
randomly choosing a cutting point where both parents
are divided in two. Then the parents exchange
information to form two offsprings that may replace
them if the children are fitter.
3) Mutation: After crossover, the strings are subjected to
mutation. Mutation prevents the algorithm to be
Fig. 5. General Scheme of Genetic Algorithm
All these operation are carried out in Genetic Algorithm
toolbox in which the following fitness function has to be
defined. The problem of computing optimal parameters of a
single power system stabilizer for different operating points
implies that power system stabilizer must stabilize the family
of N plants [1]:
, k= 1,2,3….N
(21)
Where X(t) is the state vector and U(t) is the input
stabilizing signal. A necessary and sufficient condition for the
set of plants in the system to be simultaneously stabilizable
with stabilizing signal is that Eigen values of the closed-loop
system lie in the left- hand side of the complex s-plane [1].
This condition motivates the following approach for
determining parameters Ks1, Ks2, T1 and T2 of the power
system stabilizer. Selection of Ks1, Ks2, T1 and T2 to minimize
the following fitness function,
Page 637
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
i=1,2,…N, k=1,2,..N (22)
consideration which are further used to compute the HeffronPhillips constants (A.2).
Where λi,k is the kth closed-loop eigen value of the ith plant [1]. If
a solution is found such that J<0, then the resulting Ks1, Ks2, T1
and T2 stabilize the collection of plants.
For running the GA toolbox the command gatool [10], is
to be given in command window of MATLAB and in the tool
the fitness function is to be defined in which the state matrix A
including PSS is used and the unknown PSS parameters are
taken as unknown variables which are to be optimized such
that the eigen values of the matrix lie on the left half of s-plane
i.e., in the stability region. This method of finding the
parameter is applied for the type of PSS described in section 3.
The state matrices ‘A’ and the specifications used for running
GA toolbox are mentioned in Appendix.
VIII.
APPENDIX
1.2 Calculation of Heffron-Phillips constants
All the variables with subscript ‘0’ are values of variables
evaluated at their pre-disturbance steady-state operating point
from the known values of P0 , Q0 and Vt0.
(A.2)
1.3 Modelling of System including Speed based PSS (∆ω)
When PSS of structure described in equation (13) is used as
feedback of open loop system, it forms a closed loop system.
The state equations involved are,
(A.1)
The above equations indicated in (A.1) are used to
calculate the initial conditions of the system under
Page 638
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
(A.3)
1.4 Modelling of System including Power based PSS (∆Pe)
When PSS of structure described in equation (16) is used as
feedback of open loop system, it forms a closed loop system.
The state equations involved are,
(A.4)
The wash out time constant for the both speed and power
based PSS is taken as Tw= 2sec
1.5 Modelling of System including PSS3B
The state equations of the system when PSS of structure
shown in section 3 is used in the feedback loop are derived as
below (A.3).
Page 639
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
Crossover fraction
0.7
Mutation function
Use constraint dependent default
Crossover function
Single point
Migration direction
Forward
Number of generations
300
The application of GA tool box for optimization of PSS
parameters, the following constraints on the parameters has to
be considered,
(A.5)
The state matrix ‘A’ of the system including PSS3B is shown
below (A.6) which is used in the objective function to evaluate
the fitness using GA tool box.
For speed based PSS,
10 ≤ Kc ≤ 50; 0.01 ≤ T1 ≤ 1; 0.01 ≤ T2 ≤ 0.1
For power based PSS,
0.1 ≤ K ≤ 10; 0.01 ≤ T ≤ 1
-3 ≤ Ks1 ≤ 0 ; 20 ≤ Ks2 ≤ 60 ; 0 ≤ T1 ≤ 0.3 ; 0 ≤ T2 ≤ 0.1
IX.
RESULT AND DISCUSSION
The parameters of the PSS obtained using pole placement and
Genetic Algorithm techniques are shown below.
Single input parameters:
1) Speed based PSS using Pole placement technique are,
Kc=9.6763, T1=0.285sec, T2=0.05sec
(A.6)
The washout time constants is taken as T w1=Tw2=10sec.
1.6 Specifications of Genetic Algorithm
For using GA toolbox to optimize the PSS parameters the
following specifications are used,
Parameterd obtained using GeneticAlgorithm is,
Kc=10.541, T1=0.498sec, T2=0.1sec
2) Power based PSS using Pole placement technique are,
K=0.8954, T=0.3104sec
Parameters obtained using Genetic Algorithm is,
K=3.4, T=0.498sec
Table 1. Genetic Algorithm Specifications for Toolbox
Table 2. PSS3B parameters
Population size
75
Creation function
Use constraint dependent default
Scaling function
Rank
Selection function
Roulette
PSS3B
Parameters
Ks1
Ks2
T1
T2
Pole
Placement
-0.5
48.259
0.05sec
0.25sec
Genetic
algorithm
-0.354
20.003
0.15sec
0.1sec
Page 640
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 632-641
ISSN 2078-2365
technique and are simulated to analyse the dynamic response
in both the cases.
The technique of computing parameters becomes complex
with the increase in number of machines in case of pole
placement technique, whereas the technique of Genetic
Algorithm can be used to compute optimal parameters of PSS
for wide range of operating conditions in power system and
also can be implemented for multi-machine system. The
settling time of the PSS is less in case of Genetic Algorithm
technique when compared to Pole Placement Technique.
XI.
Fig. 6. Simulation output of SMIB with GA-PSS
REFERENCES
[1] “Tuning of Power System Stabilizers via Genetic Algorithm for
Stabilization of power Systems” by MehranRashidi, FarzanRashidi, Hamid
Moaavar, 0-7803-7952-7/03/$17.00 0 2003 IEEE
[2]. M. Sreedevi and P. Jeno Paul, “Comparison of Two Power system
stabilizers for Power system stability” International Journal of Signal System
Control and Engineering Applications, 3(4): 70-76, 2010 ISSN: 1997-5422
[3]. Joe H. Chow, George E. Boukarim, and Alexander Murdoch, “Power
System Stabilizers as Undergraduate Control Design Projects”, IEEE
transactions on power systems, vol. 19, no. 1, February 2004. pp. 144-151
[4]. “Efficient pole-assignment method for designing stabilisers in
multimachine power systems” by S. Elangovan and CM. Lim, IEE
PROCEEDINGS, Vol. 134, Pt. C, No. 6, NOVEMBER 1987
[5]. G.R. Bérubé, L.M. Hajagos, Members Kestrel Power Engineering Ltd.
Accelerating-Power Based Power System Stabilizers
Fig. 7. Simulation output of SMIB with Pole Placement-PSS
The settling time of the simulation response for PSS3B are
compared in table shown below,
Settling
Time
Without PSS
Pole placement
PSS
GA PSS
Table 3. Settling time comparison
Single input PSS
Dual input
PSS
Speed based
Power
PSS3B
PSS
based PSS
56.43sec
56.43sec
56.43sec
4.14sec
5.79sec
3.66sec
3.29sec
X.
1.93sec
1.74sec
[6].K.R.Padiyar, “Power system Dynamics Stability and Control”. John Wiley;
Interline Publishing, 1996
[7].P. M. Anderson and A.A Fouad “Power system control and stability”,Iowa
state university Press 1977
[8].P.Kundur, “Power system stability and control”. McGraw-Hill, New York
1994.
[9]. E.V Larsen and D.A. Swann, “Applying Power System Stabilizers, Parts I,
I1 and III”, IEEE Trans., Vol. PAS-100, June 1981, pp. 3017- 3046
[10]. A. J. Chipperfield and P. J. Fleming “The MATLAB Genetic Algorithm
Toolbox” From IEE Colloquium on Applied Control Techniques Using
MATLAB, Digest No. 1995/014, 26/01/95
[11].
S.N.Sivanandam,
S.N .Deepa
“Introduction
Algorithms”.Springer-Verlag Berlin Heidelberg 2008.
to
Genetic
CONCLUSION
The optimal parameters of dual input conventional pss,
PSS3B is obtained using pole placement and genetic algorithm
Page 641
Pavan et. al., Dynamic analysis of Single Machine Infinite Bus system using Single input and Dual input PSS
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 738-744
ISSN 2078-2365
Capacitors Natural Voltage Balancing Mechanism Investigation in
Flying Capacitor Multicell Converters
Vahid Dargahi and Abbas Shoulaie
Department of Electrical Engineering, Iran University of Science and Technology, Tehran 16846, Iran
vdargahi@elec.iust.ac.ir, shoulaie@iust.ac.ir
Abstract—This paper presents an analytical analysis of flying
capacitors voltage balancing process in flying capacitor multicell
converters. The analysis is based upon renowned theory of
Double Fourier series which leads to modeling and state-space
representation of converters. State-space representation of
converter can be utilized to investigate the transient and steady
states of internal flying capacitors voltages. To provide
verification, experimental results acquired from a laboratory
prototype are compared against numerical solution of differential
equations of converter state-space representation and simulation
results.
Keywords Flying Capacitor Multicell Converter, Double
Fourier Series, Self balancing.
I.
INTRODUCTION
As a consequence of reaching higher power and lack of its
suitable ranged switches, multilevel converters popped up in
1975 and have been continuously developed in recent years
due to the necessity of increase in power level of industrial
applications especially high power applications such as high
power AC motor drives, active power filters, reactive power
compensation and FACTS devices. The main reason is the
capability of these topologies to handle voltage/power in the
range of kilovolts/megawatts as a result of recent developments
in the area of high power semiconductors [1]-[5].
The concept of multilevel arises from acquiring a staircase
output voltage waveform as voltage levels from input dc
voltages by means of converter appropriate configuration and
its proper switching pattern. This staircase voltage by its
resemblance to sinusoidal voltage waveform leads to primitive
advantages of utilizing switches with low-voltage ratings,
higher power quality, lower total harmonic distortion, etc [1][5].
The term multilevel starts with the three-level converter
introduced by Nabae et al. The Neutral Point Clamped (NPC)
converter, presented in the early 80’s, is a standard topology in
industry on its 3-level version. However, for a higher number
of levels, this topology has some drawbacks such as: voltage
balance of the dc-link capacitors and the number of clamping
diodes [2]-[4].
Alternatives for the NPC converters are the multicell
topologies. Different cells and approaches to interconnect them
lead to many topologies which the most important ones are the
Cascaded Multicell (CM) and the Flying Capacitor Multicell
(FCM) accompanied by its sub-topology Stacked Multicell
(SM) converters [2][3][5].
The FCM converter, and its derivative, the SM converter,
have many advantageous properties for medium voltage
applications, particularly the transformer-less operation and the
ability to naturally maintain the flying capacitors voltages at
their target operating levels. This substantial property is called
natural balancing and allows the construction of such
converters with a large number of voltage levels. Natural selfbalancing of the flying capacitors voltages occurs without any
feedback control. A necessary condition for this phenomenon
is that average currents of the flying capacitors must be zero.
As a result, each cell must be controlled with the same duty
cycle and a regular phase shifted progression along the cells.
Generally, an output RLC filter (balance booster circuit), tuned
to the switching frequency or multiple of that, is suggested to
be connected across the load in order to accelerate this self
balancing process in the transient states [2][3][5].
The FCM converter uses a series connection of “cells”
comprising a flying capacitor and its associated complimentary
switch pair and produces a switched voltage that is the sum of
the individual cell states [2][3][5].
Despite of mentioned appreciable advantages, multilevel
converters possess some following main drawbacks: increased
number of isolated dc voltages, clamping diodes, capacitors
and of power semiconductor switches accompanied by their
related gating and protection circuits which result in a
sophisticated overall system [2][3][5].
As mentioned, voltage natural balancing mechanism is a
fundamental principle in flying capacitor and stacked multicell
converters which consents to construction of voltage levels at
the converter output [8]-[11]. The main objective of this paper
is to provide a mathematical model for flying capacitor
multicell converters intended for investigation of transient and
steady state of flying capacitors voltages accompanied by
taking into account the effect of balance booster circuit.
Page 738
Vahid and Shoulaie, Capacitors Natural Voltage Balancing Mechanism Investigation in Flying Capacitor Multicell Converters
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 738-744
ISSN 2078-2365
II. INSTANTANEOUS MODELING OF FLYING CAPACITOR
MULTICELL CONVERTERS IN STATE-SPACE REPRESENTATION
A. Fundamental Concepts of Flying Capacitors Multicell
Converters
Flying capacitor multicell converters (FCMCs) which have
been proposed by T.A. Meynard are relatively new breed of
multilevel converters in comparison with conventional neutral
point clamped (NPC) and cascade H-bridge (CHB) ones. A
typical configuration of FCMC is depicted in Fig 1. As
illustrated, R cells in a FCMC are overlapped to form a
required converter’s leg. Each cell consists of one voltage
source (a dc voltage source equal to E in Rth cell and
capacitors possessing specific voltages in remaining cells) and
two power semiconductor switches which are in
complementary state to each other to avoid short-circuiting of
voltage sources. Phase shifted carrier sinusoidal pulse width
modulation (PSCSPWM) technique is the most common
control scheme which is applied to switching strategy of
FCMCs to guaranty both best harmonic performance and
voltage balancing mechanism of flying capacitors. It should be
noted that in a R-cell FC converter each switch sustains just a
fraction of DC link voltage, i.e. E/R. This R-cell configuration
leads to R+1 levels of voltage with peak to peak voltage value
of E at the converter output. Flying capacitor multicell
converters are in preference to the NPC and CHB ones as
considering advantages such as: modularity, noninterdependency of cells as fault occurs and ease of reaching
higher voltage levels just by introducing new cells [2][3][5].
Switches states of a 4-cell-5-level FCMC and output
voltage, using PSCSPWM control method, are illustrated in
Table-1.
0.5E SR
0.5E
SR
S3
+
EC( R-1)= ( R-1)E/ R
S3
S2
+
EC2= 2E/R
S2
S1
+
EC1 = E/R
S1
Cell- R
Cell-2
+
l
r
Cell-1
_
Fig. 1. R-cell flying capacitor multicell converter with maximum output voltage
value of E.
Table-1: States of switches in a 4-cell-5-level conventional FCMC.
Output
Number of
State of Switches {(S4, S3, S2, S1)}
Voltage
States
Level
+E
{(1,1,1,1)}
1
+ 0.5E
{(1,1,1,0)(1,1,0,1)(1,0,1,1) (0,1,1,1)}
4
0
{(1,1,0,0)(1,0,0,1)(0,0,1,1)(1,0,1,0)(0,0,1,1)(0,1,0,1)}
6
-0.5E
{(1,0,0,0)(0,1,0,0)(0,0,1,0)(0,0,0,1)}
4
-E
{(0,0,0,0)}
1
B. Instantaneous Model of the Converter in the State-Space
Representation
By utilizing proper switching pattern in FCMCs, their
capacitor voltages would reach to the specific values which
allow to constructing the desired output voltage levels. This
property is known as natural voltage balancing mechanism and
is achieved using phase shifted carrier pulse width modulation
(PSCPWM) switching technique in the converter [8]-[11].
Natural self-balancing process of the flying capacitors
voltages, as one of the advantages of FCM converters occurs
without any feedback control. A necessary condition for this
phenomenon is that average currents of the flying capacitors
must be zero. As a result, each cell must be controlled with the
same duty cycle and a regular phase shifted progression along
the cells. Generally, an output RLC filter (balance booster
circuit), tuned to the switching frequency or multiple of that, is
suggested to be connected across the load in order to accelerate
this self balancing process in the transient states. In this case,
the dynamic of the self-balancing process depends on the
impedance of load at the switching frequency. If the impedance
at the switching frequency is high then the natural balancing is
very slow and vice versa. The output RLC filter should be
tuned to the switching frequency as follow [8]-[11]:
Lb Cb
1
2 fSW
(1)
Where, fSW is the switching frequency, Lb and Cb are
inductance and capacitance of the output RLC booster circuit,
respectively.
iout( t )
Vout
The output voltage of a R-cell FC converter has R+1-levels
and its frequency spectrum has the harmonics around the
R k fSW th harmonic where k and fSW are the integer
number and the switching frequency, respectively [2][3][5].
However in this section a mathematical model of the
converter will be presented to verify this balancing property.
This model describes time domain differential equations of the
flying capacitors voltages and state-space representation of the
converter. Utilization a numerical solution of differential
equations leads to acquire transient and steady state response of
flying capacitors voltages. According to Fig. 1, a switching
function of a cell is defined as follows [8]-[11]:
1
H t
1
if S is on
if S is off
1: R
(2)
And its double Fourier series expansion can be expressed as
follows [6]-[7]:
Page 739
Vahid and Shoulaie, Capacitors Natural Voltage Balancing Mechanism Investigation in Flying Capacitor Multicell Converters
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 738-744
ISSN 2078-2365
H (t ) M cos r t r
4
sin m n J n m M
2 2
m
1 2
m1 n
cos m c t c
n r t r
R
(3)
Where M , r , r , J n . , c , c are modulation index,
angular frequency of reference sinusoidal waveform, phase
angle of reference sinusoidal waveform, nth order Bessel
function of first kind, triangular carrier waveform angular
frequency and triangular carrier waveform phase angle,
respectively.
Output voltage and current of the converter and associated
differential equations of the flying capacitors voltages based on
mentioned switching functions can be obtained as follows [8][11]:
E
1 R1
H R (t ) H t H 1 t Ec
2
2 1
Vout
iout t
Vout t
(5)
ZL
1
H 1 t H t iout t
dt
2
dEc
C
(4)
(6)
Where, ZL is a series connection of resistance r and
inductance l and C is capacitance of flying capacitors.
Flying
capacitor
multicell
converter’s
state-space
representation can be written as follows:
Ec A Ec B E
1
Hi 1 t H i t j 1 t j t
4Ci
Bi1
R t
H i 1 t H i t
4Ci
ZL
r
r l
M
2
2
l
cos r t r tan 1 r
r
4
sin m n
2
2
2
m r mc nr l
( 1)2
m t c
c
R
m1 n
J m M cos n t
r
r
n 2
m
n
l
r
tan 1 c
r
(11)
1: R
To consider the balance booster circuit effect on the
dynamic of the self-balancing process following modification
should be applied to the acquired state-space equations:
Aij
1
H i 1 t H i t j 1 t j 1 t j t j t
4Ci
(12)
1
R t R t H i 1 t H i t
4Ci
(13)
i, j 1: R 1
Bi1
i 1: R 1
And:
(7)
Ec Ec1 Ec 2 Ec3 ...... Ec ( R1)
Aij
H t
t
T
i, j 1: R 1
i 1: R 1
(8)
(9)
(10)
And:
Page 740
Vahid and Shoulaie, Capacitors Natural Voltage Balancing Mechanism Investigation in Flying Capacitor Multicell Converters
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 738-744
ISSN 2078-2365
t
H t
Zb
M
1
2
Rb r Lb r Cb
r t r
r Lb r Cb 1
cos
1
2
tan
Rb
4
2
1
m R 2 m n L m n C
b
c
r
b
c
r
b
sin m n J n m M
2 2
m1 n
m t ( 1)2
c
c
R
cos n r t r
1
m n L m n C
r
b
c
r
b
c
1
tan
Rb
1: R
III.
(14)
Fig. 2. Transient and steady state of internal flying capacitors voltages of a 6cell-7-level FCMC acquired from state-space numerical solution and simulation
for resistive-inductive load in parallel with RLC booster circuit.
Table-3: Parameters used in simulation and numerical solution of state-space
representation of 6-cell-7-level FCMC.
System Parameters
Values
DC voltage (E)
600 V
Internal flying capacitors (C)
2200 uF
PS-SPWM carrier frequency (fSW)
5 kHz
Fundamental output voltage frequency
50 Hz
Modulation index
0.8
Resistive-inductive load (r-l)
10 Ω – 50 mH
Booster circuit(RLC)
10 Ω – 10 uH-101.32 uF
NUMERICAL AND SIMULATION RESULTS
To provide verification to the elaborated state-space
representation of the FCMC, numerical solution is utilized to
solve differential equations of 6-cell-7-level and 4-cell-5-level
converters. Transient and steady state of internal flying
capacitors voltages of mentioned converters acquired from
state-space numerical solution are shown in Figs. 2-6. System
parameters used in numerical solution are given in Tables 2-5.
Table-2: Parameters used in simulation and numerical solution of state-space
representation of 6-cell-7-level FCMC.
System Parameters
Values
DC voltage (E)
600 V
Internal flying capacitors (C)
560 uF
PS-SPWM carrier frequency (fSW)
5 kHz
Fundamental output voltage frequency
50 Hz
Modulation index
0.8
Resistive-inductive load (r-l)
10 Ω – 50 mH
Booster circuit(RLC)
20 Ω – 10 uH-101.32 uF
Fig. 3. Transient and steady state of internal flying capacitors voltages of a 6cell-7-level FCMC acquired from state-space numerical solution and simulation
for resistive-inductive load in parallel with RLC booster circuit.
Table-4: Parameters used in simulation and numerical solution of state-space
representation of 4-cell-5-level FCMC.
System Parameters
Values
DC voltage (E)
600 V
Internal flying capacitors (C)
2200 uF
PS-SPWM carrier frequency (fSW)
5 kHz
Fundamental output voltage frequency
50 Hz
Modulation index
0.8
Resistive-inductive load (r-l)
10 Ω – 50 mH
Booster circuit(RLC)
10 Ω – 10 uH-101.32 uF
Page 741
Vahid and Shoulaie, Capacitors Natural Voltage Balancing Mechanism Investigation in Flying Capacitor Multicell Converters
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 738-744
ISSN 2078-2365
Numerical solutions are in accordance with simulation
results and admit the model of the converter. It is worthwhile
noting that the value of resistor in the balance booster circuit
plays an important role in transient state of flying capacitors
voltages and its reduction accelerates the self balancing process
in flying capacitors and vice versa.
IV.
Fig. 4. Transient and steady state of internal flying capacitors voltages of a 4cell-5-level FCMC acquired from state-space numerical solution and simulation
for resistive-inductive load in parallel with RLC booster circuit.
Table-5: Parameters used in simulation and numerical solution of state-space
representation of 4-cell-5-level FCMC.
System Parameters
Values
DC voltage (E)
600 V
Internal flying capacitors (C)
2200 uF
PS-SPWM carrier frequency (fSW)
5 kHz
Fundamental output voltage frequency
50 Hz
Modulation index
0.9
Resistive-inductive load (r-l)
10 Ω – 50 mH
Booster circuit(RLC)
10 Ω – 10 uH-101.32 uF
EXPERIMENTAL RESULTS
To verify the mathematical model of FCMC, the measured
transient and steady state voltages of flying capacitors of a 3cell-4-level converter acquired from the prototype system,
illustrated in Fig. 7 and Fig. 9, are compared against simulation
results and numerical solution of the converter time-domain
differential equations, presented in Fig. 8 and Fig.10. Also
main parameters of the converter are given in Tables 6 and 7.
The match between simulation, numerical solution and
experimental results confirms the modified mathematical
model of FCMC.
Table-6: Parameters used in simulation, numerical solution, and experimental
converter.
System Parameters
Values
DC voltage (E)
75 V
Internal flying capacitors (C)
560 uF
PS-SPWM carrier frequency (fSW)
5 kHz
Fundamental output voltage frequency
50 Hz
Modulation index
0.8
Resistive-inductive load (RL- LL)
25 Ω – 30 mH
Booster circuit
24 Ω – 10 uH- 101.32 uF
Fig. 5. Transient and steady state of internal flying capacitors voltages
of a 4-cell-5-level FCMC acquired from state-space numerical solution
and simulation for resistive-inductive load in parallel with RLC booster
circuit.
Fig. 7. Transient and steady state of internal flying capacitors voltages of
a 3-cell-4-level FCMC acquired from state-space numerical solution and
simulation for resistive-inductive load in parallel with RLC booster
circuit.
Fig. 6. Transient and steady state of internal flying capacitors voltages of
a 4-cell-5-level FCMC acquired from state-space numerical solution and
simulation for resistive-inductive load in parallel with RLC booster
circuit for two different modulation indexes.
Page 742
Vahid and Shoulaie, Capacitors Natural Voltage Balancing Mechanism Investigation in Flying Capacitor Multicell Converters
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 738-744
ISSN 2078-2365
circuit.
Fig. 8. Transient and steady state of internal flying capacitors voltages of
a 3-cell-4-level FCMC acquired from experimental prototype for
resistive-inductive load in parallel with RLC booster circuit.
Table-7: Parameters used in simulation, numerical solution, and experimental
converter.
System Parameters
Values
DC voltage (E)
75 V
Internal flying capacitors (C)
560 uF
PS-SPWM carrier frequency (fSW)
5 kHz
Fundamental output voltage frequency
50 Hz
Modulation index
0.8
Resistive-inductive load (RL- LL)
25 Ω – 30 mH
Booster circuit
24 Ω – 10 uH- 101.32 uF
Fig. 10. Transient and steady state of internal flying capacitors voltages
of a 3-cell-4-level FCMC acquired from experimental prototype for
resistive-inductive load in parallel with RLC booster circuit.
V.
CONCLUSION
Multicell converters are very interesting for highpower/medium-voltage
applications,
for
considerably
improvement of the output voltage frequency spectrum and
reduction of the conduction losses, switching ripple and value
of dV/dt.
This paper presents a modified mathematical model of
flying capacitor multicell converters. In the proposed model the
effect of balance booster circuit which is usually connected in
parallel with load to accelerate the self balancing process of
flying capacitors, is also considered. Numerical solutions,
simulation and experimental results are in accordance with
each other and confirm the validity of proposed model for
FCMCs.
REFERENCES
[1]
Fig. 9. Transient and steady state of internal flying capacitors voltages of
a 3-cell-4-level FCMC acquired from state-space numerical solution and
simulation for resistive-inductive load in parallel with RLC booster
R. H. Baker and L. H. Bannister, “Electric power converter,” U.S. Patent
3 867 643, Feb. 1975.
Page 743
Vahid and Shoulaie, Capacitors Natural Voltage Balancing Mechanism Investigation in Flying Capacitor Multicell Converters
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 738-744
ISSN 2078-2365
[2]
[3]
[4]
[5]
[6]
[7]
J. S. Lai and F. Z. Peng, “Multilevel converters—A new breed of power
converters,” in Proc. IEEE Ind. Applicat. Soc. Annu. Meeting, 1995, pp.
2348–2356.
J. Rodriguez, J. Lai, and F. Z. Peng, “Multilevel inverters: A survey of
topologies, controls and applications,” IEEE Trans. Ind. Electron., vol.
49, no. 4, pp. 724–738, Aug. 2002.
A. Nabae, I. Takahashi, and H. Akagi, “A new neutral-point clamped
PWM inverter,” IEEE Trans. Ind. Applicat., vol. IA-17, pp. 518–523,
Sept./Oct. 1981.
T. A. Meynard and H. Foch, “Multi-level choppers for high voltage
applications,” Eur. Power Electron. Drives J., vol. 2, no. 1, p. 41, Mar.
1992.
H. S. Black, Modulation Theory. New York: Van Nostrand, 1953.
D. G. Holmes and T. A. Lipo, Pulse Width Modulation for Power
Converters: Principles and Practice. Piscataway, NJ: IEEE Press, 2003.
X. Yuan, H. Stemmler, and I. Barbi, “Self-balancing of the
clampingcapacitor- voltages in the multilevel capacitor-clampinginverter under sub-harmonic PWM modulation,” IEEE Trans. Power
Electron., vol. 16, no. 2, pp. 256–263, Mar. 2001.
[9] R.Wilkinson, H. du Mouton, and T. Meynard, “Natural balance of
multicell converters : The two-cell case,” IEEE Trans. Power Electron.,
vol. 21, no. 6, pp. 1649–1657, Nov. 2006.
[10] R.Wilkinson, H. du Mouton, and T. Meynard, “Natural balance of
multicell converters: The general case,” IEEE Trans. Power Electron.,
vol. 21, no. 6, pp. 1658–1666, Nov. 2006.
[11] B. P. McGrath and D. G. Holmes, “Analytical modelling of voltage
balance dynamics for a flying capacitor multilevel converter,” in Conf.
Rec. IEEE Power Electron. Specialists Conf. (PESC), 2007, pp. 968–
974.
[8]
Page 744
Vahid and Shoulaie, Capacitors Natural Voltage Balancing Mechanism Investigation in Flying Capacitor Multicell Converters
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 731-737
ISSN 2078-2365
Capacitors Voltage Balancing Modeling in Three Phase Flying
Capacitor Converters with Booster
Vahid Dargahi and Abbas Shoulaie
Department of Electrical Engineering, Iran University of Science and Technology, Tehran 16846, Iran
vdargahi@elec.iust.ac.ir, shoulaie@iust.ac.ir
Abstract— This paper provides a mathematical model for voltage
natural balancing process in three phase capacitor-clamped
multicell converters. The analysis leads to state-space model of
the converters. State-space representation of converter can be
utilized to investigate the start-up and steady states of internal
flying capacitors voltages. To provide verification, numerical
solutions for three phase capacitor-clamped multicell converter’s
analytic model are presented.
Keywords Three Phase Flying Capacitor Multicell Converter,
Self-balancing , Voltage Balance Booster.
I.
INTRODUCTION
As a consequence of reaching higher power and lack of its
suitable ranged switches, multilevel converters were introduced
in 1975 and have been continuously developed in recent years
due to the necessity of increase in power level of industrial
applications especially high power applications such as high
power AC motor drives, active power filters, reactive power
compensation and FACTS devices. The main reason is the
capability of these topologies to handle voltage/power in the
range of kilovolts/megawatts as a result of serial connection of
power switches in these converters [1].
The concept of multilevel arises from acquiring a staircase
output voltage waveform as voltage levels from input dc
voltages by means of converter appropriate configuration and
its proper switching pattern [2]-[3].
In comparison with the conventional two-level converters,
multilevel ones excel at producing an output voltage
comprising several steps with considerable enhancements to
power quality, harmonic content, and efficiency. Multilevel
converters have other appreciable advantages such as lower
switching losses, lower voltage ratings of used semiconductor
switches, reduction of output dv/dt stress and filter inductance,
etc. [4]-[6].
The term multilevel starts with the three-level converter
introduced by Nabae et al. The Neutral Point Clamped (NPC)
converter, presented in the early 80’s, is a standard topology in
industry on its 3-level version. However, for a higher number
of levels, this topology has some drawbacks such as: voltage
imbalance issue of the dc-link capacitors and the excessive use
of clamping diodes [7].
Cascade multicell (CM) converters use a series connection
of H-bridge modules. Dc link voltage of each H-bridge module
must be an isolated one. Modularity, ease of extension the
number of output voltage levels via adding new modules,
reliability, and fault tolerant feature are the most notable
advantages of these topologies [8]-[9].
Flying capacitor (FC) based converters use ladder
connection of units called as ‘cells’. Each cell is composed of
one flying capacitor and two complimentary power switches.
Redundant switching states in flying capacitor based converters
can be implemented to stabilize the voltage across flying
capacitors at their requisite values. The difference between
voltages across two adjacent flying capacitors determines the
step value of staircase-form output voltage [10]-[12].
The FCM converter, and its derivative, the SM converter,
have many advantageous properties for medium voltage
applications, particularly the transformer-less operation and the
ability to naturally maintain the clamping capacitors voltages at
their target operating levels. This substantial property is called
natural balancing and allows the construction of such
converters with a large number of voltage levels. Natural selfbalancing of the flying capacitors voltages occurs without any
feedback control. A necessary condition for this phenomenon
is that average current of the clamping capacitors must be zero.
As a result, each cell must be controlled with the same duty
cycle and a regular phase shifted progression along the cells.
Generally, an output RLC filter (balance booster circuit), tuned
to the switching frequency or multiple of that, is suggested to
be connected across the load in order to accelerate this selfbalancing process in the transient states. The FCM converter
uses a series connection of “cells” comprising a flying
capacitor and its associated complimentary switch pair and
produces a switched voltage that is the sum of the individual
cell states [13].
Despite of mentioned appreciable advantages, multilevel
converters possess some following main drawbacks: increased
number of isolated dc voltage sources, clamping diodes,
capacitors and of power semiconductor switches accompanied
Page 731
Vahid and Shoulaie, Capacitors Voltage Balancing Modeling in Three Phase Flying Capacitor Converters with Booster
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 731-737
ISSN 2078-2365
by their related gating and protection circuits which result in a
sophisticated overall system [14].
As mentioned, voltage natural balancing mechanism is a
fundamental principle in flying capacitor and stacked multicell
converters which consents to construction of voltage levels at
the converter output [11]-[21]. The main objective of this paper
is to provide a mathematical model for three phase flying
capacitor multicell converters with intention for investigation
of start-up and steady state of clamping capacitors voltages by
taking into account the effect of balance booster circuit.
0.5E SR
0.5E
SR
S3
+
EC( R-1)= ( R-1)E/ R
S3
S2
+
EC2= 2E/R
S2
S1
+
EC1 = E/R
S1
iout( t )
+
l
Vout
Cell- R
Cell-2
r
Cell-1
_
Fig. 1. R-cell flying capacitor multicell converter with maximum output voltage
value of E.
II. INSTANTANEOUS MODELING OF THREE PHASE FLYING
CAPACITOR MULTICELL CONVERTERS IN STATE-SPACE
REPRESENTATION
A. Fundamental Concepts of Flying Capacitors Multicell
Converters
Flying capacitor multicell converters (FCMCs) which have
been proposed by T.A. Meynard are relatively new breed of
multilevel converters in comparison with conventional neutral
point clamped (NPC) and cascade H-bridge (CHB) ones. A
typical single phase configuration of FCMC is depicted in Fig
1. As illustrated, R cells in a FCMC are overlapped to form a
required converter’s leg. Each cell consists of one voltage
source (a dc voltage source equal to E in Rth cell and capacitors
possessing specific voltages in remaining cells) and two power
semiconductor switches which are in complementary state to
each other to avoid short-circuiting of voltage sources. Phase
Shifted Carrier Sinusoidal Pulse Width Modulation (PSCSPWM) technique is the most common control scheme which
is applied to switching strategy of FCMCs to guaranty both
best harmonic performance and voltage balancing mechanism
in clamping capacitors. It should be noted that in a R-cell
FCMC each switch sustains just a fraction of DC link voltage,
i.e. E/R. This R-cell configuration leads to R+1 levels of
voltage with peak to peak value of E at the converter output.
FCMCs are in preference to the NPC and CHB ones as
considering appreciable advantages such as: modularity, noninterdependency of cells as fault occurs and ease of reaching
higher voltage levels just by introducing new cells [1]-[10].
Control strategy, switches states and output voltage of a 4cell-5-level FCMC are illustrated in Fig. 2 and Table-1,
respectively.
Fig. 2. Switching pattern of a typical 4-cell 5-level FCMC, switches states and
output voltage.
Table-1: States of switches in a 4-cell-5-level conventional FCMC.
Output
Number
State of Switches {(S4, S3, S2, S1)}
Voltage
of States
Level
+0.5E
{(1,1,1,1)}
1
+ 0.25E
{(1,1,1,0)(1,1,0,1)(1,0,1,1) (0,1,1,1)}
4
0
{(1,1,0,0)(1,0,0,1)(0,0,1,1)(1,0,1,0)(0,0,1,1)(0,1,0,1)}
6
-0.25E
{(1,0,0,0)(0,1,0,0)(0,0,1,0)(0,0,0,1)}
4
-0.5E
{(0,0,0,0)}
1
The output voltage of a R-cell FCM converter has R+1levels and its frequency spectrum has the harmonics around the
R k fSW th harmonic where k and fSW are the integer
number and the switching frequency, respectively [11]-[14]. A
typical three phase clamping capacitor multicell converter.is
shown in Fig 3.
Page 732
Vahid and Shoulaie, Capacitors Voltage Balancing Modeling in Three Phase Flying Capacitor Converters with Booster
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 731-737
ISSN 2078-2365
0.5E SR,a
O
0.5E
+
EC(R-1),a= (R-1)E/R
-
SR,a
S3,a
S3,a
Cell-R,a
SR,b
+
EC(R-1),b= (R-1)E/R
-
SR,b
S3,b
S3,b
Cell-R,b
SR,c
+
EC(R-1),c= (R-1)E/R
-
SR,c
S3,c
S3,c
S2,a
+
E C2,a = 2E/R
S2,a
S1,a
+
E C1,a = E /R
S1,a
Cell-2,a
Cell-1,a
S2,b
+
E C2,b= 2E/R
S2,b
S1,b
+
E C1,b= E /R
S1,b
Cell-2,b
Cell-1,b
S2,c
+
E C2,c= 2E/R
S2,c
S1,c
+
E C1,c = E /R
S1,c
ia (t)
r
l
+
Va (t)
_
ib (t)
r
l
+
Vb(t)
_
ic (t)
r
l
+
Vc (t)
_
balancing property. In the derived model, capacitors voltages
are state variables.
Utilization a numerical solution for proposed model
differential equations leads to acquire transient and steady state
response of flying capacitors voltages. According to Fig. 3,
switching function of cell in each phase of the converter is
defined
as follows [15]-[21]:
n
1
H , x t
1
if S , x is on
if S , x is off
1: R, x a , b, c
(2)
And its Double Fourier Series expansion can be expressed
as follows [15]:
H , x (t ) M cos
4
Fig. 3. Three phase R-cell flying capacitor multicell converter with maximum output
sin m n J n m M cos m n
2 2
voltage value of E in each phase.
m 1 n m
Cell-2,c
Cell-R,c
Cell-1,c
B. Instantaneous Model of the Three Phase Converter in the
State-Space Representation
By utilizing proper switching pattern in FCMCs, their
capacitor voltages would reach to the specific values which
allow to constructing the desired output voltage levels. This
property is known as voltage natural balancing mechanism and
is achieved using PSC-SPWM switching technique in these
converters [11]-[21].
Natural self-balancing process of the clamping capacitors
voltages, as one of the advantages of FCMCs occurs without
any feedback control. A necessary condition for this
phenomenon is that average current of the flying capacitors
must be zero. As a result, each cell must be controlled with the
same duty cycle and a regular phase shifted progression along
the cells. Generally, an output RLC filter (balance booster
circuit), tuned to the switching frequency or multiple of that, is
suggested to be connected across the load in order to accelerate
this self-balancing process in the start-up states. In this case,
the dynamic of the self-balancing process depends on the
impedance of balance booster circuit at the switching
frequency. If the impedance at the switching frequency is high
then the natural balancing is very slow and vice versa. The
output RLC filter should be tuned to the switching frequency
as follow [1]:
Lb Cb
1
2 fSW
(1)
Where, fSW is the switching frequency, Lb and Cb are
inductance and capacitance of the output RLC booster circuit,
respectively.
In this section a mathematical model for three phase
capacitor-clamped converter will be presented to verify natural
c t c
1 2 , t
r
r
x
0
xa
2
x
xb
3
4
xc
3
R
(3)
where M , r , r , J n . , c , c are modulation index,
angular frequency of reference sinusoidal waveform, phase
angle of reference sinusoidal waveform, nth order Bessel
function of first kind, triangular carrier waveform angular
frequency and triangular carrier waveform phase angle,
respectively.
According to Fig. 3, output voltage and current of the
converter based on mentioned switching functions can be
obtained as follows:
Vao t Van t Vno t
Vbo t Vbn t Vno t
Vco t Vcn t Vno t
Vao t Vbo t Vco t
(4)
Van t Vbn t Vcn t 3Vno t
0
Van t
2 1 1 Vao t
1
Vbn t 1 2 1 Vbo t
3 1 1 2
Vco t
Vcn t
(5)
Page 733
Vahid and Shoulaie, Capacitors Voltage Balancing Modeling in Three Phase Flying Capacitor Converters with Booster
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 731-737
ISSN 2078-2365
ia t
T
Vao t Vbo t Vco t
i
t
.
,
,
b
ZL
ZL
ZL
ic t
Vxo t
ix t
C , x
C , a
(6)
E
1 R1
H R, x (t ) H , x t H 1, x t Ec , x t
2
2 1
Vxn t
ZL
dEc , x t
dt
dEc , a t
1
H 1, x t H , x t ix t
2
(8)
1
H 1,a t H ,a t
2 ZL
2 E
2 1 R1
H R, a (t ) H j , a t H j 1, a t Ecj , a t
3 2 j 1
3 2
1 E
1 1 R1
H R,b (t ) H j ,b t H j 1,b t Ecj ,b t
3 2 j 1
3 2
1 E
1 1 R1
H R,c (t ) H j ,c t H j 1,c t Ecj ,c t
3 2 j 1
3 2
C , a
dt
dEc , a t
R, a t
H
dt
(7)
(9)
H
1, a
t H ,a t E
R,c t
H
1, a
t H ,a t E
R 1
1
H 1, a t H , a t j ,c t j 1,c t Ecj , c t
12
j 1
12
x y :
1
H i 1, x t H i, x t j , x t j 1, x t
6Ci , x
Aij , xy t
x y :
1
H i 1, x t H i, x t j , y t j 1, y t
12Ci , x
x, y a , b, c
i, j 1: R 1
, x t
(11)
Eci , a t Bi1, a t
Eci ,b t Bi1,b t E
Eci , c t Bi1, c t
and:
R 1
1
H 1, a t H , a t j ,b t j 1,b t Ecj ,b t
12
j 1
12
Aij , ac t
Aij ,bc t
Aij , cc t
6
R 1
1
H 1, a t H , a t j , a t j 1, a t Ecj , a t
6
j 1
R,b t
Aij , ab t
Aij ,bb t
Aij , cb t
2 R, x t
1
Bi1, x t
H i 1, x t H i, x t R, y t
12Ci , x
y|
x y
i, j 1: R 1
x, y a , b, c
1, a t H , a t E
dEci , a t
dt A t
dEci ,b t ij , aa
Aij ,ba t
dt
dE t Aij , ca t
ci , c
dt
(10)
Where, ZL is a series connection of resistance r and
inductance l and C is capacitance of flying capacitors.
Three phase clamping capacitor multicell converter’s statespace representation can be written as follows:
H ,x t
ZL
(13)
M
cos tan 1 0,1
r 0,1
4
sin m n J n m M
2 2
m
r
m, n
m1 n
1
cos m n tan m, n
(12)
m, n
m, n
1
2
m, n 2
(14)
mc nr l
r
To consider the balance booster circuit effect on the
dynamic of the self-balancing process following modification
should be applied to the acquired state-space equations:
Page 734
Vahid and Shoulaie, Capacitors Voltage Balancing Modeling in Three Phase Flying Capacitor Converters with Booster
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 731-737
ISSN 2078-2365
x y :
j , x t j , x t
1
H i 1, x t H i , x t
6Ci , x
t j 1, x t
j
1,
x
Aij , xy t
x
y
:
t j , y t
1
j, y
12C H i 1, x t H i , x t
j 1, y t j 1, y t
i, x
2 R, x t R , x t
1
H i 1, x t H i , x t
R, y t R , y t
12Ci , x
y|
x y
i, j 1: R 1
x, y a , b, c
Bi1, x t
, x t
H ,x t
Zb
M
cos tan 1 0,1
Rb 0,1
m, n
m, n
1
m, n 2
mc nr Lb mc nr Cb
2
(16)
(17)
1
Rb
III.
Fig. 4. Transient and steady state of internal flying capacitors voltages of a 2cell-3-level three phase FCMC acquired from state-space numerical solution for
resistive-inductive load in parallel with RLC booster circuit.
4
sin m n J n m M
2 2
m Rb m, n
m1 n
1
cos m n tan m, n
(15)
NUMERICAL SOLUTION RESULTS
To provide verification to the elaborated state-space
representation of the three phase FCMCs, numerical solution is
utilized to solve differential equations of 2-cell-3-level and 3cell-4-level three phase converters. Transient and steady state
of internal flying capacitors voltages of mentioned converters
acquired from state-space numerical solution are shown in
Figs. 4-9. System parameters used in numerical solution are
given in Tables 2-4.
Fig. 5. Start-up state of internal flying capacitors voltages of a 2-cell-3-level
three phase FCMC acquired from state-space numerical solution for resistiveinductive load in parallel with RLC booster circuit.
Table-3: Parameters used in numerical solution of state-space representation
of 2-cell-3-level three phase FCMC.
System Parameters
Values
DC voltage (E)
600 V
Internal flying capacitors (C)
560 uF
PSCSPWM carrier frequency (fSW)
5 kHz
Fundamental output voltage frequency
50 Hz
Modulation index
0.8
Resistive-inductive load (r-l)
10 Ω – 80 mH
Booster circuit(RLC)
200 Ω – 6.97 uH-144 uF
Table-2: Parameters used in numerical solution of state-space representation
of 2-cell-3-level three phase FCMC.
System Parameters
Values
DC voltage (E)
600 V
Internal flying capacitors (C)
560 uF
PSCSPWM carrier frequency (fSW)
5 kHz
Fundamental output voltage frequency
50 Hz
Modulation index
0.8
Resistive-inductive load (r-l)
10 Ω – 80 mH
Booster circuit(RLC)
10 Ω – 6.97 uH-144 uF
Page 735
Vahid and Shoulaie, Capacitors Voltage Balancing Modeling in Three Phase Flying Capacitor Converters with Booster
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 731-737
ISSN 2078-2365
Fig. 6. Transient and steady state of internal flying capacitors voltages of a 2cell-3-level three phase FCMC acquired from state-space numerical solution
for resistive-inductive load in parallel with RLC booster circuit.
Fig. 8. Transient and steady state of internal flying capacitors voltages of a 3cell-4-level three phase FCMC acquired from state-space numerical solution for
resistive-inductive load in parallel with RLC booster circuit.
Fig. 7. Start-up state of internal flying capacitors voltages of a 2-cell-3-level
three phase FCMC acquired from state-space numerical solution for resistiveinductive load in parallel with RLC booster circuit.
Fig. 9. Start-up state of internal flying capacitors voltages of a 3-cell-4-level
three phase FCMC acquired from state-space numerical solution for resistiveinductive load in parallel with RLC booster circuit.
Table-4: Parameters used in numerical solution of state-space representation
of 3-cell-4-level three phase FCMC.
System Parameters
Values
DC voltage (E)
600 V
Internal flying capacitors (C)
560 uF
PSCSPWM carrier frequency (fSW)
5 kHz
Fundamental output voltage frequency
50 Hz
Modulation index
0.8
Resistive-inductive load (r-l)
10 Ω – 50 mH
Booster circuit(RLC)
20 Ω – 6.97 uH-144 uF
Numerical solutions predict the start-up and steady state of
clamping capacitors voltages precisely and admit the proposed
model for three phase clamping capacitor multicell converters.
It is worth noting that the value of resistor in the balance
booster circuit plays an important role in start-up state of flying
capacitors voltages and its reduction accelerates the selfbalancing process and vice versa.
IV.
CONCLUSION
Clamping capacitor multicell converters are very interesting
for high-power/medium-voltage applications, for considerably
improvement of the output voltage frequency spectrum, voltage
natural balancing of clamping capacitors and fault tolerance.
This paper presents a mathematical model for three phase
clamping capacitor multicell converters. In the proposed model
the effect of balance booster circuit which is usually connected
in parallel with load to accelerate the self-balancing process of
flying capacitors, is considered which has not been reported in
Page 736
Vahid and Shoulaie, Capacitors Voltage Balancing Modeling in Three Phase Flying Capacitor Converters with Booster
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 731-737
ISSN 2078-2365
literature. Numerical solutions confirm the validity of proposed
model for three phase flying capacitor multicell converters.
REFERENCES
A. K. Sadigh, V. Dargahi, and A. Shoulaie, “Elimination of one DC
voltage source in stacked multicell converters,” IET. Power. Electron, in
press.
[2] J. S. Lai and F. Z. Peng, “Multilevel converters—A new breed of power
converters,” in Proc. IEEE Ind. Applicat. Soc. Annu. Meeting, 1995, pp.
2348–2356.
[3] J. Rodriguez, J. Lai, and F. Z. Peng, “Multilevel inverters: A survey of
topologies, controls and applications,” IEEE Trans. Ind. Electron., vol.
49, no. 4, pp. 724–738, Aug. 2002.
[4] J. Rodriguez, S. Bernet, B. Wu, J. Pontt, and S. Kouro, “Multi-level
voltage-source-converter topologies for industrial medium-voltage
drives,” IEEE Trans. Ind. Electron., vol. 54, no. 6, pp. 2930–2945, Dec.
2007.
[5] V. Dargahi, A. Shoulaie, “Capacitors natural voltage balancing
mechanism investigation in flying capacitor multicell converters,” in
Proc. 19th Iranian Conference on Electrical Engineering (ICEE), pp. 1–
6, Tehran, 2011.
[6] A. Nabae, I. Takahashi, and H. Akagi, “A new neutral-point clamped
PWM inverter,” IEEE Trans. Ind. Applicat., vol. IA-17, pp. 518–523,
Sept./Oct. 1981.
[7] S. Kouro, M. Malinowski, K. Gopakumar, J. Pou, L. G. Franquelo,
B.Wu, J. Rodriguez, M. A. Perez, and J. I. Leon, “Recent advances and
industrial applications of multilevel converters,” IEEE Trans. Ind.
Electron., vol. 57, no. 8, pp. 2553–2580, Aug. 2010.
[8] T. A. Meynard and H. Foch, “Multi-level choppers for high voltage
applications,” Eur. Power Electron. Drives J., vol. 2, no. 1, p. 41, Mar.
1992.
[9] L. G. Franquelo, J. Rodríguez, J. I. Leon, S. Kouro, R. Portillo, and M.
A. M. Prats, “The age of multilevel converters arrives,” IEEE Ind.
Electron. Mag., vol. 2, no. 2, pp. 28–39, Jun. 2008.
[10] A. Shukla, A. Ghosh, and A. Joshi, “Natural balancing of flying
capacitor voltages in multicell inverter under PD carrier-based pwm,”
IEEE Trans. Power Electron., vol. 26, no. 6, pp. 1682– 1693, Jun. 2011.
[1]
[11] M. Salehifar, V. Dargahi, and A. Shoulaie, “A series active power filter
using a flying capacitor multicell inverter,” in Proc. 1st Power
Electronic & Drive Systems & Technologies Conference (PEDSTC), pp.
191–195, Tehran, 2010.
[12] V. Dargahi, M. Salehifar, M. Abarzadeh, and A. Shoulaie, “Grid
interaction of DG units with a modified mixed cascade flying capacitor
multicell inverter,” in Proc. 2nd Power Electronic & Drive Systems &
Technologies Conference (PEDSTC), pp. 1–6, Tehran, 2011.
[13] V. Dargahi, M. Abarzadeh, M. Salehifar, and A. Shoulaie, “Voltage
compensation by a modified mixed cascade flying capacitor multicell
inverter,” in Proc. 1st Power Quality Conference (PQC), pp. 1–5,
Tehran, 2010.
[14] V. Dargahi, M. J. Zandzadeh, M. Salehifar, and A. Shoulaie, “Utilization
a flying capacitor multicell converter based adaptive shunt active power
filter to enhance power quality,” in Proc. 25th International power
System Conference (PSC), pp. 1–9, Tehran, 2010.
[15] D. G. Holmes and T. A. Lipo, Pulse Width Modulation for Power
Converters: Principles and Practice. Piscataway, NJ: IEEE Press, 2003.
[16] X. Yuan, H. Stemmler, and I. Barbi, “Self-balancing of the
clampingcapacitor- voltages in the multilevel capacitor-clampinginverter under sub-harmonic PWM modulation,” IEEE Trans. Power
Electron., vol. 16, no. 2, pp. 256–263, Mar. 2001.
[17] R.Wilkinson, H. du Mouton, and T. Meynard, “Natural balance of
multicell converters : The two-cell case,” IEEE Trans. Power Electron.,
vol. 21, no. 6, pp. 1649–1657, Nov. 2006.
[18] R.Wilkinson, H. du Mouton, and T. Meynard, “Natural balance of
multicell converters: The general case,” IEEE Trans. Power Electron.,
vol. 21, no. 6, pp. 1658–1666, Nov. 2006.
[19] B. P. McGrath and D. G. Holmes, “Analytical modelling of voltage
balance dynamics for a flying capacitor multilevel converter,” IEEE
Trans. Power Electron., vol. 23, no. 2, pp. 543–550, Mar. 2008.
[20] B. P. McGrath and D. G. Holmes, “Enhanced voltage balancing of a
flying capacitor multilevel converter using phase disposition (PD)
modulation,” IEEE Trans. Power Electron., vol. 26, no. 7, pp. 1933–
1942, Jul. 2011.
[21] B. McGrath and D. Holmes, “Analytical determination of the capacitor
voltage balancing dynamics for three phase flying capacitor converters,’’
in Conf. Rec. 42nd IEEE IAS Annu. Meeting, Sep. 23–27, 2007, pp.
1974–1981.
Page 737
Vahid and Shoulaie, Capacitors Voltage Balancing Modeling in Three Phase Flying Capacitor Converters with Booster
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 757-763
ISSN 2078-2365
Estimation of Re-striking Transient Over voltages in a
132KV Gas insulated Substation
M. Kondalu1, P.S. Subramanyam2
Electrical & Electronics Engineering, JNT University. Hyderabad.
1
Kondalu_m@yahoo.com
2
subramanyamps@gmail.com
Joginpally B.R. Engineering College,
Moinabad, Hyderabad
Abstract: --This paper presents the most significant results of a Very fast
transient Overvoltages generated due to switching operations have been
analyzed and presented. Since the contact speed of dis-connector switches
is low, re-striking occurs many times before the current interruption is
completed .Each re-striking generates transient overvoltage with level of
magnitude. These transient have travelling wave behaviour, they travel to
the external systems through enclosures, bushings, cable joints etc.. and
cause damage to the outside equipment. They can lead secondary break
downs in GIS and may give rise to electromagnetic interference. The
Earth faults give rise to TEV which can interfere with the operation and
control of secondary equipment in a 3-phase 132kv GIS. Thus it is
important to develop a suitable MATLAB7.8 models for estimation of
these overvoltages.
Keywords—Gas Insulated Substation (GIS), very fast Transient
overvoltages, 3phase faults, MATLAB 7.8 software and Control
circuitry
I. INTRODUCTION
For accurate analysis of transients, it is essential to
find the VFTO’s and circuit parameters. Due to the travelling
nature of the transients the modelling of GIS makes use of
electrical equivalent circuits composed by lumped elements
and especially by distributed parameter lines, surge impedances
and travelling times. The simulation depends on the quality of
the model of each individual GIS component. In order to
achieve reasonable results in GIS structures highly accurate
models for each internal equipment and also for components
connected to the GIS are necessary.
The dis-connector spark itself has to be taken into
account by transient resistance according to the Toepler’s
equation and subsequent arc resistance of a few ohms. The
wave shape of the over voltage surge due to dis-connector
switch is affected by all GIS elements. Accordingly, the
simulation of transients in GIS assumes an establishment of the
models for the Bus, Bushing, Elbow, Transformers, Surge
Arresters, Breakers, Spacers, dis-connectors, and Enclosures
and so on.
A GIS system comprising of an Input Cable, Spacer,
Dis-connector Switch, Bus bar of 10mts length and load has
been considered for modelling into electrical network and
analysis. The Fast Transient Over voltage waveform generated
during Closing and Opening operation of Dis-connector Switch
and 3-phase faults has been considered for calculations.
Spacers are simulated by lumped Capacitance. The
Inductance of the bus duct is calculated from the diameters of
Conductor and Enclosure. Capacitances are calculated on the
basis of actual diameters of inner and outer cylinders of central
conductor and outer enclosure. Cone Insulators used for
supporting inner conductor against outer enclosure are assumed
to be disk type for approximate calculation of spacer
capacitance.
The busduct can be modelled as a series of Pi-network
or as sequence parameters. However in this model, it is
considered as distributed Pi-network. The Schematic Diagram
of a Typical Gas Insulated System (GIS) is shown in below
figure 3.
Frequency nature, the VFTO imposed on the transformers
connected directly to the GIS would not be distributed evenly
on all transformer windings. Some windings, e.g. the first few
turns connecting to the 132kV GIS, would be subject to a
higher magnitude of overvoltage, posing a potential risk of
insulation breakdown of the transformers[9][10].
Page 757
Kondalu and Subramanyam, Estimation of Re-striking Transient Overvoltages in a 132KV Gas insulated Substation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 757-763
ISSN 2078-2365
II. MODELLING
OF 132KV GAS INSULATED
SUBSTATION
During the current operation of dis-connector switch in a
GIS, re-strikes(pre-strikes) occur because of low speed of the
dis-connector switch moving contact, hence Very fast
Transient Over voltage are developed. These VFTO’s are
caused by switching operations and 3-phase fault
When a dis-connector switch is opened on a floating
section of switchgear, trapped charge may be left on the
floating section. In the opening operation of dis-connector
switch, transients are produced and the magnitude of these
transients and rise times depends on the circuits parameters.
When there is a fault occurs, there is a short circuit in the
system. Transients are also produced due to the faults in the
system. Due to this VFTO’s are caused by switching operation
can also lead to secondary breakdown with in GIS. Re-striking
surges generated by the dis-connector switches at GIS
generally possess extremely high frequencies ranging from
several hundred KHz to several MHz . For the development of
equivalent circuit low voltage step response measurements of
the main GIS components have been made. Using MATLAB
7.8 of the equivalent models is developed.
During opening of Dis- connector switch (DS),
transients are produced due to internal oscillations. The
magnitude of transients and rise times depends on the circuit
parameters like inductance, Capacitance and connected Load.
Assuming that some trapped charge is left is left during
operating operation; transients can be calculated during closing
operation of DS.
Fast transient over voltages generated during Dis-connector
Switch operation are a sequence of voltage of voltage steps
created by voltage collapse across the gas at re-striking specific
over voltage shape is formed by multiple reflections and
refractions. Operation of Dis-connector Switch (DS) can be
shown by using the fig 1
Where
L1 = Inductance of Source
C1 = Capacitance of Source
C2 = Capacitance of GIS Open part
U1 = Power Frequency Voltage
U2 = Power GIS Voltage
The more frequent service situation of the isolator is its
use to connect or dis-connect unloaded parts of the installation
as is shown in fig 1 for example apart of the of the GIS is disconnected by an isolator from an overhead supply line. Where
by the self-capacitance C2 of this part of circuit can be upto
several nF, depending on its length. First re-strike across the
gap occurs when voltage across the gap exceeds the breakdown
voltage. The occurrence of re-strikes is described with the
following Fig 2
The voltage across the gap is the difference between U1
and U2, if it is assumed that the breakdown voltage UB of the
gap increases with increasing separation and therefore with
time as shown in fig 2.Then the curve U2 can be constructed as
follows. At the instant of mechanical contact separation, U1
and U2 have the same value, the voltage U2 continues to retain
this value, while U1 changes with power frequency, the voltage
(U2-U1)
Across the gap of the isolator also changes. As soon as,
(U2-U1) exceeds the dielectric strength UB of the gap, a
breakdown and thus first re-strike occurs. Both electrodes are
there by electrically connected by conducting spark, whereby
GIS section with initial voltage U2 is very rapidly charged to
instantaneous value of U1. The transient current flowing
through the spark then interrupts as soon as the GIS have been
charged to U1 and spark extinguishes.
The voltage U2 now remains constant with time,
while the voltage U1.on the side of supply keeps changing.
This continues until the second re- strike occurs with an
increased breakdown voltage UB as a consequence of larger
separation. Hence U2 follows U1, until finally at the end of the
switching process the gap no longer can be broken down.
Transients are also produced due to faults in the system. When
there is a fault, there will be short circuit in the system. Due to
this, oscillations occur due to presence of inductance and
capacitance on both sides of the fault section causing
transients.
Dis-connector Switch (DS) operation typically
involves slow moving contacts which results in numerous
discharges during operation .For example, a floating section of
switchgear between a disconnect switch and an open breaker
(load side may be disconnected from an energized Gas
insulated system (supply side).
For capacitive currents below—1 amp, are-strike occurs
every time the voltage between the connects exceeds the
dielectric strength of the gaseous medium between them.
Each re-strike generates a spark, which equalizes the
potentials between the switch contacts. Following spark
extinction, the supply and load side potentials will deviate
according to the AC supply voltage variation and the discharge
characteristics of the load side respectively. Another spark will
result when the voltage across the electrode gap dependent
breakdown voltage UB and the potential difference of the load
and supply side U.
Page 758
Kondalu and Subramanyam, Estimation of Re-striking Transient Overvoltages in a 132KV Gas insulated Substation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 757-763
ISSN 2078-2365
Each Dis-connector Switch (DS) operation generates a
large number of ignitions between the moving contacts. The
number of ignitions depends on the speed of the contacts. The
largest and steepest surge voltages are generated only by those
breakdowns at the largest contact gap. Therefore, only by those
breakdowns (10-50) need be considered for dielectric purpose.
A. Calculation of variable arc resistance
The Variable arc resistance is calculated using the
formula:
∫
Where,
= Toepler’s constant
= 0.005 volt.sec/mt for SF6 under uniform field
conditions
L = spark length in meters
= Initial charge or charge at the instant of breakdown
T = spark collapse time in sec.
Fig. 1 Electric Circuit for explaining re-strikes
Fig 2 Voltage of the open –ended GIS side of the Isolator
Page 759
Kondalu and Subramanyam, Estimation of Re-striking Transient Overvoltages in a 132KV Gas insulated Substation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 757-763
ISSN 2078-2365
Fig. 3 Schematic diagram of a typical Gas Insulated Substation
Fig. 4. 132kv GIS MATLAB implemented circuit when CB in Re-Striking Condition
The value of time varying spark resistance R(t) is calculated three post type spacers and a 132kv gas bushing containing
until it reaches a value of 1 to 3 ohms. The integral in the stress capacitor.
denominator sums up the absolute value of current ‘i’ through
.
the resistance R(t) over the time beginning at breakdown
inception. Thus, it corresponds to the charge conducted through
the spark channel up to time‘t’.
III. 3-PHASE EQUIVALENT CIRCUIT FOR 132KV
The busduct can be modelled as a series of Pi-network
GIS SYSTEM FOR 10MTS LENGTH
or as sequence parameters. However in this model, it is
considered as distributed Pi-network. The Schematic Diagram
The bus duct is dividing into three sections of length
of a Typical Gas Insulated System (GIS) is shown in below
1mts, 4mts, 5mts from load side. The GIS bushing is
figure 3.
represented by a capacitance of 125pf. The resistance of 1 ohm
Assuming that some trapped charge is left on the floating
spark channel is connected in series with circuit breaker.
section of switchgear during opening operation of disMATLAB Circuit for 10 mts. length in a 3-phase 132kv GIS
connector switch, a voltage of certain value is considered
shown in the fig. 4.
during MATLAB
Due to trapped charge some voltage remains on the floating
The apparatus as disconnected with an earthing switch, three section which can create severe conditions because the first redisc type Spencer’s , a load bus bar above to 10mts long width strike can occur at the peak of power frequency voltage giving
a voltage of 2.0 p.u. On re-strike the voltage on each side will
Page 760
Kondalu and Subramanyam, Estimation of Re-striking Transient Overvoltages in a 132KV Gas insulated Substation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 757-763
ISSN 2078-2365
collapse initially zero and hence creating two 1.0pu voltage
steps of opposite polarities. In this, it is assumed that restriking is created at 1.0 p.u. respectively on either side of disconnector switch (DS). The transients due to closing of the
circuit breaker are calculated and maximum voltage obtained
with a rise time.
This method implemented on MATLAB 7.8. the voltage
before and after circuit breaker is taken to be 1.0 pu and -1.0pu
as the most enormous condition but depending on the time of
closing of circuit breaker the magnitude of the voltage on the
load side changes.
For different values of voltages on the load side the
magnitudes and rise time of the voltage wave are calculated
keeping source side voltages as constant as 1.0p.u the values
are tabulated in table I.
Similarly by changing the magnitudes of the voltage on the
source side, keeping voltage on load side constant at 1.0p.u.
Then the transient due to variation of voltage on source side
obtained. The values are tabulated in Table II.
TABLE I
7
8
9
10
0.4
0.3
0.2
0.1
1.79
1.71
1.42
1.39
9
12
11
9
IV. RESULTS AND DISCUSSION
The phenomenon that occurs during the DS closing into
a capacitive load is very nearly the reserve of processes that
occur during its opening. Here, the first restrike occurs due to
the residual voltage left behind by a previous opening on the
load side. Circuit breaker or load break switch closing or
openings also generate VFTO in the case of re-strikes but the
number of such VFTO is much lower than those generated by
DS operation. The various transient voltage and current at
different positions in a 3 phase 132kv GIS for the first
switching operation presented in results.
Assumed that there is a second re-strike another switch is
connected in parallel to the circuit breaker for simulation in
MATLAB modeling. Transients are calculated by closing this
switch when voltage difference across the contacts of the
circuit breaker reaches maximum value.
TRANSIENT DUE TO VARIATION OF RE-STRIKE VOLTAGE ON LODE SIDE
S.no
1
2
3
4
5
6
7
8
9
10
Load side Voltage
(p.u)
-1.0
-0.9
-0.8
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
Magnitude of the
voltage (p.u)
2.45
2.39
2.19
2.15
2.03
1.96
1.82
1.77
1.53
1.45
Rise Time
(Nanos)
10
13
12
10
12
11
10
13
12
9
During Re-strike operation (source & Load) the voltage
through the resistance of the circuit breaker is shown in fig.5
and fig.6. From the graph it was found the maximum current is
25A at a rise time of 13ns.
TABLE II
TRANSIENTS DUE TO VARIATION OF RE-STRIKE VOLTAGE ON SOURCE SIDE
S.no
1
2
3
4
5
6
Source side
Voltage (p.u)
1.0
0.9
0.8
0.7
0.6
0.5
Magnitude of the
voltage (p.u)
2.43
2.37
2.17
2.11
2.01
1.91
Rise Time
(nanos)
9
10
11
10
9
10
Fig. 5 Transient voltage waveform during Re-strike for 10mts
from source side in a 3-phase 132kv GIS
Page 761
Kondalu and Subramanyam, Estimation of Re-striking Transient Overvoltages in a 132KV Gas insulated Substation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 757-763
ISSN 2078-2365
Fig 6Transient voltage waveform during Re-strike for
10mts from Load side in a 3-phase 132kv GIS
Fig.8 Current waveform during Re-strike for 10mts from
Load side in a 3-phase 132kv GIS
TABLE III
THE ANALYSIS VALUES ARE TABULATED AS FOLLOWS:
Mode of operation
Magnitude of
voltage(p.u)
Rise time
(Nano sec)
During closing
operation
2.46
69
During opening
operation
1.22
56
During second restrike
2.45
112
Fig 7 Current waveform during Re-strike for 10mts from
source side in a 3-phase 132kv GIS
V. CONCLUSION
To introduce the current chopping, the circuit breaker is
opened remains. Hence to calculate transients due to opening
operation the CB is opened at 12ns. The transients are obtained
and show in fig.7.
The transient calculated due to re-strike gives the peak
voltage of 2.45p.u at a rise time of 112ns show in fig.8.
A model is Developed for the prediction of the VFTO
phenomena in the circuit of voltage and current transformers in
GIS. The main advantage of such model is to enable the
transient analysis of GIS.A spark collapse time was correctly
simulated by the variable resistor. By this spark collapse time,
resistance of the VFTO is extended, and the component caused
by short surge impedance discontinuities such as spacers, disconnectors and short bus branches were damped.
A GIS system comprising of spacers, bus bar and disconnectors has been considered for modeling into electric
network. The inductance of the bus bar is calculated from
Page 762
Kondalu and Subramanyam, Estimation of Re-striking Transient Overvoltages in a 132KV Gas insulated Substation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 757-763
ISSN 2078-2365
diameters of conductors and enclosure using standard formulae.
Cone insulators used for supporting inner conductor against
outer enclosure are assumed to be disk type for approximate
calculation of spacer capacitance is calculated using formulae
for concentric cylinders. The entire bus length is modeled as
distributed pi-network. The peak magnitude of fast transient
currents generated during switching event changes from one
position to another in a 132kv GIS for a particular switching
operation. These transients over voltages are reduced by
connecting suitable resistor in an equivalent circuit during
closing and opening operation.
VI. REFERENCES
[1]
Boggs SA., Chu F.Y. and Pujimotor N. (IYXZ), 'Disconnect Switch
Induced Transients and Trapped Charge in GIs', EEE Trans. PAS, Vol.
PAS-101, No.IO, pp3593-3601.
[2] MohanaRao M., Naidu M.S. (199% 'Estimation of Fast Transient
Overvoltages in the case of Disconnnector operation in a GIS', 3d
workshop & conference on EHV Technology, IISC Bangalore.
[3] J.B. Kim,M.S. Kim,K.S.Park, W.P.Son.,.D.S. Kim, G.S. Kil.
Development of monitoring and diagnostic system for SF6 gas
insulated switchgear.
IEEE Conference Record of the 2002 IEEE International Symposium
on Electrical Insulation. Boston, Massachusetts, UnitedStates, pp.453456, 2002.
[4] M.kondalu, G.Sreekanthreddy, Dr. P.S. subramanyam,” Estimation
Transientover voltages in gas insulated bus duct from 220kv gas
insulated substation”, International journal of Computer applications,
(0975-8887) volume 20-no.8 april 2011.
[5] M.kondalu, G.Sreekanthreddy, Dr. P.S. subramanyam,” Analysis and
Calculation of very fast transient over voltages in 220kv gas insulated
substationIinternational Journal of Engineering &techsciencevol 2(4)
2011
[6] Li Liu-ling, Hu Pan-feng, Qiu Yu-chang, Analysis of Very Fast
Transient Overvoltage Calculation Affected by Different Transformer
WindingModels, Journal of Xi’an Jiaotong University, 2005 ′ 10 ″
Vol.39 No.10:1160-1164.
[7] Yang Linghui, Zhang Jiamin, Research on transient overvoltage during
Operation of 500kV GIS disconnecting switch, East China Electric
Power, Jan.2004, Vol.32 No.1:37-41
[8] Shibuya Y, Fujita S, Shimomura T. Effects of very fast transient overVoltages on Transformer [J].IEE Proceedings, Part C,1999,146(4): 459
464.
[9] G. Ecklin, D. Schlicht, and A. Plessel. Over voltages in GIS caused by
the Operation of isolators in Surges in High Voltage Networks.
K.Ragaller, Ed. New York City: Plenum Press, 1980.
[10] H. Hiesinger,RWitzmann. Very fast Transient Breakdown at a
needle Shaped Protrusion, IX Int. Conf. on Gas Dis. and Their
Appli. Sep 1988.
Page 763
Kondalu and Subramanyam, Estimation of Re-striking Transient Overvoltages in a 132KV Gas insulated Substation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 784-790
ISSN 2078-2365
Different Braking Techniques Employed to a
Brushless DC Motor Drive used in Locomotives
M.Rakesh, P.V.R.L. Narasimham
Department of EEE., Gudlavalleru Engineering College, Gudlavalleru-521356, A.P, India
Abstract— Brushless Direct current (BLDC) motors are
gaining attention these days in many applications because of
their simplicity in its control and high power density. Due to
their usage many advances are taking place in field of
automobiles in general and locomotives in particular. To have
an effective control over the locomotive it is desirable to have
control over starting, running and braking of a bldc drive. This
paper mainly describes about different types of braking that
can be applied to a bldc drive used in locomotives. Various
braking methods are introduced and described. The braking of
bldc motor is simulated in MATLAB/Simulink. The simulation
results are presented and comparative study is made.
Key words: BLDC drive, braking methods, locomotives,
Matlab/Simulink.
I.
INTRODUCTION
Brushless Direct current (BLDC) motors are one of the
motor types rapidly gaining popularity in industry such as
appliances, automotives, aerospace, consumer, medical,
industrial automation equipment and instrumentation. Recent
trend in automobile industry is using these BLDC motors as
electric vehicles as these are energy efficient and pollutant
free. Simulation studies indicate that a 15% longer driving
range is possible for an electric vehicle with PM brushless
motor drive systems compared with induction types.
As the name implies, BLDC motors do not use brushes
for commutation; instead they are electronically commutated.
In BLDC motor since the back emf is non sinusoidal, the
inductance do not vary sinusoidally in the abc frame and it
does not seem advantageous to transform the equations to
d-q frame since inductances will not be constant after
transformation [5].
The braking of BLDC motors is quite easier as these
machines employ a permanent magnet as its rotor. The
braking methods of a BLDC motor are similar to that of a
direct current machine. This paper deals with different types
of braking applicable to a BLDC drive. The performance of
locomotive is examined for dynamic braking, plugging and
regenerative braking and simulation results are presented.
II.
MATHEMATICAL MODELING OF BLDC MOTOR
In modeling a BLDC motor, abc phase variable model is
preferred to d-q axis model as the mutual inductance between
stator and rotor is non-sinusoidal[1]. The mathematical
modeling is done in abc phase variable model and is
expressed in state-space form.
Following assumptions are made in modeling the BLDC
motor[10].
The motor is not saturated.
Stator resistances of all the windings are equal and
self and mutual inductances are constant.
The power semiconductor devices are ideal.
The voltage equations of the three phase stator windings are
va
R 0 0
ia
L M M
ia
ea
vb = 0 R 0
ib + p M L M
ib + eb
vc
ic
ic
0 0 R
M M
L
(1)
ec
The generated electro-magnetic torque equation is
Te = (eaia+ebib+ecic)/ωm
(2)
The equation of motion is
pωm =(Te - Tl – Bωm)/J
(3)
These voltage equations are transformed to state-space form
and are arranged as follows:
ia
p ib
ic
-R/L 0
0
= 0 -R/L 0
0
0
-R/L
ia
1/L
0 0
va
ib +
0
1/L 0
vb
ic
0
0 1/L vc
(4)
Where
va, vb, vc are the voltages of the three phases a,b and c in volts
R is resistance of each phase of motor in ohms
ia, ib, ic are the currents of the three phases a,b and c in
amperes.
p is the derivative
Page 784
Rakesh and Narasimham, Different Braking Techniques Employed to a Brushless DC Motor Drive used in Locomotives
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 784-790
ISSN 2078-2365
L is self inductance of each phase of motor in henrys
M is mutual inductance between respective phases in henrys
ea, eb, ec are the back emfs of the three phases a,b and c in
volts
Te is the electromagnetic torque in Newton meters
ωm is the mechanical speed of the motor in radians per second
Tl is load torque in Newton meters
B is damping constant in newtons per radian per second
J is inertia of rotor in kg – m2
L is difference of self and mutual inductances in henrys
The modeling of the machine during motoring operation is
presented above and is modified for different braking
operations.
III.
BRAKING AND TYPES OF BRAKING
which opposes the motion. If the input current is in phase
with back emf, motoring torque is developed otherwise if
the input current is in out of phase with back emf then
braking torque is developed.
The electro magnetic torque developed in phase ‘a’ is
Tea = (ea*ia)/ωr
(5)
Braking during forward motoring is called forward braking
while it is in reverse rotation is called reverse braking. In
Speed-Torque plane, forward braking will result in second
quadrant operation where as reverse braking results in fourth
quadrant operation.
Figure.1. is a basic bldc motor drive used in locomotives.
a three phase inverter is used for exciting the three phase
bldc motor. For the motor control, Commutation logic and
control block takes the rotor position, torque command and
current feed back as inputs for switching the gate drives of
the switches.
In locomotives, precise control over stopping of machine is
important along with start. In such a case to stop the
machine quickly and accurately, braking methods are useful.
Braking is nothing but stopping the machine at a desired
position. Ideal braking is bringing the machine to rest in no
time.
Braking of locomotive can be done as electric braking or
mechanical braking. In mechanical braking the motion is
restricted by the friction applied by mechanical brakes
which is preferred during low speeds. In electric braking the
motor works as a generator developing a negative torque
which restricts the motion. The purpose of electrical braking
is to restrict the motion of the machine as quick as possible.
Electric braking cannot replace the ordinary mechanical
brakes, as the vehicle cannot be held stationary by it. In
locomotives, for the braking to be done perfectly and
smoothly, electric braking in conjugation with mechanical
braking is used. This is done by applying electrical braking
to slow down the locomotive to a lower speed and then
mechanical brakes are applied.
Electric braking to above drive can be implemented in
three ways namely Dynamic Braking, Plugging and
Regenerative Braking.
During electric braking the motor torque will reverse and the
machine will work as a generator, absorbing mechanical
energy from load and converting it into electrical energy.
The mechanical energy is obtained from the load either from
the energy stored in the inertia of the motor load system or
from the active load torque when the locomotive is moving
down gradient. Electrical braking reduces the wear of the
brake shoes and gives higher rate of braking retardation,
thus brings the vehicle quickly to rest and shortens the
running time to a considerable extent.
This braking can be implemented by disconnecting the
power supply to windings and short circuiting them. The
short circuited windings carrying higher magnitudes of
current will damage the windings. To limit the short circuit
current flowing in these windings, an external resistance R is
connected in series with the windings. This resistance is
used to limit current and is called braking resistance. Thus,
the power generated in the three stator windings during
Figure.1 Basic BLDC Motor drive during normal operation
A. DYNAMIC BRAKING
Dynamic braking is bringing the machine to rest position by
dissipating the kinetic energy possessed by the rotor of
motor in the form of heat energy through some external
resistance.
Braking action can be achieved by generating a torque of
opposite polarity (braking torque) to that of motoring torque
Page 785
Rakesh and Narasimham, Different Braking Techniques Employed to a Brushless DC Motor Drive used in Locomotives
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 784-790
ISSN 2078-2365
braking
is
wasted
in
the
external
resistance.
Figure.2 Circuit configuration of bldc motor drive during
dynamic braking
Fig.2 is an implementation of dynamic braking to the drive
in fig.1. The generated power is dissipated in the resistor R
through the fed back diodes. When fast braking is desired,
that is in order to keep the braking torque at a fixed value
the resistance has to be decreased with time. This is done by
controlling switch So [4].
The equation of speed during normal operation is
ω= (V/K) - (R*T)/K2
(6)
For dynamic braking V=0, then the speed equation becomes
ω= - (R*T)/K2
(7)
Here R is the resistance of winding after insertion of braking
resistance. So during dynamic braking the speed-torque
relation is in the form of a straight line which passes through
the origin with a negative slope of –R/K2.
B. PLUGGING
Plugging is a method of braking obtained by reversing
the applied supply voltage, so that the input voltage assists
the back emf in forcing armature current in reverse
direction. This reversed current will have impact on torque,
thus producing deceleration. Plugging provides faster
braking response because braking torque is high as the
magnitude of current during this braking is high. Even
though plugging provides faster braking response it is highly
in-efficient because in addition to generated power, the
power supplied by the source is also wasted in resistances
and plugging increases the inverter rating also. Plugging
can be implemented to the drive in figure.1 by reversing the
voltage (by high speed switches) and a braking resistance is
connected just as in dynamic braking.
The speed equation now becomes
ω = (-V/K) - (R*T)/K2
(8)
So plugging provides torque at zero speed. when reverse
voltage is applied for stopping the locomotive the supply
must be disconnected at the instant where speed is close to
zero. Otherwise it will rotate in reverse direction (reverse
motoring takes place).
The speed-torque relation is of the form of a straight line
with a negative intercept. Thus the speed-torque plot during
plugging doesn’t pass through origin, it is a straight line
having a slope of –R/K2 and having a negative intercept of
–V/K.
C. REGENERATIVE BRAKING
In regenerative braking, instead of wasting the power in
external resistance the power generated during retardation is
fed back towards the source i.e., the motor works as a
generator developing a negative torque which opposes the
motion and the generated energy is supplied to the source.
For the generated energy to be supplied to the source two
conditions should be satisfied
i) back emf should be greater than supply voltage (E > V)
for all speeds
ii) Current has to reverse its direction
For the above two conditions to be satisfied, increase the
back emf so that it is greater than the supply voltage. In
order to increase the back emf, increase the speed. The
speed increases when the locomotive is moving down the
gradient or by increasing the field flux. But increasing the
field flux beyond rated is not possible as the permanent
magnets are used in field system. So, for a source of fixed
voltage of rated value regenerative braking is possible only
for speeds higher than rated value and for a variable voltage
source it is possible for below rated speeds also.
During regeneration if the generated power is not absorbed
by the load, it will be supplied to the line and the line
voltage will rise to dangerous values leading to insulation
break down. Hence regenerative braking should be used
only when there are loads connected to absorb regenerated
power.
Figure.3 basic four-quadrant bldc motor drive for
regeneration
Figure.3 shows a basic four-quadrant electronically
commutated motor drive which provides regenerative
braking. During regeneration the capacitor C stores the
energy recovered from the load through the feed back diodes
Page 786
Rakesh and Narasimham, Different Braking Techniques Employed to a Brushless DC Motor Drive used in Locomotives
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 784-790
ISSN 2078-2365
across the switches. To limit the capacitor C voltage to a
safer value Switch S0 is used to dissipate the excess energy
through the resistor.
The speed equation during regeneration becomes
ω = (V/K) - (R*T)/K2
(9)
The above equation is of the form of a straight line with a
positive intercept. As we keep on decreasing the voltage
with reference to back emf, the voltage becomes zero
finally. Thus the speed equation becomes
ω = (V/K) + (R*T)/K2
(10)
So during regenerative braking the speed-torque relation is
in the form of a straight line which passes through the origin
with a slope of R/K2.
IV.
reversal in current is observed (at 0.05 sec) but not in back
emf, this produces negative torque called braking torque.
The magnitude of current during braking will decide the
braking time and it exists till the kinetic energy possessed by
rotor is dissipated completely. This current has to be
restricted in order to protect the circuitry from damage. To
limit the current we can use fixed or variable resistance, but
variable resistance is employed in order to decrease the
braking time.
SIMULATION RESULTS
The locomotive system is developed using the motor
parameters listed in below table:
Voltage
200
Torque constant
0.66
Vdc,(volts)
Kt(Nm/A)
No of Poles, 2P
4
Rotor inertia
0.79
J(kg-m )
Winding
1.4
Noload Speed
5400
resistance Rs(Ω)
Nnl(rpm)
Winding
8.90
Stalling Torque
8
inductance
Tstall, (Nm)
Ls(mH)
2
In my model, initially the machine is started as a motor and
is subjected to braking after the locomotive reaches a steady
speed. Three different types of brakings are performed in
MATLAB / SIMULINK.
A. DYNAMIC BRAKING
During simulation, machine is allowed to attain a steady
speed initially and after that at 0.05sec dynamic braking
is applied. The simulation results are obtained.
Figure.5 a) Torque and b) Speed waveforms during dynamic
braking
Initially the vehicle attained a steady speed in 0.03sec
during motoring. At time t =0.05sec, the dynamic brake is
applied which developed a negative torque that opposed the
motion and tended the locomotive to rest. The time taken by
the locomotive to reach zero speed is 37.5 milliseconds.
Figure.4 Current and Back emf waveforms during dynamic
braking
Figure.4 shows the variation of current and back emf
waveforms from motoring to dynamic braking. Here a
Page 787
Rakesh and Narasimham, Different Braking Techniques Employed to a Brushless DC Motor Drive used in Locomotives
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 784-790
ISSN 2078-2365
Figure.6 Speed-Torque plot of locomotive from motoring to
dynamic braking
When Dynamic brake is applied, the operating point has
shifted to second quadrant with same magnitude of speed.
Speed can’t be changed abruptly but the torque maintains
the same magnitude but with opposite in sign (negative
torque) as shown in figure.6. This negative torque means the
negative power; i.e. power generated which is dissipated in
the braking resistance. The locomotive stops when the
operating point is origin.
B. PLUGGING
During simulation, machine is allowed to attain a steady
speed initially and after that at 0.05sec reverse voltage is
applied to stop the vehicle. The simulation results are
obtained. Figure.7 shows the variations in current and back
emf waveforms from motoring to braking
Figure.8 a) Torque b) Speed waveforms during plugging
In plugging, time taken by the locomotive to reach zero
speed is 31.2 milliseconds (figure.8), which is 37.5
milliseconds in case of dynamic braking for the same
machine. So plugging gives quick response compared to
dynamic braking.
Figure.7 Current and Back emf waveforms during plugging
At a time of 0.05sec, when plugging is applied current has
changed its direction. In this case of braking, the magnitude
of current has increased to 2.67 times to that of steady value
with out braking resistance. Figure.8 shows the waveforms
of torque and speed during plugging. With out braking
resistance, the braking torque has increased with respect to
current, and it also increased by 2.7 times to that of steady
value in magnitude but with a negative sign. So to
incorporate plugging the inverter has to be redesigned and
proper care is to be taken for the windings to with stand this
much of current.
Figure.9 Speed-Torque plot of locomotive from motoring to
braking for plugging
When plugging is applied for the machine operating in first
quadrant the operating point has shifted to second quadrant
with same magnitude of speed but with a change in the
magnitude of torque with an opposite sign (negative torque).
The path of the curve is a straight line as shown in figure.9.
When speed is zero there does torque exists. So mechanical
Page 788
Rakesh and Narasimham, Different Braking Techniques Employed to a Brushless DC Motor Drive used in Locomotives
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 784-790
ISSN 2078-2365
brakes are applied at that instant to prevent reverse
motoring.
C. REGENERATIVE BRAKING
During simulation, at 0.05sec to stop the locomotive
regenerative braking is applied. The simulation results are
shown. Figure.10 shows the variations in current and back
emf waveforms from motoring to braking during
regeneration.
Figure.11 a)Torque b) Speed waveforms during regenerative
braking
The time taken by the machine to reach zero speed is 38.3
milliseconds (figure.11), which is almost same during
dynamic braking. This time interval is the time for which
regeneration took place. Till 0.05seconds, the three windings
of the machine had consumed the power and after that
instant the power has become negative because the three
windings are generating power which is fed to an external
load.
Figure.10 Current and Back emf waveforms during
regenerative braking
In this braking, in order to keep the magnitude of current
with in the bounds proper care has to be taken in varying
voltage. Figure.11 shows the torque and speed waveforms
during regenerative braking.
Figure.12 Speed-Torque plot of locomotive from motoring
to regenerative braking
The speed-torque plot of the machine during its operation
from motoring to regeneration is shown in figure.12. When
Regenerative braking is applied, the power generated is
delivered to an external load and the operating point has
shifted to second quadrant, this quadrant is called Forward
regeneration. The locomotive stops regeneration when the
operating point tends to origin.
V.
CONCLUSION
From the simulation results analysis of the three brakings,
i) Regenerative braking is more useful as no power is
wasted but this process is costlier as this requires some
external circuitry for regeneration.
Page 789
Rakesh and Narasimham, Different Braking Techniques Employed to a Brushless DC Motor Drive used in Locomotives
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 784-790
ISSN 2078-2365
ii) Dynamic braking can be used where stopping the
machine is important not wasted power i.e., where economy
is a factor.
iii) Plugging is the most in-efficient method as this will
damage the windings, though this gives faster braking to
incorporate this the inverter has to be redesigned. Power
supplied by the source is wasted along with the power
generated.
Hence regenerative braking is preferred in locomotives. For
consumers using bldc motor drive regenerative braking is
suggested, if they can afford the extra cost for the
regenerative circuitry else dynamic braking is preferred.
REFERENCES
[1] P.Pillay and R.Krishnan, “Modeling, simulation and analysis of
permanent-magnet motor drives, part-II: the brushless DC motor drives,”
IEEE Trans. on Industry Applications, vol. 25, pp.274-279,
March/April1989.
[2] T. Jahns, R. C. Bccerra, and M. Ehsani, “lntcgrated current
regulation for a brushless ECM drive,” IEEE Trans. Pother( -Iron.,
vol. 6, no. I, pp. 118-126, Jan. 1991.
[3] Roger C.Becerra, “Four-Quadrant Brushless ECM Drive with Integrated
Current Regulation,” IEEE Trans. on Industry Applications, vol. 28, No.4,
July/August 1992.
[4] Gopal K Dubey “Fundamentals of Electrical Drives”, Narosa Publishing
House, New Delhi, Second Edn, 2001, Chapter 7, pp271-277
[5] Pragasan Pillay and R.Krishnan, ‘‘Modelling of permanent magnet
motor drives’’ IEEE Transactions on industrial electronics, Vol.35, No.4,
November 1988.
[6] Krishnan R “motor Drives Modeling, Analysis and Control”, Prentice
Hall of India, First Edn, 2002.
[7] Pragasan Pillay and R.Krishnan, ‘‘Control Characteristics and Speed
Controller Design for a High Performance Permanent Magnet
Synchronous Motor Drive’’ IEEE Transactions on Power Electronics Vol.5
No.2 April 1990.
[8] P.Pillay and R.Krishnan, “An investigation into the torque behaviour of
a brushledd dc motor drive”,IEEE Transactions 1988
[9] Pragasan Pillay and R.Krishnan, ‘‘Application characteristics of
permanent magnet synchronous and brushless dc motors for servo drives’’
IEEE Trans. on Industry Applications, vol. 27, No.5, september/october
1991.
[10] Vinatha U, Swetha Pola, Dr K.P.Vittal, ‘‘Simulation of Four Quadrant
Operation & Speed Control of BLDC Motor on MATLAB / SIMULINK’’.
[11] Padmaraja yedmale,‘‘Brushless dc motor fundamentals’’ Microchip
AN885- Microchip technology Inc
Page 790
Rakesh and Narasimham, Different Braking Techniques Employed to a Brushless DC Motor Drive used in Locomotives
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 751-756
ISSN 2078-2365
Step-Up Dc/Dc Converter for Distributed Power
Generation Systems
T. Karthikeyan, B.Gowdhami and. Sathishkumar M.E.
1
PG Student, 2PG Student and 3Assitant professor
EEE Mailam Engineering College, Villupuram, India
E-mail :karthimec2011@gmail.com,gowdhamiped@gmail.com and mecposk@yahoo.co.in
can be easily transported and converted to other forms for
the benefit of society. In terms of power they come in
various sizes from mW to MW ranges.
Abstract: This paper presents new step-up dc/dc converter
topologies intended for distributed power generation
systems. The topologies contain a voltage-fed quasi-Zsource inverter with continuous input current on the
primary side, a single-phase isolation transformer, and a
voltage doublers rectifier (VDR). To increase the power
density of the converter, a three-phase auxiliary ac link (a
three-phase inverter and a three-phase isolation
transformer) and a three-phase VDR are proposed to be
implemented. This paper describes the operation principles
of the proposed topologies and analyzes the theoretical and
experimental results
Keywords: QZSI, voltage doublers rectifier, DC/DC
converter, PV panel.
1. INTRODUCTION
To interconnect a low-dc-voltage-producing PV
(typically 40–80 Vdc) to residential loads (typically 230Vac single phase or 3 × 400 Vac), a special voltage
matching converter is required. Due to safety and dynamic
performance requirements, the interface converter should be
realized within the dc/dc/ac concept. This means that low
voltage from the PV first passes through the front-end stepup dc/dc converter with the galvanic isolation; subsequently,
the output dc voltage is inverted in the three-phase inverter
and filtered to comply with the imposed standards and
requirements (second dc/ac stage).
Solar energy is the most abundant renewable
resource. The electromagnetic waves emitted by the sun are
referred to as solar radiation. The amount of sunlight
received by any surface on earth will depend on several
factors including; geographical location, time of the day,
season, local landscape and local weather. The light's angle
of incidence on a given surface will depend on the
orientation since the Earth's surface is round and the
intensity will depend on the distance that the light has to
travel to reach the respective surface. The radiation received
by a surface will have two components one which is direct
and will depend on the distance the rays travel (air mass).
The other component is called diffuse radiation and is
illustrated in figure 2.1. The range of wavelengths of light
that reach the earth varies for 300nm to 400nm
approximately. The spectrum outside the atmosphere, which
closely resembles 'black body' radiation, since the
atmosphere selectively, absorbs certain wavelengths. They
can directly convert the sun's energy into electricity which
The design of the front-end isolated dc/dc converter is most
challenging because this stage is the main contributor of
interface converter efficiency, weight, and overall
dimensions. The low voltage provided by the PV is always
associated with high currents in the primary part of the dc/dc
converter (switching transistors and primary winding of the
isolation transformer). These high currents lead to high
conduction and switching losses in the semiconductors and
therefore reduce the efficiency. Moreover, the large voltage
boost factor requirement presents a unique challenge to the
dc/dc converter design . This specific requirement could be
fulfilled in different ways: by use of an auxiliary boost
converter before the isolated dc/dc converter or by use of an
isolation transformer with a large turns ratio for effective
voltage step-up. A direct step-up dc/dc converter without
input voltage preregulation is simpler in control and
protection. Due to the reduced number of switching devices,
the converter tends to have better efficiency and reliability.
The varying voltage from the PV passes through the high-
Page 751
Karthikeyan et. Al., Step-Up Dc/Dc Converter for Distributed Power Generation Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 751-756
ISSN 2078-2365
frequency inverter to the step-up isolation transformer. The
magnitude of the primary winding voltage is controlled by
the duty cycle variation of inverter switches in accordance
with the PV output voltage and converter load conditions.
The isolation transformer should have an increased turns
ratio (approximately 1 : 17) to provide effective voltage
step-up in the whole range of input voltage and load
variations. The choice of dc/dc converter topology in that
case can be broadly categorized as a push–pull or a singlephase full-bridge topology. Because of the symmetrical
transformer flux and minimized stress of primary inverter
switches, the fullbridge topology has been found to be most
useful in terms of cost and efficiency, particularly when
implemented for power levels higher than 3 kW. This paper
is devoted to a new power circuit topology to be
implemented in the front-end dc/dc converter for distributed
power generation. The topology proposed contains a
voltage-fed quasi-Z-source inverter (qZSI) with continuous
input current at the converter input side, a high-frequency
stepup isolation transformer, and a voltage doubler rectifier
(VDR). In contrast to earlier presented topologies, the novel
converter provides such advantages as increased reliability,
isolation transformer with reduced turns ratio, and reduced
impact on the PV due to continuous input current. To
improve the power density of the converter, the topology
with a three phase intermediate ac link is discussed in the
final section of this paper.
2. DESCRIPTION OF PROPOSED TOPOLOGY
The voltage-fed qZSI with continuous input current
implemented at the converter input side has a unique
feature: It can boost the input voltage by utilizing extra
switching state—the shoot-through state. The shoot-though
state here is the simultaneous conduction of both switches of
the same phase leg of the inverter. This operation state is
forbidden for the traditional voltage source inverter (VSI)
because it causes the short circuit of the dc-link capacitors.
In the discussed qZSI, the shoot-through state is used to
boost the magnetic energy stored in the dc-side inductors
(L1 and L2) without short-circuiting the dc capacitors. This
increase
in
inductive
energy,
in
turn,
presented in as a modification of a currently popular
voltage-fed Z-source inverter (ZSI). The drawback
associated with the conventional ZSI is substantial—
discontinuous input current during the boost mode that
could have a negative influence on the PV.
The discussed qZSI features continuous current
drawn from the FC as well as lower operating voltage of the
capacitor C2, as compared to the ZSI topology.
Uc1= 1-Ds/1-2Ds.UIN…………1
Uc2=Ds/1-2Ds.UIN…………….2
Fig.1 Proposed power circuit diagram of Quasi Z-source
converter
provides the boost of voltage seen on the transformer
primary winding during the traditional operating states
(active states) of the inverter. Thus, the varying output
voltage of the PV is first preregulated by adjusting the
shoot-through duty cycle; afterward, the isolation
transformer is being supplied with a voltage of constant
amplitude value. Although the control principle of the qZSI
is more complicated than that of a traditional VSI, it
provides a potentially cheaper, more powerful, reliable, and
efficient approach to be used for FC powered systems. The
voltage-fed qZSI with continuous input current was first
where Ds is the duty cycle of the shoot-through state
Ds= ts/T…………….3
where tS is the duration of the shoot-through state and T is
the operation period.
When the input voltage is high enough, the shoot-through
states are eliminated, and the qZSI starts to operate as a
traditional VSI, thus performing only the buck function of
the input voltage.
Thus, the qZSI could realize both the voltage boost
and the buck functions without any additional switches
using a special control algorithm only.
Page 752
Karthikeyan et. Al., Step-Up Dc/Dc Converter for Distributed Power Generation Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 751-756
ISSN 2078-2365
3.Voltage Boost Control Method of qZSI-Based SinglePhase DC/DC Converter
Fig. 2 shows the control principle of the singlephase qZSI in the shoot-through (voltage boost) operating
mode. Fig. 2(a) shows the switching pattern of the
traditional single-phase VSI. These switching states are
known as active states when one and only one switch in
each phase leg conducts. To generate the shoot-through
states, two reference signals (Up and Un) were introduced
[Fig. 2(b)]. If the triangle waveform is greater than Up or
lower than Un, the inverter switches turn into the shootthrough state [Fig. 2(b)]. During this operating mode, the
current through the inverter switches reaches its maximum.
Depending on the control algorithm, the shoot through
current could be distributed between one or both inverter
legs. The dc-link voltage and the primary winding voltage
waveforms of the isolation transformer during shoot-through
are shown in Fig. 2(c) and (d), respectively. According to
the presented control methodology (Fig. 2), the shootthrough states are created during the zero states of the fullbridge inverter, where the primary winding of the isolation
transformer is shorted through either the top (T1 and T3) or
bottom (T2 and T4) inverter switches. To provide a
sufficient regulation margin, the zero-state time tZ should
always exceed the maximum duration of the shoot-through
states tS,max per one switching period.
Thus, each operating period of the qZSI during the
shoot-through always consists of an active state tA, shootthrough state tS, and zero state tZ.
T = tA + tS + tZ………4
DA+ DS+ DZ=1…………..5
where DA is the duty cycle of an active state, DS is
the duty cycle of a shoot-through state, and DZ is the duty
cycle of a zero state. It should be noted that the duty cycle of
the shoot-through state must never exceed 0.5. It should be
noted here that, in the presented control scheme, the shootthrough time interval is evenly split into two intervals of
half the duration.
Fig.2. Proposed operating principle and resulting voltages of
the single-phase qZSI in the shoot-through (voltage boost)
mode.
In that case, the operating frequency of the quasiZ-source (qZS) network will be two times higher, and the
resulting switching frequency of the power transistors will
be up to three times higher than the fundamental harmonic
frequency of the isolation transformer. That fact is very
relevant for proper component and operating frequency
selection.
In the operating points, when the input voltage is
high enough, the shoot-through states are eliminated, and
the qZSI operates as a traditional VSI. Thus, the qZSI
discussed could provide both the voltage boost and buck
functions by the single stage energy conversion.
4.Power Circuit Design Considerations
This section provides an overview of the design
process of the proposed dc/dc converter. In the given
application, the desired value selected for the dclink voltage
UDC was 80 V. It is assumed that the converter is always
operating with the rated load and between two boundary
operating points, which correspond to the minimal UIN,min
and maximal UIN,max input voltages. In the first case, the
shoot-through states should be used to boost the input
Page 753
Karthikeyan et. Al., Step-Up Dc/Dc Converter for Distributed Power Generation Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 751-756
ISSN 2078-2365
voltage to the predefined dc-link voltage level. In the second
case, when the input voltage is equal to the desired dc-link
voltage, no shoot-through is applied, and the qZSI operates
as a traditional VSI. The design of the power converter
should be performed for the operating point with a minimal
possible input voltage and at rated power, when the shootthrough duty cycle reaches its maximum. As a consequence,
the boost ratio of the input voltage is also maximal
Bmax= UDC/UIN,min= 80/40= 2
To achieve proper efficiency of the converter and better
transformer utilization, in real designs, proper balance
between the boost ratio and the transformer turns ratio
should be found. In the current application, the maximal
duty cycle of the shootthrough
state is
Fig.3 Proposed Quasi Z-source Series-parallel resonant
converter
Ds,max=0.25
During the active states, the transformer primary
winding is being supplied from the inverter by a voltage
with an amplitude value UTR,pr = UDC = 80 V. To reduce
the turns ratio n of the isolation transformer, a VDR was
implemented on the secondary side of the converter. In
contrast to the traditional full-bridge rectifier, two diodes of
one leg in the VDR topology are replaced by the capacitors.
Since each capacitor charges to the peak secondary voltage
UTR,sec, the output voltage from this circuit will be the sum
of the two capacitor voltages or twice the peak voltage of
the secondary winding. This circuit then produces an output
voltage that is twice the transformer secondary voltage. Due
to the voltage doubling effect, the VDR enables the use of
the isolation transformer with a reduced secondary turns
ratio, i.e., 1 : 3.75 for the application discussed.
Furthermore, the VDR improves the rectification efficiency
due to minimized voltage drops in the components (twice
reduced number of rectifying diodes and full elimination of
a smoothing inductor).
Fig.4 Simulation diagram
Page 754
Karthikeyan et. Al., Step-Up Dc/Dc Converter for Distributed Power Generation Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 751-756
ISSN 2078-2365
Fig5 dc link
voltage
Fig.5 Output voltage current waveform
5.CONCLUSION
This paper has presented two new isolated step-up
dc/dc converter topologies with qZSIs. The topologies are
intended for applications with widely varying input voltage
and stabilized output voltage and when the galvanic
separation of the input and output sides is required. The
high-frequency transformer stack is responsible for
providing the input/output galvanic isolation demanded in
many applications. This paper has focused on an example of
the step-up dc/dc converter with high-frequency isolation
for the distributed power generation systems.
The proposed converters have the following key
features in comparison to traditional topologies.
1) The qZSI implemented on the primary side of the
converter could provide both the voltage boost and buck
functions with no additional switches, only by use of a
special control algorithm.
2) The qZSI has an excellent immunity against the cross
conduction of the top- and bottom-side inverter switches.
Moreover, the qZSI implemented can boost the input
voltage by introducing a shoot-through operation mode,
which is forbidden in traditional VSIs.
3) The qZSI implemented has the continuous input current
(input current never drops to zero) during the shoot-through
(voltage boost) mode.
4) The high-frequency step-up isolation transformer
provides the required voltage gain as well as input–output
galvanic isolation demanded in several applications.
5) The VDR implemented on the converter secondary side
has the improved rectification efficiency due to the reduced
voltage drop (twice reduced number of rectifying diodes and
full elimination of the smoothing inductor).
6) The turns number of the secondary winding of the
isolation transformer could be reduced by 62% (turns ratio
of 1 : 3.75 in the case of VDR instead of 1 : 10 of traditional
full-bridge rectifiers) due to the voltage doubling effect
available with the VDR.
6.REFERENCE
.
[1] A. F. Zobaa and C. Cecati, “A omprehensive review on
distributed power generation,” in Proc. SPEEDAM, 2006,
pp. 514–518.
[2] J. Padulles, G.W. Ault, and J. R.McDonald, “An
approach to the dynamic modelling of fuel cell
characteristics for distributed generation operation,”in Proc.
IEEE Power Eng. Soc.Winter Meeting, 2000, vol. 1, pp.
134–138.
[3] W. Choi, P. Enjeti, and J. W. Howze, “Fuel cell powered
UPS systems:Design considerations,” in Proc. IEEE 34th
PESC, Jun. 15–19, 2003,vol. 1, pp. 385–390.
[4] M. H. Todorovic, L. Palma, and P. N. Enjeti, “Design of
a wide inputrange DC–DC converter with a Robust power
control scheme suitable forfuel cell power conversion,”
IEEE Trans. Ind. Electron., vol. 55, no. 3,pp. 1247–1255,
Mar. 2008.
[5] S. K. Mazumder, R. K. Burra, and K. Acharya, “A
ripple-mitigating andenergy-efficient fuel cell powerconditioning system,” IEEE Trans. PowerElectron., vol. 22,
no. 4, pp. 1437–1452, Jul. 2007.
Page 755
Karthikeyan et. Al., Step-Up Dc/Dc Converter for Distributed Power Generation Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 751-756
ISSN 2078-2365
[6] J. S. Yu and P. N. Enjeti, “A high frequency link direct
dc-ac converterfor residential fuel cell power systems,” in
Proc. IEEE 35th PESC,Jun. 20–25, 2004, vol. 6, pp. 4755–
4761.
[7] J. C. Han and P. N. Enjeti, “A new soft switching direct
converter for residential fuel cell power system,” in Conf.
Rec. 39th IEEE IAS Annu.Meeting, Oct. 3–7, 2004, vol. 2,
pp. 1172–1177.
[8] S. K. Mazumder, R. Burra, R. Huang, M. Tahir, K.
Acharya,G. Garcia, S. Pro, O. dodrigues, and E. Duheric, “A
high-efficiency universalgrid-connected fuel-cell inverter
for residential application,” IEEETrans. Power Electron.,
2009, to be published.
Page 756
Karthikeyan et. Al., Step-Up Dc/Dc Converter for Distributed Power Generation Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 625-631
ISSN 2078-2365
Study of Reverberation Time Series and Echo
Detection Algorithm in Reverberation Limited
Scenarios
Sowmya S.T.V
Department of ECE, GITAM Institute of Technology, GITAM
University, Visakhapatnam,India
Email: sowmy482@gmail.com
Abstract— An innovative approach to the generation of
reverberation time series and echo detection algorithms is
presented The time series approach utilizes recent developments in
linear spectral prediction research in which the spectra of
stochastic process are modelled as rational functions and
algorithms are used to efficiently compute optimal estimates of
coefficients which specify the spectra. The approach taken in this
paper is to detect echo signal in two steps . In the first part the
reverberation time series is generated using autoregressive
formulation and in the second part echo is detected using order
partition prewhiten algorithm.
Keywords- Active sonar, Reverberation, Autoregressive model,
Reverberation spectrum, pre-whiten
I. INTRODUCTION
S
ONAR is an acronym for sound navigation and ranging
.sonar is a system that uses transmitted and reflected
underwater sound waves to detect and locate submerged
objects or measure the distance of underwater target.
Sonar Reverberation (and radar clutter) has been modeled in a
variety of ways and for a diversity of applications. Expected
reverberation power (intensity) level models are perhaps the
most common. In these, the expected reverberation power
level at the input, output, or some intermediate point in the
sonar (radar) system is estimated as a function of the
environmental and system parameters. The models are useful
in evaluating system performance for signal processing
approaches which depend primarily upon power level, such as
single beam energy detectors and matched filters. The time
series simulation models can be developed to generate
P.Chandra sekhar
Department of ECE, GITAM Institute of Technology, GITAM
University, Visakhapatnam,India
Email: chandrasekhar.au@gmail.com
complex (in phase and quadrature) reverberation voltage levels
in time series form, as they would occur at some point in the
sonar system. The time series data can be run through
emulations of alternative signal processing algorithms to
evaluate relative performance of the various processes.
Reverberation is caused by seabed, sea surface and the In
homogeneity of the granule in the seawater. As a
noise,reverberation can influence the detection performance of
target echo and cause some serious problems to active sonar .
Due to the fact that reverberation and target echo are
correlative and their spectrums are close, how to restrain
reverberation is a problem necessary to be solved for active
sonar. Inorder to restrain the reverberation signal we are using
pre whiten method here. The principle of this method is that an
AR model is established from reverberation and then a whiten
filter is designed using the power spectrum of this model.
II. TIME SERIES MODELING
Recent developments in linear spectral prediction (UP)
techniques allow stochastic processes to be modeled in a
straightforward manner. Common time series models of
sampled stochastic processes, which are basic to many LSP
techniques, include autoregressive (AR), moving average
(MA), and autoregressive moving average (ARMA) processes.
These processes can be realized as the outputs of linear digital
filters driven by white noise processes, where the filter transfer
functions have all-pole, all-zero, and pole-zero realizations for
AR, MA, and ARMA processes, respectively. The digital
filters can be implemented as recursive infinite impulse
response (IIR) filters for AR and ARMA processes, and as a
transversal finite impulse response (FIR) filter for an MA
process. If the stochastic process to be modelled is non
Page 625
Sowmya and Chandra,
Study of Reverberation Time Series and Echo Detection Algorithm in Reverberation Limited Scenarios
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 625-631
ISSN 2078-2365
stationary: but quasi-stationary, the digital filter realization is
time varying, with time-varying poles and/or zeros.
in bk i k ,n
p
k 1
on bk kn 2
i=1,2,…….p
(6a)
p
A. Autoregressive Process
In an auto regressive model of order p, the value
xn of the
complex stochastic process at time n given as a linear
combination of past values and a random input n , such that
xn bk xn k n
p
(6b)
k 1
where
i k ,n is
the time-varying autocorrelation function of
the non-stationary process.
i k,n E{xnk xn*i }
(7)
is a complex white noise process with zero mean
If the spectrum (which in general is time-varying) of the
process is known, then the auto-correlation function can be
found by inverse Fourier transformation and the parameters
{ bk , } evaluated. If the spectrum is not known, but one has
and unity standard deviation. The system transfer function
H(Z),between input and output is represented in terms of
model parameters as an all pole function.
the past p values of the process, as in the above Wiener
filtering formulation, then the auto covariance function can be
used as a local estimate of the autocorrelation function.
k 1
Where
n
H ( z)
(1)
1 bk z k
p
k j
(2)
The poles of H(Z) are the zeros of the polynomial in the
denominator and the number of poles p is referred to as the
model order. The discrete power density spectrum is
pm
2 t
1 bk e
p
k 1
Where
j 2 mk / M
2
t is the sampling interval and pm is the power at
(3)
radial frequency.The problem of modeling an arbitrary
stochastic process {x,, n = e-, -1, 0, 1, .-} as an AR process
reduces to the selection of the model parameters. The manner
in which they are selected will depend on a priori information
about {x,;. A standard formulation is to select the model
parameters such that the linear estimate of the process {x,} at
the present time n, given the past p values of the process { X n
- k , k = 1: 2, --, p } , is best in a least squares sense. That is,
defining the linear estimate of order p as
n 1 k
*
x j k x j i ,
j n p i
i<k
i k ,n n 1i
*
x j k x j i ,
j
n p j
i>k
(8)
The auto covariance function can then replace the
autocorrelation function in the normal equations. With this
replacement, are known as the Yule-Walker equations and will
be referred to in this way in this paper, regardless of whether
the auto covariance or autocorrelation functions are used in
them. A variety of ways to solve the Yule-Walker equations
are available. We will use the Levinson-Durbin approach .
Having solved for estimates of the model parameters in this
way, one can evaluate the spectral estimate or the linear
estimation filter transfer function &z) by using the parameter
estimates in expressions (3) and (2), respectively.
The filter will produce a statistical realization of the process
{x} when driven by white noise. The spectral estimate is
sometimes referred to as the maximum entropy spectral
estimate.
(4)
the parameters are selected such that the estimation error is
minimized in a mean square sense
min E{ } min E{( xn bk xn k ) 2 }
p
{a k }
2
n
{a k }
k 1
(5)
This is the digital Wiener optimal one-step prediction filtering
problem, and it leads to the specification of the { bk , } as
the solution of the normal equations
Page 626
Sowmya and Chandra,
Study of Reverberation Time Series and Echo Detection Algorithm in Reverberation Limited Scenarios
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 625-631
ISSN 2078-2365
.
Fig-1. Linear Spectral Prediction Approach To Modeling Stochastic
Quasi-Stationary Coherent Time Series.
The Levinson-Durbin algorithm is utilized to solve the YuleWalker equations to obtain the optimal set of non stationary
poles, or more precisely, the optimal set of non stationary filter
coefficients for each beam.
Fig-2. Processing Flow Of Reverberation Simulation Model.
The generation of coherent reverberation which is correlated
between beams and overlapping spectra is illustrated in Fig.3,
where Ha, Hb, .-, Hn,, represent transfer functions of the linear
filters associated with beams, a, b, ...,n, respectively.
evaluation scheme based on a spatial grid approach of
Ackerman for a general formulation of reverberation
developed by Faure, Ol'shevskii and Middleton. This leads to
the computation of surface, volume, and/or bottom
reverberation spectra and Dower levels as functions of:
a) transmit signal waveform and power level,
b) spreading and absorption propagation losses,
c) backscattering strength,
d) transmit and receive beam patterns,
e) sonar platform-ocean geometry, and
f) sonar and scatter motion.
The formulation represents reverberation spectra at the input
of a receiver after beam forming, but prior to signal processing
operations. The reverberation is non stationary, in that the
spectra vary with time (and range), both in their spectral shape
and their power levels. The reverberation is quasi stationary,in
that the rate of change with time is assumed to be small
relative to the transmit pulse duration (or the reverberation
correlation time).Although the formulation can be extended to
relax some of the following , intrinsic assumptions include:
a) primary scattering only,
b) iso-sound-speed ocean,
c) direct propagation path,
d) narrow-band transmit signals,
e) back scattering is from a large number of randomly
distributed weak discrete scatters,
f) radial velocity distribution of scatters is spatially uniform.
A. Faure, Olshevskii and Middleton Formulation:
The power density spectrum Pf ( f , r ) of the reverberation
envelope at the receiver input from scatters at range r and
frequency f, letting * designate the convolution operations
given by
Pf ( f , r ) 2 (r )Yf ( f , r ) * STf ( f ) * D f ( f )
2
2 (r )
(9)
Where
= Total reverberation power from scatters at range
r,
Yf ( f , r ) = Sonar motion envelope (power density) spectrum
Fig-3. Parallel Structure of Coherent Multiple-Beam Time Series
Filter
resulting from sonar motion and stationary (non moving)
scatters at range r,
Y ( f , r )df 1
III. REVERBERATION SPECTRUM MODELING
The generation of a time series based on the existance of a
spectrum was discussed. Here, an approach to generating the
expected spectrum for sonar reverberation as a function of
pertinent sonar system parameters and environmental
conditions is presented. The approach is a numerical
STf ( f )
f
(10)
= Transmit signal envelope (energy density)
Spectrum,
Page 627
Sowmya and Chandra,
Study of Reverberation Time Series and Echo Detection Algorithm in Reverberation Limited Scenarios
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 625-631
ISSN 2078-2365
STf ( f ) df 1
Df ( f )
(11)
= Scatter motion spectrum resulting from random
motion of the scatters. This is equal to the probability density
function of Doppler frequencies of backscattered signals from
a large number of scatters with random motion and received by
a stationary (non moving) sonar which transmitted a pure tone
signal. Since it is a probability function.
D f ( f )df 1
(12)
The terms
, Yf and D f
2
depends on whether the
p02cTsv102r
(r )
2r 2
S2 (r )
b
S
B 0
TR
( , )dd
(13)
p02cTsS 102r
0 bTR ( S , ) cos S d
2r 3
B2 (r ) same as S2 (r ) , except s B
S respectively.
and
B
replace
(14)
s S and
(15)
Where
p02
=
transmit source mean square pressure,
C
= speed of sound in water (m/s)
T
= time duration of transmit signal
(s),
r
=
range from mono static sonar to center of
set of discrete scatterers (m),
= absorption loss ,
sv , s B , s S
=
= transmit power beam pattern
= receive power beam pattern
The surface backscattering strength can be represented as a
function of surface grazing angle w i nd speed, and frequency
(f) through models such as the Chapman-Harris model
SS 10 log sS
3.3 log( S 30) 42.4 log 2.6
158(f 0.33 ) 0.58
(18)
SB 10 log sB
10 log(sin 2 B ) 27 .
(19)
The bottom backscattering strength can be represented as
reverberation is from volume, surface or bottom scatters.
The reverberation power levels are given by
2
v
bT ( , )
bR ( , )
Where
2
volume (m-3), surface (m-') bottom (m-2)
bTR ( , ) = transmit-receive product power beampattern.
= azimuthal angle (rad),
= elevation angle (rad),
S
= sin^-1 (z/r) = elevation angle to surface at
The sonar motion envelope spectrum for volume, surface, and
bottom reverberation, for a sonar moving at constant speed v0
and constant direction are given by
j 4 T
S
0 )dd
b ( , ) exp(
TR
Y ( f , r ) F B 0
vf
S
bTR ( , )dd
B 0
j 4v T
0 cos cos ) cos dd
b ( , ) exp(
TR
S
b
Y ( f ,r) F 0
sf
(
,
)
cos
b
d
TR S
S
0
(20)
(21)
YBf ( f , r ) =same as Ysf ( f , r ) except S replace b .
(22)
backscattering strength, respectively,
B
range r and sonar depth z (rad),
Fig- 4. Spatial Division Of Reverberation Field.
=sin^-1(zb-z/r) = elevation angle to bottom, at
range r and sonar vertical distance above bottom of (zB - Z )
(rad).
bTR ( , ) [bT ( , ) bT ( , )].[bR ( , ) bR ( , )]
The transmit-receive beam pattern is further defined by
(16)
The power density spectrum Pf(f, r) of the reverberation
envelope given by (9) is evaluated numerically dividing space
into a set of cells, as illustrated in Fig. 6. The ocean is divided
into spherical shells which represent the portion of the ocean
that is ensonified by the signal wavefront at particular instants
of time after transmission (and corresponding ranges). The
Page 628
Sowmya and Chandra,
Study of Reverberation Time Series and Echo Detection Algorithm in Reverberation Limited Scenarios
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 625-631
ISSN 2078-2365
spherical shells are subdivided into a grid with three types of
cells corresponding to those which contribute to surface,
volume, and bottom reverberation. The location of each cell is
defined by the range and the azimuth and elevation angles to
the center of the cell relative to the platform velocity vector.
When the parameters ai of this AR model have been estimated,
the system function of pre-whiten filter is as follows:
H ( z)
1 ai z
1
p
i 1
i
(25)
Output data y(n) is obtained by data x(n) passing the above
system function
y(n) x(n) * z1[ H ( z)]
(26)
The parameters of this AR model can be estimated through
autocorrelation method, Burg method, Lattice recursive
algorithm and so on.
Fig- 5. Reverberation Spectrum Model
IV. DETECTION OF ECHO SIGNAL IN REVERBERATION
BACKGROUND
Reverberation is caused by seabed, sea surface and the in
homogeneity of the granule in the seawater. As a noise,
reverberation can influence the detection performance of target
echo and cause some serious problems to active sonar . Due to
the fact that reverberation and target echo are correlative and
their spectrums are close, how to restrain reverberation is a
problem necessary to be solved for active sonar. Inorder to
restrain the reverberation signal we are using pre whiten
method here.
B. Order Partition Pre-Whiten Algorithm
Order partition pre-whitens algorithm means that the
sampled reverberation data is processed according to time
order.
Firstly, reverberation data is partitioned into several
segments. Supposing that the echo signal s(n) is included in
the signal x(n) which is received by hydrophone during the
observation time T. The pulse width of s(n) is TP . Data x(n) is
partitioned into several segments and the width of each
segment
is
A. Pre-whiten method
The principle of this method is that an AR model is
established from reverberation and then a whiten filter is
designed using the power spectrum of this model.
Reverberation data is considered as Gaussian color noise. And
the spectrum transformations between adjacent data segments
are not obvious. The rationality of this hypothesis has been
verified. The performance of this method is nicer even echotoreverberation ratio is comparative low. On the premise of
this hypothesis, a method using all-pole pre-whiten filters is
proposed based on AR model at first. Then ,based on this
model ,order partition algorithim is brought forward in detail.
The AR model of reverberation data is shown in equation
r (n) a i r (n i) w(n)
p
i 1
(23)
where w(n) is a Gaussian white noise with an average of 0,
{a1, a2, …, ap} are estimated using the adjacent segment data.
And its power spectrum is shown in equation (2).
s( w)
2
1 a i e jwt
(24)
Fig 6. Sketch Map Of Data Partition
Firstly, reverberation data is partitioned into several segments.
Supposing that the echo signal s(n) is included in the signal
x(n) which is received by hydrophone during the observation
time T. The pulse width of s(n) is TP . Data x(n) is partitioned
into several segments and the width of each segment is TD as
manifested in Fig.1, where dT denotes offset between the
adjacent kth and k+1th data segments. The following
conditions should be satisfied when partitioning the data into
several segments:
(1) The width of each data segment should be equivalent to the
usable signal s(n) in order to satisfy the requirement of local
stationary.
p
i 1
Page 629
Sowmya and Chandra,
Study of Reverberation Time Series and Echo Detection Algorithm in Reverberation Limited Scenarios
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 625-631
ISSN 2078-2365
(3) To ensure that the echo signal s(n) is situated in the next
adjacent data segment
when s(n) is not completely included
in a certain segment, following in equation should be satisfied:
(28)
dT TD TP
power spectrum of transmitted signal
50
power spectrum of transmitted signal
45
40
power spectrum values in db
(2) To ensure that the echo signal s(n) is fully included in one
segment, the width of s(n) should be smaller than each data
segment. The following in equation should be satisfied:
(27)
TD TP
30
25
20
15
10
5
According to the actual situation, the width of each data
segment is double of the usable signal and the data
overlapping rate is ½.
(29)
TD 2TP
dT 1 / 2(TD )
35
0
50
100
150
sample values
200
250
300
Fig-8 Transmitted Signal Power Spectrum
scatter motion spectrum,
0.08
scatter motion spectrum
0.07
0.06
(30)
power spectrum
0.05
0.04
0.03
0.02
0.01
Secondly, the AR model of the kth data segment is
established. In the method, the system function of the
prewhiten filter which is based on the kth data segment for the
k+1th data segment is as follows:
1 a k ,i z i
pk
0
50
100
150
no of samples
200
250
300
Fig-9. Scattering Motion Spectrum
23
18
power density spectrum of reverberation
x 10
power density spectrum of reverberation
16
(31)
14
12
power density values
H k ( z)
1
0
i 1
10
8
6
Lastly, the output data
yk1 (n) is obtained by passing data
4
2
xk1 (n) through above system function:
0
0
(32)
The flow chart of order partition pre-whiten algorithm is
shown in below figure.
100
200
300
no.of samples
400
500
600
Fig-10. Power Density Spectrum of Reverberation
80
AUTOCORRELATION FUNCTION
yk 1 (n) xk 1 (n) * z1[ H k ( z)]
60
40
20
0
-20
0
10
20
30
40
50
FREQUENCY IN HZ
60
70
80
90
100
0
10
20
30
40
50
TIME SAMPLE
60
70
80
90
100
NORMAL TIME SERIES
80
60
40
20
0
-20
-40
Fig-11. Time Series from autocorrelation function
Fig 7:Flow Chart Of Order Partition Pre-Whiten Algorithm
7
4
time series signal
x 10
time series signal
3
2
V. SIMULATIONS AND RESULTS
The reverberation power spectrum using faure,olshevskii, and
middeleton formulation is obtained and according to the block
diagram and the time series using autoregressive model is obtained.
normal time series
1
0
-1
-2
-3
-4
0
100
200
300
time sample
400
500
600
Fig-12. Time Series from the reverberation spectra
Page 630
Sowmya and Chandra,
Study of Reverberation Time Series and Echo Detection Algorithm in Reverberation Limited Scenarios
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 625-631
ISSN 2078-2365
partition algorithm is the better approach for detecting the
echo signal in reverberation background.
echo signal
1
0.8
0.6
REFERENCES
0.4
amplitude
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0
200
400
600
time samples
800
1000
1200
Fig 13: Echo Signal
reverberation signal
1
0.5
amplitude
0
-0.5
-1
-1.5
0
200
400
600
time samples
800
1000
1200
Fig 14:ReverberationSignal
echo+revereberation signal
1
0.5
amplitude
0
[1]. S. G. Chamberlain and J. C. Galli, “A Model for Numerical Simulation
of Non stationary Sonar Reverberation Using Linear Spectral Prediction,”
IEEE journal of oceanic engineering VOL. OE-8, NO. 1, Jan. 1983, pp 21-36.
[2]. H. Weinbeg, "Navy interim surface ship model(NISSM) 11,"Naval
Underwater Svstems Ctr., New London, CT. NUSC Tech. Publ. 372 and
NUSC Tech. Rep. 4527, 1973.
[3]. P. C. Etter and R. S. Plum, "A survey of underwater acoustic models and
environmental acoustic data banks." Naval Anti- Submarine Warfare
Syst.Proj. Office, Dept. of Navy.Washington DC, ASWR Tech. Rep.80-1
15,1980.
[4] . C. L. Ackerman and R. L. Kesser. "Reverberation spectrum model for
matched filter homing systems." Pennsylvania State Univ. Applied Research
Lab, Tech. Memo. File TM73-285, Dec. 4, 1973.
[5]. P. Faure, "Theoretical model of reverberation noise,'' J.
Acoust.Soc.Amer.,vol. 36. no. 2. pp. 259-266.Feb. 1964.
[6]. V. V. Ol'shevskii, Characteristics of Sea Reverberation. English
translation by V. M. Albers. New York: Consultants Bureau,1967.
[7]. A. D. Waite, Sonar for Practising Engineers, 3rd ed., John Wiley & Sons,
Ltd., 2002.
[8]. Robert J. Urick, Principles of Underwater Sound, 3rd ed., New York
McGraw-Hill, 1983.
[9]. Manuel Aineto, Stuart Lawson, “Narrowband signal detection in a
reverberation-limited environment”, Oceans 97, MTS/IEEE Proceedings,
Nova Scotia, 1997, pp.27-32.
[10]. Liang Hong, Li Zhishun, “An Adaptive Method to Detect Moving
Target
in a Reverberation Background”, Applied Acoustics, Vol.22, No.2, 2003,
pp.26-29.
-0.5
-1
-1.5
0
200
400
600
time samples
800
1000
1200
Fig 15 : Echo+Reverberation Signal
restraining the reveraberation
1
0.8
0.6
0.4
amplitude
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0
200
400
600
time samples
800
1000
1200
Fig 16: Restraining The Reverberation Signal
VI. CONCLUSION
This paper describes the approach to the numerical
generation of reverberation time series and echo detection
algorithm. on the premise of local stationary of reverberation,
a method using all pole partition pre whiten filters which is
based on AR model is proposed. The simulation results
indicate that this method is effective even in the background of
low echo to reverberation ratio of input. Moreover, the order
Page 631
Sowmya and Chandra,
Study of Reverberation Time Series and Echo Detection Algorithm in Reverberation Limited Scenarios
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 620-624
ISSN 2078-2365
Climate Change Prediction by Wireless Sensor
Technology
Chin-Yuan Hsieh
Graduate School of Information Technology and
Applications,
Kao Yuan University, Kaohsiung 821, Taiwan
Email: cyhsieh@cc.kyu.edu.tw
Abstract – The prediction of the temperature microchange in the
forest becomes critically important due to the study of global
change. In this paper we develop a model to predict and modify the
measured temperature by the Zigbee and wireless sensor network
technology. The modification model is developed by a pair of
integral equations. The tangential surface electric and magnetic
fields are induced by the surface current flowing in the tree skin.
The surface current is generated by the electromagnetic wave
hinging on the rough tree surfaces. The scatter field including the
Kirchhoff and complementary fields from the tangential surface
electric and magnetic fields is measured by the Zigbee receiver with
the wireless sensor network. The wireless sensor network (WSN)
measurement system includes temperature sensor, zigbee and
Tmote Sky chip. From the measured temperature in the tree skin we
found the temperature data of 24 hours in a day show an
exponential function distribution. We also found the temperature
reaches peak at different time from the measured data. The time
shift is in the range of 1 to 2 hours. The time shift could be from the
different dielectric constant of tree skin and the tree density in the
forest. After comparing the model prediction with the measured
data, the correlation coefficient is 0.94. The excellent prediction
reaches in this model prediction.
Keywords: temperature microchange, wireless sensor network,
global change, integral equation, WSN.
I. INTRODUCTION
The study of wireless sensor networks is challenging in
that it requires an enormous breadth of knowledge from an
enormous variety of disciplines. The temperature microchange
of forest tree cannot be measured by the thermometer. The
microchange of forest temperature is a critical factor to the
growth and the type of trees, therefore the forest density,
distribution and the temperature will affect the global
temperature change. In general, forests are sensitive to the
variability and change of climate. The Climatic factors
influence forest health-temperature, rainfall, atmospheric
levels of carbon dioxide, other greenhouse gases and extreme
weather. The microchange of tree temperature cannot be
Chen-Yu Hsieh
Dept. of Mechatronics, School of Engineering Science,
Simon Fraser University, BC, Canada
Email: chen-yu_hsieh@sfu.ca
measured by the traditional thermometer and need to be
measures and studied by the wireless sensor network
technology [1-4]. In this paper a wireless sensor network
based on a ZigBee technique for the temperature microchange
was proposed. The wireless sensor network consists of
spatially distributed autonomous sensors to monitor the
environmental or physical conditions, such as temperature,
humidity, light, pollutants, or vibration and to pass their data
through the network to a main location of coordinator,
notebook, smart phone or ipad cooperatively. The ZigBee
device is a specification for a suite of high level
communication protocols using small, low-power digital
radios based on an IEEE 802 standard for personal area
networks. The technology defined by the ZigBee specification
is intended to be simpler and less expensive than other
wireless personal area network, such as Bluetooth. ZigBee is
targeted at radio-frequency applications that require a low data
rate, long battery life, and secure networking. ZigBee has a
defined rate of 250 kbps best suited for periodic or intermittent
data or a single signal transmission from a sensor or input
device. The proposed modern measurement technique is able
to offers the change rate of forest temperature by the wireless
communication in real time. The architecture of a wireless
sensor network for measuring the forest temperature
microchange includes the sensor, electromagnetic wave
transmitter and receiver, hardware circuitry and software
system. The hardware circuitry of the network node is
designed based on a Tmote Sky chip. The The Tmote Sky
platform is a wireless sensor module IEEE 802.15.4 compliant
and has the characteristics of Ultra low power based on a TI
MSP430 and Chipcon CC2420 radio. From the data collected
by the system, the model is developed to predict the
temperature microchange. In last section we compare the
model prediction and the measured data to evaluate the
prediction of model.
Page 620
Chin and Chen, Climate Change Prediction by Wireless Sensor Technology
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 620-624
ISSN 2078-2365
II. METHODOLOGY
The aim of the paper is to apply the wireless sensor
technologies and wireless communications technology for the
temperature microchange measurement of trees in the forest. A
wireless sensor communication network is composed of nodes,
each of which can compute power, transmit and receive
messages over communication links, wireless or cabled. A
single wireless sensor communication network may consist of
several interconnected subnets of different topologies. In this
measurement the mesh network is applied. Mesh nets can be
good models for large-scale networks of wireless sensors that
are distributed over a geographic region. Mesh networks are
regularly distributed networks that generally allow
transmission only to a node’s nearest neighbors. The nodes in
these networks are generally identical, so that mesh nets are
also referred to as peer-to-peer nets.
An advantage of mesh nets is that, although all nodes may be
identical and have the same computing and transmission
capabilities, certain nodes can be designated as ‘group leaders’
that take on additional functions. If a group leader is disabled,
another node can then take over these duties. The required
transmission power increases as the square of the distance
between source and destination. Therefore, multiple short
message transmission hops require less power than one long
hop. In fact, if the distance between source and destination is
R, the power required for single-hop transmission is
proportional to R2. If nodes between source and destination are
taken advantage of to transmit n short hops instead, the power
required by each node is proportional to R2/n2.
In this paper the measurement system focuses on ZigBee
devices based on the Wireless Sensor Networks. A general
Wireless Sensor Network protocol consists of the application
layer, transport layer, network layer, data link layer, physical
layer, power management plane, mobility management plane
and the task management plane [1-4]. In the paper we applied
the standard WSN with zigbee technology operating in the
Industrial Scientific and Medical (ISM) frequency band of 2.4
GHz. The ISM frequency band provides license free
operations, huge spectrum allocation and worldwide
compatibility. For monitoring the long distance data the
architecture of the WSN technology with Wi-Fi (IEEE 802.11)
and PC-based systems is developed in Fig. 1.
Fig. 1 The layout structure of multi-hop in wireless sensor
network technology system
For measuring the tree temperature we install the zigbee
sensor node in the fir and gingko to collect the temperature by
the wireless sensor network technology. The data are
retrieved, processed and transmitted for the real time
monitoring. For avoiding the human interference the height of
sensor node is installed in 3 meters far from the ground.
Meantime we consider the sunlight will affect the data
transmission status, the directions of sensor nodes faces north
to make the measurement precisely. For the tree routing of
Zigbee, a Zigbee node transmitting a packet to the destination
is set up to follow the tree mesh topology. The direction of
zigbee antenna is directional radiation pattern to increase the
received signal strength and the communication distance. All
the nodes are distributed and installed along the zigzag path to
enhance the successful transmission rate.
The TmoteSky wireless system can transmit the power
consumption status by wireless at contant time period. Tmote
Sky is the wireless sensor module for high data-rate sensor
network applications with ultra low-power, high-reliability and
ease of development. Tmote Sky also offers a number of
integrated peripherals including a 12-bit ADC and DAC,
Timer, and UART bus protocols, and a performance boosting
DMA controller. The TmoteSky wireless sensor with USB
connector is developed by the TinyOS operation system in
Fig. 2.
Fig. 2 The TmoteSky wireless sensor network chip with USB
connector
The operating voltage is in the range of 2.1V to 3.6V. The
time interval can be adjusted as requirement. To save the
consumption of power the transmission period is set up 5
minutes for each transmission. The memory for the data
storage is 1MB. If the system is busy, the data to be
transmitted will be stored in a short time. The Zigbee is a
wireless network standard based on IEEE 802.15.4. The
transmission range of a wireless sensor network is usually low,
less than 100 meters, therefore the characteristics of WSN are
low data rate, short distance, low price and low power
consumption. The compute capability is limited based on the
low complexity, and limited resources.
Page 621
Chin and Chen, Climate Change Prediction by Wireless Sensor Technology
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 620-624
ISSN 2078-2365
III. MODEL DEVELOPMENT
For adjusting the measured temperature the electromagnetic
wave scattering from the rough surface of a tree will be
predicted. The prediction model is developed based on a pair
of integral equations. The integral equation is to solve the
scatter strength of electromagnetic wave. The scatter strength
is induced by the surface current along the tree skin, and the
surface current is generated by electric and magnetic fields
from the transmitter. In medium 1 (incident plane) the
governing integral equations for the tangential surface fields
on a dielectric surface are [5]
2
nˆ E 2nˆ E i
nˆ E ds
4
[1]
i
2
nˆ H 2nˆ H
nˆ H ds
4
2
nˆ t Et
nˆ t Et ds
4
2
nˆ t H t
nˆ t H t ds
4
2Re Eqp c Eqp k Eqp c Eqpc
*
*
*
(7)
where Re is the real part operator and * is the symbol for
complex conjugate.
(b)
Fig. 3 The scatter and rescatter fields from random rough
surfaces of a tree
To obtain the coherent and incoherent power, we have to
subtract the mean-squared power from the total power. That is,
[2]
Eqp s Eqp s Eqp s Eqp s
*
(3)
H t jkt (nˆ t Et )Gt / t (nˆt H t ) Gt (nˆt H t )Gt
(4)
The spectral representation for the Green's function and its
gradient are
1
j
) exp[ ju(x x) iv(y y) jq z z]dudv
2 q
1
g
G
( ) exp[ ju(x x ) iv(y y) jqz z]dudv
2 q
G (
2
2
2
where q = k - u - v and g xˆu yˆv zˆq .
The tangential surface field includes two components, the
Kirchhoff component (scatter field) and its complementary
component (rescatter field), and is described in Fig. 3. Two
corresponding components for the scattered fields are [6]
[1] E qp E qp E qp
(6)
Where s means the tangential surface field, k means the
Kirchhoff field and c means the complimentary field in
c
Eqp k Eqpk
(a)
Et jkt t (nˆ H t )Gt (nˆt Et ) Gt (nˆt Et )Gt
k
*
(2)
where
E jk (nˆ H )G (nˆ E ) G (nˆ E )G
H jk (nˆ H )G / (nˆ H ) G (nˆ H )G The fields
in the lower medium (medium 2) can be written in terms of the
fields in the upper medium (medium 1) by applying the
boundary conditions on the continuity of the tangential fields
as follows:
s
Eqp s Eqp s
(1)
In medium 2, we have
(5)
equation (6). In terms of the surface tangential field for the
dielectric surfaces, the far-zone scattered fields can be derived.
Consequently the average scattered power and the scattering
coefficients can be found in terms of the far-zone scattered
field. With the field expression the average scattered power is
given by
Eqp Eqp
k
k*
Eqp
k
*
Eqp
k *
Eqp Eqp
c
2 Re[ Eqp c Eqp k Eqp c Eqp k ]
*
c*
E qp
c
c *
Eqp
*
(8)
To carry out the ensemble average operation we make an
assumption the surface with Gaussian height distribution.
After the operation of ensemble average and integrating, the
scattered power can be obtained. The bistatic scattering
coefficient,
Pqp
qp0
, is related to the average power expression ,
qp (4R2 Pqp ) /( E0 2 A0 )
, as
0
(9)
Where Eo is the incident field, R is the distance, Ao is the
illuminated area of a tree and P qp represents the polarized and
depolarized scatter power. The subscript
represents the
polarization of incident electromagnetic wave and subscript
represents the polarization of scatter electromagnetic wave.
From the complex calculation we summarized the scatter
coefficient of the summation of Kirchhoff and complementary
scatter strength from the rough tree skin and becomes
qp s
exp{
2
(kL) 2 ( k )2 (cos s 2 cos 2 )
(k ) 2 n I qp
e
4
n 1
(kL) 2
[(sin s cos s sin cos ) 2 (sin s sin s sin sin ) 2 ]}
4n
n n!
(10)
Page 622
Chin and Chen, Climate Change Prediction by Wireless Sensor Technology
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 620-624
ISSN 2078-2365
I qp (cos s cos ) 2 n fqp exp[ 2(k ) 2 cos s cos
2
2
{[cos s (cos s cos )] f Fqp (kx ,k y )
n
*
qp
[cos (cos s cos )]n fqp Fqp (ksx ,ksy )]} exp[ (k ) 2 (cos cos s )]
*
2
1
2n
*
[(cos s ) Fqp (kx ,k y ) (cos s cos ) n Fqp (kx ,k y ) Fqp (ksx ,ksy )
4
(cos cos s ) n Fqp (kx ,k y ) Fqp (ksx ,ksy ) (cos 2 n ) Fqp (ksx ,ksy ) ]
2
*
where
The
f
Kirchhoff and the complementary field coefficients qp and
Fqp are given in appendix of [6]. The measured is modified by
the prediction of a pair of integral equations. To evaluate the
prediction model we further apply the correlation coefficient
in this paper. The correlation coefficient is a measure of how
well trends in the predicted values. It is a measure of how well
the predicted values from a forecast model "fit" with the real
measured data. The correlation coefficient, , is a quantity that
gives the quality of a least squares fitting to the original data.
2
The coefficient of determination, r , is the square of the
correlation coefficient. The coefficient of determination is
useful because it gives the proportion of the variance of one
variable that is predictable from the other variable. The
coefficient of determination is the ratio of the variation of the
measured average temperature to the total variation and
represents the percent of the measured average temperature
that is the closest to the line of best fit. The coefficient of
determination is computed as:
r 2 1 (SSE / SST )
SSE (Ti Tˆi ) 2
(13)
The SST (Total Sum of Square) measures the deviations of the
measured data from their mean and is the sum of squared
deviations of individual measurements from the mean:
SST (Ti T ) 2
Ti
20
Temperature
15
10
5
0
10
20
30
Time (Hour)
means the
temperature measured. In (14) T means the average value of
the measured temperature. The smaller sum of square error,
SSE, the more reliable the predictions obtained from the
model. The coefficient of determination takes on values
between 0 and 1. The higher the coefficient of determination,
the more available the model prediction is.
IV. RESULT AND DISCUSSION
Temperature
25
0
(14)
In (13) T̂ means the model prediction value and
Table 1 the average temperature during 24 hours among 25
days
o
o
o
time Temp.( C tim Temp.( C tim Temp.( C
)
e
)
e
)
1
11
9
11.82
17
15.06
2
12
10
14.01
18
14.01
3
11
11
15.06
19
11.82
4
12
12
18.7
20
12.27
5
11
13
20.04
21
11.07
6
12.01
14
22
22
12.01
7
11.07
15
20.04
23
11
8
12.27
16
18.7
24
12
Temperature (degree c)
(12)
Where SSE (Error Sum of Square) measures the deviations of
measured data from their predicted values. In general the SSE
is the sum of the squared differences between each observation
and its group's mean. It can be used as a measure of variation
within a cluster:
From the measured data the change rate of temperature is
proportional to the time. During the time closing the noon
the change of temperature is large each day. The data are
measured 25 days in a month. The average temperatures of
24 hours are listed in table 1. From the measured data in Fig.
3 the change rate of temperature is related with time and
show the exponential relation. From table 1 the peak value
of average temperature is at 14:00 instead of 12:00. The
time shift may be from the forest density and the dielectric
permittivity of tree skin.
Fig.4 The measured average temperature v.s. hours
From Fig. 4 the minimum average temperature of trees in the
forest is 11oC, but the maximum average one is 22oC. The
average temperature in a month is 13.83 oC and the variance
of average temperature is 11.7oC. It is interesting to find the
time of temperature peak is 22oC at 14:00 instead of at noon.
The time shift may be caused by the dielectric coefficient of
the tree skin and the tree density in the forest. From the
measured data we also found the change rate of temperature
is proportional to its second-order differentiation. In other
Page 623
Chin and Chen, Climate Change Prediction by Wireless Sensor Technology
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 620-624
ISSN 2078-2365
words the second-order differentiation of average
temperature collected is proportional to the measured average
temperature. It means the average temperature has the
characteristics of the exponential function distribution:
(10)
T k k1e
It is interesting to find the peak value of average
temperature in not at noon in a whole day, but time shift
happens in Fig. 5. The measurement is taken 25 days in a
month. The mean value of temperature measured is 13.83.
The model of exponential function shows the temperature
change rate is faster in the daytime, but slower at night. The
predicted exponential equation (10) with both of k1= 10 and
k2= 50 and modified by constant k= 3. The value of m is the
hour on maximum average temperature. In this study the
value of m is 14 in Fig. 5. After comparing the measured
average temperature of the ginkgo tree with that of the
model prediction, the correlation coefficient is 0.935736.
The model prediction is excellent.
( t m) 2 / k2
V.
CONCLUSION
The microchange of forest temperature is a critical factor to
the growth of trees. Due that the forest temperature and
density will affect the global change the forest temperature
microchange becomes critically important for studying the
global climate change. In the paper we develop a novel model
to study the forest temperature microchange during a month.
We also construct a novel measurement technology by
wireless sensor network. From the measurement we found the
microchange of tree average temperature has an exponential
function distribution 24 hours in a day. It is attractive that the
peal value of tree temperature reaches at the different time
interval instead of noon in a day. From the comparisons of the
measured temperature with the model predictions, we found
the correlation coefficient between the prediction and the
measurement is 0.94. From the model analysis, the dielectric
constant of tree skin and density of forest trees maybe the
main factors to cause the time shift of the temperature peak. In
the future we will investigate it.
Fig. 5 the comparisons of model prediction with measured
temperature of ginkgo tree
VI.
REFERENCES
[1]
Yoo, S.; Kim, J.; Kim, T.; Ahn, S.; Sung, J.; Kim, D. A2S:
Automated agriculture system based on WSN. In ISCE 2007. IEEE
International Symposium on Consumer Electronics, 2007, Irving, TX,
USA, 2007.
[2]
Goense, D.; Thelen, J. Wireless sensor networks for precise
phytophthora decision support, ASAE Annual International Meeting;
Tampa, FL, USA, 2005.
[3]
Lea-Cox, J.D.; Kantor, G.; Anhalt, J.; Ristvey, A.; Ross,
D.S. A wireless sensor network for the nursery and greenhouse
industry. In Southern Nursery Association Research Conference, Vol.
52, 2007.
[4]
Ruiz-Garcia, L.; Barreiro, P.; Rodríguez-Bermejo, J.;
Robla, J.I. Monitoring intermodal refrigerated fruit transport using
sensor networks: a review. Span. J. Agric. Res., 5, 142-156, 2007.
[5]
Poggio, A.J., and E.K. Miller, “Integral Equation Solution
of Three Dimensional Scattering Problems,” Computer Techniques
for Electromagnetics, Pergamon, New York, Chapter 4, 1973.
[6]
Chin-Yuan Hsieh, Adrian K. Fung, Giuseppe Nesti, A.
Sieber and Peter Coppe, A Further Study of the IEM Surface
Scattering Model, IEEE Transactions On Geoscience and Remote
Sensing Society, PP. 901-909, Vol. 35, N0. 4, July 1997
[7]
Chin-Yuan Hsieh, Polarimetric Bistatic Scattering From
Randomly very Rough Surfaces, Microwave and Optical Technology
Letters, Vol.25, No. 4, May 20, 2000
25
20
15
10
T(Ginkgo)
Data(Ginkgo)
5
0
1
3
5
7
9
11
13
15
17
19
21
23
Page 624
Chin and Chen, Climate Change Prediction by Wireless Sensor Technology
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Robust Neuronal Adaptive Control for a Class of
Uncertain Nonlinear Complex Dynamical
Multivariable Systems
Farouk Zouari
LAboratoire de Recherche en Automatique (LARA)
École Nationale d’Ingénieurs de Tunis (ENIT), B.P. 37, 1002 Tunis
Email:zouari.farouk@gmail.com
Abstract: In this paper, we proposed the development of
neural adaptive controls to ensure the robustness of
uncertain nonlinear multivariable systems. We used two
techniques: Robust neural adaptive control and neural
indirect adaptive control. The study of the stability and
robustness of both techniques was performed by Lyapunov
theory. To validate these techniques and discover their
effectiveness, a simulation example was considered. The
simulation results obtained by these two control techniques
have shown the effects of disturbance compensation, good
performance tracking data paths and stability control
systems. Comparative studies between these two techniques
show that the neural indirect adaptive control cannot
mitigate the effect of disturbances compared to the robust
neural adaptive control.
Keywords: Neural adaptive controls; Robust neural
adaptive control; Neural indirect adaptive control;
Lyapunov theory; uncertain nonlinear multivariable
systems.
I.
INTRODUCTION
In recent decades, the robust neuronal adaptive control
of complex nonlinear dynamic systems has been
studied in several research works which we quote [1-6].
It is used in several industrial applications and
particularly in the cases where we are confronted with
complex nonlinear dynamics and inaccuracies due to
Page 671
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
uncertainties attached to the system to be controlled.
The use of the method of robustification is essential to
improve the tracking performance and ensure the
robustness of the closed loop system in front of the
structural uncertainties and external disturbances. This
control technology adds to the main control signal a
supervisory signal by sliding mode or of type.
Several studies of robustification of sliding mode
adaptive neural control are based on the use of adaptive
neural networks for modeling the process or to
calculate the desired control law [7-11]. Generally, the
control laws were derived from the examination of
stability. Overall, the constructed command ensures
stability and good tracking performance. The
disadvantage of sliding mode adaptive neural control is
the existence of the sign function that causes sudden
and rapid changes of the control signal, which can
excite the high frequency of the process and cause
damage it. Many solutions have been proposed in
literature in particular Lie Slotine and added a transition
band around the sliding surface to transform the sign
function in saturation and thus remove the abrupt
changes [12].
Several research works have studied the technique like
[13-16]. This technique aims at determining the
tracking performances based on a criterion connecting
on the one hand the norms of prosecution errors and
on the other hand the desired level of disturbance
attenuation. This criterion can be interpreted in the state
space by obtaining a positive definite matrix unique
solution of the Riccati equation.
The contribution of this paper is to propose two
adaptive controls neuronal structures of nonlinear
dynamical multivariable systems rested on the theory of
Lyapunov. The architecture and learning algorithm of
these two neural adaptive control structures require the
modeling of system to be controlled, that is to say the
determination of its state equations using the concepts
of neural networks. In this sense, we proposed
linearization technique inputs- outputs of the system to
be controlled based on neural networks. This technique
consists in finding linear relationships between inputs
and outputs of the system. The neural model of the
system obtained online by this linearization technique
is used to calculate the commands laws. In fact, the first
proposed control structure which is the neural indirect
adaptive control uses the Jacobian matrix of the neural
model during the calculation of the parameters of
neural controller. by cons the second proposed control
structure which is the robust neural adaptive control,
use the state equations derived from the neural model to
determine the neuronal controller and add to main
control signals supervision signals by the technique .
This document is organized as follows: In Section 2 we
present the proposed architecture of recurrent neural
network and its learning algorithm based on Lyapunov
theory, used in the calculations of the model parameters
of a complex nonlinear dynamic multivariable system.
In Section 3 we show the architecture of the neuronal
controller and its proposed learning algorithm in the
structure of neuronal indirect adaptive control. The
proposed structure of robust neural adaptive control by
the technique is described in Section 4. The numerical
results and discussions of these two commands
mentioned previously are presented in Section 5.
Finally, the conclusion is given in Section 6.
II.
NEURAL NETWORK MODELING APPROACH
Neural network modeling of a system from samples
affected by noise usually requires four steps:
Page 672
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
- The first step is the choice of the architecture of the
neural network, that is to say the number of neurons in
the input layer which is a function of past values of
inputs and outputs, the number of hidden layers, the
number of neurons in each hidden layer, the number of
neurons in the output layer, the activation functions of
each neuron and organization of these neurons between
themselves [17-28].
-The second step is the normalization or the
transformation performed on the data inputs-outputs to
distribute them uniformly and adapt them to an
acceptable level for the neural network [29-31]. All
data values must be between or .
a. Proposed architecture
neural network:
of
recurrent
The figure 1 shows the architecture of the neural
network used during the identification phase of an
uncertain and perturbed nonlinear complex dynamic
multivariable system (with inputs and outputs).
- The third step is learning or in other words the
calculation of network parameters from samples inputsoutputs system to be identified [32-34].
- The fourth step is the validation of the neural network
obtained by using the tests measurements performances
criteria.
The structure of this neural network is composed of
three parts: two linear parts to model the behavior
linear of the system and a nonlinear part to approximate
the nonlinear dynamics.
Page 673
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
u1 k 1
u p k 1
u1 k 2
u1 k nb1
u p k 2
u p k nbp
f1
y1 k 1
y1 k na1
f1
yp k 1
f2
l1 k 1
f2
yp k nap
l k n
f1
l1 k
lp k
z
f2
f2
y1 k
yp k
Page 674
1
1
Farouk Zouar, 1 Robust cNeuronal
Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
z1
l p k 1
l p k ncp
z1
z1
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Figure 1- Proposed architecture of the neural network
y k y1 k ,
with:
, y p k is the vector of the neural network outputs at instant k .
T
u k 1 u1 k 1 ,
y k y1 k ,
, u p k 1 is the vector of the system inputs.
T
, yp k is the vector of the system outputs.
U k 2 u1 k 2 ,
T
, u1 k nb1 ,
, u j k 2 ,
, u j k nbj ,
(1)
Y k 1 y1 k 1 ,
, y1 k na1 ,
, y j k naj ,
, u p k 2 ,
, y j k 1 ,
, yp k 1 ,
, u p k nbp
, yp k nap
T
T
such as 2 nbj , 1 j p
such as 1 naj , 1 j p
(2)
Page 675
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
k l1 k , , l p k is the vector of the second hidden layer outputs of neural model
T
Yr k 1 l1 k 1 ,
, l1 k nc1 ,
, l j k ncj ,
, l j k 1 ,
, l p k 1 ,
, l p k ncp
T
such as 1 ncj , 1 j p
(3)
f1 x
e x 1
and f2 x x are the activation functions of neurons.
e x 1
nh the number of neurons in the first hidden layer.
The coefficients of the vector of the neural model parameters w are decomposed into eight groups, formed
respectively by:
w111
w1
1
w nh 1
w11nr
the weights between neurons of the input layer and neurons of the first hidden layer,
w1nh nr
w311
w3
3
w p1
w31nh
3
w pnn the weights between neurons of the first hidden layer and neurons of the second hidden layer,
w2
11
w2
the bias of neurons in the first hidden layer,
2
w nh 1
w4
11
w4
4
w p1 the bias of neurons in the second hidden layer,
Page 676
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
w511
w5
5
w p1
w51nr
5
w pnr
w611
w6
6
w nh 1
the weights between neurons of the input layer and neurons of the second hidden layer,
w61nh
the weights between neurons of the first hidden layer,
6
w nh nh
w711
w7
7
w p1
w71 p
the weights between neurons of the second hidden layer,
7
w pp
w811
w8
8
w p1
w81 p
the weights between neurons of the input layer and output layer.
8
w pp
xh k
11
x k
the first hidden layer outputs of neural model,
h
x nh 1 k
h
nr naj ncj nbj p
p
p
p
j 1
j 1
j 1
(4)
The vector of the neural model parameters is as follows:
w w111 ,
,w1nh nr ,w211 ,
w311 ,
,w2 nh 1 ,
,w3 pnh ,w411 ,
,w5 pnr ,w611 ,
,w4 p1 ,w511 ,
,w6 nh nh ,w711 ,
,w7 pp ,w811 ,
,w8 pp
T
(5)
T
T
T
T
(k) U k 2 , Y k 1 , Yr k 1 1 k , , n k
T
r
(6)
The vector of the first hidden layer outputs is in the following form:
Page 677
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
xh1 k f1 s1 k
xh k
h
x n k
h
f1 snh k
(7)
Such as:
1
s1 k w 11 k
S k
1
snh k w nh 1 k
w11nr k 1 k w611 k
1
w nh nr k nr k w6 nh 1 k
w61nh k xh1 k 1 w211 k
w6 nh nh k xh nh k 1 w2 nh 1 k
(8)
The vector of neural model outputs y k is given by:
y k w7 k
1
11
7
y p k w p1 k
3
w71 p k l1 k 1 w 11 k
7
w pp k l p k 1 w3 p1 k
w511 k
w5 k
p1
w31nh k xh1 k
3
h
w pnn k x nh k
w51nr k 1 k w411 k w811 k
w5 pnr k nr k w4 p1 k w8 p1 k
w81 p k u1 k 1
w8 pp k u p k 1
(9)
Assuming that:
Page 678
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
h1 (k) w711 k
hp (k) w7 p1 k
w511 k
w5 k
p1
3
w71 p k x1 k 1 w 11 k
7
w pp k xp k 1 w3 p1 k
w51nr k 1 k w411 k
w5 pnr k nr k w4 p1 k
w31nh k xh1 k
w3 pnn k xh nh k
(10)
Equation (9) then becomes:
y k h (k) w8 k
11
1
1
h (k) w8 k
p1
y p k p
w81 p k u1 k 1
w8 pp k u p k 1
(11)
, p,qs max nb1 ,
Equation (11) can be rewritten in the following state representation:
xi k 1 xi 1 k i n j ,
(12)
, mj 1, j 1,
, nbp , na1 ,
, nap , nc1 ,
, ncp , n j j 1 * qs 1 mj j * qs
xmj k 1 h j (k qs ) w8 ji k qs ui k qs 1
p
i 1
y j k xn j k
Where:
n p * qs
T
x k x1 k , , xn k
n
(13)
We can write equation (12) as follows:
Page 679
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
x k 1 Ax k B H k qs g k qs u k qs 1
T
y k C x k
(14)
With:
M1 0
A 0
0
A
j 1,
0 1
0 0
M j
0
0
Mj
0
n n
0
0
M p
,p
0
1
0
0
0
0
0
1
0
(15)
qs qs
Page 680
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
0
b1 0
B 0
0
0 bp
0
B n p
j 1, p
0
b
j
0
1
b j qs
(16)
0
c1 0
0
C
0
0 c p
0
C n p
j 1, , p
1
0
c
j
0
c j qs
(17)
Page 681
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
h1 (k qs )
H ( k qs )
hp (k qs )
p
H ( k qs )
j 1, , p
h j (k qs )
(18)
w811 k qs
w81 p k qs
g k qs
8
w8 p1 k qs
w
k
q
pp
s
g ( k qs ) p p
i 1, , p j 1, , p
w8ij k qs
(19)
Define the modeling errors by:
ei k yi k yi k i
p
(20)
e k yy k x k
and the modeling errors vector of all states is defined by:
(21)
with:
yy k y1 k , , y1 k qs 1 , , y j k , , y j k qs 1 ,
yp k , , yp k qs 1
T
n
,
1 j p
(22)
Combining (14) and (21), the dynamic of modeling errors is then given by:
Page 682
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
e k 1 Ae k B H k qs g k qs u k qs 1 y k qs
emoo k C T e k
(23)
where :
emoo k e1 k ,
, e p k
T
(24)
III.
NORMALIZATION TECHNIQUES OF DATA
There are two techniques of normalization:
-Min-Max normalization:
This technique performs a linear transformation on the
original data so that all values are in the interval a , b .
The formula of the normalization min-max is the
following:
min
'
b a a
max min
'
with :
is the data value to normalize.
' is the new data value after the
normalization, ' a , b .
10
(26)
is the smallest integer such as max ' 1 .
where :
IV.
(25)
LEARNING ALGORITHM OF NEURAL
NETWORK
In this section, we propose a theorem 1 that can be
used during the learning phase of a neural network.
min and max are respectively the minimum and the
maximum of data value to normalize .
0
w(k 1) 1
p
2 yi k
i
w k
i 1
- Normalization by decimal scaling:
The data are normalized by the following formula:
Theorem 1:
The learning procedure of a neural network may be
given by the following equation:
p
y k
0 i i
ei k
m
w k
i 1
w k
i w k i
2
2
p
i 1
yi k
2 i
w k
i 1
(27)
m1
such as :
(28)
Page 683
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
i 0i 0, m
(29)
1
m
i 0
i
0
(30)
0
(31)
0 i i 1, p
(32)
i i 1, p
(33)
(34)
(35)
p
i 1
i
(36)
the Euclidean norm.
T designates the transpose operator.
Proof:
V k
e k 2 e k
2
i
i
Using the following Lyapunov function:
p
i 1
2
i
p
i 1
i
2
2
w k
2
2
w k
2
(37)
ei k ei k ei k 1
with :
w k w k w k 1
(38)
(39)
Page 684
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
V k i ei k ei k i ei k w k w k w k 0
The learning procedure of the neural network is stable if:
p
p
i 1
i 1
T
2
2
(40)
V k i ei k ei k i ei k w k
Using the previous equation, we can write:
p
p
i 1
i 1
2
T
w k
w k 1
2
Such as: 1 0 .
(41)
therefore:
w k
2
p
ei k
i
w k
i 1
2
p
e k
T
w k w k i ei k i
w k 1 0
i 1
(42)
e k
w k i ei k i
i 1
w k
1
2
p
ei k
4 i
w
k
i 1
For equation (42) has a unique solution, it is necessary that:
p
2
The term w k can be written as follows:
(43)
p
ei k
w k i ei k
w
k
1
i
w k
2
p
ei k
2 i
w k
i 1
(44)
Like:
lim i zi 1 1
m
z1
i 0
(45)
Page 685
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
We can write:
w k 1
m
i 1
i z
w k 1
i 0
0 w k i w k i
m
i 1
0
p
ei k
2 i
w k
i 1
0
p
yi k
2
i
w k
i 1
(46)
therefore :
p
e k
i i
ei k m
0
w k
i 1
w k
i w k i
2
2
p
i 1
ei k
2 i
w k
i 1
p
y k
0 i i
e k
w k i
m
i 1
w k
i w k i
2
2
p
i 1
yi k
2 i
w k
i 1
0
w k 11
p
2 yi k
i
w k
i 1
p
yi k
e k
0 i
w k i
m
i 1
w k
i w k i
2
2
p
i 1
yi k
2 i
w k
i 1
(47)
The choice of the initial synaptic weights and the bias can influence the convergence speed of the learning algorithm
of neural network [35-46]. According to [47], the weights can be initialized by a random number generator with a
uniform distribution between and or a normal distribution N 0, 2 .
- If the weights are initialized by a random number generator with a uniform distribution:
s
3
nr
2
nr 1 1 m 0
m1
(48)
Page 686
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
- If the weights are initialized by a random number generator with a normal distribution:
s
1
nr
2
nr 1 1 m 0
m1
(49)
where :
V.
s 2.29
VALIDATION TESTS OF THE NEURAL
MODEL
Most validation tests use a set of samples inputsoutputs which have not been used in learning. Such a
test set or validation should if possible cover the same
range operating as the set of training samples.
The residues
ei k i 1,
, p obtained from the
estimated model parameters represent non-measurable
disturbance presented within the system. The residues
must constitute independent random sequences thus
assimilating the prediction errors to white noise.
Qi 100% 1
N
k 1
yi (k ) yi (k )
1
yi (k )
N
k 1
N
N
k 1
2
yi (k )
2
i 1,
Various tests called whiteness tests residues were
developed to validate this property. These validation
tests of a model are based on analysis of prediction
errors, on the Nash-Sutcliffe criterion, on the autocorrelation of residues, on the cross-correlation
function between the residues and other inputs in the
system [31, 48-52].
The Nash-Sutcliffe criterion relating to on each output
is given by the following relation:
,p
(50)
The correlation functions are:
- Autocorrelation functions of the residues:
e (k) N e (k) e (k ) N e (k)
N
Rei ei ( )
k 1
1
i
N
k 1
i
1
e (k)
N
i
k 1
N
1
ei (k)
k 1
N
2
N
k 1
i
i
i 1, , p
(51)
- Cross-correlation function between the residues and previous inputs:
Page 687
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
u (k) N u (k) e (k ) N e (k)
N
Rui e j ( )
k 1
1
i
1
ui (k)
N
k 1
N
N
k 1
i
ui (k)
k 1
N
2
1
N
k 1
j
1
e j (k)
N
k 1
N
j
e j (k)
k 1
N
2
i 1,
, p j 1,
,p
(52)
N is the number of samples.
1, 0 R ( ) 0
, ui e j
Rei ei ( )
0, 0
(53)
Ideally, if the model is validated, the results of these correlation tests and criterion Nash lead to the following results:
and
Qi 100%
,
i 1,
,p,
j 1,
,p
Typically, it is verified that Qi 100% and the functions R are null for the interval 20, 20 with a confidence
interval 95% , that is to say:
1.96
N
R
1.96
N
(54)
- Average error on each output is defined as follows:
AREi
1 N
ei k i 1,
N k 1
, p
(55)
- Mean absolute error on each output is:
AAREi
1 N
ei k i 1,
N k 1
, p
(56)
- Root mean square error on each output:
RMSi
2
1 N
ei k i 1,
N k 1
, p
(57)
- MAE (Mean Absolute Error):
MEA
p N
1
ei k
p * N
i 1 k 1
(58)
The desired value of AREi , AAREi , RMSi and MEA is zero.
Page 688
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
VI.
NEURONAL INDIRECT ADAPTIVE CONTROL
In this section, we propose a neuronal indirect adaptive
control structure of a complex dynamic multivariable
system (with p inputs and p outputs) and a learning
algorithm of a neural controller.
VII.
PROPOSED STRUCTURE OF THE NEURONAL
INDIRECT ADAPTIVE CONTROL
In this work, the structure of the neural indirect
adaptive control is given in Figure 2. The architecture
On-line update of
ANN by
Lyapunov-Base
algorithm
y k 1
rm k
y k na
r k
Referenc
e model
r k nc
u k 2
u k nb
z1
Recurrent
Neural
Network
u k 1
z1
of connections of the neurons between them in the
neural controller is shown in Figure 3. The real-time
adjustment of the controller parameters is performed in
two steps. The first step is the estimation of the model
neuronal parameters of the system, using Theorem 1
and from the knowledge of several inputs-outputs
couples. The second step is the online calculation of
controller parameters based on Theorem 2 and the
yk
.
u k 1
Jacobian matrix of the neural model
+
-
z1
System
yk
z1
Online calculation of the
Jacobian matrix of the
neural model
yk
u k 1
w k , H k , g k
Online estimation of neural model parameters of system
Page 689
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Figure 2. Structure of the neuronal indirect adaptive control
y1 k 1
y1 k na
yp k 1
yp k na
f1
r1 k
r1 k nc
f2
rp k
rp k nc
u1 k 2
u1 k nb
u1 k 1
f1
f1
u p k 1
f2
z1
Page 690
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical
Multivariable
z1
u p k 2
Systems
u p k nb
z1
z1
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Figure 3- Proposed architecture of the neural controller
VIII.
LEARNING ALGORITHM OF NEURAL CONTROLLER
The learning algorithm of neural controller of a nonlinear complex uncertain and perturbed multivariate system (with
p inputs and p outputs) can use Theorem 2.
Theorem 2:
The learning of the neural controller can be made by the following equation:
Page 691
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
c0 c
wc(k 1) 1
2
p
p
ui k 1
y
k
i
2 c c
ci
i
wc k
wc k
i 1
i 1
c0 ci
p
i 1
yi k
eci k
w k
p
yi k
2 c ci
wc k
i 1
2
2
ui k 1
ci
wc k
i 1
p
wc k
2
ci wc k i
mc
i 1
(59)
Such as:
mc 1
ci 0i 0, mc
(60)
c
(61)
mc
i 0
i
1
c 0
(62)
c 0
(63)
0 ci i 1, p
(64)
ci i 1, p
(65)
c c
(66)
(67)
Page 692
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
c
p
i 1
i
c
i 1, p
ci 0
(68)
i 1, p
eci k ri k yi k
(69)
nrc p na nb nc
(70)
(71)
wc wc111 , ,wc1nhc nrc ,wc 211 , ,wc 2nhc 1 ,
wc311 , ,wc3 pnhc ,wc 411 , ,wc 4 p1 ,wc511 ,
,wc5 pnrc ,wc611 , ,wc6nhc nhc ,wc711 , wc7 pp
is the weights vector of neural controller.
T
nhc is the number of neurons in the hidden layer of neural controller.
Proof:
V k
ci
ec k
ci
ec k
From the following Lyapunov function:
ns
i 1
2
ns
2
i 1
i
2
i
2
c
2
wc k
2
c
2
wc k
2
nu
i 1
ci
2
u k 1
2
i
(72)
The adjustment parameters procedure of the neural controller is stable if:
V k ci eci k ei k ci eci k
ns
ns
i 1
i 1
c wc k
T
2 0
(73)
2
wc k c wc k
2
ci ui k 1 2
nu
2
i 1
with :
Equation (73) then becomes:
Page 693
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
wc k
2
2
ns
nu
eci k
ui k 1
c ci
ci
wc k
wc k
i 1
i 1
2
ns
eci k
T
wc k c wc k ci eci k
2 0
i 1
wc k
If the previous equation has a unique solution, the term 2 is as follows:
(74)
ec k
ci eci k i
c wc k
i 1
wc k
ns
2
2
2
ns
nu
eci k
ui k 1
ci
4 c ci
wc k
wc k
i 1
i 1
(75)
2
ns
eci k
c wc k ci eci k
i 1
wc k
The adjustment parameters equation of the controller neural can be written:
wc k
2
ns
nu
eci k
ui k 1
ci
2 c ci
wc k
wc k
i 1
i 1
2
(76)
Like:
lim ci zi 1 1
m
z1
i 0
(77)
We can write:
Page 694
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
m
wc k 1 ci zi 1 wc k 1
i 0
eci k
m
i 1
wc k
ci wc k i
2
2
ns
nu
1
i
eci k
ui k 1
2 c ci
ci
wc k
wc k
i 1
i 1
c0 c wc k ci eci k
ns
(78)
therefore :
wc k 1 wc k wc k 1
c0 c
1
2
n
nu
s
eci k
ui k 1
2 c ci
ci
wc k
wc k
i 1
i 1
ns
ec k
c0 ci eci k i
i 1
wc k
wc k
2
2
ns
nu
eci k
ui k 1
ci
2 c ci
wc k
wc k
i 1
i 1
2
ci wc k i
m
i 1
If all conditions are met proper identification, the neural model outputs y k are good approximations of the system
(79)
outputs y k , which allow writing:
yi k
yi k
i 1,
,p
(80)
Equation (79) then becomes:
c0 c
wc k 1 1
2
ns
nu
2 c c yi k c ui k 1
i
i
wc k
wc k
i 1
i 1
(81)
The calculation of the term
ns
yi k
c
c
ec
k
i
i
0
wc k
i 1
wc k
2
2
ns
nu
yi k
u k 1
ci i
2 c ci
wc k
wc k
i 1
i 1
yi k
can be determined through the neural model as follows:
wc k
2
ci wc k i
m
i 1
Page 695
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
y1 k y1 k
wc k u1 k 1
y k y k
p
p
wc
k
u
1 k 1
y1 k u1 k 1
u p k 1 wc k
y p k u p k 1
u p k 1 wc k
(82)
yi k
u j k 1
We can write:
yi k yi k 1
, i 1,
u j k 1 u j k 2
, p , j 1,
,p
(83)
IX.
ROBUST NEURONAL ADAPTIVE CONTROL
The structure of the proposed robust neural adaptive control is given in Figure 4.
Control
additive
R k qs
r k
ec k qs
K
Neural controller
(Recurrent
Neural Network)
ur k 1
uc k 1
u k 1
Nonlinear
multivariable
system
TDL
Neural model
(Recurrent
Neural
Network)
H (k)
yy k qs
y k qs
CT
+ emoo k qs
-
g (k)
Online Learning
TDL
Page 696
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Figure 4. Proposed structure of robust neuronal adaptive control
With:
TDL : designates delays.
The vectors of reference signals are:
r k r1 k ,
(84)
R k r1 k ,
, rp k
T
, r1 k qs 1 ,
, rj k,
, r j k qs 1 ,
, rp k ,
, rp k qs 1 , 1 j p
T
(85)
The command applied to the system is given by:
uc k 1 u k 1 ur k 1
(86)
The architecture of the neuronal controller (Figure 5) is deduced from the architecture of the neural model (Figure 1).
From equation (23), we can write:
u k qs 1 g k qs
1
H k q Ke k y k q
s
s
(87)
For equation (87) is realizable, the matrix g k must be invertible.
Such as:
The matrix K K1 ,
, K p
T
pn
is calculated so that the matrix
A BK
has all its eigenvalues strictly less
than 1.
The neural controller equation is then:
Page 697
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
u k 1 g k
1
r k H k Ke k q
c
s
(88)
It is also assumed:
g k
1
G1 k ,
, G p k
T
(89)
r1 k
u1 k 2
rp k
e k q
u1 k nb1
u p k 2
u p k nbp
y1 k 1
y1 k na1
yp k 1
yp k nap
l1 k 1
l1 k nc1
T
c
f1
z1
K1
f1
f2
f1
s
f2
z1
l1 k
lp k
+
-
G1 k
+
Gp k
+
-
+
u p k 1
z1
Page 698
-
z1
u1 k 1
Kp
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
z1
1
z
Systems
l k 1
l p k ncp
p
z1
z1
e k q
T
c
s
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Figure 5- Proposed Architecture of the neural controller
In reality, there are always modeling errors during the identification phase and disturbance which may affect the
system. The equation of the dynamics of tracking errors without the additive component of control may be as follows:
ec k 1 qs A BK ec k qs B l k qs
(90)
With l k represents the set of disturbances and uncertainty estimates
l k emoo k
(91)
If we put:
A0 A BK
(92)
Page 699
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
er k C T ec k
(93)
The transfer function between the term l z and prosecution errors er z is as follows:
H er ,l z C T zI A0 B
1
(94)
Theorem 3:
The additive component ur can be calculated using techniques based on optimization H . It can be given by:
ur k 1 g k
1
3
1
BT P ec k qc
(95)
where :
H er ,l z
2
such as 2 0
(96)
3 is a positive scalar and P a symmetric positive definite matrix verifying the following Riccati equation:
1
2
T
P A0 I A0 I P CC T
PBBT P Q 0
2
3
2
(97)
With: Q 0 and
2
1
2
3
2
0
Proof:
The dynamic of the system prosecution errors with the additive command ur is in the following form:
Page 700
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
ec k 1 qs A BK ec k qs B l k qs g k ur k 1
A0 ec k qs B l k qs Bg k ur k 1
(98)
To verify the stability of the system, define the following candidate Lyapunov function:
V k
T
1
ec k qs P ec k qs
2
(99)
The term V k is given by:
T
T
1
1
ec k qs P ec k qs ec k qs P ec k qs
2
2
T
T
1
T
ec k qs A0 I P P A0 I ec k qs ec k qs PB l k qs g k ur k 1
2
(100)
V k
Using the Riccati equation (97), we obtain:
T
T
1
1
1
2
PBBT P ec k qs
ec k qs Q ec k qs ec k qs
2
3
2
2
2
T
T
1
ec k qs CC T ec k qs ec k qs PBg k ur k 1
2
V k
ec k qs PB l k qs
T
(101)
Replacing the value of ur k 1 in (95), equation (101) becomes:
V k
T
T
1
1
ec k qs Q ec k qs
e k qs PBBT P ec k qs
2 c
2
2 2
T
T
1
ec k qs CC T ec k qs ec k qs PB l k qs
2
(102)
Page 701
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
If we assume:
a
T
T
1
1
ec k qs Q ec k qs
e k qs PBBT P ec k qs
2 c
2
2 2
T
T
1
ec k qs CC T ec k qs ec k qs PB l k qs
2
2
T
T
1
1
1
2
ec k qs Q ec k qs ec k qs CC T ec k qs 2 l k qs
2
2
2
T
ec k qs PB k q 2 k q 2
T
1 1
T
2
e
k
q
PBB
P
e
k
q
l
s
2 l
c
s
c
s
s 2
2 2 2
2
(103)
V k a
T
1
1
2
ec k qs CC T ec k qs 2 l k qs
2
2
According to the previous equation, we can write:
2
(104)
then:
V k V 0
k
T
1 k
1
2
T
e
i
q
CC
e
i
q
c s 2 2 l i qs
c s
2 i 0
i 0
2
As V k 0 , in this case inequality (105) can be written as follows:
(105)
k
2
T
1 k
1
2
T
e
i
q
CC
e
i
q
l i qs V 0
c
s
c
s
2
2 i 0
2
i 0
(106)
If V 0 0 , we can write:
e i q
k
i 0
k
r
s
i 0
l
s
i q
2
2
2
2
(107)
Finally, we can write:
Page 702
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
H er ,l z
2
(108)
X.
NUMERICAL RESULTS AND DISCUSSION:
Either the nonlinear system described by the equations system:
x1 x2
2
x2 6 x2 3 x1 1 sin u1 2 x1 1
x3 x4
2
x4 5 25 5 53 x1 5 1 sin u1 2 x1 25 25 5 6 x2
x3 3x4 4 2 4 4 u2 2 4 2 4 u2 4u2 2
y1 x1
y x
3
2
(109)
with :
u1 and u2 are the system inputs.
y1 and y2 are the system outputs.
1 and 2 are noises such as:
1 max y1
2 10 max y1
(110)
(111)
Figure 6 shows the evolution of the system parameters.
Page 703
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
0.155
9.5
0.15
0.145
9
Amplitude
Amplitude
0.14
8.5
8
0.135
0.13
0.125
7.5
0.12
7
0.115
0
1
2
3
Time(s)
4
5
0
1
2
5
x 10
(a)
4
5
5
x 10
(b)
0.7
10.5
0.65
Amplitude
10
Amplitude
3
Time(s)
9.5
0.6
9
0.55
8.5
0.5
0
1
2
3
Time(s)
(c)
4
5
0
5
x 10
1
2
3
Time(s)
4
5
5
x 10
(d)
Page 704
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
16
9.8
9.6
14
9.4
12
9
Amplitude
Amplitude
9.2
8.8
8.6
8.4
10
8
6
8.2
4
8
7.8
2
0
1
2
3
Time(s)
4
5
0
1
2
5
x 10
(e)
3
Time(s)
4
5
5
x 10
(f)
Figure 6 – Evolution of system parameters: (a) parameter 1 ; (b) parameter 2 ; (c) parameter 3 ;
(d) parameter 4 ; (e) parameter 5 ; (f) parameter 6
studied system is given in Figure 2. For online learning of
this model we used Theorem 1. The maximum number of
iterations is 1000 during this learning phase.
1
1
0.8
0.8
0.6
0.6
0.4
0.4
input u2 sequences
input u1 sequences
The figures 7 and 8 represent respectively the training
sequences and assessment performance sequences (or test
sequences) which are normalized by the technical of MinMax normalization. The neural model structure of the
0.2
0
-0.2
0.2
0
-0.2
-0.4
-0.4
-0.6
-0.6
-0.8
-0.8
-1
0
1
2
3
Time(s)
(a)
4
-1
5
5
x 10
0
1
2
3
Time(s)
4
5
5
x 10
(b)
Page 705
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
1
1
0.8
0.8
0.6
0.6
0.4
0.4
output y2 sequences
output y1 sequences
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
0.2
0
-0.2
0.2
0
-0.2
-0.4
-0.4
-0.6
-0.6
-0.8
-0.8
-1
0
1
2
3
Time(s)
4
-1
5
0
1
2
5
x 10
(c)
3
Time(s)
4
5
5
x 10
(d)
Figure 7 – Training sequences: (a) control input u1 ; (b) control input u2 ; (c) desired output y1 ; (d) desired output
1
1
0.8
0.8
0.6
0.6
0.4
0.4
input u2 sequences
input u1 sequences
y2
0.2
0
-0.2
0.2
0
-0.2
-0.4
-0.4
-0.6
-0.6
-0.8
-0.8
-1
0
1
2
3
Time(s)
(a)
4
-1
5
5
x 10
0
1
2
3
Time(s)
4
5
5
x 10
(b)
Page 706
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
1
1
0.8
0.8
0.6
0.6
0.4
0.4
output y2 sequences
output y1 sequences
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
0.2
0
-0.2
0.2
0
-0.2
-0.4
-0.4
-0.6
-0.6
-0.8
-0.8
-1
0
1
2
3
Time(s)
4
-1
5
0
1
2
5
x 10
3
Time(s)
(c)
4
5
5
x 10
(d)
Figure 8 – Test sequences: (a) input u1 ; (b) input u2 ; (c) output y1 ; (d) output y2
Tables 1 and 2 show the obtained test results from different candidate neural models. Note that to obtain a neural
model of the system studied of a satisfactory accuracy, it requires that: m 2 , nb1 2 , nb 2 2 , na1 2 , na 2 2 ,
nc1 1, nc 2 1 , nh 8 , 1 0.7 , 2 0.8 , 1 2 , 2 2.4 , 1.8 , 0 0.96 , 1 0.025 , p 2 ,
0.4 and 2 0.015 .
Tableau 1. Evolution of MEA of different candidate neural models in the case m 1
nb1
nb 2
na 1
na 2
nc1
nc 2
nh
1
2
1
2
0
1
1
1
1
1
1
1
1
0.2
0.3
0.05
0.1
0.2
0.6
0.1
0.9
0.84
1
2
2
1
1
2
2
0.6
0.5
0.1
0.7
0.9
1.2
0.2
0.8
0.76
MEA
Page 707
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
1
3
3
2
2
1
3
0.3
0.4
0.2
1
0.98
0.8
0.3
0.7
0.67
1
1
2
1
2
2
4
0.15
0.7
0.3
1.1
1
0.9
0.4
0.6
0.43
2
2
2
1
3
1
5
0.3
0.3
0.4
1.3
1.5
0.7
0.5
0.5
0.38
2
3
3
2
2
1
6
0.5
0.6
0.5
1.4
1.7
1.2
0.6
0.4
0.27
2
1
2
1
2
2
7
0.6
0.7
0.6
1.5
2
1.4
0.7
0.3
0.14
2
2
2
2
1
1
8
0.7
0.8
0.4
2
2.4
1.8
0.96
0.04
0.06
3
3
3
1
2
2
9
0.7
0.9
0.5
2.1
2.5
1.8
0.9
0.1
0.12
Tableau 2. Evolution of MEA of different candidate neural models in the case m 2
nb1
nb 2
na 1
na 2
nc1
nc 2
nh
1
2
1
2
0
1
2
1
1
1
1
1
1
1
0.2
0.05
0.3
0.1
0.2
0.6
0.1
0.8
0.1
0.772
1
2
2
1
1
2
2
0.6
0.1
0.5
0.7
0.9
1.2
0.2
0.4
0.4
0.093
1
3
3
2
2
1
3
0.3
0.2
0.4
1
0.98
0.8
0.3
0.3
0.4
0.086
1
1
2
1
2
2
4
0.15
0.3
0.7
1.1
1
0.9
0.4
0.4
0.2
0.063
2
2
2
1
3
1
5
0.3
0.4
0.3
1.3
1.5
0.7
0.5
0.3
0.2
0.045
2
3
3
2
2
1
6
0.5
0.5
0.6
1.4
1.7
1.2
0.6
0.2
0.2
0.037
2
1
2
1
2
2
7
0.6
0.6
0.7
1.5
2
1.4
0.7
0.15
0.15
0.022
2
2
2
2
1
1
8
0.7
0.4
0.8
2
2.4
1.8
0.96
0.025
0.015
0.001
3
3
3
1
2
2
9
0.7
0.5
0.9
2.1
2.5
1.8
0.9
0.05
0.0.5
0.081
The autocorrelation functions of residuals and crosscorrelation functions between inputs and residues
(figure 9) are within the confidence intervals,
MEA
validating the use of neural network of characteristics
( nb1 2 , nb 2 2 , na1 2 , na 2 2 , nc1 1 , nc 2 1 ,
nh 8 ) as a model of the studied system.
Page 708
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
1.2
1.2
1
1
Auto-correlation of residuals e2
Auto-correlation of residuals e1
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
0.8
0.6
0.4
0.2
0
0.6
0.4
0.2
0
0
5
10
15
20
-0.2
25
0
10
15
lag
(a)
(b)
0.08
0.08
0.06
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
-0.08
-25
5
lag
Cross correlation function
between input u1 and output residues e2
Cross correlation function
between input u1 and output residues e1
-0.2
0.8
20
25
0.04
0.02
0
-0.02
-0.04
-0.06
-20
-15
-10
(c)
-5
0
lag
5
10
15
20
25
-0.08
-25
-20
-15
-10
-5
0
lag
5
10
15
20
25
(d)
Page 709
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
0.08
0.08
0.06
0.06
Cross correlation function
between input u2 and output residues e2
Cross correlation function
between input u2 and output residues e1
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
0.04
0.02
0
-0.02
-0.04
-0.06
-0.08
-25
0.04
0.02
0
-0.02
-0.04
-0.06
-20
-15
-10
-5
0
lag
5
10
15
20
25
(e)
-0.08
-25
-20
-15
-10
-5
0
lag
5
10
15
20
25
(f)
Figure 9 – Validation tests of the chosen neural model: (a) Autocorrelation function of the prediction error e1 ; (b)
Autocorrelation function of the prediction error e2 ; (c) Cross-correlation function between the input u1 and the
residues e1 ; (d) Cross-correlation function between the input u1 and the residues e2 ; (e) Cross-correlation function
between the input u2 and the residues e1 ; (f) Cross-correlation function between the input u2 and the residues e2 .
After the determination phase of a neural network
capable to best approximate the desired relationships
of inputs-outputs of the studied system, the proposed
structures of neural adaptive control (indirect adaptive
control and robust neural adaptive control by the
technique H ) are applied to this system.
First, we will control the system by neural indirect
adaptive control. The structure of this command is that
presented in Figure 2. The architecture of the neural
controller is that given by Figure 3. The online
calculation procedure of the controller parameters uses
the theorem 2. The maximum number of iterations is
1000 during the phase of the calculation of the
parameters. The evolution of mean values of the
absolute differences between reference signals and
outputs of the system as a function to the parameters
( na , nb , nc and nhc ) are presented in Tables 3 and 4.
Page 710
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
From the results of these evolutions, the chosen neural controller
has the characteristics:
na 2
n 2
b
nc 1
nhc 7
and during learning of this controller, the chosen parameters ( c1 , c2 , c1 , c2 , c , c0 , c1 , c2 , m ) are as
(112)
c1 0.8
c 0.9
2
c1 2.1
c2 2.7
c 1.8
c 1.7
m 2
c0 0.96
c 0.021
1
c2 0.019
follows:
(113)
The average value of absolute differences between reference signals and the outputs of the system is given by the
following equation:
VME
1 P 1 N
eci k
p i 1 N k 1
1 P 1 N
ri k yi k
p i 1 N k 1
(114)
Page 711
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Tableau 3. Evolution of the average values of absolute differences between the reference signals and the system outputs of different
candidate neuronal controllers in the case m 1
na
nb
nc
nhc
c1
c2
c
c1
c2
c
c0
c1
VME
1
1
1
2
0.4
0.2
0.4
0.1
0.2
0.7
0.1
0.9
0.0943
1
2
1
3
0.5
0.3
0.7
0.5
0.8
0.9
0.2
0.8
0.0741
2
1
2
4
0.6
0.4
1.2
1.3
0.9
1.3
0.2
0.8
0.0526
2
2
1
5
0.7
0.5
1.3
1.4
1.3
1.4
0.3
0.8
0.0431
2
1
1
6
0.8
0.7
1.4
1.5
2.1
1.6
0.8
0.2
0.0291
2
2
1
7
0.8
0.9
1.7
2.1
2.7
1.8
0.96
0.04
0.0048
3
2
1
8
0.9
1
1.8
2.5
2.9
2.1
0.82
0.18
0.0069
2
2
2
9
1.1
1.4
2.05
2.7
2.8
2.7
0.8
0.2
0.0087
2
2
1
10
1.2
1.5
2
2.8
2.9
2.9
0.7
0.3
0.0091
Tableau 4. Evolution of the average values of absolute differences between the reference signals and the system outputs of different
candidate neuronal controllers in the case m 2
na
nb
nc
nhc
c1
c2
c
c1
c2
c
c0
c1
c2
VME
1
1
1
2
0.4
0.2
0.4
0.1
0.2
0.7
0.1
0.2
0.7
0.0874
1
2
1
3
0.5
0.3
0.7
0.5
0.8
0.9
0.2
0.2
0.6
0.0532
Page 712
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
2
1
2
4
0.6
0.4
1.2
1.3
0.9
1.3
0.2
0.3
0.5
0.0214
2
2
1
5
0.7
0.5
1.3
1.4
1.3
1.4
0.3
0.4
0.4
0.0172
2
1
1
6
0.8
0.7
1.4
1.5
2.1
1.6
0.8
0.1
0.1
0.0091
2
2
1
7
0.8
0.9
1.7
2.1
2.7
1.8
0.96
0.021
0.019
0.0016
3
2
1
8
0.9
1
1.8
2.5
2.9
2.1
0.82
0.015
0.165
0.0054
2
2
2
9
1.1
1.4
2.05
2.7
2.8
2.7
0.8
0.1
0.1
0.0076
2
2
1
10
1.2
1.5
2
2.8
2.9
2.9
0.7
0.15
0.15
0.0083
The results obtained by the proposed neural indirect adaptive control applied to the system are defined in Figures 10,
11 and 12.
2
80
60
Control signal u2
applied to the system
Control signal u1
applied to the system
1.5
1
0.5
40
20
0
-20
-40
0
-60
0
1
2
3
Time(s)
(a)
4
5
0
5
x 10
1
2
3
Time(s)
4
5
5
x 10
(b)
Page 713
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
6
Signal reference r1
System output y1
70
5
50
Signal reference r2
and System output y2
Signal reference r1
and System output y1
Signal reference r2
System output y2
60
4
3
2
40
30
20
10
0
-10
1
-20
-30
0
0
1
2
3
Time(s)
4
5
0
1
2
5
x 10
3
2
1
0
-1
-2
-3
0
1
2
3
Time(s)
(e)
4
5
5
x 10
(d)
Evolution of the difference
between the reference signal r2 and the system output y2
Evolution of the difference
between the reference signal r1 and the system output y1
(c)
3
Time(s)
4
5
30
20
10
0
-10
-20
-30
0
5
x 10
1
2
3
Time(s)
4
5
5
x 10
(f)
Figure 10 – Results of neuronal indirect adaptive control where r1 and r2 are respectively a signal with random
uniformly distributed amplitudes and a triangular signal: (a) control signal u1 applied to the system ; (b) control
signal u2 applied to the system; (c) reference signal r1 and the system output y1 ; (d) reference signal r2 and the
system output y2 ; (e) Evolution of the difference between the reference signal r1 and the system output y1 ; (f)
Evolution of the difference between the reference signal r2 and the system output y2 .
Page 714
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
14
0.7
12
0.6
10
Control signal u2
applied to the system
Control signal u1
applied to the system
0.5
0.4
0.3
0.2
8
6
0.1
4
0
2
-0.1
0
0
1
2
3
Time(s)
4
5
0
1
2
5
x 10
(a)
3
Time(s)
4
5
5
x 10
(b)
Signal reference r1
System output y1
Signal reference r2
System output y2
30
2.5
2
Signal reference r2
and System output y2
Signal reference r1
and System output y1
25
1.5
1
0.5
15
10
5
0
0
20
0
1
2
3
Time(s)
(c)
4
0
5
5
x 10
1
2
3
Time(s)
4
5
5
x 10
(d)
Page 715
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
1
Evolution of the difference
between the reference signal r2 and the system output y2
Evolution of the difference
between the reference signal r1 and the system output y1
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0
1
2
3
Time(s)
4
5
6
4
2
0
-2
-4
-6
-8
0
1
2
5
x 10
(e)
3
Time(s)
4
5
5
x 10
(f)
Figure 11 – Results of neuronal indirect adaptive control where r1 and r2 are respectively a sinusoidal signal and a
sinusoidal signal : (a) Control signal u1 applied to the system ; (b) Control signal u2 applied to the system ; (c)
reference signal r1 and the system output y1 ; (d) reference signal r2 and the system output y2 ; (e) Evolution of
the difference between the reference signal r1 and the system output y1 ; (f) Evolution of the difference between the
reference signal r2 and the system output y2 .
1.4
1.2
30
25
0.8
Control signal u2
applied to the system
Control signal u1
applied to the system
1
0.6
0.4
0.2
20
15
10
0
5
-0.2
0
-0.4
0
1
2
3
Time(s)
4
5
0
5
x 10
1
2
3
Time(s)
4
5
5
x 10
Page 716
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
(a)
(b)
60
Signal reference r2
System output y2
Signal reference r1
System output y1
5
4.5
50
Signal reference r2
and System output y2
Signal reference r1
and System output y1
4
3.5
3
2.5
2
40
30
20
1.5
10
1
0.5
0
0
0
1
2
3
Time(s)
4
0
5
2000
4000
5
x 10
6000
8000
Samples
0
-1
-2
-3
-4
-5
0
1
2
3
Time(s)
12000
14000
(d)
Evolution of the difference
between the reference signal r2 and the system output y2
Evolution of the difference
between the reference signal r1 and the system output y1
(c)
10000
4
5
5
0
-5
-10
-15
-20
-25
-30
-35
-40
0
5
x 10
(e)
1
2
3
Time(s)
4
5
5
x 10
(f)
Figure 12 – Results of neuronal indirect adaptive control where r1 and r2 are respectively a triangular signal and a
triangular signal: (a) control signal u1 applied to the system; (b) Control signal u2 applied to the system; (c)
reference signal r1 and the system output y1 ; (d) reference signal r2 and the system output y2 ; (e) Evolution of
the difference between the reference signal r1 and the system output y1 ; (f) Evolution of the difference between the
reference signal r2 and the system output y2 .
Page 717
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Secondly, the proposed robust neural adaptive control by the technique H is applied to the system using the
following procedure:
- We chose:
2 0.1
3 0.02
(115)
- The offline calculation of the matrix A , B , C , K , P , Q :
Based on the neural identification results of the system, we have:
0
0
A
0
0
1 0 0
0
1
0 1 0
, B
0
0 0 1
0 0 0
0
0
1
0
0
, C
0
0
1
0
0
0
1
0
(116)
So that all the eigenvalues of matrix A0 are less than 1:
0.002 0.05 1 0.01
K
0.007 0 0.003
0
(117)
The resolution of the Riccati equation (97) gives:
10 0 0 0
0 10 0 0
P
0 0 10 0
0 0 0 10
(118)
If :
Page 718
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
10.2 0
0
10
10.2
21
0 0.17
Q
0
0
19 10
0.17 10 20.6
0
(119)
The figures (13), (14) and (15) show the results of the robust neural adaptive control by technical H . We notice that
the specified constraint of attenuation is verified:
- where r1 and r2 are respectively a signal with random uniformly distributed amplitudes and a triangular signal:
e i
N
i 0
2
450.37
2
5629.82
r
(120)
i
k
i 0
l
(121)
H er ,l z
0.08 0.1
(122)
- Where r1 and r2 are respectively a sinusoidal signal and a sinusoidal signal:
e i
N
i 0
2
516.23
2
5735.82
r
(123)
i
k
i 0
l
(124)
H er ,l z
0.09 0.1
(125)
Page 719
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
- Where r1 and r2 are respectively a triangular signal and a triangular signal:
e i
N
i 0
325.17
2
r
(126)
i
k
i 0
2
l
5419.82
(127)
H er ,l z
0.06 0.1
(128)
2
Control signal u1
applied to the system
1.5
1
0.5
0
0
1
2
3
Time(s)
4
3
Time(s)
4
5
5
x 10
80
Control signal u2
applied to the system
60
40
20
0
-20
-40
-60
0
1
2
5
5
x 10
Page 720
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
(a)
6
(b)
Signal reference r1
System output y1
Signal reference r1
and System output y1
5
4
3
2
1
0
0
70
1
2
3
Time(s)
4
3
Time(s)
4
5
5
x 10
Signal reference r2
System output y2
60
Signal reference r2
and System output y2
50
40
30
20
10
0
-10
-20
-30
0
1
2
(c)
5
5
x 10
(d)
Page 721
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Evolution of the difference
between the reference signal r1 and the system output y1
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
Evolution of the difference
between the reference signal r2 and the system output y2
0
1
2
3
Time(s)
4
3
Time(s)
4
5
5
x 10
10
8
6
4
2
0
-2
-4
-6
-8
0
1
2
(e)
5
5
x 10
(f)
Figure 13 – Results of the robust neural adaptive control by technique H where r1 and r2 are respectively a signal
with random uniformly distributed amplitudes and a triangular signal : (a) control signal u1 applied to the system;
(b) Control signal u2 applied to the system; (c) reference signal r1 and the system output y1 ; (d) reference signal
r2 and the system output y2 ; (e) Evolution of the difference between the reference signal r1 and the system output
y1 ; (f) Evolution of the difference between the reference signal r2 and the system output y2 .
Page 722
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
14
0.7
12
0.6
10
Control signal u2
applied to the system
Control signal u1
applied to the system
0.5
0.4
0.3
0.2
8
6
4
0.1
0
2
-0.1
0
1
2
3
Time(s)
4
0
5
0
1
2
5
x 10
(a)
3
Time(s)
4
3
Time(s)
4
5
5
x 10
(b)
Signal reference r1
System output y1
Signal reference r2
System output y2
30
2.5
2
Signal reference r2
and System output y2
Signal reference r1
and System output y1
25
1.5
1
0.5
0
20
15
10
5
0
1
2
3
Time(s)
(c)
4
0
5
0
1
2
5
x 10
5
5
x 10
(d)
Page 723
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Evolution of the difference
between the reference signal r1 and the system output y1
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
Evolution of the difference
between the reference signal r2 and the system output y2
0
1
2
3
Time(s)
4
3
Time(s)
4
5
5
x 10
10
8
6
4
2
0
-2
-4
-6
-8
0
1
2
5
5
x 10
(e)
(f)
Figure 14 – Results of the robust neural adaptive control by technique H where r1 and r2 are respectively a
sinusoidal signal and a sinusoidal signal : (a) Control signal u1 applied to the system; (b) Control signal u2 applied
to the system; (c) reference signal r1 and system output y1 ; (d) reference signal r2 and system output y2 ; (e)
Evolution of the difference between the reference signal r1 and system output y1 ; (f) Evolution of the difference
between the reference signal r2 and the system output y2 .
Page 724
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
1.4
30
1.2
25
0.8
Control signal u2
applied to the system
Control signal u1
applied to the system
1
0.6
0.4
0.2
20
15
10
0
5
-0.2
0
-0.4
0
1
2
3
Time(s)
(a)
4
5
0
1
2
5
x 10
3
Time(s)
4
5
5
x 10
(b)
Page 725
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Signal reference r1
System output y1
5
4.5
Signal reference r1
and System output y1
4
3.5
3
2.5
2
1.5
1
0.5
0
0
1
2
3
Time(s)
4
3
Time(s)
4
5
5
x 10
60
Signal reference r2
System output y2
Signal reference r2
and System output y2
50
40
30
20
10
0
0
1
2
(c)
5
5
x 10
(d)
Page 726
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
Evolution of the difference
between the reference signal r1 and the system output y1
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
0
-1
-2
-3
-4
-5
Evolution of the difference
between the reference signal r2 and the system output y2
0
1
2
3
Time(s)
4
5
5
x 10
5
0
-5
-10
-15
-20
-25
-30
-35
-40
0
1
2
3
Time(s)
4
5
5
x 10
(e)
(f)
Figure 15 – Results of the robust neural adaptive control by technique H where r1 and r2 are respectively a
triangular signal and a triangular signal : (a) Control signal u1 applied to the system; (b) Control signal u2 applied
to the system; (c) Reference signal r1 and system output y1 ; (d) Reference signal r2 and system output y2 ; (e)
Evolution of the difference between the reference signal r1 and the system output y1 ; (f) Evolution of the difference
between the reference signal r2 and the system output y2 .
Page 727
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
Based on the results of the proposed neural adaptive
controls, we can conclude:
- Robust neural adaptive control by technique
- The control signals are bounded.
uncertainties compared with neural indirect adaptive
control.
H reduces the effect of disturbances and / or
- Abrupt changes of system parameters involve sudden
changes of the amplitudes of commands laws and the
outputs of the controlled system.
- The proposed neural adaptive control guarantees the
stability of control structures and show robustness in
the presence of parameter changes of the controlled
system.
XI.
CONCLUSION
In this work, the purpose of the command is solving
problems tracking given trajectories. The principal
contribution of this work lies in developing new
methodologies of adaptive control based on neural
network. Two techniques of neural adaptive control
have been proposed, developed and tested
successfully. The first technique which is indirect
neural adaptive control has the advantage of being
simple to the use. It uses the neural model of the
system to be controlled and Lyapunov theory for make
online learning of neural controller and to maintain
stability of the controlled system. On the other hand,
this technique risks not to mitigate the effects of
disturbance and therefore the controlled system cannot
follow the trajectories of references of good
performance. To solve this problem, we propose a
robust neural adaptive control by the technique H . In
the second technique, the control law implemented is
[1]
[2]
[3]
[4]
Ge, S.S., Jing Wang, “Robust adaptive neural control for a
class of perturbed strict feedback nonlinear systems,” Neural
Networks, IEEE Transactions on, Vol. 13, No. 6, 2002, pp.
1409 - 1419.
Ruliang Wang, Chaoyang Chen, “Robust adaptive neural
control for a class of stochastic nonlinear systems,”
Computational Intelligence and Security (CIS), 2010
International Conference on, Nanning, 11-14 Dec. 2010, pp.
511 - 514.
Hu, Tingliang, Zhu, Jihong, Sun, Zengqi, “Robust adaptive
neural control of a class of nonlinear systems,” Tsinghua
Science and Technology, Vol. 12, No. 1, 2007, pp. 14 - 21.
Mou Chen, Shuzhi Sam Ge, Bernard Voon Ee How, “Robust
adaptive neural network control for a class of uncertain mimo
the sum of two terms. Indeed, the first term is an
approximation of the inputs-outputs linearization
control of the system to be controlled based on
neural network. On the other hand, the second
term is a term of robustification. The two
proposed techniques of neural adaptive control
ensure the stability of control structure and present
robustness in the presence of parameter changes
of the system to be controlled. Robust neural
adaptive control by the technique H reduces the
effect of disturbances and / or uncertainties compared
with the neuronal indirect adaptive control. The robust
neuronal adaptive control
by the technique H
allows to the controlled system outputs to follow
correctly the reference signals relative to the adaptive
control neuronal indirect.
References
[5]
[6]
[7]
nonlinear systems with input nonlinearities,” IEEE
Transactions on Neural Networks, Vol. 21, No. 5, 2010, pp.
796-812.
Xiaoou Li and Wen Yu, “Robust adaptive control using
neural networks and projection,” Advances in neural
networks, Vol. 3174, 2004.
Y. Zuo, Y. N. Wang, Y. Zhang, Z. L. Shen, Z. S. Chen, J.
Chen, Q. Y. Xie, “RBF neural network robust adaptive
control for wind generator system,” MECHANIKA, Vol. 17,
No. 5, 2011, pp. 557-561.
Lewis, F.L., Liu, K., Yesildirek, A, “Multilayer neural net
robot controller with guaranteed tracking performance,”
Intelligent Control, 1993., Proceedings of the 1993 IEEE
International Symposium on, Chicago, IL , USA, 25-27 Aug
Page 728
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
[8]
[9]
[10]
[11]
[12]
[13]
1993 , pp. 225 - 231.
Yangmin Li and Yugang Liu, “Sliding Mode Adaptive
Neural-network Control for Nonholonomic Mobile Modular
Manipulators,” 2nd International conference on Autonomous
Robots and Agents, Palmerston North, New Zealand, 13-15
Dec. 2004, pp. 511 - 514.
Tairen Sun, Hailong Pei, Yongping Pan, Hongbo Zhou,
Caihong Zhang, “Neural network-based sliding mode
adaptive control for robot manipulators,” Neurocomputing,
Vol.74, No. 14–15, 2011, pp. 2377-2384.
Tsung-Chih Lin, “Based on interval type-2 fuzzy-neural
network direct adaptive sliding mode control for SISO
nonlinear systems,” Communications in Nonlinear Science
and Numerical Simulation, Vol.15, No. 12, 2010, pp. 40844099.
A. Nied, S.I. Seleme Jr., G.G. Parma, B.R. Menezes, “Online neural training algorithm with sliding mode control and
adaptive learning rate,” Neurocomputing, Vol.70, No. 16-18,
2007, pp. 2687-2691.
Slotine JJ. and Lie W, “Applied Nonlinear Control,”
Prentice-Hall.Inc, 1991.
Yi Zuo, Yaonan Wang, Xinzhi Liu, Simon X. Yang, Lihong
Huang, Xiru Wu, Zengyun Wang, “Neural network robust
H tracking control strategy for robot manipulators,”
Applied Mathematical Modelling, Vol. 34, No. 7, 2010, pp.
1823-1838.
[14] Xingjian Jing, “An H control approach to robust learning
of feedforward neural networks,” Neural Networks, Vol.24,
No. 7, 2011, pp. 759-766.
[15] Miyasato, Y., “Adaptive H formation control for EulerLagrange by utilizing neural network approximators,”
American Control Conference (ACC), June 29 2011-July 1
2011, San Francisco, CA, pp. 1753 - 1758.
[16] Sitthidet Vachirasricirikul, Issarachai Ngamroo, “Robust
controller design of heat pump and plug-in hybrid electric
vehicle for frequency control in a smart microgrid based on
specified-structure mixed
[17]
[18]
[19]
[20]
[21]
H2 / H
control technique,”
Applied Energy, Vol. 88, No. 11, 2011, pp. 3860-3868.
Psaltis D., Sideris A.,Yamamura, A. A., “A multilayered
neural network control,” IEEE Control Systems Magazine,
1988, Vol. 8, No. 2, pp. 17-21.
Baltersee J, Chambers JA, “Nonlinear adaptive prediction of
speech with a pipelined recurrent neural network,” IEEE
Trans Signal Process, 1998, Vol. 46, No. 8, pp. 2207–2216.
Stavrakoudis DG, Theocharis JB, “Pipelined recurrent fuzzy
neural networks for nonlinear adaptive speech prediction,”
IEEE Trans. Systems, Man and Cybernetics, (Part B), 2007,
Vol. 37, No. 5, pp. 1305-1320.
Haiquan Zhao, Jiashu Zhang, “A novel adaptive nonlinear
filter based pipelined feedforward second-order Volterra
architecture,” IEEE Trans Signal Process, 2009, Vol. 57, No.
1, pp. 237–246.
Po-Rong Chang, Jen-Tsung Hu, “Optimal nonlinear
adaptive prediction and modeling of MPEG video in ATM
networks using pipelined recurrent neural networks,” IEEE J
Select Areas Commun, 1997, Vol. 15, No. 6, pp. 1087–1100.
[22] Yih-Shen Chen, Chung-Ju Chang, Yi-Lin Hsieh, Nat.
Chiao Tung Univ, Hsinchu, Taiwan, “A channel effect
prediction-based power control scheme using PRNN/ERLS
for uplinks in DS-CDMA cellular mobile systems,” IEEE
Trans Wireless Commun, 2006, Vol. 5, No. 1, pp. 23–27.
[23] Mandic DP, Chambers JA, Toward, “an optimal PRNN
based nonlinear prediction,” IEEE Transactions on Neural
Networks, 1999, Vol. 10, No. 6, pp. 1435– 1442.
[24] Mandic DP, Chambers JA, “On the choice of parameters of
the cost function in nested modular RNNs,” IEEE
Transactions on Neural Networks, 2000, Vol. 11, No. 2, pp.
315–322.
[25] Haiquan Zhao, Jiashu Zhang, “Pipelined Chebyshev
functional link artificial recurrent neural network for
nonlinear adaptive filter,” IEEE transactions on systems man
and cybernetics Part B Cybernetics a publication of the IEEE
Systems Man and Cybernetics Society, 2010, Vol. 40, No. 1,
pp. 162–172.
[26] Williams RJ, Zipser D, “A learning algorithm for continually
running fully recurrent neural networks,” Neural
Computation, 1989, Vol. 1, No. 2, pp. 270–280.
[27] Barak A. Pearlmutter, “Dynamic recurrent neural networks,”
Technical report CMU-CS-90-196, 1990.
[28] B. A. Pearlmutter, “Gradient calculations for dynamic
recurrent neural networks: A survey,” IEEE Transactions on
Neural networks, 1995, Vol. 6, No. 5, pp. 1212 – 1228.
[29] Irina Klevecka, Janis Lelis, “Pre-Processing of Input Data of
Neural
Networks:
The
Case
of
Forecasting
Telecommunication Network Traffic,” Telektronikk, 2008,
Vol. 3, No. 4, pp. 168 – 178.
[30] Haykin, S., “Neural Networks - A Comprehensive
Foundation,” 2nd Ed.Prentice-Hall, Upper Saddle River, NJ.
1998.
[31] Jalil Asadisaghandi,Pejman Tahmasebi, “Comparative
evaluation of back-propagation neural network learning
algorithms and empirical correlation for prediction of oil
PVT propreties in Iran oilfields,” Journal of Petroleum
Science and Engineering, 2011, Vol. 78, No. 2, pp. 464-475.
[32] H. Al-Duwaish, M.N. Karim, and V. Chandrasekar, “Use of
multilayer feedforward neural networks in identification and
control of wiener model,” IEE Proceedings, Control theory
and applications, 1996, Vol. 143, No. 3, pp. 255-/258.
[33] K. S. Narendra and K. Parthasarathy, “Identification and
Control of Dynamical Systems Using Neural Networks,”
IEEE Transactions on Neural Networks, 1990, Vol. 1, No. 1,
pp. 4-27.
[34] Chao-Chee Ku and Kwang Y. Lee, “Diagonal recurrent
neural networks for dynamic systems control,” IEEE
Transactions on Neural Networks, 1995, Vol. 6, No. 1, pp.
144-156.
[35] T. Denoeux, R. Lengellé, “Initializing back propagation
networks with prototypes,” Neural Networks, 1993, Vol. 6,
No. 3, pp. 351-363.
[36] G.P. Drago, S. Ridella, “Statistically controlled activation
weight initialization (SCAWI),” Neural Networks, IEEE
Transactions on, 1992, Vol. 3, No. 4, pp. 627-631.
Page 729
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 671-730
ISSN 2078-2365
[37] J.-P. Martens, “A stochastically motivated random
initialization of pattern classifying MLPs,” Neural
Processing Letters, 1996, Vol. 3, No. 1, pp. 23-29.
[38] T. Masters, “Practical Neural Network Recipes in C++,”
Academic Press, Boston, 1993.
[39] D. Nguyen, B. Widrow, “Improving the learning speed of 2layer neural networks by choosing initial values of the
adaptive weights,” International Joint Conference on Neural
Networks, San Diego, CA , USA, 1990, Vol. 3, pp. 21-26.
[40] S. Osowski, “New approach to selection of initial values of
weights in neural function approximation,” Electronics
Letters, 1993, Vol. 29, No. 3, pp. 313-315.
[41] J.F. Shepanski, “Fast learning in artificial neural systems:
multilayer perceptron training using optimal estimation,”
Neural Networks, IEEE International Conference on, San
Diego, CA , USA, 1988, Vol. 1, pp. 465-472.
[42] H. Shimodaira, “A weight value initialization method for
improving learning performance of the back propagation
algorithm in neural Networks,” Tools with Artificial
Intelligence, Proceedings, Sixth International Conference on,
New Orleans, LA, USA, 1994, pp. 672-675.
[43] Y.F. Yam, T.W.S. Chow, “Determining initial weights of
feedforward neural networks based on least squares method,”
Neural Processing Letters, 1995, Vol. 2, No. 2, pp. 13-17.
[44] Y.F. Yam, T.W.S. Chow, C. T. Leung, “A new method in
determining initial weights of feedforward neural networks
for training enhancement,” Neurocomputing, 1997, Vol. 16,
No. 1, pp. 23-32.
[45] L.F.A. Wessels, E. Barnard, “Avoiding false local minima by
proper initialization of connections,” Neural Networks, IEEE
Transactions on, 1992, Vol. 3, No. 6, pp. 899-905.
[46] N. Weymaere, J.P. Martens, “On the initialization and
optimization of multilayer perceptrons,” Neural Networks,
IEEE Transactions on, 1994, Vol. 5, No. 5, pp. 738-751.
[47] Jim Y. F. Yam and Tommy W. S. Chow, “A weight
initialization method for improving training speed in
feedforward neural network,” Neurocomputing, 2000, Vol.
30, No. 1-4, pp. 219-232.
[48] J.E. Nash, J.V. Sutcliffe, “River flow forecasting through
conceptual models part I — A discussion of principles,”
Journal of Hydrology, 1970, Vol. 10, No. 3, pp. 282-290.
[49] S.A. Billings, Q.M. Zhu, “Nonlinear model validation using
correlation tests,” International Journal of Control, 1994,
Vol. 60, No. 6, pp. 1107-1120.
[50] S.A. Billings, H.B. Jamaluddin, S. Chen, “Properties of
neural networks with applications to modelling non-linear
dynamical systems,” International Journal of Control, 1992,
Vol. 55, No. 1, pp. 193-224.
[51] Dokla, Mahmoud, Osman, Mohammed, UAE U,
“Correlation of PVT Properties for UAE Crudes,” SPE
Formation Evaluation, 1992, Vol. 7, No. 1, pp. 41-46.
[52] Glaso, Oistein, SINTEF/NTH Trondheim, “Generalized
Pressure-Volume-Temperature Correlations,” Journal of
Petroleum Technology, 1980, Vol. 32, No. 5, pp. 785-795.
Page 730
Farouk Zouar, Robust Neuronal Adaptive Control for a Class Of Uncertain Nonlinear Complex Dynamical Multivariable
Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 642-649
ISSN 2078-2365
Dynamic Analysis of A Fossil-Fueled Steam
Electric Power Plant Using Fuzzy PID Controller
BEELA.RAJESH
Department of Electrical and Electronics Engineering
GIT, GITAM University
Visakhapatnam, 530045, Andhra Pradesh, India
Email: rajesh38@gmail.com
Abstract – Boiler-steam turbine generation units are used in the
power system especially for some special features of fossil-fuelled
power plants. When a load disturbance occurred in the system, a
frequency variation will cause a primary regulation action on
generation units. The units will automatically adjust their outputs
to fit for the new load demand. Variation of the governing valve
position may exceed to the outlet pressure of the related boiler but
boiler often has a long control time cycle after the pressure error
was observed. Fuzzy PID controller is usually used to speed up the
regulation procedure of boiler and to improve the stability of the
steam parameters upstream of steam turbine.
Model parameter identification is one of the most reliable tools to
estimate the model parameters. In the proposed work, a general
model of power plant with PID & fuzzy PID control system is built
for power system dynamic analysis. The model responses will be
compared to the model without a PID & fuzzy PID model to
evaluate the impact of PID & fuzzy PID model on system frequency
stability.
Keywords: PID controller, Fuzzy PID control system, dynamic
model, Parameter identification, power plant
I.
INTRODUCTION
Boiler-steam turbine generation units are used in the
power system for some special features of fossil-fuelled power
plants. When a load disturbance occurred in the system,
frequency variation will cause a primary regulation action on
generation units. The units will automatically adjust their
outputs to fit for the new load demand. Variation of the
governing valve position may exceed to the outlet pressure of
the related boiler but boiler often has a long control time cycle
after the pressure error was observed. PID controller and FPID
controller is usually used to speed up the regulation procedure
of boiler and to improve the stability of the steam parameters
upstream of steam turbine. Many researchers have studied the
SMT. T. PADMAVATHI
Department of Electrical and Electronics Engineering
GIT, GITAM University
Visakhapatnam, 530045, Andhra Pradesh, India
Email: tadipadma@gmail.com
mathematic models of power plant for power system dynamic
analysis. According to their research, low order models for
turbine units are more popular for power system dynamic
analysis. According to huge test experiences, single turbine
model is not without a consideration of a main stream pressure
variation. Boiler model is also needed for some circumstances.
Control System of boiler and the PID controller and FPID
controller acting on both the boiler and turbine systems will
have great impact on the pressure stability even output power
of turbine units. But these control systems are not well
considered in relative research .In this paper, a fossil-fuel
power plant model is presented with PID controller and FPID
controller power system analysis. The model parameters are
identified for a turbine coal fired generation unit. The model
responses are compared to the model without PID controller
and FPID controller model to evaluate the impact of PID
controller and FPID controller model on system frequency
stability. Frequency response models have received limited
treatment in the literature. The basic concept of the model
derived here is based on the idea of uniform or average
frequency, where synchronizing oscillations between
generators are filtered out, but the average frequency
behaviour is retained. The synchronizing oscillations are,
taken from the simulations of reference [l].We seek to average
these individual machine responses with a smooth curve that
can be used to represent the average frequency for the system.
Such a filtered or average frequency.. Similar and related
approaches have been pursued more recently [3, 4]through
work on energy functions. The basic ideas are also important
in the work on system Area control simulators [5, 6],as well as
the work on long term dynamics [7, 8].In addition to these
resources, certain ideas have also been adopted from the work
on coherency based dynamic equivalents [9, 10], as well as the
work on transient energy stability analysis. A Genetic
Algorithm (GA) represents a heuristic search technique based
on the evolutionary ideas of natural selection and genetics.
Page 642
Beela and Padmavathi, Dynamic Analysis of A Fossil-Fueled Steam Electric Power Plant Using Fuzzy PID Controller
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 642-649
ISSN 2078-2365
Although randomized, using the historical information they
direct the search into the region of better performance within
the search space.
.
In this paper, the PID and FPID controller is
developed and compared with respect to their overshoot or
undershoot and settling time under various operating
conditions for a two area steam turbine and boiler model.
II.
MODELING OF STEAM TURBINE
In a steam turbine the stored energy of high temperature and
high pressure steam is converted into mechanical (rotating)
energy, which then is converted into electrical energy in the
generator. The original source of heat can be a furnace fired by
fossil fuel (coal, gas, or oil) or biomass. The turbine can be
either tandem compound or cross compound. In a tandem
compound unit all sections are on the same shaft with a single
generator, while a cross compound unit consists of two shafts
each connected to a generator. The cross compound unit is
operated as one unit with one set of controls. The power
output from the turbine is controlled through the position of
the control valves, which control the flow of steam to the
turbines. The valve position is influenced by the output signal
of the turbine controller. High Pressure (HP), Intermediate
Pressure (IP) and Low Pressure (LP) are the different turbine
sections. The turbine considered for study in this paper is
reheating type .Reheating improves efficiency [8].The effects
of steam chest; reheated and nonlinear characteristics of
control valve are considered. The fraction of turbine power
generated by intermediate section is assumed as negligible on
base value
Fig.2.1 steam turbine model
Steam flow entered into steam turbine gs is
proportional to sum of the product of governing valve position
variation PGV and steam pressure variation of superheater Ps
and two variations themselves. TH, TR and TL are time
constants of three equivalent steam volume as high pressure
volume, reheated volume and crossover volume, and Pg, Pr
and Pc are average steam pressures of three volumes. Output
power is a sum of output by three kinds of turbine cylinder.
Power of each cylinder is considered to be proportion to its
inlet steam pressure due to high pressure ratio. Relative with
the rated output power the output portions of three cylinders
are KH, KR and KL respectively.
III.
MODELING OF BOILER
The modelling of a general fossil fuelled boiler. Two
equivalent storage volumes and an equivalent resistant
component are assumed for the steam transmission process.
The output power of a steam turbine generation unit is critical
to the power system analyses. It depends greatly on the steam
flow gs entered into steam turbine, which may be changed
rapidly due to a variation of governing valve position P GV or
the super heater steam pressure PS.PGV is controlled by
governing system of steam turbine while P S will be affect by
PGV and/or the firing command PF controlled by a boiler
governing system. The supplied steam pressure P S respond fast
to the variation of governing valve position P GV but very
slowly to the change in firing command P F because of huge
time lag of the heat transfer and steam transmission processes.
Then the relationship between pressure drop and flow is
assumed to be linear for a small disturbance.
Page 643
Beela and Padmavathi, Dynamic Analysis of A Fossil-Fueled Steam Electric Power Plant Using Fuzzy PID Controller
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 642-649
ISSN 2078-2365
controller (PID controller) is a generic control loop feedback
mechanism controller) widely used in industrial control
systems – a PID is the most commonly used feedback
controller. A PID controller calculates an "error" value as the
difference between a measured process variables and a desired
set point. The controller attempts to minimize the error by
adjusting the process control inputs. The PID controller
calculation algorithm involves three separate constant
parameters, and is accordingly sometimes called three-term
control: the proportional, the integral and derivative values,
denoted P,I, and D. Heuristically, these values can be
interpreted in terms of time: P depends on the present error, I
on the accumulation of past errors, and D is a prediction of
future errors, based on current rate of change. The weighted
sum of these three actions is used to adjust the process via a
control element.
Fig.3.1 Simplified boiler dynamic model
Where K is the proportion gain, PD is the drum steam
pressure Variation and gD is the flow rate variation discharge
from the drum. Pressure in the equivalent storage volume is
generally proportional to the integral of the mass flow
difference between its input and output interfaces. Where
gsand gW are the flow rate variation discharge from the super
heater and the water wall respectively, while Ts and T D are
their relative time constants. The dynamic process of the fuel
feed and burring system and water walls are both considered
as first-order inertia and pure delay. The effect on reheated
pressure of fuel feed variation is ignored. According to the
mathematic models shown before, the whole boiler system can
be merged as a model. Where gQ is the release heat variation
from the fired fuel or the Mass flow rate variation of the fuel.
Tw and TF are the time Constants of water walls and fuel feed
system respectively.
The effect on reheated pressure of fuel feed variation is
ignored. According to the mathematic models shown before,
the whole boiler system can be merged as a model. Where gQ
is the release heat variation from the fired fuel or the Mass
flow rate variation of the fuel. T w and TF are the time
Constants of water walls and fuel feed system respectively.
IV.
PID CONTROLLER
One of the usual compensators that are widely used is PID
controller. The combination of lead and lag compensators is
used to achieve desired transient behaviour and low steady
state error. The structure of this compensator that has been
used in this study A proportional–integral–derivative
In this tutorial we assume the controller is used in closed
loop unity feedback system the variable de note the tracking
error which is send to the PID controller. The controller single
u form the controller to plant equal to the Proportional gain
( ) time magnitude error gain Integral gain ( ) time the
integral of the time pulse the Derivative gain ( ) times the
derivative of the error.
Fig.4.1 block diagram of a PID controller
V.
FUZZY PID CONTROLLER
Although it is possible to design a fuzzy logic type of PID
controller by a simple modification of the conventional ones,
via inserting some meaningful fuzzy logic IF- THEN rules into
the control system, these approaches in general complicate the
overall design and do not come up with new fuzzy PID
controllers that capture the essential characteristics and nature
Page 644
Beela and Padmavathi, Dynamic Analysis of A Fossil-Fueled Steam Electric Power Plant Using Fuzzy PID Controller
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 642-649
ISSN 2078-2365
of the conventional PID controllers. Besides, they generally do
not have analytic formulas to use for control specification and
stability analysis. The fuzzy PID controllers to be introduced
below are natural extensions of their conventional versions,
which preserve the linear structures of the PID controllers,
with simple and conventional analytical formulas as the final
results of the design. Thus, they can directly replace the
conventional PID controllers in any operating control systems
(plants, processes). The conventional design of PID controller
was somewhat modified and a new hybrid fuzzyPID controller
was designed. Instead of summation effect a mamdani based
fuzzy inference system is implemented. The inputs to the
mamdani based fuzzy inference system are error and change in
error.
The main difference is that these fuzzy PID controllers are
designed by employing fuzzy logic control principles and
techniques, to obtain new controllers that possess analytical
formulas very similar to the conventional digital PID
controllers.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
e
PB
PB
PB
PB
PB
PS
PS
PS
PS
PS
ZO
ZO
ZO
ZO
ZO
NS
NS
NS
NS
NS
NB
NB
NB
NB
NB
ec
PB
PS
ZO
NS
NB
PB
PS
ZO
NS
NB
PB
PS
ZO
NS
NB
PB
PS
ZO
NS
NB
PB
PS
ZO
NS
NB
Kp
PB
PB
PB
PB
PB
PM
PM
PB
PB
PB
PM
PM
PM
PM
PB
PB
PM
PM
PM
PB
PB
PB
PB
PB
PB
Ki
ZO
ZO
ZO
ZO
ZO
PS
PS
ZO
ZO
ZO
PS
PS
PS
PM
PM
PM
PM
PM
PM
PM
ZO
ZO
ZO
ZO
ZO
Kd
ZO
PS
ZO
PS
ZO
PS
PS
PM
PS
PS
ZO
PS
PM
PS
ZO
ZO
PS
PB
PS
ZO
ZO
PS
PB
PS
PS
VI.
1.
RESULTS AND DISCUSSION
TWO
AREA
STEAM
SIMULATIONS RESULTS:
TURBINE
MODEL
Fig.5.1Structure
of FUZZY
Table 1 Rule base
for
FPID
Controller
Fig 6.1Frequency deviation inarea2with1%distrubancearea1
Page 645
Beela and Padmavathi, Dynamic Analysis of A Fossil-Fueled Steam Electric Power Plant Using Fuzzy PID Controller
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 642-649
ISSN 2078-2365
Fig
Fig 6.2Frequency deviation inarea1with1%distrubancearea2
6.4Frequency deviation inarea1with2%distrubancearea2
In the above graphsfirstly, a step load disturbance occurs in
two areas (area1&area2) with step load increasing of 0.01p.u.
With the use of a PID controller with a step load disturbance
of 0.01p.u rise time has been reduced to 0.33s, settling time to
9.97 and overshoot is reduced to 146.With the use of a fuzzy
PID controller with a step load disturbance of 0.01p.u rise time
has been reduced to 0.31s, settling time to 9.97 and overshoot
is reduced to 132.6.
In the above graphs firstly, a step load disturbance occurs in
two areas (area1&area2) with step load increasing of 0.02p.u.
With the use of a PID controller with a step load disturbance
of 0.02p.u rise time has been reduced to 0.33s, settling time to
9.97 and overshoot is reduced to 142.With the use of a fuzzy
PID controller with a step load disturbance of 0.02p.u rise time
has been reduced to 0.3s, settling time to 9.97 and overshoot is
reduced to 128.3.
Fig 6.5Frequency deviation inarea2with3%distrubancearea1
Fig 6.3Frequency deviation inarea2with2%distrubancearea1
Page 646
Beela and Padmavathi, Dynamic Analysis of A Fossil-Fueled Steam Electric Power Plant Using Fuzzy PID Controller
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 642-649
ISSN 2078-2365
Fig 6.6Frequency deviation inarea2with3%distrubancearea1
Fig 6.8Frequency deviation inarea1with4%distrubancearea2
In the above graphs firstly, a step load disturbance occurs in
two areas (area1&area2) with step load increasing of 0.03p.u.
With the use of a PID controller with a step load disturbance
of 0.03p.u rise time has been reduced to 0.32s, settling time to
9.97 and overshoot is reduced to 137.With the use of a fuzzy
PID controller with a step load disturbance of 0.03p.u rise time
has been reduced to 0.3s, settling time to 9.97 and overshoot is
reduced to 125.2.
In the above graphs firstly, a step load disturbance occurs in
two areas (area1&area2) with step load increasing of 0.04p.u.
With the use of a PID controller with a step load disturbance
of 0.04p.u rise time has been reduced to 0.31s, settling time to
9.97 and overshoot is reduced to 132.With the use of a fuzzy
PID controller with a step load disturbance of 0.03p.u rise time
has been reduced to 0.3s, settling time to 9.97 and overshoot is
reduced to 123.
Fig 6.9Frequency deviation inarea2with5%distrubancearea1
Fig 6.7Frequency deviation inarea2with4%distrubancearea1
Page 647
Beela and Padmavathi, Dynamic Analysis of A Fossil-Fueled Steam Electric Power Plant Using Fuzzy PID Controller
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 642-649
ISSN 2078-2365
super heater steam pressure is controlled by the governing
system.
TW Flow rate variation
TD Drum steam temperature
TS Steam temperature
K Proportion gain
TF Firing command
Fig 6.10Frequency deviation inarea2with5%distrubancearea1
In the above graphs firstly, a step load disturbance occurs in
two areas (area1&area2) with step load increasing of 0.05p.u.
With the use of a PID controller with a step load disturbance
of 0.05p.u rise time has been reduced to 0.33s, settling time to
9.97 and overshoot is reduced to 128.With the use of a fuzzy
PID controller with a step load disturbance of 0.05p.u rise time
has been reduced to 0.31s, settling time to 9.97 and overshoot
is reduced to 120.5.
2. SYSTEM DATA
2.1 Steam turbine model:
The steam turbine is mostly large power reheat units.
There are multi low pressure cylinders and even multi
intermediate pressure cylinders. The intermediate pressure
cylinders can be considered as for the dynamic analysis.
TH High pressure temperature
TR Reheat temperature
TL Lower pressure temperature
Table7. 1 Time constant value for turbine model
TH
0.26s
TR
18.5s
TL
0.69s
Table.2.1 Time constant values of boiler model
TW
22.3s
TD
46.1s
TS
0.9s
TF
1.2s
K
1.8
2.2s
Table.2.2 Parameters fortwo area steam turbine model with
PID&FUZZY PID controller
KP
-1.5
KI
-1
KD
-1.5
Table.2.3 Parameters for two area steam turbine&boiler
modelwith PID&FUZZYPIDcontroller
KP
-7.67
KI
-0.99
KD
-3.5
Normal operating conditions
T Turbine
G Governor
KP Constant Power
R Feedback gain
TT1=0.03, TG1=0.08
TP1=20, R1=2.4
KP1=120, T12=0.545
B1=0.425, K1=1, a12=-1
2.2 Boiler model:
The out power of the steam generation units is critical
to the power system analyses. Its depends greatly on the steam
flow entered in to the steam turbine, which may be changed
rapidly due to a variation of governing valve position. And
Page 648
Beela and Padmavathi, Dynamic Analysis of A Fossil-Fueled Steam Electric Power Plant Using Fuzzy PID Controller
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 642-649
ISSN 2078-2365
4.
Table 7.2. Parameter Comparison of a Two area system for
PID and FPID Controllers.
Operating
Conditions
0.01
0.02
0.03
0.04
0.05
Type of
Controller
PID
FPID
PID
FPID
PID
FPID
PID
FPID
PID
FPID
V.
Rise
Time
0.33
0.31
0.33
0.3
0.32
0.3
0.31
0.3
0.33
0.31
Settling
Time
9.97
9.97
9.97
9.97
9.97
9.97
9.97
9.98
9.97
9.97
5.
Overshoot
146
132.6
142
128.3
137
125.2
132
123
128
120.5
6.
7.
8.
9.
CONCLUSIONS:
10.
For supplying stable and reliable electric power, load
frequency control is an important issue in power system
operation and control. Automatic load frequency control is
used to maintain the generator power output and frequency
within the prescribed values. In this work the two area load
frequency controller is considered. The simulated study shows
the frequency response and steady state response of two area
systems by using fuzzyPID& PID controller.
The conventional fuzzyPID is compared withPIDcontroller;
two similar areas are given with a disturbance of 0.05p.u. The
simulation study shows that the stability of the system
improved the frequency response and less settling time and
steady state responses. Hence from the results we conclude
that the fuzzy PID controller is said to be better compensating
then conventional PID controller
11.
12.
13.
14.
VI.
REFERENCES
1.
Power
Prabhakundur.
2.
3.
Power System Engineering byNagrath Kothari
IEEE Committee Report, "Dynamic Models for Steam
and Hydro Turbines in Power System Studies", IEEE
system
Stability
and
Control
Trans Power Apparatus & Systems, Vol. 92, No. 6, 1973,
pp. 1904- 1915.
L. N. Bize, J. D. Hurley, "Frequency Control
Considerations for Modem Steam and
Combustion
Turbines", IEEE Engineering Society Winter Meeting,
1999, pp. 548-553.
Dai Yipping, Zhao Ting, TianYunfeng, Gao Lin,
"Research on the influence of primary frequency
control distribution on power system security and
stability", Second IEEE Conference on Industrial
Electronics and Applications, 2007, pp.222-226.
P. M. Anderson, M. Mirheydar, "A low-order system
frequency response model", IEEE Transaction on Power
Systems, Vol. 5, No. 3, 1990, pp. 720-729
K. J. Astrom, K.Eklund, "A simplified non-linear model
of a drum boiler-turbine unit", International Journal of
Control, Vol. 16, No. I, 1972,pp. 145-169.
de Mello, F.P., "Dynamic models for fossil fuelled steam
units in power system studies", IEEE Transactions on
Power Systems, Vol. 6, No. 2,1991, pp. 753-761.
Dai Yipping, Zhao Ting, TianYunfeng, Gao Lin,
"Research on the primary frequency control
characteristics of generators in power system“,Second
IEEE Conference on Industrial Electronics and
Applications,2007, pp. 569-574.
QH WU, "Learning coordinated control of power
systems using interconnected learning automata",
Electrical power & Energy systems, Vol. 17, No. 3, 1995,
pp.91-99.
Gao Lin, Dai Vi-ping, Xia lun-rong, "A New Framework
for Power System Identification Based on an Improved
Genetic Algorithm", 2009 4th IEEE Conference on
Industrial Electronics and Applications, May 25-27,2009,
Xi'an, China.
Gao Lin, Dai Vi-ping, Xia Jun-rong, "Parameter
Identification of Hydro Generation System with Fluid
Transients Based on Improved Genetic Algorithm",
2009 Fifth International Conference on Natural
Computation, 14-16 August 2009, Tianjin, China
H.R.Berenji, Fuzzy logic controllers, in: R.R.
Yager,L.A.Zadeh(Eds.), An Introduction to Fuzzy Logic
Application in Intelligent System, Kluwer Academic
Publisher, Boston, MA, 1992.
W.R.HwangandS.ZeinSabatto.“FuzzyControllerDesignUs
ing Genetic Algorithm ” in Engineering newcentury,
Proceeding, IEEE, April 1997
by
Page 649
Beela and Padmavathi, Dynamic Analysis of A Fossil-Fueled Steam Electric Power Plant Using Fuzzy PID Controller
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 764-768
ISSN 2078-2365
Solid State Lighting Reliability from Failure
Mechanisms Perspective: A Review of Related
Literature
Shailesh K R, Ciji Pearl Kurian, Savitha G Kini
Department of Electrical & Electronics Engineering,
Manipal Institute of Technology,
Manipal University – 576104, India
shailesh9348@rediffmail.com
ABSTRACT - Remarkable long-life makes LED
lighting systems a long-term investment, and great
energy and maintenance savings easily give good
reason for the higher initial cost. LEDs are similar in
construction to microelectronics devices, but there are
functional requirements, materials, and interfaces in
LEDs that make their failure modes and mechanisms
distinctive. Over the last few years, considerable effort
has gone into the study of the failure mechanisms and
reliability of Solid State lighting systems (SSL).
Although still very incomplete, our knowledge of the
reliability issues relevant to SSL is growing. This
paper provides an overview of SSL failure modes and
mechanisms that are commonly encountered. It
focuses on the reliability issues of LED devices.
KEYWORDS: Solid State Lighting, LED Reliability,
SSL reliability, SSL Failure modes, SSL failure
mechanisms
1. INTRODUCTION
Remarkable long-life makes LED lighting systems a
long-term investment, and great energy and
maintenance savings easily give good reason for the
higher initial cost. All LED systems do not perform
equally over their years of operation. Substandard
quality products can prematurely fail or degrade in light
output far below initial claims - so much so that they
fail to provide the value initially promised. The
construction of LEDs is somewhat similar to other
semiconductor devices, but their applications and
construction make their failure modes and mechanisms
distinctive.
Hindrance to the large scale adoption of LEDs in
traditional applications is the lack of information
available on their reliability. Another obstacle is the
lack of worldwide accepted standards, because all
commercial properties of an LED lighting system, such
as luminous flux output, chromaticity, and lifetime, are
functions of the junction temperature. All LED systems
do not perform equally over their years of operation.
Substandard quality products can prematurely fail or
degrade in light output far below initial claims - so
much so that they fail to provide the value initially
promised. Accelerated Life testing in a short time can
predict the life characteristics of LED products under
the conditions of normal stress; it is the effective way
of the reliability evaluation of LED lighting products
for long-term use.
The literature review presented in this paper helps
LED lighting designers and LED product
manufacturers to understand LED failure mechanisms
and reliability thus helping them to design efficient
LED lighting products.
2. LED PRINCIPLE OF WORKING
The LED-chip is the main component of the LED
device. This chip is a semiconductor that generates
light in a PN-junction by electron p-hole
recombination. The active region in the LED-chip is a
complex structure of epitaxial layers. For different
colours differential material-combinations are used:
InAlGaP - red, InGaN - blue, GaAlAs - IR, AlGaN UV. The material and the quality of the epitaxial layers
effectively determine the efficiency factor of the
generation of light.
Page 764
Shailesh et. Al., Solid State Lighting Reliability from Failure Mechanisms Perspective: A Review of Related Literature
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 764-768
ISSN 2078-2365
The principal parameters for the function are
forward driving current IF and junction temperature Tj
within the active layer, influencing both the power
consumption and the colour, which significantly
determine the lifetime as well. LEDs are commercially
available in different technologies. Critical and
significant criteria are a stable current path through
bonding- solder- and glued connection, an appropriate
heat sink for sufficient cooling of the chip by a good
thermal and a high extraction of light from the LED by
optical elements and areas of reflection. LEDs are
encapsulated in general with transparent material like
silicone or epoxy.
temperature within LED and reducing the life of LEDs
[9].
Under Electromigration, high drive currents or
excessive current density can cause contact migration
between the electrical contact and the surface of the
LED die, which causes short circuit [10].Researchers
have reported electromigration of contact metals
occurrence along crystalline defects [11]. Proper
thermal management and improvement of thermal
conductivities of interface materials must be improved.
As it is seen that contact resistances of interface
materials is responsible for overall thermal resistance
[8].
3. CAUSES FOR LED FAILURE
4.3. Metal and dopant diffusion
LEDs fail as there is a gradual lowering of light
output and loss of optical efficiency due to aging.
Catastrophic failures, however rare, can occur as well.
LED failure modes can be broadly classified as
Semiconductor-related and Packaging-related.
4. SEMICONDUCTOR RELATED FAILURES
4.1. Nucleation and growth of dislocations
In this type of failure degradation happens in the
active region where the radiative recombination occurs
[1]. This type of failure happens, if there is an existing
defect in the crystal and this defect is accelerated by
heat or high current density or emitted light [2-5].
GaAs and AlGaAs are more susceptible to this
mechanism than GaAsP and InP. Due to different
properties of the active regions, GaN and InGaN are
almost insensitive to this kind of defect [6, 7]. Ionizing
radiations are also responsible for defect creation.
Future research should focus on improved internal
thermal management handling of thermal resistance
from junction to the package to decrease the formation
of crystal defects and dislocation movements caused by
high-current-induced thermal effects and high ambient
temperature [8].
4.2. Electromigration
It caused by high current density can move atoms
out of the active regions, leading to emergence of
dislocations and point defects, acting as nonradiative
recombination centers and producing heat instead of
light. Improperly designed LEDs may develop areas of
uneven thermal resistances leading to current crowding,
causing thermal runaway resulting in increasing
Movement of metal atoms from the electrodes into
the active region is caused by high electrical currents or
voltages at high temperatures can move metal atoms
this is metal diffusion [12, 13]. In some cases,
especially with GaN/InGaN diodes, a barrier metal
layer is used to hinder the electromigration effects [14].
Some materials, notably indium tin oxide and silver,
are subject to electromigration which causes leakage
current and non radiative recombination along the chip
edges. It is frequently observed in LEDs diffusion of
dopants into the active region during operation can
cause reduction in light output [15]. The main reason
for light degradation is current density, temperature,
and current distribution, which causes an increase in
series resistance [16-19]. LEDs with low dopant
concentration in active region degrade most rapidly.
Contaminants like oxygen can be intentionally
introduced to form complexes to prevent doping
migration [20].
4.4. Cracking of Die
Severe thermal shocks can cause breaking of dies
of LEDs. Due to differences in material properties,
LED packages can be subjected to mechanical stress
when a high drive current is applied or when high
ambient temperature conditions are suddenly applied.
The high electrical stress and extreme thermal shock
are the causes of die cracking [21]. It is necessary to
control die cracking by fine-tuning thermal expansion
coefficients between the substrate and epitaxial layers.
The growth of the optimal medium layer between the
substrate and the epitaxial layer is a important
breakthrough to prevent cracking of die [22]. It is
reported that cases light output and electrical
degradation were due to die cracking. The way in
Page 765
Shailesh et. Al., Solid State Lighting Reliability from Failure Mechanisms Perspective: A Review of Related Literature
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 764-768
ISSN 2078-2365
which die are made has a very critical impact on their
cracking. Early defects caused by the sawing or
grinding process may act as a starting point for die
cracking [23, 24].
5. PACKAGE RELATED FAILURES
5.1. Epoxy / Encapsulant degradation
Prolonged exposure to light from LEDs can cause
epoxy materials to be degraded [25, 26]. Yellowing of
epoxies is due to prolonged exposure to UV light. This
type of discoloration results in a reduction in the
transparency of the encapsulant and causes a decrease
in LED light output [27]. Further, it has been
demonstrated that degradation and the associated
yellowing increases exponentially with exposure
energy. The thermal effects associated with excessive
junction temperature also play a role in encapsulant
yellowing [27, 28].
Yellowing is also due to a combination of ambient
temperature and LED self-heating. It is found that that
a temperature of around 150oC was sufficient to change
the transparency of the epoxy, causing the attenuation
of the light output of LEDs [29]. While phosphors are a
necessary component for producing white light, their
presence causes a decrease in reliability [28].
5.2. Thermal stress
Sudden failures are most often caused by thermal
stresses and shocks [21]. Researchers have found
number of cracks introduced from thermal expansion in
the centre of the lens surface and on the inside of the
polymer encapsulation when high power LED samples
aged at different temperatures [30].Prolonged exposure
to high condensing moisture causes cloudiness of the
epoxy lenses in LEDs due to hygro-mechanical stresses
[31].
5.3. Phosphor degradation
The different phosphors used in white LEDs tend to
degrade with heat and age, but at different rates causing
changes in the produced light colour. The driving
forces are high drive current and excessive junction
temperature, which are attributed to increases in
temperature of the inside of the package [10] there by
efficiency of the phosphor is degraded when the
temperature rises.
5.4. Encapsulant carbonization
Studies indicate that carbonization of the plastic
encapsulation material on the diode surface leads to the
formation of a conductive path across the LED and
subsequently to the destruction of the diode itself [10,
35]. Carbonisation is responsible for light output
degradation. Carbonization of the encapsulant
decreases the encapsulant’s insulation resistance,
significantly inhibiting its ability to provide electrical
insulation between adjacent bond wires and leads [35].
The loss in insulation resistance at temperatures above
threshold temperature can initiate a thermal runaway
process leading to carbonization of the encapsulant. In
this process, the fusing of the bond wires at high
current causes the current to be shunted through the
plastic, leading to joule heating of the plastic [36].
Under carbonization of the encapsulant there will be
light output degradation.
5.5. Other failures
Further reliability of solder interconnects in a LED
package is influenced by environmental loads, solder
material properties, and the intermetallics formed
within the solder and the metal surfaces where the
solder is bonded [32, 33]. The reliability of the
interconnects between packages and circuit boards
connections depends on the magnitude of the
temperature swing, electrical power of LED packages
and board design [34]. Higher electrical power in LEDs
accelerated the rate of interconnect failures at solder
joints. Using an active cooling device improved the
cycles to failure and made them longer than did passive
cooling methods [34].
Delamination happens when repeated cyclic stresses
can cause the material layers of LED packages to
separate, causing significant loss of mechanical
toughness. Delamination can either occur between the
die and silicone encapsulant, between the encapsulant
and packaging lead frame, or between the LED die and
die attach [8]. Delamination causes decreased light
output. Delamination increases the thermal resistance
of the delamination layer leading to increased junction
temperature, which also affects many other types of
failures and eventually decreasing the life of LEDs.
Interface contamination during the LED manufacturing
can result in poor bonding of interfaces, which can
initiate delamination.
Page 766
Shailesh et. Al., Solid State Lighting Reliability from Failure Mechanisms Perspective: A Review of Related Literature
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 764-768
ISSN 2078-2365
6. CONCLUSIONS
As LED prices fall, designers are increasingly using
them in their product designs, especially for lighting.
LEDs have a reputation for being tireless workers that
that never need replacing and require little payment in
terms of power consumption. The main LED failure
mechanisms are mechanical and thermal in nature.
They involve thermal cycles, thermal shock, and LEDs
operating at high temperatures so the wire bond ages.
As the metal oxidizes and becomes brittle over time,
the likelihood of an LED failure increases. Another
cause of LED open circuits is electrostatic discharges.
Better understanding of the causes responsible for
failures in LEDs with respect to improving material
properties and fabrication technology is the need of the
hour. A deeper understanding of various process
variables and associated environments critical for LED
quality must form part of LED reliability studies.
Failure analysis of LEDs has been performed through
conventional
microelectronics
failure
analysis
approaches and off-line analysis techniques. There is a
need to develop advanced failure analysis techniques
for LEDs.
[7]
[8]
[9]
[10]
[11]
[12]
Collaboration between standards bodies and
professional societies is required to arrive at
internationally accepted standards to ensure a fair
comparison of published performance and reliability
data. Reliability study of complete LED luminaires is
the need of the hour.
[13]
[14]
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
Uddin A, Wei AC, Andersson TG. “Study of
degradation mechanism of blue light emitting
diodes”. Thin Solid Films 2005;483:378–81.
Tharian J. “Degradation- and failure mode
analysis of III-V nitride devices”, IPFA 2007.
14th international symposium; 2007. p. 284–7.
Hoang T, LeMinh P, Holleman J, Schmitz J.
“The effect of dislocation loops on the light
emission of silicon LEDs”. 35th European solidstate device research conference; 2005. p. 359–
62.
Lu G, Yang S, Huang Y. “Analysis on failure
modes and mechanisms of LED”. 8th ICRMS;
2009. p. 1237–41.
Yanagisawa T. “The degradation of GaAlAs red
light-emitting diodes under continuous and lowspeed pulse operations”. Microelectronics
Reliability 1998; 38: 1627–30.
Sugiura L. “Comparison of degradation caused
[15]
[16]
[17]
[18]
[19]
[20]
by
dislocation
motion
in
compound
semiconductor light-emitting devices”. Applied
Physics Letters 1997; 70: 1317–9.
Sugiura L. “Dislocation motion in GaN lightemitting devices and its effect on device
lifetime”. Journal of
Applied Physics
1997;81:1633–8
M.-H. Chang et al. “Light emitting diodes
reliability review”, Microelectronics Reliability,
2012, vol: 52,p.762–782
Arnold J. “When the light goes out: LED failure
mode and mechanisms”. DfR solutions; 2004. p.
1–4.
Barton DL, Osinski M, Perlin P, Eliseev PG, Lee
J. “Single-quantum well InGaN green light
emitting diode degradation under high electrical
stress”.
Microelectronics
Reliability,
1999;39:1219–27.
Barton DL, Zeller J, Phillips BS, Chiu P-C,
Askar S, Lee D-S, et al. “Degradation of blue
AlGaN/InGaN/GaN LEDs subjected to high
current pulses”. 33rd Annual Proceedings of
IEEE
international
Reliability
Physics
Symposium; 1995. p.191–9.
Ott M.“Capabilities and reliability of LEDs and
laser diodes”. International NASA Parts and
Packaging Publication; 1996. p. 1–7.
Hwang N, Naidu PSR, Trigg A. “Failure analysis
of plastic packaged optocoupler light emitting
diodes”. 5th conference EPTC; 2003. p. 346–9.
Mobile Reference, “Electronics Quick Study
Guide for Smartphones and Mobile Devices”,
2007
Pavesi M, Manfredi M et al. “Optical evidence of
an electro thermal degradation of InGaN-based
light-emitting diodes during electrical stress”.
Applied Physics Letters 2004;84:3403–5.
Meneghesso G, Levada S, Zanoni E. “Failure
mechanisms of GaN-based LEDs related with
instabilities in doping profile and deep levels”.
IEEE 42nd Annual International Reliability
Physics Symposium; 2004. p. 474–8.
Kozodoy P, DenBaars SP, Mishra UK.
“Depletion region effects in Mg-doped GaN”.
Journal of Applied Physics, 2000;87:770–5.
Meneghini M, Trevisanello L-R, Levada S,
Meneghesso G, Tamiazzo G, Zanoni E, et al.
“Failure mechanisms of gallium nitride LEDs
related with passivation”. IEEE IEDM Technical
Digest 2005:1009–12.
ASM International, “Microelectronics Failure
Analysis Desk Reference”, 2011
Altieri-Weimar P, Jaeger A, Lutz T, Stauss P,
Page 767
Shailesh et. Al., Solid State Lighting Reliability from Failure Mechanisms Perspective: A Review of Related Literature
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 764-768
ISSN 2078-2365
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
Streubel K, Thonke K, et al. “Influence of doping
on the reliability of AlGaInP LEDs”. Journal of
Materials Science: Materials in Electronics
2008;19:S338–41.
Lumileds. Luxeon reliability, Reliability
Datasheet RD25, Philips Lumileds; 2006.
Barton DL, Osinski M, Perlin P, Helms CJ, Berg
NH. “Life tests and failure mechanisms of
GaN/AlGaN/InGaN light emitting diodes”. IEEE
35th Annual Proceedings of Reliability Physics
Symposium; 1997. p. 276–81.
Wu JD, Huang CY, Liao CC. “Fracture strength
characterization and failure analysis of silicon
dies”. Microelectronics Reliability 2003;43:269–
77.
Iksan H, Lin K-L, Hsieh J. “Fracture analysis on
die crack failure”. IMAPS Taiwan; 2001. p. 35–
43.
Narendran N, Gu Y, Freyssinier JP, Yu H, Deng
L. “Solid-state lighting: failure analysis of white
LEDs”.
Journal
of
Crystal
Growth
2004;268:449–56.
Torikai A, Hasegawa H. “Accelerated
photodegradation of poly(vinyl chloride)”.
Polymer Degradation and Stability,1999;63:441–
5.
Zhang Q, Mu X, Wang K, Gan Z, Luo X, Liu S.
“Dynamic mechanical properties of the transient
silicone resin for high power LED packaging”.
ICEPT-HDP; 2008. p. 1–4.
Meneghini M, Trevisanello L-R, Meneghesso G,
Zanoni E. “A review on the reliability of GaNbased LEDs”. IEEE Transactions on Device and
Materials Reliability, 2008;8: 323–31.
Barton DL, Osinski
M. “Degradation
mechanisms in GaN/AlGaN/InGaN LEDs and
LDs”. Proceedings of the 10th SIMC-X; 1998. p.
259–62.
Hsu Y-C, Lin Y-K, Chen M-H, Tsai C-C, Kuang
J-H, Huang S-B, et al. “Failure mechanisms
associated with lens shape of high-power LED
modules in aging test”. IEEE Transaction on
Electronics Devices 2008;55:570–1.
Hewlett Packard. “Reliability of precision optical
performance AlInGaP LED lamps in traffic
signals and variable message signs”. Application
Brief I-004; 1997.
Choubey A, Yu H, Osterman M, Pecht M, Yun
F, Yonghong L, et al. “Intermetallics
characterization of lead-free solder joints under
isothermal aging”. Journal of Electronic
Materials, 2008;37: 1130–8.
Li GY, Chen BL. “Formation and growth
[34]
[35]
kinetics of interfacial intermetallics in Pb-free
solder joint”. IEEE Transactions on Component
Packing Technology 2003;26:651–8
Chang M-H, Das D, Lee SW, Pecht M.
“Concerns
with
interconnect
reliability
assessment of high power light emitting diodes
(LEDs)”. SMTA China south technical
conference; 2010. p. 63–9.
McCluskey P, Mensah K, O’Connor C, Lilie F,
Gallo A, Pink J. “Reliability of commercial
plastic
encapsulated
microelectronics
at
temperatures
from 125oC to 300oC”.
Proceedings of HITEN 1999; 1999. p. 155–62.
Page 768
Shailesh et. Al., Solid State Lighting Reliability from Failure Mechanisms Perspective: A Review of Related Literature
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 776-783
ISSN 2078-2365
Grid Connected DFIG With Efficient Rotor Power
Flow Control Under Sub & Super Synchronous
Modes of Operation
D.Srinivasa Rao
T.Ajaykumar
EEE Department
Gudlavalleru Engineering College, Gudlavalleru
Andhra Pradesh, INDIA
E-Mail:dsrinivasarao1993@yahoo.com
EEE Department
Gudlavalleru Engineering College, Gudlavalleru
Andhra Pradesh, INDIA
E-Mail:tajay.gec@gmail.com
Abstract—To harness the wind power efficiently the most
reliable system in the present era is grid connected Doubly
fed induction generator (DFIG). The DFIG brings out the
advantage of utilizing the turns ratio of the machine and
hence the converter does not need to be rated for the
machine’s full rated power. Depending on wind speed, a
DFIG based variable speed wind turbine is capable of
operating in sub-synchronous or super-synchronous mode
of operation using power electronic converters. The power
flow in the rotor circuit is controlled for maintaining the
stator power constant by effecting rotor voltage through
IGBT in sub-synchronous mode and in the case of supersynchronous mode it is controlled by current sequence
through LCI. The operation of the proposed scheme is
illustrated in different operating conditions i.e. above and
below synchronous speeds using computer simulations.
Keywords—
DFIG, Sub&Super synchronous,
Commutated Inverter (LCI), Sinusoidal PWM Inverter.
I.
W
Line
INTRODUCTION
ind energy has become one of the most important
and promising sources of renewable energy. With increased
penetration of wind power into electrical grids, Doubly-Fed
Induction Generator (DFIG) based wind turbines are largely
deployed due to their variable speed feature and hence
influencing system dynamics. This has created an interest in
developing suitable models for DFIG to be integrated into
power system studies. In standalone induction generator,
both the terminal voltage and frequency will vary with
variation in wind speed and load and an excitation capacitor
will be required. In grid connected induction generator,
control of the terminal voltage and frequency under change
in load and wind speed, is possible and reactive power can be
supplied by the grid.
With this DFIG based Variable-speed wind turbines, an
increased energy capture, improved power quality and
reduced mechanical stress on the wind turbine. It consists of a
wound rotor induction machine with slip rings, and power
electronic converters between the rotor slip-rings and the grid.
In this paper how we can obtain constant power for variable
wind speeds under sub & super synchronous speed operation
of a DFIG is investigated. The stator of DFIG is directly
connected to the grid while the rotor fed at variable frequency
through converter cascade (AC/DC/AC) via slip rings and
brushes to allow the DFIG to operate at variable wind speeds
in response to changing wind speeds. Both the stator and rotor
windings are able to supply power to the grid. The direction
of the power flow in the rotor circuit depends on the variation
of the wind speed. The power electronic converters control
both the direction and magnitude of the power flow of the
machine. In sub-synchronous mode, the converter feeds the
rotor windings from the grid, whereas the rotor supplies
power to the grid in super-synchronous mode of operation. To
ensure variable speed operation, and maintain the stator
power constant both converters need to be controlled under
sub- synchronous and super-synchronous modes of operation
[3].
Most, if not all, of the published papers on the
application of DFIG for wind energy conversion systems
using force commutated inverters in the rotor circuit and d-q
axis control for maintain stator power is constant. However,
in this paper another approach is used which is the power
flow approach and a very simple control technique by
employing line commutated SCR inverter in the rotor circuit
of the DFIG. In this approach the inter relations among the
Page 776
Srinivasa and Ajaykumar, Grid Connected DFIG With Efficient Rotor Power Flow Control Under Sub & Super Synchronous Modes of
Operation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 776-783
ISSN 2078-2365
rotor power (slip power sPs), the air gap power Ps and the
mechanical power Pm are used for analysis of DFIG based
wind energy conversion system.
This paper is organized as follows. In section II
Power flow in DFIG wind energy conversion system and
steady state model of DFIG are described. The operation of
the open and closed loop systems of the proposed scheme
employing sub-synchronous and super-synchronous modes by
using power electronic converters for the grid interface has
been analyzed in section III. And the development of
simulation models of the proposed scheme along with
simulation results is presented in section IV. Finally main
observations are concluded in section V.
II.
parameters. Fig.2 illustrates the standard per-phase equivalent
circuit of DFIG in which rotor circuit parameters are referred to
the stator frequency, so that all machine reactances are
determined at supply frequency.
Fig.2. Per-phase equivalent circuit of a DFIG
POWER FLOW & STEADY STATE MODEL OF DFIG
A. Power flow in DFIG
DFIG can be operated in two modes of operation namely;
sub-synchronous and super-synchronous mode depending on
the rotor speed below and above the synchronous speed. The
power flowing in the rotor of a doubly fed induction machine
(i.e. of the wound rotor type) has three components. These are
a) the electromagnetic power transferred between the stator and
the rotor through the air gap which is known as the air gap
power Ps; b) the mechanical power Pm transferred between the
rotor and shaft; c) the slip power Pr which is transferred between
the rotor and any external source or load (e.g. a converter)
through the rotor slip-rings. These three components of rotor
power are interrelated, under sub- and super-synchronous
modes of operation, as shown in figure.1
When machine is doubly-fed, the per unit power into
the rotor circuit comes from two sources
Pr, in1 = Re ([V2'(I2')*])
(1)
Pr, in2 = T (ωr/ωb) = T (1-S)
And
(2)
Where (*) denotes the complex conjugate operator. Since the
machine is a generator, positive ‘T ‘denotes generator
operation.
The power lost in the rotor circuit is
2
Pr, loss= │I2'│ Rr'
(3)
The power output of the circuit is
*
Pr, out= Re [E (I2') ]
(4)
Conservation of power requires that
Pr, in1+Pr, in2 = Pr, loss+ Pr, out
(5)
So that
*
*
2
Re [V2'(I2') ] + T (1-S) = Re [E (I2') ] +│Is│ Rr'
Or
*
*
2
T (1-S) = Re [E (I2') ] - Re [V2'(I2') ] +│Is│ Rr'
But
(8)
̅
=
–I2'
+
j
(6)
(7)
Xlr'
Substituting Eq. (8) into Eq.(7) ,
Fig.1. Power flow in DFIG wind energy conversion system
B. Steady State Model
The standard steady-state per-phase equivalent circuit can be
utilized for assessing the performance of doubly fed induction
machine subject to the usual assumptions of a three-phase
balanced supply, fixed rotor speed, and constant machine
T (1-S) = Re
-1 V2'(I2')* +│I2'│2 Rr' 1-
(9)
Or
T (1-S) = Re
V2'(I2')
*
- │I2'│2 Rr'
(10)
Page 777
Srinivasa and Ajaykumar, Grid Connected DFIG With Efficient Rotor Power Flow Control Under Sub & Super Synchronous Modes of
Operation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 776-783
ISSN 2078-2365
Having obtained the rotor current from Eq. (17) it is
now possible to obtain the air gap voltage E from Eq. (8). The
stator current can then be found from,
Cancelling out the (1-s) term
(I2')
T=Re
*
-│I2'│
2
This resulting equation represents the basic torque equation
for a doubly fed induction generator.
Solution of Eq.11 in terms of the rotor current has been
developed by Smith et. al [4]. Expanding Eq. (11),
T=
'
'
I2, re +
' 2
I2, im - (I2, re )
-(I2, im')2
(12)
In general, the phase position of the rotor voltage is typically
defined as its relative phase position with respect to the stator
terminal voltage V1. Hence,
and
can be assumed to
be known or specified quantities. Assuming that T and S are
also specified, Eq. (12) can be solved for the currents by also
assuming that their ratio (power factor) is specified. An
alternative approach to solving Eq. 12 is to assume that the
phase position of the rotor current is known rather than the
rotor voltage. In this case, assuming the real part of the stator
current as reference,
'
I2, im = 0
(13)
And
'
I2, re = I2
'
(14)
Eq. (12) becomes
'
'2
I2 – (I2 )
(15)
'
Which is simply a quadratic in terms of I2 .Upon solving
Eq. (15)
T=
I 2' =
√
(16)
Or
I2' =
√(
)
I 1 = I 2' – E
(11)
(17)
The voltage
can also be written as V2'cosΦ2 where Φ2
represents the phase angle of the rotor terminal voltage V2'
with respect to the rotor input current I2'. Hence the rotor
current I2' can be determined as a function of slip for any
desired torque and specified value of rotor voltage and phase.
(18)
The stator voltage can then be obtained by the stator loop
equation
V1 = E-I1 (Rs+jXls)
(19)
In general the voltage obtained will not be identical to the
available terminal voltage except for specific combinations of
rotor voltage and slip. Hence, iteration is necessary to
converge on the correct values which correspond to the specified
stator terminal voltage.
III.
OPERATION UNDER SUB AND SUPER SYNCHRONOUS
MODES
Depending on wind speed, a doubly fed induction generator
(DFIG) based variable speed wind turbine is capable of
operating in sub-synchronous or super-synchronous mode of
operation using power electronic converters. Traditional
Wound Rotor Induction Generator (WRIG) will never
produce power at sub-synchronous mode of operation. In this
mode, it produces motoring torque which can be utilized by
controlling rotor voltage or current. The component of rotor
side converter must need to be controlled properly for reliable
operation of the machine under sub-synchronous and supersynchronous modes. Rotor side converter controls the
imposed voltage and current for the rotor circuit of the
machine. The control of imposed current is necessary for
creating generating torque in sub-synchronous mode of
operation. The control of voltage or current is necessary to
utilize extra generating torque in super-synchronous mode.
During sub-synchronous mode, the speed of the rotor is
less than the machine synchronous speed. As a result the slip
is positive (s > 0), and a motoring torque is produced. To
utilize this torque, negative power (according to the positive
slip) is required in the rotor circuit of the machine. These can
be achieved by the changing the magnitude of the injected
voltage to the rotor circuit and the rotor receives power form
the grid through grid side converter and DC-link. In supersynchronous mode, the rotor speed is greater than the
machine synchronous speed and slip is negative (s< 0). The
rotor voltage/current sequence has to be reversed to supply
extra generating power to the grid through DC-link and grid
side converter. The magnitude of the rotor current and
voltage is changing according to the wind variations.
The mechanical power and the stator electric power output
are computed as follows:
Pr Tm *r
Page 778
Srinivasa and Ajaykumar, Grid Connected DFIG With Efficient Rotor Power Flow Control Under Sub & Super Synchronous Modes of
Operation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 776-783
ISSN 2078-2365
Ps Tem * s
dr
Tm Tem
dt
For a loss-less generator, the mechanical equation is:
J
In steady-state at fixed speed for a loss-less generator
Tm Tem and pm Ps Pr
And it follows that
pr Pm Ps Tmr Tems sPs
where
s (s r ) / s
is defined as the slip of the generator.
Generally the absolute value of slip (s) is much
lower than 1 and, consequently, P r is only a fraction of Ps.
Since Tm is positive for power generation and since ωs is
positive and constant for a constant frequency grid voltage,
the sign of Pr is a function of the slip sign. Pr is positive for
negative slip (speed greater than synchronous speed) and it
is negative for positive slip (speed lower than synchronous
speed). For super-synchronous speed operation, Pr is
transmitted to DC bus capacitor and tends to raise the DC
voltage. For sub-synchronous speed operation, Pr is taken
out of DC bus capacitor and tends to decrease the DC
voltage. PCgrid is used to generate or absorb the power P g in
order to keep the DC voltage constant as shown in Fig.3. In
steady-state for a lossless AC/DC/AC converter Pg is equal
to Pr and the speed of the wind turbine is determined by the
power Pr absorbed or generated by PCrotor. The phasesequence of the AC voltage generated by PCrotor is positive
for sub-synchronous speed and negative for super
synchronous speed. The frequency of this voltage is equal to
the product of the grid frequency and the absolute value of
the slip. PCrotor and PCgrid have the capability for generating
or absorbing reactive power and could be used to control the
reactive power or the voltage at the grid terminals.
Between the two converters, a dc-link capacitor is
placed, as energy storage, in order to keep the voltage
or the speed of the DFIG and also the power factor at the
stator terminals, while the main objective for the grid-side
converter is to keep the dc-link voltage constant.
IV.
SIMULATION STUDIES OF PROPOSED SCHEME
This section discusses the modeling of DFIG, power
electronic converters and the simulation results of the overall
scheme in both sub-synchronous and super-synchronous
modes of operation.
A. Open Loop Super-Synchronous Mode
The block schematic for open loop super-synchronous
mode is shown in Fig.3. Ratings of DFIG used in the
proposed scheme are: Nominal power (P) = 2.65kW, V L-L =
400V, f = 50Hz, synchronous speed (Ns) = 1000 rpm,
number of poles (P) = 6 [7]. In open loop super-synchronous
mode firing angle (
) of the line commutated inverter
is varied manually to maintain the stator power constant at
2.65kW for speeds varying from 1050 rpm to 1200 rpm. As
the speed varies, the rotor power delivered to the grid is
varied but stator power is maintained constant.
The parameters chosen for the simulation study are :
stator resistance
: 0.8285Ω
stator leakage inductance
: 3.579 mH
rotor resistance
: 0.7027Ω
rotor leakage inductance
: 3.579 mH
magnetizing inductance
: 62.64 mH
The simulation model for this mode of operation is developed
and the simulation results obtained are given in Table 1.
Table 4.1 Simulation results for open loop super-synchronous mode
Speed
(Nr)
in rpm
1200
1175
1150
1125
1100
1075
1050
Firing
angle
(α) in
deg.
99.87
98.52
97.18
95.83
94.49
93.14
91.83
Stator
power
(Ps) in
watts
2642
2666
2678
2694
2647
2631
2600
Rotor
power
(Pr) in
watts
548.1
474.0
400.3
323.3
247.5
171.3
93.79
Rect.volt
age
in volts
93.17
80.56
68.23
55.40
42.72
31.23
17.96
LCI
current
(Iact) in
amp
5.977
5.990
6.005
5.988
5.984
5.899
5.800
Fig.3. DFIG system with power electronic converters
variations(or ripple) in the dc-link voltage small. With the
machine-side converter it is possible to control the torque
Page 779
Srinivasa and Ajaykumar, Grid Connected DFIG With Efficient Rotor Power Flow Control Under Sub & Super Synchronous Modes of
Operation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 776-783
ISSN 2078-2365
Fig. 4.2 Variation of active power delivered at the rotor side
(a)Nr=1200 rpm
(b) Nr=1100 rpm
Fig. 4.1 Variation of active power delivered at the stator side
Fig. 4.1 shows the variation of active power of the stator for
varying rotor speeds of 1200 rpm and 1100 rpm. It can be
seen that the stator power is delivered to the grid and is
maintained at around 2.65kW for both speeds by controlling
the firing angle of line commutated inverter.
Similarly from Fig. 4.2 we observe that the rotor power
delivered to the grid is maintained at slip times the stator
power in both speeds i.e., 1200 rpm and 1100 rpm by
controlling the firing angle of line commutated inverter.
Speed(Nr)
in rpm
1200
1175
1150
1125
1100
1075
1050
Firing
angle
(α) in
deg.
99.87
98.52
97.18
95.83
94.49
93.14
91.83
Stator
power
(Ps) in
watts
2653
2667
2679
2688
2687
2660
2650
Rotor
power
(Pr) in
watts
550.1
474.6
399.1
324.0
249.8
176.1
99.5
Rect.voltage
in volts
93.17
80.55
68.00
55.18
42.63
30.58
17.57
B. Closed Loop Super-Synchronous Mode
Fig. 4.3 shows the closed loop super synchronous
mode, in which the firing angle (
) of the line
commutated inverter is varied automatically i.e., the actual
DC link current, Iact is compared with the reference current,
Iref and any mismatch is used to change the firing angle α, of
the inverter as follows α = (Iref - Iact)*[Kp+KI/s] where Kp and
KI are the proportional and integral stage gains respectively.
The optimum values for Kp and KI have been arrived at by
trial and error method [6]. The values have been chosen
taking into account the range of mechanical torque of the
wind turbine. This range will represent the variation in wind
speed with which the system has to operate. In the proposed
scheme, the P and I controller gains (KP = 0.5 and KI = 100)
have been chosen for operating the system with rotor speed
varying from 1050 rpm to 1200 rpm, to maintain the stator
power constant at 2.65kW.
LCI
current
(Iact) in
amp
6.000
5.998
5.991
6.007
6.050
6.010
6.020
Fig. 4.3 Simulation model of the closed loop super-synchronous
mode
The simulation model for this mode of operation is
developed and is shown in Fig. 4.3. The simulation results
obtained are given in Table 4.2.
Table 4.2 Simulation results for closed loop super-synchronous mode
(a)Nr=1200 rpm
(b) Nr=1100 rpm
Page 780
Srinivasa and Ajaykumar, Grid Connected DFIG With Efficient Rotor Power Flow Control Under Sub & Super Synchronous Modes of
Operation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 776-783
ISSN 2078-2365
Fig. 4.4 shows the variation of active power of the stator for
varying rotor speeds of 1200 rpm and 1100 rpm. It can be seen
that the stator power is delivered to the grid and is maintained
at around 2.65kW for both speeds by controlling the firing
angle of line commutated inverter.
(a)Nr=1200 rpm
(b) Nr=1100 rpm
Fig. 4.4 Variation of active power delivered at the stator side
Similarly from Fig. 4.5 we observe that the rotor power
delivered to the grid is maintained at slip times the stator
power for both speeds i.e., 1200 rpm and 1100 rpm by
controlling the firing angle of line commutated inverter.
Speed
(Nr)
in rpm
Modula-tion
index (m)
800
825
850
875
900
925
950
0.2500
0.2195
0.1893
0.1594
0.1299
0.1008
0.0720
(a)Nr=800 rpm
Stator
power
(Ps) in
watts
2648
2650
2650
2650
2652
2655
2647
Rotor
power
(Pr) in
watts
651.1
574.6
498.0
421.5
345.8
270.1
193.6
Rotor
freq
(fr) in
Hz
10.0
8.75
7.50
6.25
5.00
3.75
2.50
Rotor
voltage
(RMS)
in volts
85.62
74.96
64.99
54.86
44.73
34.79
24.88
(b) Nr=900 rpm
Fig. 4.6 Variation of active power delivered at the stator side
(a)Nr=1200 rpm
(b) Nr=1100 rpm
Fig. 4.5 Variation of active power delivered at the rotor side
C. Open Loop Sub-Synchronous Mode
In open loop sub-synchronous mode, modulation index
of the sinusoidal pulse width modulation inverter is varied
manually to maintain the stator power constant at 2.65kW
for speeds varying from 800 rpm to 950 rpm. As the speed
varies, the rotor power absorbed from the grid is varied but
stator power is maintained constant.
The simulation model for this mode of operation is
developed and the simulation results obtained are given in
Table 4.3.
Fig. 4.6 shows the variation of active power of the stator for
varying rotor speeds of 800 rpm and 900 rpm. It can be seen
that the stator power delivered to the grid is maintained at
2.65kW for both speeds by controlling the modulation index
of the sinusoidal PWM inverter.
Table 4.3 Simulation results for open loop sub-synchronous mode
(a)Nr=800 rpm
(b) Nr=900 rpm
Fig. 4.7 Variation of active power absorbed from the grid at the rotor
side
Similarly from Fig. 4.7 we observe that the rotor power
absorbed from the grid is maintained at slip times the stator
power for both speeds i.e., 800 rpm and 900 rpm by
controlling the modulation index of the sinusoidal PWM
inverter.
Page 781
Srinivasa and Ajaykumar, Grid Connected DFIG With Efficient Rotor Power Flow Control Under Sub & Super Synchronous Modes of
Operation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 776-783
ISSN 2078-2365
D. Closed Loop Sub-Synchronous Mode
In closed loop sub-synchronous mode, the modulation
index of the sinusoidal pulse width modulation inverter is
varied automatically i.e., the actual rotor voltage, V2 is
compared with the reference voltage, V2 ref = s*V1 and any
mismatch is used to change the modulation index m, of the
inverter as follows. m = (V2 – V2 ref)*[Kp+KI/S]. The
optimum values for Kp and KI have been arrived at by trial
and error method. The values have been chosen taking into
account the range of mechanical torque of the wind turbine.
This range will represent the variation in wind speed with
which the system has to operate. In the proposed scheme, the
P and I controller gains (KP = 0.05 and KI = 2.38) have been
chosen for operating the system with rotor speed varying
from 800 rpm to 900 rpm, to maintain the stator power
constant at 2.65kW, though the rotor power absorbed from
the grid is varied.
The simulation model for this mode of operation is
developed and is shown in Fig. 4.8. The simulation results
obtained are given in Table 4.4
(a) Delivered to the grid
(b) Absorbed from the grid
Fig. 4.9 Variation of active power for Nr = 800 rpm
Fig. 4.9 (a) shows the variation of active power of the stator
for speed of 800 rpm. It can be seen that the stator power is
delivered to the grid and is maintained at 2.65kW by
controlling the modulation index of the sinusoidal PWM
inverter. Similarly from Fig. 4.9 (b), we observe that the rotor
power absorbed from the grid is maintained at slip times the
stator power.
V.
CONCLUSION
In this paper the operation of a double-fed woundrotor induction machine, coupled to a wind turbine, as a
generator at different speeds is investigated. A very simple
and easy to implement configurations of DFIG for wind
driven applications have been demonstrated. The power
flow in the rotor circuit has been controlled for maintaining
the stator power constant in both sub & super-synchronous
modes of operation.
The simulation results depict the smooth control of
active power fed to the grid with variation in rotor speed of
the DFIG. Such a system allows the utilization of wind
power in different operating conditions i.e. above and
below synchronous speeds, thus leading to higher power
harvesting and consequently higher efficiency of wind
energy conversion system.
Fig. 4.8 Block diagram for closed loop sub-synchronous mode
Table 4.4 Simulation results for closed loop sub-synchronous mode
Speed
(Nr)
in rpm
Modulation
index
(m)
0.2497
Stator
power
(Ps) in
watts
2610
Rotor
power
(Pr) in
watts
641.7
Rotor
freq
(fr) in
Hz
10.0
Rotor
voltage
in volts
825
0.2187
2560
554.6
8.75
75.0
850
0.1894
2652
498.5
7.50
64.9
875
0.1603
2748
437.7
6.25
55.0
900
0.1320
2690
318.0
5.00
44.7
800
85.5
REFERENCES
[1] T.A.Lipo University of Wisconsin Madison WI USA,
“A Super synchronous Doubly Fed Induction
Generator Option for Wind Turbine Applications”
IEEE Conference proceedings, 24-26 June 2009, pages
1-5.
[2] The Electric Generators Handbook “Variable Speed
Generators” by Ion Boldea Polytechnic Institute
Timisoara, Romania, 2006.
[3]
M.Aktarujjaman,
M.E.Haque,
K.M.Muttaqi,
M.Negnevitsky, and G.Ledwich, sch.of.Eng., Univ.of
Tasmania, Hobart, TAS, “Control Dynamics of a
Doubly Fed Induction Generator Under Sub and Super-
Page 782
Srinivasa and Ajaykumar, Grid Connected DFIG With Efficient Rotor Power Flow Control Under Sub & Super Synchronous Modes of
Operation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 776-783
ISSN 2078-2365
Synchronous Modes of Operation” IEEE Conference
Proceedings, 20-24 July 2008, pages 1-9.
[4] “Design and Test of DC Voltage Link Conversion
System and Brushless Doubly-Fed Induction Generator
for Variable-Speed Wind Energy Applications” by
T.A.Lipo, D.Panda, and D.Zarko University of
Wisconsin Madison, Wisconsin, August 1999 - May
2003.
[5] W. Leonhard, Control of Electrical Drives, 2nd ed.
Berlin, Germany: Springer- Verlag, 1996.
[6] “A Simple Controller using Line Commutated Inverter
with Maximum Power Tracking for Wind-Driven
Grid-Connected Permanent Magnet Synchronous
Generators” by V.Lavanya, N.Ammasai Gounden, and
Polimera Malleswara Rao Department of Electrical and
Electronics Engineering, NIT, Tiruchirappalli, IEEE
Conference Proceedings, 2006.
[7] “An Improved Control Strategy of Limiting the DC-Link
Voltage Fluctuation for a Doubly Fed Induction Wind
Generator” by Jun Yao, Hui Li, Yong Liao, and Zhe
Chen, Chongqing University, Chongqing, China. IEEE
transactions on power electronics, vol. 23, no. 3, may
2008.
[8] “Design and Test of DC Voltage Link Conversion
System and Brushless Doubly-Fed Induction Generator
for Variable-Speed Wind Energy Applications” by
T.A.Lipo, D.Panda, and D.Zarko University of
Wisconsin Madison, Wisconsin, August 1999 - May
2003.
Page 783
Srinivasa and Ajaykumar, Grid Connected DFIG With Efficient Rotor Power Flow Control Under Sub & Super Synchronous Modes of
Operation
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
The Effects of the Bad Weather on the
Transmission and Performance Efficiency of
Optical Wireless Communication Systems
Abd El–Naser A. Mohamed1, Ahmed Nabih Zaki Rashed2*, and Amina E. M. El-Nabawy3
1,2,3
Electronics and Electrical Communications Engineering Department
Faculty of Electronic Engineering, Menouf 32951, Menoufia University, EGYPT
1
E-mail: Abd_elnaser6@yahoo.com, 2*E-mail: ahmed_733@yahoo.com
Abstract: Optical wireless links offer gigabit per second
data rates and low system complexity. For ground space
and or terrestrial communication systems, these links
suffer from atmospheric loss mainly due to fog,
scintillation and precipitation. Optical Wireless link
provides high bandwidth solution to the last mile access
bottleneck. However, an appreciable availability of the
link is always a concern. Wireless Optics (WOs) links
are highly weather dependent and fog is the major
attenuating factor reducing the link availability. Optical
wireless links offer gigabit per second data rates and low
system complexity. For ground space and or terrestrial
communication scenarios, these links suffer from
atmospheric loss mainly due to fog, scintillation and
precipitation signals and then to upgrade the
transmission bit rate distance product for ultra long
transmission links. This paper has presented the bad
weather effects such as rain, fog, snow, and scattering
losses on the transmission performance of wireless
optical communication systems. It is taken into account
the study of bit error rate, maximum signal to noise
ratio, maximum transmission optical path lengths and
maximum transmission bit rates under these bad
operating conditions.
Keywords: Wireless Optics (WOs), Specific attenuation,
Visibility, Rain Scattering, Rain Attenuation, Empirical
model, and Bad weather effects.
I. INTRODUCTION
The optical wireless communication (OWC) system
has attracted significant interest because it can solve the
last mile problem in urban environments. The last mile
problem is when Internet providers cannot connect the
fiber optic cables to every household user because of the
high installation costs. The only disadvantage of the
OWC system is that its performance depends strongly on
weather conditions. Fog and clouds scatter and absorb the
optical signal, which causes transmission errors. Most
previous studies consider only single-scattering effects
and assume that the received signal has no intersymbol
interference (ISI), which is true only for light-fog
conditions [1]. Maintaining a clear line of sight (LOS)
between transmit and receive terminals is the biggest
challenge to establish optical wireless links in the free
space especially in the troposphere [2]. The LOS is
diminished due to many atmospheric influences like fog,
rain, snow, dust, sleet, clouds and temporary physical
obstructions like e.g., birds and aeroplanes [3]. Moreover,
the electromagnetic interaction of the transmitted optical
signal with different atmospheric effects results in
complex processes like scattering, absorption and
extinction that are a function of particle physical
parameters. Hence the local atmospheric weather
conditions mainly determine the availability and
reliability of such optical wireless links since there is
always a threat of downtime of optical wireless link
caused by adverse weather conditions [4]. Optical
wireless links are also influenced by atmospheric
temperature that varies both in spatial and temporal
domains. The variation of temperature in the optical
wireless channel is a function of atmospheric pressure and
the atmospheric wind speed. This effect is commonly
known as optical turbulence or scintillation effect and
causes received signal irradiance or power fades in
conjunction with the variation of temperature along the
propagation path. As a result of this scintillation
phenomenon, the optical wireless channel distance and
the capacity are reduced. Thereby restricting the regions
and times where optical wireless links can be used
potentially. In order to take full advantage of the
tremendous usefulness of optical wireless technology
require a proper characterization of different atmospheric
effects influences and a meaningful interpretation of the
filed measurements in such adverse conditions [5].
Optical Wireless communication, also known as free
space optical (FSO), has emerged as a commercially
viable alternative to radio frequency (RF) and millimeter
wave wireless for reliable and rapid deployment of data
and voice networks. RF and millimeter wave technologies
Page 650
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
allow rapid deployment of wireless networks with data
rates from tens of Mbit/sec (point-to-multipoint) up to
several hundred Mbit/sec (point-to-point). Though
emerging license free bands appear promising, they still
have certain bandwidth and range limitations [6]. Optical
wireless can augment RF and millimeter wave links with
very high (>1 Gbit/sec) bandwidth. In fact, it is widely
believed that optical wireless is best suited for multi
Gbit/sec communication. The general acceptance of free
space laser communication (lasercom) or optical wireless
as the preferred wireless carrier of high bandwidth data
has been hampered by the potential downtime of these
lasercom systems in heavy, visibility limiting, weather.
There seems to be much confusion and many
preconceived notions about the true ability of lasercom
systems in such weather. There still is some confusion
over how different laser wavelengths are attenuated by
different types of weather [7].
In the present study, optical wireless communication
is now a well established access technology, better known
for its robustness in transmitting large data volumes in an
energy efficient manner. However the bit error rate (BER)
performance of a wireless optical communication ground
link is adversely affected by cloud coverage, harsh
weather conditions, and atmospheric turbulence. Fog,
clouds and dry snow play a detrimental role by
attenuating optical energy transmitted in terrestrial free
space and thus decrease the link availability and
reliability.
II. BLOCK DIAGRAM OF OPTICAL WIRELESS
COMMUNICATION SYSTEM
There are three key function elements of optical
wireless communication system as shown in Fig. 1. The
transmitter, the atmospheric channel and the receiver. The
transmitter converts the electrical signal into light signal.
The light propagates through the atmosphere to the
receiver, which converts the light back into an electrical
signal. The transmitter includes a modulator, a laser
driver, a light emitting diode (LED) or a laser, and a
telescope [8].
Fig. 1. General block diagram of optical wireless communication system.
The modulator converts bits of information into
signals in accordance with the chosen modulation method.
The driver provides the power for the laser and stabilizes
its performance, it also neutralizes such effects as
temperature and aging of the laser or LED. The light
sources convert the electrical signal into optic radiation.
The telescope aligns the laser LED radiation to a
collimated beam and directs it to the receiver. In the
atmospheric channel, the signal is attenuated and blurred
as a result of absorption, scattering and turbulence. This
channel maybe the traversed distance between a ground
station and a satellite or a path of a few kilometers
through the atmosphere between two terrestrial
transceivers [9].
The receiver includes a telescope, filter, photo
detector, an amplifier, a decision device, and a clock
recovery unit. The telescope collects the incoming
radiation and focus it onto filter. The filter removes
background radiation and allows only the wavelength of
the signal to pass through the electronic signal. The
decision unit determines the nature of the bits of
information based on the time of arrival and the amplitude
of the pulse. The clock recovery unit and synchronizes the
data sampling to the decision making process.
Page 651
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
III. SYSTEM MODEL ANALYSIS
In a terrestrial FSO, the communication transceivers
are typically located in the troposphere. Troposphere is
home to all kinds of weather phenomena and plays a very
detrimental role for FSO communications in lower
visibility range conditions mainly due to rain, snow, fog
and clouds. The estimated of fog, snow and rain
attenuation effects using empirical models as mentioned
in Ref. [10]:
fog
3.912
V 55x104
q
,
(1)
Where V is visibility range in km,
is transmission
wavelength in m. αfog( ) is the total extinction
coefficient and q is the size distribution coefficient of
scattering related to size distribution of the droplets. In
case of clear or foggy weather with no rain or snow, Ref.
[11, 12] approximations of the q parameter to compute the
fog attenuation, that are very accurate for the narrow
wavelength range between 1.3–1.65 m.
1.6 V 50 km
q 1.3 6 km V 50 km ,
0 V 0.5 km
(2)
Transmitted optical pulses in free space are mainly
influenced by two main mechanisms of signal power loss,
absorption and scattering. Absorption is mainly due to
water vapours and carbon dioxide, and depends on the
water vapour content that is dependent on the altitude and
humidity. By appropriate selection of optical wavelengths
for transmission the losses due to absorption can be
minimized. It was found that scattering (especially Mie
scattering) is the main mechanism of optical power loss as
the optical beam looses intensity and distance due to
scattering. The beam loss due to scattering can be
calculated from the following empirical, visibility range
dependent formula [13]:
sca t
17 550
V
0.195V
, dB/km
(3)
Where V is visibility range in km,
is transmission
wavelength in m. Then the total attenuation of wireless
medium communication system can be estimated as:
fog snow rain scat , dB/km
(4)
When the optical signal passes through the atmosphere, it
is randomly attenuated by fog and rain. Although fog is
the main attenuation factor for optical wireless links, the
rain attenuation effect cannot be ignored, in particular in
environments where rain is more frequent than fog. As the
size of water droplets of rain increases, they become large
enough to cause reflection and refraction processes. These
droplets cause wavelength independent scattering [13]. It
was found that the resulting attenuation increases linearly
with rainfall rate; furthermore the mean of the raindrops
size is in the order of a few millimeters and it increases
with the rainfall rate [14]. Let R be the rain rate in mm/h,
the specific attenuation of wireless optical link is given by
[15]:
ra in 1.076 R0.67 dB/km
(5)
If S is the snow rate in mm/h then specific attenuation in
dB/km is given by [16, 17] as:
snow a Sb dB/km
(6)
If is the wavelength, the parameters a and b for dry
snow are given as the following:
a 5.42x104 5.495876, b 1.38
(7)
The parameters a and b for wet snow are as follows [18,
19]:
a 1.023x104 3.7855466, b 0.72
(8)
In order to estimate the coverage at millimeter
wavelengths under direct Line of Sight (LOS) conditions,
the free space propagation model is used. The SNR
requirements for modulation scheme at a fixed data rate
of one Gbit/sec is obtained from the following formula
[20]:
4
SNR PT 30 GT GR 20log
c
10 logkB B.W T NF Fm ,
dB
(9)
Where PT is the transmitter power, GT is the transmitter
antenna gain, GR is the receiver antenna gain, c is the
carrier wavelength, kB is the Boltzmann’s constant
(1.38X10-23 J/K), Receiver Bandwidth (B.W= 1 MHz), T
is the ambient temperature in K, , Receiver Noise Figure,
Fm is the Fade margin, and α is the total attenuation in
dB/km. The maximum propagation distance (L) for
meeting the SNR requirements [21]:
L 10 / 20 ,
(10)
The transmitter and receiver antenna gains can be
expressed as the following:
Page 652
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
Dr
GR
,
2
GT
32
2
div
,
(11)
Where θdiv is the transmitter divergence of the beam in
radians can be expressed as follows:
4
div
,
Dt
(13)
The basic formula for a typical optical link is an
exponential decaying function as function of the path
length L as the following expression [22, 23]:
PR PT
Dt div L2
Dr
e L ,
(14)
Where PR is the received power after traveling the path
length L through the lossy medium, PT is the initial
transmitted power, and α is the total attenuation
coefficient of the medium. The bit error rate (BER)
essentially specifies the average probability of incorrect
bit identification. In general. The higher the received
SNR, the lower the BER probability will be. For most
PIN receivers, the noise is generally thermally limited,
(12)
which independent of signal current. The bit error rate
(BER) is related to the signal to noise ratio (SNR) as
follows [24, 25]:
2
SNR
. exp
BER
,
SNR
.
8
(15)
The maximum transmission bit rates BRmax. which is a
losses limited one, and is given by [26]:
BRmax . Bu exp L m)
(16)
Where Bu is the maximum available transmission bit rate
without any limitations, and αm is the system marginal
loss.
IV. Simulations Results and Performance Evaluation
The model have been deeply investigated to present
the bad weather effects on the transmission performance
and system operation characteristics of wireless optical
communication systems for different visibility ranges
over wider range of the affecting parameters.
Table 1: Proposed operating parameters for wireless optical communication systems [2, 5, 13, 20].
Operating parameter
Room temperature, T=T0
Signal transmitted power, PT
Operating signal wavelength,
System marginal loss, αm
Transmitter lens diameter, Dt
Bit rate (max.) without any limitations, Bu
Detector electronic bandwidth (B.W)
Value and unit
300 K
100 m Watt
1.3 ≤ , µm ≤ 1.65
3 dB
100 cm
1 Gbit/sec
1 MHz
High visibility, Vhigh
50 ≤ Vhigh, km ≤ 80
Medium visibility, Vmedium
6 ≤ Vmedium, km ≤ 50
0 ≤ Vlow, km ≤ 0.5
Low visibility, Vlow
Receiver aperture diameter (antenna size) Dr
50 cm
System marginal loss, αm
3 dB
Receiver noise figure, NF
5 dB
Fade margin, Fm
20 dB
Snow rate, S
0.2 mm/h
Rain rate, R
Carrier wavelength,
1 mm/h
c
1.55 m
Page 653
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
windows under both bad weather effects of dry and
wet snow. It is observed that second optical
transmission window has presented higher
maximum propagation distance compared to third
optical transmission window. As well as low
visibility range has presented the highest
propagation distance compared to both medium
and high visibility.
Based on the modeling equations analysis and the
assumed set of the operating parameters as shown in
Table 1. The following facts are assured as shown in the
series of Figs. (2-31):
i) Figs. (2-7) have assured that maximum
propagation distance decreases with increasing
visibility ranges for both optical transmission
Maximum propagation distance, L, km
45
42.5
2' nd o ptical transmissio n windo w =1.3 m
40
3' rd o ptical transmissio n windo w =1.55 m
37.5
35
32.5
30
27.5
25
22.5
Dry snow
20
17.5
15
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 2. Maximum propagation distance in relation to low visibility at different optical transmission windows at the
assumed set of the operating parameters.
Maximum propagation distance, L, km
46
42
Wet snow
38
34
30
26
2' nd o ptical transmissio n windo w =1.3 m
3' rd o ptical transmissio n windo w =1.55 m
22
18
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 3. Maximum propagation distance in relation to low visibility at different optical transmission windows at the
assumed set of the operating parameters.
Page 654
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
Maximum propagation distance, L, km
12
11
2' nd o ptical transmissio n windo w =1.3 m
10
3' rd o ptical transmissio n windo w =1.55 m
9
8
7
6
5
Dry snow
4
3
2
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 4. Maximum propagation distance in relation to medium visibility at different optical transmission windows at the
assumed set of the operating parameters.
Maximum propagation distance, L, km
14
13
Wet snow
12
11
10
9
8
7
6
2' nd o ptical transmissio n windo w =1.3 m
5
4
3' rd o ptical transmissio n windo w =1.55 m
3
2
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 5. Maximum propagation distance in relation to medium visibility at different optical transmission windows at the
assumed set of the operating parameters.
Page 655
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
Maximum propagation distance, L, km
4
2' nd o ptical transmissio n windo w =1.3 m
3.5
3' rd o ptical transmissio n windo w =1.55 m
3
2.5
2
1.5
1
Dry snow
0.5
0
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 6. Maximum propagation distance in relation to high visibility at different optical transmission windows at the
assumed set of the operating parameters.
Maximum propagation distance, L, km
4.5
2' nd o ptical transmissio n windo w =1.3 m
4
3' rd o ptical transmissio n windo w =1.55 m
3.5
3
2.5
2
1.5
Wet snow
1
0.5
0
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 7. Maximum propagation distance in relation to high visibility at different optical transmission windows at the
assumed set of the operating parameters.
Page 656
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
100
Received signal power, PR, Watt
2' nd o ptical transmissio n windo w =1.3 m
90
3' rd o ptical transmissio n windo w =1.55 m
80
70
60
50
Dry snow
40
30
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 8. Received signal power in relation to low visibility at different optical transmission windows at the assumed set
of the operating parameters.
Received signal power, PR, Watt
85
80
2' nd o ptical transmissio n windo w =1.3 m
75
3' rd o ptical transmissio n windo w =1.55 m
70
65
60
55
50
45
40
Wet snow
35
30
25
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 9. Received signal power in relation to low visibility at different optical transmission windows at the assumed set
of the operating parameters.
Page 657
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
Received signal power, PR, Watt
200
185
2' nd o ptical transmissio n windo w =1.3 m
170
3' rd o ptical transmissio n windo w =1.55 m
155
140
125
110
95
Dry snow
80
65
50
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 10. Received signal power in relation to medium visibility at different optical transmission windows at the assumed
set of the operating parameters.
180
Received signal power, PR, Watt
2' nd o ptical transmissio n windo w =1.3 m
160
3' rd o ptical transmissio n windo w =1.55 m
140
120
100
80
Wet snow
60
40
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 11. Received signal power in relation to medium visibility at different optical transmission windows at the assumed
set of the operating parameters.
Page 658
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
475
Received signal power, PR, Watt
450
2' nd o ptical transmissio n windo w =1.3 m
425
3' rd o ptical transmissio n windo w =1.55 m
400
375
350
325
300
275
250
225
Dry snow
200
175
150
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 12. Received signal power in relation to high visibility at different optical transmission windows at the assumed set
of the operating parameters.
Received signal power, PR, Watt
400
375
2' nd o ptical transmissio n windo w =1.3 m
350
3' rd o ptical transmissio n windo w =1.55 m
325
300
275
250
225
200
Wet snow
175
150
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 13. Received signal power in relation to high visibility at different optical transmission windows at the assumed set
of the operating parameters.
Page 659
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
16
2' nd o ptical transmissio n windo w =1.3 m
Signal to noise ratio, SNR, dB
15
3' rd o ptical transmissio n windo w =1.55 m
14
13
12
11
10
9
8
Dry snow
7
6
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 14. Signal to noise ratio in relation to low visibility at different optical transmission windows at the assumed set of
the operating parameters.
14
2' nd o ptical transmissio n windo w =1.3 m
Signal to noise ratio, SNR, dB
13
3' rd o ptical transmissio n windo w =1.55 m
12
11
10
9
8
7
Wet snow
6
5
4
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 15. Signal to noise ratio in relation to low visibility at different optical transmission windows at the assumed set of
the operating parameters.
Page 660
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
Signal to noise ratio, SNR, dB
25
22.5
20
Dry snow
17.5
15
2' nd o ptical transmissio n windo w =1.3 m
12.5
3' rd o ptical transmissio n windo w =1.55 m
10
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 16. Signal to noise ratio in relation to medium visibility at different optical transmission windows at the assumed
set of the operating parameters.
23.5
2' nd o ptical transmissio n windo w =1.3 m
Signal to noise ratio, SNR, dB
22
3' rd o ptical transmissio n windo w =1.55 m
20.5
19
17.5
16
Wet snow
14.5
13
11.5
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 17. Signal to noise ratio in relation to medium visibility at different optical transmission windows at the assumed
set of the operating parameters.
Page 661
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
Signal to noise ratio, SNR, dB
35
34
2' nd o ptical transmissio n windo w =1.3 m
33
3' rd o ptical transmissio n windo w =1.55 m
32
31
30
29
28
27
Dry snow
26
25
24
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 18. Signal to noise ratio in relation to high visibility at different optical transmission windows at the assumed set of
the operating parameters.
33
Signal to noise ratio, SNR, dB
32
2' nd o ptical transmissio n windo w =1.3 m
31
3' rd o ptical transmissio n windo w =1.55 m
30
29
28
27
26
25
24
Wet snow
23
22
21
20
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 19. Signal to noise ratio in relation to high visibility at different optical transmission windows at the assumed set of
the operating parameters.
Page 662
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
0.007
2' nd o ptical transmissio n windo w =1.3 m
Bit error rate, BERx10-9
0.006
3' rd o ptical transmissio n windo w =1.55 m
0.005
0.004
0.003
0.002
Dry snow
0.001
0
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 20. Bit error rate in relation to low visibility at different optical transmission windows at the assumed set of the
operating parameters.
0.007
2' nd o ptical transmissio n windo w =1.3 m
0.006
Bit error rate, BERx10-9
3' rd o ptical transmissio n windo w =1.55 m
0.005
0.004
0.003
0.002
Wet snow
0.001
0
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 21. Bit error rate in relation to low visibility at different optical transmission windows at the assumed set of the
operating parameters.
Page 663
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
0.12
3' rd o ptical transmissio n windo w =1.55 m
Bit error rate, BERx10-10
0.108
2' nd o ptical transmissio n windo w =1.3 m
0.096
0.084
0.072
0.06
0.048
0.036
Dry snow
0.024
0.012
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 22. Bit error rate in relation to medium visibility at different optical transmission windows at the assumed set of the
operating parameters.
0.16
3' rd o ptical transmissio n windo w =1.55 m
0.14
Bit error rate, BERx10-10
2' nd o ptical transmissio n windo w =1.3 m
0.12
0.1
0.08
0.06
Wet snow
0.04
0.02
0
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 23. Bit error rate in relation to medium visibility at different optical transmission windows at the assumed set of the
operating parameters.
Page 664
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
0.018
0.016
Bit error rate, BERx10-12
Dry snow
0.014
0.012
0.01
0.008
2' nd o ptical transmissio n windo w =1.3 m
0.006
3' rd o ptical transmissio n windo w =1.55 m
0.004
0.002
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 24. Bit error rate in relation to high visibility at different optical transmission windows at the assumed set of the
operating parameters.
0.035
3' rd o ptical transmissio n windo w =1.55 m
Bit error rate, BERx10-12
0.03
2' nd o ptical transmissio n windo w =1.3 m
0.025
0.02
0.015
Wet snow
0.01
0.005
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 25. Bit error rate in relation to high visibility at different optical transmission windows at the assumed set of the
operating parameters.
Page 665
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
Maximum bit rate, BRmax., Gbit/sec
0.11
0.1
Dry snow
0.09
0.08
0.07
0.06
2' nd o ptical transmissio n windo w =1.3 m
0.05
3' rd o ptical transmissio n windo w =1.55 m
0.04
0.03
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 26. Maximum transmission bit rate in relation to low visibility at different optical transmission windows at the
assumed set of the operating parameters.
Maximum bit rate, BRmax., Gbit/sec
0.1
0.09
2' nd o ptical transmissio n windo w =1.3 m
0.08
3' rd o ptical transmissio n windo w =1.55 m
0.07
0.06
0.05
0.04
0.03
Wet snow
0.02
0.01
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low visibility, VLow, km
Fig. 27. Maximum transmission bit rate in relation to low visibility at different optical transmission windows at the
assumed set of the operating parameters.
Page 666
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
Maximum bit rate, BRmax., Gbit/sec
1.4
2' nd o ptical transmissio n windo w =1.3 m
1.2
3' rd o ptical transmissio n windo w =1.55 m
1
0.8
0.6
Dry snow
0.4
0.2
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 28. Maximum transmission bit rate in relation to medium visibility at different optical transmission windows at the
assumed set of the operating parameters.
Maximum bit rate, BRmax., Gbit/sec
1
0.9
2' nd o ptical transmissio n windo w =1.3 m
0.8
3' rd o ptical transmissio n windo w =1.55 m
0.7
0.6
0.5
0.4
Wet snow
0.3
0.2
0.1
6
10
14
18
22
26
30
34
38
42
46
50
Medium visibility, VMedium, km
Fig. 29. Maximum transmission bit rate in relation to medium visibility at different optical transmission windows at the
assumed set of the operating parameters.
Page 667
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
Maximum bit rate, BRmax., Gbit/sec
3.25
2' nd o ptical transmissio n windo w =1.3 m
3
3' rd o ptical transmissio n windo w =1.55 m
2.75
2.5
2.25
2
1.75
1.5
Dry snow
1.25
1
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 30. Maximum transmission bit rate in relation to high visibility at different optical transmission windows at the
assumed set of the operating parameters.
2.75
Maximum bit rate, BRmax., Gbit/sec
2' nd o ptical transmissio n windo w =1.3 m
2.5
3' rd o ptical transmissio n windo w =1.55 m
2.25
2
1.75
1.5
1.25
1
Wet snow
0.75
0.5
50
52.5
55
57.5
60
62.5
65
67.5
70
72.5
75
77.5
80
High visibility, VHigh, km
Fig. 31. Maximum transmission bit rate in relation to high visibility at different optical transmission windows at the
assumed set of the operating parameters.
observed that second optical transmission window
ii) As shown in Figs. (8-13) have proved that received
has presented lower signal to noise ratio compared
signal power increases with increasing visibility
to third optical transmission window. As well as
ranges for both optical transmission windows
high visibility range has presented the highest
under both bad weather effects of dry and wet
signal to noise ratio compared to both medium and
snow. It is also observed that second optical
low visibility.
transmission window has presented lower received
iv) As shown in Figs. (20-25) have assured that bit
signal power compared to third optical
error rate increases with increasing visibility ranges
transmission window. As well as high visibility
for both optical transmission windows under both
range has presented the highest received signal
bad weather effects of dry and wet snow. It is
power compared to both medium and low
observed that second optical transmission window
visibility.
has presented higher bit error rate compared to
iii) Figs. (14-19) have indicated that signal to noise
third optical transmission window. As well as low
ratio increases with increasing visibility ranges for
visibility range has presented the highest bit error
both optical transmission windows under both bad
rate compared to both medium and high visibility.
weather effects of dry and wet snow. It is also
Page 668
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
v) Figs. (26-31) have indicated that maximum
transmission bit rate increases with increasing
visibility ranges for both optical transmission
windows under both bad weather effects of dry and
wet snow. It is also observed that third optical
transmission window has presented higher
transmission bit rate compared to second optical
transmission window. As well as high visibility
range has presented the highest transmission bit rate
compared to both medium and low visibility.
V. Conclusions
In a summary, the wireless optical communication
systems have deeply investigated under the bad
weather of rain, fog, scattering dry and wet snow over
wide range of the affecting parameters. Maximum
propagation distance, received signal power, signal to
noise ratio, bit error rate, and transmission rates for
different visibility ranges are the major interesting
design parameters as a measurement of the system
performance under different optical transmission
windows. It is theoretically found that wet snow has
presented bad effects on the wireless optical
communication systems compared to dry snow. As
well as optical wireless communication systems have
presented the highest received signal power, signal to
noise ratio, transmission bit rates, and the lowest
propagation distance and bit error rate for different
visibility ranges at third optical transmission window
compared to second optical transmission window.
REFERENCES
[1] A. Nkansah and N. J. Gomes, “A WDM/SCM
Star/Tree Fiber Feed Architecture for Pico-cellular
Broadband Systems”, International Topical Meeting
on Microwave Photonics, pp. 271-274, Sept. 2003.
[2] Abd El-Naser A. Mohammed, Mohamed M. E. ElHalawany, Ahmed Nabih Zaki Rashed, and
Mohammed S. F. Tabour “High Transmission
Performance of Radio over Fiber Systems over
Traditional Optical Fiber Communication Systems
Using Different Coding Formats for Long Haul,”
International Journal of Computer Science and
Telecommunications (IJCST), Vol. 2, No. 3, pp. 2942, June 2011.
[3] C. Lim, D. Novak, “Capacity Analysis for WDM
Fiber-Radio Backbones with Star Tree and Ring
Architecture Incorporating Wavelength Interleaving”,
Journal of Lightwave Tech., Vol. 21, No. 12, pp.
3308-3315, Dec. 2003.
[4] S. Sarkar, S. Dixit, and B. Mukherjee, “Hybrid
Wireless Optical Broadband Access Network
(WOBAN)μ A Review of Relevant Challenges,”
IEEE/OSA Journal on Lightwave Technology, , Vol.
25, No. 11, pp. 1233-1239, Nov. 2007.
[5] H. Hennier, O. Wilfert, “An Introduction to Free space
Optical Communications,” Radio Engineering
Journal, Vol. 19, No. 2, pp. 203-212, June 2010.
[6] Abd El-Naser A. Mohammed, Ahmed Nabih Zaki
Rashed, and Mohammed S. F. Tabour “Transmission
Characteristics of Radio over Fiber (ROF)
MillimeterWave Systems in Local Area Optical
Communication Networks,” International Journal of
Advanced Networks and Applications, Vol. 2, No. 6,
pp. 876-886, May/June 2011.
[7] H. Manor and S. Arnon, “Performance of an Optical
Wireless Communication System As A function of
Wavelength,” Applied Optics , Vol. 42, No. 21, pp.
4285–4294, July 2003.
[8] C. C. Davis, I. I. Smolyaninov and S. D. Milner,
“Flexible Optical Wireless Links and Networks”,
IEEE Communication Magazine, Vol. 3, No. 2, pp.
51-57, Mar. 2003.
[9] A. K. Majumdar and J. C. Ricklin, “Free Space Laser
Communications, Principles and advantages”,
Springer Science 2008.
[10] J. C. Juarez, A. Dwivedi, A. R. Hammons, S. D.
Jones, V. Weerackody and R. A. Nichols, “Free
Space Optical Communications for Next Generation
Military
Networks”,
IEEE
Communication
Magazine, Vol. 2, No. 1, pp. 46-51, November 2006.
[11] D. Giggenbach, J. Horwath and B. Epple, “Optical
Satellite Downlinks to Optical Ground Stations and
High-Altitude Platforms,” IST Mobile & Wireless
Communication Summit, Budapest, Hungary, July
2007.
[12] M. Aliba, K. Oawa, and S. Ito, “Measurement and
Simulation of the Effect of Snow Fall on Free Space
Optical Propagation,” Applied Optics, Vol. 47, No.
31, pp. 5736-5743, 2008.
[13]
H.
Hemmati,
“Deep
Space
Optical
Communications”, John Wiley & Sons, 2006.
[14] Ibrahim M. El-dokany, Abd El–Naser A. Mohamed,
Ahmed Nabih Zaki Rashed, and Amina M. ElNabawy, “Upgrading Efficiency and Improvement of
the Performance of Broadband Wireless Optical
Access Communication Networks” International
Journal of Communication Networks and Information
Security (IJCNIS), Vol. 3, No. 2, pp. 149-162, Aug. 2011.
Page 669
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 650-670
ISSN 2078-2365
[15] S. Muhammad, B. Flecker, E. Leitgeb and M.
Gebhart, “Characterization of Fog Attenuation in
Terrestrial Free Space Optical Links”, Journal of
optical Engineering, Vo. 46, No. 4, pp. 1-9, June
2007.
[16] M. Akiba, K. Ogawa, K. Wakamori, K. Kodate and
S. Ito, “Measurement and Simulation of the Effect of
Snowfall on Free Space Optical Propagation”,
Journal of Applied Optics, Vol. 47, No. 31, pp. 57365743, Nov. 2008.
[17] S. Betti, V. Carrozzo, E. Duca, “Over Stratospheric
Altitude Optical Free Space Links: System
Performance
Evaluation Transparent
Optical
Networks”, in ICTON, International Conference on
Transparent Optical Networks, July 2007.
[18] P. F. J. Lermusiaux, C.-S. Chiu, G. G.
Gawarkiewicz, P. Abbot, A. R. Robinson, R. N.
Miller, P. J. Haley, W. G. Leslie, S. J. Majumdar, A.
Pang and F. Lekien, “Quantifying Uncertainties in
Ocean Predictions” in Oceanography, special issue
on “Advances in Computational Oceanography”, T.
Paluszkiewicz and S. Harper, Eds., Vol. 19, No. 1,
pp. 92-105, 2006.
[19] Ahmed Nabih Zaki Rashed, “High Transmission Bit
Rate of Multi Giga Bit per second for Short Range
Optical Wireless Access Communication Networks”
International Journal of Advanced Science and
Technology, Vol. 32, pp. 23-32, July 2011.
[20] D. Giggenbach, J. Horwarth, M. Knapek, “Optical
Data Downlinks From Earth Observation Platforms”,
Free Space Laser Communication Technologies XXI,
Proc. of SPIE, Vol. 19, No. 3, pp. 12-25, 2009.
[21] P. Corrigan, R. Martini, E. A. Whittaker, C. Bethea,
“Quantum Cascade Lasers and the Kruse model in
Free Space Optical Communication”, Optics Express,
Vol. 17, No. 6, pp. 432-444, 2009.
[22] Z. Bielecki, W. Kolosowski, J. Mikolajczyk, “Free
Space Optical Data Link Using Quantum Cascade
Laser”, PIERS Proceedings, Cambridge, USA, July
2-6, 2008.
[23] Abd El-Naser A. Mohammed, Mohamed M. E. ElHalawany, Ahmed Nabih Zaki Rashed, and
Mohamoud M. Eid “Optical Add Drop Multiplexers
with UW-DWDM Technique in Metro Optical
Access Communication Networks,” International
Journal
of
Computer
Science
and
Telecommunications (IJCST), Vol. 2, No. 2, pp. 513, April 2011.
[24] P. S. Andre, A. N. Pinto, J. L. Pinto, T. Almeida, and
M. Pousa,
“Selective Wavelength Transparent
Optical Add-Drop Multiplexer Based on Fiber Bragg
Gratings ,” J. Opt. Commun., Vol. 24, No. 3, pp. 222229, 2006.
[25] M. S. Ab-Rahaman, S. Suliana, K. Mat, and B. Ng,
“The Hybrid Protection Scheme in Hybrid
OADM/OXC/MUX ,” Australian J. of Basic and
Applied Sciences, Vol. 2, No. 4, pp. 968-976, 2008.
[26] A. A. Aboul Enein, F. Z. El-Halafawy, M. H. A.
Hassan, A. A. Mohammed, “Thermal Environmental
Effects,” Alex. Eng. J., Alex. Univ., Vol. 28, No. 2,
pp.169-183, 1989.
Page 670
Abd El–Naser et. Al., The Effects of the Bad Weather on the Transmission and Performance Efficiency of Optical
Wireless Communication Systems
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 769-775
ISSN 2078-2365
Combined Economic and Emission Dispatch for
a Wind Integrated System Using Particle Swarm
Optimization
K.S. Linga Murthy1, G.V.S.Subramanyam2, K.SriChandan3
E.E.E. Department, Gitam University, India.
(Email: 1 hod_eee@gitam.edu, 2subbu.venkatasatya@gmail.com, 3srichandank@gmail.com)
Abstract—This paper deals with the problems of Economic
Dispatch, Emission Dispatch and Combined Economic and
Emission Dispatch problems for an integrated system having
thermal and wind units. Particle Swarm Optimization and
Genetic Algorithm methods are used to solve the problems of
Economic dispatch, Emission dispatch and Combined Economic
and Emission Dispatch problems. The effectiveness of PSO and
GA methods are demonstrated by comparing the results obtained
with both the methods.
Keywords —Emission level, Fuel cost, Particle Swarm
optimization, Wind integration.
I. INTRODUCTION
A
Power system is a mix of different types of generations,
out of which thermal, hydro and nuclear power
generations have the maximum contribution. However,
economic operation has conveniently been considered by
proper scheduling of thermal and hydro-generation only. The
nuclear stations are run at their base loads keeping safety in
mind [13].
The purpose of economic dispatch is to find out the most
economical schedule of the generating units while satisfying
load demand and operational constraints. Economic dispatch is
a familiar problem pertaining to the allocation of the amount
of power to be generated by different units in the system on an
optimum economy base [1]. This problem has been tackled by
many researchers in the past. Recently the problem which has
attracted much attention is pollution minimization due to
pressing public demand for clean air. Environmental pollution
is a direct consequence of industrial advancement.
Technology, which has made economic development possible,
produces enormous quantities of harmful by-products and
wastes.
Thermal power stations are major causes of atmospheric
pollution, because of high concentration of pollutants they
cause. It is utmost important to protect our environment from
harmful emissions out from thermal power plants. Power
utilities using fossil fuels as a primary energy source, give rise
to particulates and gaseous pollutants apart from heat. The
particulates as also the gaseous pollutants such as carbon
dioxide (CO2), oxides of sulphur (SOX) and oxides of nitrogen
(NOx) cause detrimental effects of CO2 on the environment is
not yet precisely known. Pollution control agencies
(Municipal/Governmental regulatory bodies) restrict the
amount of emission of pollutants depending upon their relative
harmfulness to human beings [1]. So, the emission dispatch
has been formulated. Of the pollutants emitted, NOx is of
major concern and hence it has been considered. The objective
of emission dispatch is to minimize the total environmental
degradation or the total pollution emission due to burning of
fuels for production to meet the load demand [14]. Hence,
there is a need to formulate the combined emission and
economic dispatch (CEED) problem.The idea behind
combined emission and economic dispatch is to compute the
optimal generation for individual units of the power system by
minimizing the fuel cost and emission levels simultaneously,
Page 769
Linga et. al., Combined Economic and Emission Dispatch for a Wind Integrated System Using Particle Swarm Optimization
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 769-775
ISSN 2078-2365
subjected to the system constraints.
E. Constraints
1. Power Balance Constraint
P
ng
II. Problem formulation
A. Economic dispatch
The generation cost function is usually expressed as a
quadratic polynomial and can be represented as below for an
ith generator.
+ bi Pi +ci Rs/hr
( 1)
In the expression above, Pi is the output power in MW and ai
, bi, ci are the fuel cost-coefficients of the ith generating unit.
B. Emission Dispatch
Pollutant Emissions from the generating units such as
oxides of nitrogen can be expressed as a quadratic polynomial
and can be represented as below for an ith generator.
Ei = αi Pi2 + βi Pi + i Kg/hr
(2)
In the expression above, Pi is the output power in MW and
αi, βi, i are the emission coefficients of the ith generating unit.
C. Constraints
1. Power Balance Constraint
P
ng
gi
= PD + PL
(3)
i1
Where, ng = number of generating units, P gi is the power
generated by ith unit in MW, Pw is the wind power that is
available in MW, PD is the load demand in MW, PL is the
transmission loss in MW.
2. Generating Limits
Pgimin ≤ Pgi ≤ Pgimax (i=1,2,3,....ng)
ng= number of generators.
D. CEED Cost Function
In this formulation both fuel cost objective and emission
level objective are combined to form a single objective with
the introduction of factor called ‘The Price Penalty Factor’,
hm (Rs/Kg).
Minimize
FT =( a i P i2 + bi P i + ci ) + hm( αi P i2 + i P i + i)
Where, FT is the total cost of generation(RS/hr).
(4)
(5)
Where, ng = number of generating units, Pgi is the power
generated by ith unit in MW, PD is the load demand in MW, PL
is the transmission loss in MW.
Losses can be calculated by B coefficients, which can be
expressed as
PL =
Fi =ai Pi2
= PD + PL
gi
i1
ng
ng
i 1
j 1
Pi Bij Pj
(6)
Where, Bij is generation loss coefficient. Pi and Pj are
The real power injections at ith and jth buses respectively.
2. Generating Limits
Pgimin ≤ Pgi ≤ Pgimax (i=1,2,3,....ng)
ng= number of generators.
(7)
III. WIND INTEGRATION
This paper deals with a multi-objective generation dispatch
problem that considers environment and fuel cost under
substantial penetration of wind energy has been proposed [11].
Wind plants are different from conventional generation plants
in that their fuel supply is neither steady nor controllable, and
as a result, they exhibit greater uncertainty and variability in
their output [10].
Wind plants naturally operate when the wind blows, and
their power levels vary with the strength of the wind. The
turbine power output is controlled by pitching the blades. With
each new generation of wind turbines, the size has increased
and reductions in the life-cycle cost of energy have been
achieved through economies of turbine scale and a larger rotor
to increase energy capture. However, there are constraints to
this continued growth in size. At some point, it will cost more
to build a larger turbine than the benefit of increased energy
benefit is worth. In addition, land transport restrictions, cost as
well as crane requirements, can impose size limits for wind
turbines installed on land. A misconception about wind power
[9] is that wind plants will cause the entire power system to
collapse. But, because abrupt wind-related changes in plant
output do not occur, this fear is unfounded. In fact, a modern
wind plant will actually help a power system handle a major
outage or contingency elsewhere on the system. Reactivepower control and low-voltage ride-through capabilities of
modern wind plants actually improve system stability.
Wind-energy generation only occurs when the wind is
Page 770
Linga et. al., Combined Economic and Emission Dispatch for a Wind Integrated System Using Particle Swarm Optimization
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 769-775
ISSN 2078-2365
blowing [12]. Wind power is therefore not dispatchable like
conventional energy sources and delivers a variable level of
power depending on the wind speed. Wind is primarily an
energy resource and not a capacity resource. Its primary value
is to offset fuel consumption and the resulting emissions. The
output of output of wind power plant, or multiple wind power
plants, is variable over time. Each megawatt generated by wind
reduces the required generation of other units. Therefore, the
remaining nonwind generation units only need to supply the
load that is not supplied by the wind. This remaining load is
often called the net load. Therefore, the non-wind portion of
the power system is operated to the net load, which is the
difference between load and wind. Although wind is a variable
resource, operating experience and detailed wind integration
studies have yet to find a credible and firm technical limit to
the amount of wind energy that can be accommodated by
electrical grids. Some countries already receive a significant
amount of electricity from wind power. There is not a
technical limit to increased penetration of wind energy but
there might be an economic limit, a point at which it is deemed
too expensive to accommodate more energy from wind in
comparision with the value that it adds to the system.
IV. Problem formulation with Wind Integration
A.CEED problem formulation with WIND Integration
Minimize
FT = (ai Pi2 + bi Pi +ci) + hm ( αi Pi2 + βi Pi + i)
Where, FT is the total cost of generation (RS/hr).
B. Constraints
1 Power Balance Constraint
P
ng
gi
+Pw = PD + PL
(8)
i1
where, ng = number of generating units, Pgi is the power
generated by ith uint in MW, Pw is the wind power that is
available in MW, PD is the load demand in MW, PL is the
transmission loss in MW.
2. Generating Limits
Pgimin ≤ Pgi ≤ Pgimax (i=1,2,3,....ng)
ng= number of generators
V.
GENETIC ALGORITHM
The aim of optimization is to determine the best-suited
solution to a problem under a given set of constraints. Since
the beginning of the nineteenth century, a significant evolution
in optimization theory has been noticed [15]. Classical linear
programming and traditional non-linear optimization
techniques such as Lagrange’s Multiplier was prevalent until
this century. Unfortunately, these derivative based
optimization techniques can no longer be used to determine the
optima on rough non-linear surfaces. One solution to this
problem has been put forward by the evolutionary algorithms
research community. Genetic algorithm (GA), enunciated by
Holland, is one such popular technique which comes under
evolutionary algorithms.
Genetic Algorithm consists of a string representation of points
in the search space, a set of genetic operators for generating
new search points, a fitness function to evaluate the search
points and a stochastic assignment to control the genetic
operations [12]. It typically consists of three phases.
1. Initialization
2. Evaluation
3. Genetic Operation
Initialization is the generation of initial population of
chromosomes i.e. initial search points. Fitness function is so
selected that the most fit solution is the nearest to the global
optimum point. For minimization type problems, fitness
function can be function of variables that bear inverse
proportionality relationship with the objective function. The
genetic operators are reproduction, crossover, and mutation.
Reproduction is simply an operator where by an old
chromosome is copied into a mating pool according to its
fitness value. The commonly used method for selecting
chromosomes for parents to cross over is Roulette Wheel
selection, in roulette wheel selection technique, selection is
usually implemented as a linear search through roulette wheel
with slots weighed in proportion to string fitness values. The
crossover is mainly responsible for the global search property
of the GA. It is recombination operation. Here the gene
information (information in a bit) contained in the two selected
parents is utilized in certain fashion to generate two children
who bear some of the useful characteristics of parents and
expected to be more fit than parents. Usually, the probability
of Crossover (PC) is high and chosen to be in between (0.6 to
0.8). Mutation operator is capable of creation new genetic
material in the population to maintain the population diversity.
It is nothing but random alteration of a bit value at a particular
bit position in the chromosome. Usually, the probability of
Mutation (PM) is very less and is chosen to be in between
(0.001 to 0.01). We have another operator in GA, called
Elitism. The copying of best population to next population is
called Elitism. If the probability is high, then the convergence
rate increases. Usually, the probability of Elitism (P E) is
chosen to be 0.15
Page 771
Linga et. al., Combined Economic and Emission Dispatch for a Wind Integrated System Using Particle Swarm Optimization
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 769-775
ISSN 2078-2365
A. Implementation of CEED with Genetic Algorithm:
Step7: Repeat the procedure from Step 2 until chromosome
count > population size.
Proposed Algorithm for solving CEED problem:
Step1: Generate initial population of chromosome of binary
bits using random generation technique.
Step2: Implementation of a problem in a GA starts from the
parameter encoding. Proposed approach uses the equal system
λ (equal system incremental cost) criterion as its basis[11]. The
only encoded parameter is the normalized system incremental
cost, λnm. Decode the chromosomes of the population and
determine normalized system incremental cost, λnm
yj =
l
i 1
2i-1 *bij (j=1,β,…L)
(9)
(10)
(11)
(12)
(13)
obtained
(14)
Step4:Calculate the generation output of all the units for each
chromosome from its λf,e value and enforce P i limits.
Pi =
(bi + hm *βi) ng
- 2 Bij Pj
λ f,e
i=1
2α +βhmαi
( i
)
λ f,e
(15)
Step5: Calculate transmission losses using B-coefficient
equation (5) and compute the error ε
P
ng
ε = |PD-PL-
gi
|
(16)
i1
Step6: Calculate the fitness value of the chromosome, using
the equation
fitness=
1/ (1 50* ferror )
ferror= abs(Pgi-PD-Ploss)
Step10: Copy the PE % chromosomes of old population to new
population starting from the best ones from the top.
Step11: Perform crossover on selected parents and generate
new child chromosomes, repeat it to get required number of
chromosomes.
Step13: Perform mutation on all chromosomes.
Step14: Replace old population with new population.
l=length if string.
bij =ith binary digit of jthstring.
L=population size.
1-
Step9: Check if the error is less than ε. if yes, go to Step 15.
Step12: Add all the generated child chromosomes to new
population.
Step3:Calculate the actual system incremental cost λf,e
Initial point in search space, λf,e is calculated as,
λf,e = λmin + λnm(λmax – λmin)
Calculating λmin and λmax values:
α(i)=b(i)+hm*β(i)
β(i)=β*(a(i)+hm*α(i))
λmin=α(i)+(β(i)*Pmin(i))
λmax=α(i)+(β(i)*Pmax(i))
The equivalent decimal integer of binary string λ is
from:
Step8: Sort the chromosomes and all their related data in the
descending order of fitness.
(17)
(18)
Step15: Calculate the total cost, fuel cost, emission release,
emission cost, power generated by units.
VI. Particle Swarm Optimization
The aim of optimization is to determine the best-suited
solution to a problem under a given set of constraints. Since
the beginning of the nineteenth century, a significant evolution
in optimization theory has been noticed [15]. Classical linear
programming and traditional non-linear optimization
techniques such as Lagrange’s Multiplier was prevalent until
this century. Unfortunately, these derivative based
optimization techniques can no longer be used to determine the
optima on rough non-linear surfaces. One solution to this
problem has been put forward by the evolutionary algorithms.
Genetic algorithm (GA), enunciated by Holland, is one such
popular algorithm which is a guided search technique. When it
comes to evolutionary programming, techniques like Particle
Swarm optimization (PSO) and Differential Evolution (DE)
have been proposed. These algorithms are inspired by
biological and sociological motivations and can take care of
optimality on rough, discontinuous and multimodal surfaces.
Particle Swarm Optimization (PSO) has been developed
through simulation of simplified social models [2]. This
algorithm is motivated by the behavior of organisms such as
bird flocking and fish schooling and it utilizes a population
based search procedure. The algorithm searches a space by
adjusting the trajectories of individual vectors, called
“particles” as they are conceptualized as moving points in
multidimensional space. The individual particles are drawn
stochastically towards the positions of their own previous best
Page 772
Linga et. al., Combined Economic and Emission Dispatch for a Wind Integrated System Using Particle Swarm Optimization
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 769-775
ISSN 2078-2365
performance and the best previous performance of their
neighbors. The particles are thought of as collision-proof birds
and the original intent is to graphically simulate the graceful
but unpredictable choreography of a bird flock.
PSO is initialized with a group of random particles and then
searches for optima by updating generations. Each particle in
PSO represents a feasible solution. In other words, each
particle represents a point in multi-dimensional search space,
in which optimal point is to be determined. Each particle
changes its state by ‘flying’ around the multi-dimensional
search space until a relatively unchanging state (optimal state)
has been obtained. In every iteration, each particle is updated
by following two “best” values. The first one is the best
solution it has achieved so far. This value is called “localbest”.
Another “best” value that is tracked by the particle swarm
optimizer is a global best and called “globalbest”.
In PSO, the coordinates of each particle represent a possible
solution that has two vectors associated with it, position (xi)
and velocity (vi) vectors [16]. The size of the vectors xi and vi
is equal to the problem space dimension. Each particle updates
its position based on its own best exploration, best swarm
overall experience, and its previous velocity vector according
to the following equations:
vik+1= wvik+c1r1(localbesti-xik)+c2r2(globalbesti-xik)
(19)
xik+1 = xik+vik+1
(20)
where, c1and c2 are two positive constants, r1and r2 are two
randomly generated numbers with a range of [0,1]. The first
term of right-hand side of (19) corresponds to global search.
The second and third terms of equation (19) corresponds to
local search. So, this method has a well-balanced mechanism
to utilize global and local search efficiently [2].
VII. Algorithm for solving Combined Economic and
Emission Dispatch problem using PSO
Power outputs from each generator are taken as the particles
of the PSO [2].
The PSO algorithm for dispatch problems is stated as follows:
Step1:The particles are randomly generated between the
maximum and minimum operating limits of the generators.
Step2:The particle velocities are generated randomly.
Step3:Objective function values of the particles are evaluated.
Penalties are given for violations of demand constraint
(2).These values are set the localbest value of the particle.
Step4:The best value among all the localbest values
(globalbest) is identified.
Step5:New velocities for the particles are calculated using
(19).
Step6:The positions for each particle are updated using (20).
Step7:New objective function values are calculated for new
positions of the particles. If the new value is better than the
previous localbest, the new value is set to localbest. If the
stopping criterion is met, the positions of the particles
represent the optimum solution. If the stopping criteria is not
met, the procedure is repeated from Step4.
VIII. Results
In GA, the population size is taken as 60,String length =16, P c
=0.70, Pm =0.01, Pe =0.15
In the PSO technique, the population is taken as 40 and the
values of c1 and c2 are c1=2 and c2=2.
The techniques are tested on IEEE 30-bus system [17], having
6 generators and a total demand of 900MW. The cost
coefficients for the generators and their capacities, the
corresponding emission coefficients for the generators and the
B-coefficients considered in[17], are mentioned in the
appendix. The Economic Dispatch problem is solved using
GA and PSO and the results are tabulated in Table I. The
Emission Dispatch problem is solved using GA and PSO and
the results are tabulated in Table II. Later, CEED problem is
solved using GA and PSO and the results are tabulated in
Table III.
The CEED problem is solved using PSO taking into
account wind integration [11]. For a total demand of
PD=900MW, the Transmission losses, Fuel Costs, Emission
release and Total cost are calculated without and with wind
integration. Results are compared in Table IV. The parameters
mentioned above are calculated with increasing wind
penetration and the results are tabulated in Table V.
Computations have been carried out in MATLAB 7.5
environment.
Table I. Results comparing GA and PSO for Economic
Dispatch problem
Method
Transmission
losses (MW)
Fuel cost
(Rs/hr)
Emission
release
(Kg/hr)
GA
27.0153
50691.60
793.319
PSO
20.5518
45464.00
708.54
Page 773
Linga et. al., Combined Economic and Emission Dispatch for a Wind Integrated System Using Particle Swarm Optimization
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 769-775
ISSN 2078-2365
Table II. Results comparing GA and PSO for Emission
Dispatch problem
Emission
release(Kg/hr)
Method
Total cost(Rs/hr)
Transmission
losses (MW)
Emission
release
(Kg/hr)
Fuel cost
(Rs/hr)
452.228
406.069
61518.00
58022.00
IX. Conclusions
GA
26.12
647.11
49634.30
PSO
16.96
646.13
48069.00
Table III. Results comparing GA and PSO for Combined
Economic and Emission Dispatch problem
Method
Transmission
losses (MW)
Fuel cost
(Rs/hr)
Emission
release
(Kg/hr)
Total
cost
(Rs/hr)
GA
23.41
48029.00
687.52
78983.7
PSO
20.55
45464.00
603.61
72651.0
Table IV: Results comparing without and with wind
integration.
Parameters
Transmission
losses (MW)
Fuel
Cost(Rs/hr)
Emission
release(Kg/hr)
Total cost
(Rs/hr)
Without wind
integration
With
PW = 90 MW
20.55
16.728
45464.00
41150.00
603.61
452.228
72651.00
61518.00
Table V: Results with increasing wind penetration.
WindPower
integrated(MW)
Transmission
losses (MW)
Fuel Cost(Rs/hr)
PW = 90
PW = 120
16.728
15.533
41150.00
39732.00
In this paper, GA and PSO techniques are used to solve the
Economic Dispatch problem, the Emission Dispatch problem
and the Combined Economic and Emission Dispatch
problems. The results showed that in Economic Dispatch
problem, the objective being to reduce the fuel cost, the
emission level was higher, also, PSO yielded in better results
when compared with GA. Similarly, in Emission Dispatch
problem, the objective being to reduce the pollution level, the
fuel cost was higher, also, PSO yielded in better results when
compared to GA. Hence, the CEED problem is formulated
whose objective is to reduce both fuel cost and emission
release. Results have shown that PSO yielded in better results
when compared to GA.
Later, when wind integration was taken into account, the
results have shown that the transmission losses, fuel cost,
emission levels and total cost can be reduced. Also, it was
observed that the transmission losses, fuel cost, emission level
and total cost can be reduced further with increase in wind
penetration level.
However, there may be a limit for the amount of wind
power that can be integrated to a system.
X. References
[1]J.Nanda, D.P.Kothari, K.S.Linga Murthy ‘Economic-Emission
load dispatch through goal programming techniques’, IEEE Trans.
on Energy conversion, Vol.3, No 1, March1988.
[2]T.Thakur, Kanik Sem, Simedha Saini and Sudhanshu Sharma ‘A
Particle Swarm Optimization solution to NO2 and SO2 Emissions for
environmentally constrained Economic dispatch problem’, β006
IEEE.
[γ]James Kennedy and Russell Eberhart ‘Particle Swarm
optimization’, 1995 IEEE.
[4]J.W.Lamont and E.V.Obessis ‘Emission dispatch models and
algorithms for the 1990’s’, IEEE Trans. on Power systems, Vol.10,
No.2, May 1995.
[5]A.A.El-Keib, H.Ma, J.L.Hart ‘Environmentally constrained
Economic dispatch using Lagrangian relaxation method’, IEEE
Trans. on Power systems, Vol.9, No.4, Nov. 1994.
[6]Yong-Lin Hu and William G.Wee ‘A Hierarchical system for
Economic dispatch with environmental constraints’, IEEE Trans. on
Power systems,Vol. 9,No.2, May 1994.
Page 774
Linga et. al., Combined Economic and Emission Dispatch for a Wind Integrated System Using Particle Swarm Optimization
International Electrical Engineering Journal (IEEJ)
Vol. 3 (2012) No. 2, pp. 769-775
ISSN 2078-2365
[7]P.Venkatesh, R.Gnanadass and Narayana Prasad Padhy
‘Comparison and application of Evolutionary programming
techniques to Combined Economic Emission dispatch with line flow
constraints’ IEEE Trans. on Power systems, Vol. 8, No.β, Mayβ00γ.
[8]Chao-Lung Chiang ‘Improved genetic algorithm for power
Economic dispatch of units with valve-point effects and multiple
fuel’, IEEE Trans. on Power systems, Vol.β0, No.4, Nov.β005.
[9]‘Wind plant integration’, IEEE power and energy magazine,
Nov./Dec.2005.
[10]‘A mighty wind’, IEEE power and energy magazine,
March/April 2009.
[11]H.T.Jadhav, Ananya Deb, Ranjit Roy ‘A craziness based
Differential Evolution algorithm for thermal-wind generation
dispatch considering Emission and Economy with valve-point effect’,
2011 IEEE.
[12]‘Wind power myths debunked’, IEEE power and energy
magazine Nov./Dec. 2009.
[1γ] ‘Power system optimization’ D.P.Kothari, J.S.Dhillon
[14]Dr.M.Sudhakaran,Dr.S.M.R.Slochanal,R.Sreeram,
N.Chandrasekhar ‘Application of refined Genetic algorithm to
Combined Economic and Emission dispatch’, β004 IEEE.
[15]Swagatam Das, Ajith Abraham, Amit Konar ‘Particle Swarm
optimization and Differential evolution algorithms: technical
analysis, applications and hybridization perspectives’
Springer-Verlag Berlin Heidelberg 2008.
[16] M.A.AlRashidi and M.E. El-Hawary ‘Economic dispatch with
environmental considerations using Particle Swarm optimization’,
2006 IEEE.
[17]M.Rajkumar,S.Kannan,K.Mahadevan,S.Ramesh
3
4
5
6
0.00683
0.00683
0.00461
0.00461
-0.54551
-0.54551
-0.51116
-0.51116
40.26690
40.26690
42.89553
42.89553
The Transmission loss coefficients matrix
0.00014 0.000017 0.000015 0.000019 0.000026 0.000022
0.000017 0.000060 0.000013 0.000016 0.000015 0.000020
0.000015
0.000019
0.000026
0.000022
0.000013
0.000016
0.000015
0.000020
0.000065
0.000017
0.000024
0.000019
0.000017
0.000071
0.000030
0.000025
0.000024
0.000030
0.000069
0.000032
0.000019
0.000025
0.000032
0.000085
‘A Nondominated sorting genetic algorithm-II technique for
environmental/economic power dispatch’, β010 IEEE.
APPENDIX
Fuel Cost Coefficients and Generator Capacity Limits
Generator
1
2
3
4
5
6
a
b
0.1524
0.1058
0.0280
0.0354
0.0211
0.0179
c
38.5397
46.1591
40.3965
38.3055
36.3278
38.2704
756.798
451.325
1049.99
1243.53
1658.55
1356.65
Pmin
(MW)
10
10
35
35
130
125
Pmax
(MW)
125
150
225
210
325
315
Emission Coefficients
Generator
1
2
α
0.00419
0.00419
0.3276
0.3276
13.85932
13.85932
Page 775
Linga et. al., Combined Economic and Emission Dispatch for a Wind Integrated System Using Particle Swarm Optimization