Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
0% found this document useful (0 votes)
41 views

Control of An LLC Resonant Converter Using Load Feedback Linearization

Uploaded by

hsky19748
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
41 views

Control of An LLC Resonant Converter Using Load Feedback Linearization

Uploaded by

hsky19748
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 12

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO.

1, JANUARY 2018 887

Control of an LLC Resonant Converter Using Load


Feedback Linearization
Zhijian Fang , Junhua Wang, Shanxu Duan, Kaipei Liu, and Tao Cai

Abstract—LLC resonant converter is a nonlinear system, limit- voltage switching (ZVS) or zero current switching (ZCS) [10],
ing the use of typical linear control methods. This paper proposed the resonant converters can work at a very high switching fre-
a new nonlinear control strategy, using load feedback linearization quency, reducing the transformer and filter’s dimensions and
for an LLC resonant converter. Compared with the conventional PI
controllers, the proposed feedback linearized control strategy can weights and achieve a high efficiency. Therefore, resonant con-
achieve better performance with elimination of the nonlinear char- verters are receiving more and more attentions.
acteristics. The LLC resonant converter’s dynamic model is built Current papers report several popular types of resonant con-
based on fundamental harmonic approximation using extended de- verters. Series resonant converters (SRC), consisting of series
scribing function. By assuming the dynamics of resonant network inductor and capacitor, can regulate the output voltage by con-
is much faster than the output voltage and controller, the LLC
resonant converter’s model is simplified from seven-order state trolling the switching frequency. High efficiency can be obtained
equations to two-order ones. Then, the feedback linearized control due to achieving ZVS. However, the SRCs have poor charac-
strategy is presented. A double loop PI controller is designed to teristics under no-load and short-circuit conditions. They are
regulate the modulation voltage. The switching frequency can be usually used for the high voltage and low current applications.
calculated as a function of the load, input voltage, and modulation Parallel resonant converters (PRC) can solve these problems us-
voltage. Finally, a 200 W laboratory prototype is built to verify the
proposed control scheme. The settling time of the LLC resonant ing the parallel capacitor. But the appearance of large circulating
converter is reduced from 38.8 to 20.4 ms under the positive load resonant currents will result in large power losses even under
step using the proposed controller. Experimental results prove the no-load condition. They are usually used for narrow voltage
superiority of the proposed feedback linearized controller over the range applications. In order to overcome the drawbacks of the
conventional PI controller. SRC and PRC, series–parallel resonant converters are proposed,
Index Terms—Extended describing functions (EDFs), feedback which add a parallel capacitor. They can operate in a wide volt-
linearization, LLC resonant converter, model. age range with a high efficiency. But the circulating resonant
current is still large especially when the converter works away
I. INTRODUCTION
from the resonant point. The LLC resonant converters are mul-
C/DC converters have been widely used in a variety of
D applications for many years, such as telecommunications
equipment [1], power supply of computer [2], adapter of laptop
tiple resonant converters containing series resonant inductor,
capacitor, and parallel magnetic inductor. ZVS at the primary
side and ZCS at the secondary side are realized. High efficiency
[3], [4], and DC motor drivers [5]. Nowadays, dc/dc convert- and power density are achieved. Moreover, LLC resonant con-
ers are being employed for renewable energy applications like verter can work in a wide voltage range with a comparatively
Photovoltaics (PV) [6], [7], dc microgrid [8], and HVdc systems small switching frequency’s variation. These are very advan-
[9]. In all these applications, high efficiency and high power den- tageous especially when dealing with renewable energy and
sity are the main objectives for a dc/dc converter. Pulse width battery charging applications. Hence, the LLC resonant con-
modulation (PWM) and resonant modulation are two main types verters are becoming quite popular in industrial electronics and
of the dc/dc converters. Due to the capability of achieving zero front-end applications.
However, the LLC resonant converter is a nonlinear system.
Manuscript received June 11, 2016; revised September 15, 2016 and Decem- Controlling an LLC converter is more difficult than a PWM
ber 12, 2016; accepted February 13, 2017. Date of publication February 22, converter due to the complex nonlinear model. The state space
2017; date of current version October 6, 2017. This work was supported in part
by the National Natural Science Foundation of China under Project 51507114,
averaging method is widely used to model the PWM converters
and in part by the Natural Science Foundation of Hubei Province of China un- [11]. But the modulation signals’ frequency should be much
der Project 2014CFB272. Recommended for publication by Associate Editor smaller than switching frequency in these methods, which are
J. A. Oliver. (Corresponding author: J. Wang.)
Z. Fang, J. Wang, and K. Liu are with the School of Electrical Engi-
not satisfied in the LLC resonant converter. Fundamental har-
neering, Wuhan University, Wuhan 430072, China (e-mail: fzj@whu.edu.cn; monic approximation (FHA) and steady-state models can pro-
junhuawang@whu.edu.cn; kpliu@whu.edu.cn). vide good insights into the converter behavior and overall the
S. Duan and T. Cai are with the State Key Laboratory of Advanced Elec-
tromagnetic Engineering and Technology, School of Electrical and Electronic
steady-state gain. But they cannot provide dynamic character-
Engineering, Huazhong University of Science and Technology, Wuhan 430074, istics to guide the controller design [12]–[15]. Some researches
China (e-mail: duanshanxu@hust.edu.cn; caitao@hust.edu.cn). use the simulation results to build the converter model [16],
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
[17]. However, they cannot provide enough model informa-
Digital Object Identifier 10.1109/TPEL.2017.2672731 tion for designing the controller of LLC converter. Small signal

0885-8993 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
888 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 1, JANUARY 2018

models of the LLC converters are built in [18] and [19], based
on the extended describing function (EDF) method. In order
to provide a more accuracy model, the third and fifth har-
monics are considered in [20]. However, these small signal
models cannot be used when the converter operation condi-
tion varies. Jinhaeng et al. [21], [22] build small signal models
of the LLC converter under the worst condition to guide the
controller’s design. But these models limit the dynamic per-
formance of most operation conditions. Resonant current feed- Fig. 1. Circuit diagram of an LLC resonant converter.
back control is proposed to decrease the operation influence
of the small signal model [23]. The large signal models based
on EDF have been proposed for the LLC converter in [24]–
[26]. These models provide enough dynamic information of the
LLC converter when large signal transient disturbance occurs.
But these models are still nonlinear making the control design
difficult.
Linear controllers like PI and PID are applied in the LLC con-
verter to achieve expected performance. Due to the limitations Fig. 2. Conversion stages and subsystems in an LLC resonant converter.
of small signal models, these controllers are only valid near the
particular operation point. A three-pole two-zero compensation
is designed in [27]–[30] to improve the dynamic characteristics
in a wide input and load range. Buccella et al. [31] propose an ob-
server based controller, which improves the dynamic response
compared with the traditional PID controller. But these lin-
ear compensations performances are limited by the converter’s
nonlinearity. Some nonlinear control methods are proposed like
sliding-mode control [32], bang-bang control [33], and robust
control [34], [35]. They have strong robustness against parame-
ter uncertainties and load variations. But the performance is still Fig. 3. Equivalent circuit diagrams of ac subsystem.
undesirable compared to the linear compensation. The optimal
trajectory control is proposed to achieve the best dynamic per- voltage. In order to analyze and design LLC resonant converter,
formance in [36]–[39]. The variable equal load resistor values the mathematical model needs to be built.
need be calculated at every instantaneous time. Meanwhile, the The LLC resonant converter can be divided into two parts:
online computation is nonlinear and complicated. Thus, it is dc/ac stage and ac/dc stage, as shown in Fig. 2. The ac sub-
very difficult to implement. system contains three resonant components. Using FHA, each
In this paper, a feedback linearization controller of an LLC of the three ac state variables can be decomposed into the sine
resonant converter is proposed using the load feedback strat- and cosine components. Then, the ac subsystem is a six-order
egy. A double loop controller is designed to get the modulation dynamic model. In the dc subsystem, the rectifier current flows
voltage. Then, the switching frequency is calculated as a func- through the output capacitor and load. DC subsystem is a one-
tion of the modulation voltage, input voltage, and load. The order dynamic model.
instantaneous equal load is estimated based on the output volt-
age and current feedback. In Section II, a nonlinear model of A. AC Subsystem Modeling
LLC resonant converter is built. Some assumptions are given in
Section III and the complex LLC resonant converter’s model is Fig. 3 shows the equivalent circuit of the ac subsystem. The
simplified from seven to two orders. In Section IV, the proposed square wave voltage generated from the full bridge is applied
feedback linearization control strategy is presented. Finally, the to the resonant tank. The rectifier voltage is equal to the square
experimental results are obtained, validating the effectiveness wave voltage due to the large output capacitor. Based on the
of the proposed method. Kirchhoff’s laws, the sate equations of LLC resonant converter
can be obtained as follows:

II. NONLINEAR MODEL OF LLC RESONANT CONVERTER Lr i̇r = − vc + vab − vm


Fig. 1 shows the circuit diagram of an LLC resonant con- Cr v̇c = ir
verter containing three resonant components Lr , Cr , and Lm .
Lm i̇m = vm (1)
The power MOSFETs of the full bridge are controlled in com-
plementary at duty 50%, ignoring the dead time and providing where ir , vc , and im are the inductor current, capacitor voltage,
a square input voltage. The full-wave rectifier, together with and magnetic current, respectively. vab and vm are the input and
the output capacitor, acts as a nonlinear load to provide the dc output square wave voltage.

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
FANG et al.: CONTROL OF AN LLC RESONANT CONVERTER USING LOAD FEEDBACK LINEARIZATION 889

Fig. 4. Equivalent circuit diagrams of dc subsystem.


Fig. 5. Nonlinear combined system model of an LLC resonant converter.

Using the FHA method, the state variables, input and output and analyzing the dc subsystem. The state equations of the dc
voltage can be approximated as sinusoidal states with switching subsystem are given as follows:
angle frequency ω s , as shown in the following equation. The
derived process of magnetic voltage vm is given in the Appendix C v̇o = iB r − vo /R (5)

ir = ir s sin ωs t + ir c cos ωs t where C is the output capacitor and R is the load.

vc = vcs sin ωs t + vcc cos ωs t C. Combined System Modeling of AC and DC


im = im s sin ωs t + im c cos ωs t Fig. 5 shows the combined system model of the LLC resonant
vab = (4vi /π) sin ωs t converter. The ac and dc subsystems are combined through the
transform and rectifier. iB r can be calculated as follows:
vm = [(ir s − im s ) /ip ] (4nvo /π) sin ωs t 
iB r = 2nip /π = 2n (ir s − im s )2 + (ir c − im c )2 /π. (6)
+ [(ir c − im c ) /ip ] (4nvo /π) cos ωs t (2)
where ir s , ir c , vcs , vcc , im s , im c , vm s , vm c are the sine and
Equations (3)–(6) give the state equations of the LLC resonant
cosine components of the inductor current, capacitor voltage,
converter.
and magnetic current and voltage. vi and vo are the input and
output dc voltage, respectively. ip is the primary rectifier cur-
III. SIMPLIFIED MODEL OF LLC RESONANT CONVERTER
rent. n is the transformer ratio. By making use of (2) from (1),
the equations of fundament components can be obtained. Since In Section II, the mathematical model of an LLC resonant
these equations are correct at every time, the coefficients of sine converter is built. But this model is still difficult to analyze
and cosine components must be equal, respectively. Then, the and design the controller due to the seven-order nonlinear state
state equation of the ac subsystem can be derived equations. As shown in (3)–(6), the complexity and nonlinearity
⎡ ⎤ ⎡−4 (i − i ) nv / (πL i ) − v /L + ω i ⎤
of the model are caused by the resonant network and rectifier
i̇r s rs ms o r p cs r s rc circuit in the LLC resonant converter. In order to simplify the
⎢ ⎥ ⎢ ⎢+4vi / (πLr ) ⎥
⎥ model, the influences of resonant network and rectifier must be
⎢ i̇r c ⎥ ⎢
⎢ ⎥ ⎢−4 (ir c − im c ) nvo / (πLr ip ) − vcc /Lr − ωs ir s ⎥ ⎥ reduced. As shown in Fig. 4, the resonant network and rectifier
⎢i̇ ⎥ ⎢ ⎥
⎢ ms⎥ ⎢ 4 (ir s − im s ) nvo / (πLm ip ) + ωs im c ⎥ can be equal as a controlled current source, the controlled current
⎢ ⎥= ⎢ ⎥
⎢i̇m c ⎥ ⎢ ⎥ ip can be calculated as follows:
⎢ ⎥ ⎢ 4 (ir c − im c ) nvo / (πLm ip ) − ωs im s ⎥
⎢ ⎥ ⎢ ⎥
⎣ v̇cs ⎦ ⎣ ir s /Cr + ωs vcc ⎦ ip 2 = (ir s − im s )2 + (ir c − im c )2 . (7)
v̇cc i /C − ω v
rc r s cs The state equation of ip can be obtained as

ip = (ir s − im s )2 + (ir c − im c )2 . (3) i̇p = (ir s − im s ) i̇r s − i̇m s /ip + (ir c − im c ) i̇r c − i̇m c /ip .
(8)
Equation (3) can be simplified as follows: Combine (3) with (8), the state equation can be derived as
follows:
ẋ = f (x, ωs )
i̇p = − 4 (1/Lr + 1/Lm ) nvo /π
ip = h (x)
x = [ir s , ir c , vcs , vcc , im s , im c ] . (4) 4 (ir s − im s ) vi /π−
+ / (Lr Lp ) . (9)
(ir s − im s ) vcs − (ir c − im c ) vcc
B. DC Subsystem Modeling
The resonant state variables ir s , im s , ir c , im c , vcs , and vcc
Fig. 4 shows the equivalent circuit of the dc subsystem. The in the right part of (9) are decided by (3), which make the
rectifier is equal to the current source, providing the energy for controlled current source very complex. As shown in (3), the
the output filter and load. iB r is the average rectifier output resonant network’s model has six poles and two double zeros.
current. Since the switching frequency is very high and the The dominant poles are the one beat frequency double poles. As
output capacitor is large, the average value is valid for modeling the switching frequency moves to resonant frequency, the beat

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
890 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 1, JANUARY 2018

Fig. 7. Simplified equivalent circuit diagrams of an LLC resonant.

output equal inductor Ls can be derived as follows:


Ls = π 2 /8n2 (1/Lr + 1/Lm ) . (11)

The equal rectifier voltage vn is controllable and decided by


the switching frequency, resonant network, and output voltage.
Fig. 6. Pole-zero placement of the LLC resonant converter small signal model. Since the dynamic of resonant network is much faster than the
output voltage and controller, the controllable rectifier voltage
vn can reached at the steady state during every control period.
frequency double poles will move to the lower frequency. When Since the switching frequency of LLC resonant converter is
the switching frequency is very close to resonant frequency, the near the resonant frequency, the resonant state variables can be
beat frequency will eventually split and becomes two real poles assumed as sinusoidal signals. The dc subsystem’s load can be
[40]. In the dc subsystem, there is only one pole ignoring the equal to the resister Req . Based on FHA method, the equivalent
ESR zero, as shown in (5). In the usual LLC resonant converter, resister is
the resonant network’s parameters are usually very small to
Req = 8n2 R/π 2 . (12)
achieve high resonant frequency. The output filter’s parameters
are usually very large to maintain the output voltage and cur- Then, the expressions for the steady-state secondary rectifier
rent. So the dynamic of resonant network is much faster than the voltage vn can be achieved based on FHA analysis
output voltage and controller. Therefore, the dynamics of the res-  
2 2 2 2
onant converters can be divided into the low-frequency behavior vn = vi / n (1 + h − h/f ) + Q (f − 1/f ) . (13)
and high-frequency behavior. The low-frequency dynamics are

determined by the output capacitor, resonant impedance, and where h = Lr /Lm , f = ωs /ωr , ωr = 1/ Lr Cr , Q =

load. The high-frequency dynamic is around the beat frequency, Lr /Cr /Req .
which is the difference value between the switching frequency Therefore, the steady-state value for the dc subsystem input
and the resonant frequency. The pole-zero placement of the LLC vn is calculated as the function of the input voltage vi , output
resonant converter of the small signal model is shown in Fig. 6. load R, and switching frequency fs . Then, the simplified model
As the switching frequency increases, the six high frequency of an LLC resonant converter can be achieved based on the equal
poles and four zeros move to the imaginary axis. As shown circuit in Fig. 7
in the figure, the high frequency poles and four zeros caused
by the resonant network are far away from the imaginary axis. Ls i̇B r = − vo + vn
Therefore, these high frequency poles and zeros can be ignored vo
C v̇o = iB r − . (14)
and the LLC resonant converter’s model can be simplified into R
a two-order model.
Assume the dynamic of resonant network is much faster IV. FEEDBACK LINEARIZATION CONTROL OF AN LLC
than the output voltage and controller. The resonant state RESONANT CONVERTER
variables can reach the steady state at the ending time of In Section III, the simplified model of an LLC resonant con-
each control period. The resonant state variables, such as verter is built. It is a simple two-order model. But it is still
ir c , ir s , im c , im s , vcc , vcs , in (9) are constant values during a nonlinear system due to the nonlinear function between the
the control period. Therefore, the last part of (9) can be equal to switching frequency and the rectifier voltage. Conventional lin-
a voltage source ear control methods such as PI controllers are difficult to be
applied. In order to eliminate the nonlinear characteristics, a
i̇p = 4 (1/Lr + 1/Lm ) n (vn − vo ) /π (10) linear feedback controller needs to be developed. As shown
in (13), the rectifier voltage vn is a nonlinear function of the
where vn is the equal secondary side rectifier voltage. The ac switching frequency. If we can formulate an inverse function
subsystem can be equal to an ideal controllable voltage source of the rectifier voltage, the nonlinear factors can be eliminated.
and an inductor. The simplified equivalent circuit of LLC reso- Then, the reconstructed system can be a linear system. The con-
nant converter is shown in Fig. 7. Combining (6) and (10), the ventional PI controller can be used to achieve the modulation

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
FANG et al.: CONTROL OF AN LLC RESONANT CONVERTER USING LOAD FEEDBACK LINEARIZATION 891

voltage. This paper proposes a nonlinear control strategy using 1) If Δ > 0, one root is real and two are complex conjugates
the load feedback linearization to achieve a great performance    
of the LLC resonant converter. 3 q q 2 p 3 3 q q2 p3 b
x1 = − + + + − − + −
First, we should formulate an inverse function of rectifier 2 4 27 2 4 27 3a
voltage. Define the input variable of the inverse function is the    
modulation voltage vr n . The output variable is the switching fre- 3 q q2 p3 3 q q2 p3
quency. The formulated function should be the inverse function x2 = − + + ω+ − − + ω2
2 4 27 2 4 27
of the rectifier voltage to make the rectifier voltage vn linearly
varying with the modulation voltage vr n . b

Since the input voltage and output load can be measured, (13) 3a
   
can be expressed as
3 q q2 p3 2 3 q q2 p3
vn = M (ωs ) . x3 = − + + ω + − − + ω
(15) 2 4 27 2 4 27
The switching frequency is calculated from the modulation b

voltage. Then, define the transfer function between the switching 3a
frequency and the modulation voltage as follows: √
−1 + 3i
ω= . (23)
ωs = g (vr n ) . (16) 2
In order to eliminate the nonlinearity, the functions need meet 2) If Δ = 0, all roots are real, and at least two are equal
the requirement of the following equation:  
q q b
x1 = − + 3 − −
3

vn = M (ωs ) = M (g (vr n )) = kvr n (17) 2 2 3a


 
where g(vr n ) is the inverse function of M (ωs ), and k is the q
x2 = x3 = − ω + 3 − ω 2 − .
3
q b
(24)
proportion coefficient. The value of k does not affect the lin- 2 2 3a
earization process. In this paper, we assume k = 1. Then, the 3) If Δ < 0, all roots are real and unequal
modulation voltage vr n needs to meet the equation as follows: 
  p θ b
x1 = 2 − cos −
vr n = vi / n (1 + h − h/f 2 )2 + Q2 (f − 1/f )2 . (18) 3 3 3a

p θ b
The equation can be extended as follows: x2 = 2 − cos( + 120◦ ) −
  3 3 3a
f 6 Q2 + f 4 (1 + h)2 − 2Q2 − (vi /nvr n )2 
p θ b
  x3 = 2 − cos( − 120◦ ) −
+ f 2 Q2 − 2h (1 + h) + h2 = 0. (19) 3 3 3a

−q −27p
Define θ = arccos . (25)
2p2
2
x=f
Finally, the switching frequency can be derived. Since the
a = Q2 solution is the square value of the switching frequency, all the
roots need to be positive. If the roots are negative or complex,
b = (1 + h)2 − 2Q2 − (vi /nvr n )2 no solution is existed to achieve the modulation voltage. Then,
c = Q2 − 2h (1 + h) the switching frequency is forced to be resonant frequency to
achieve the maximum output voltage. When the equation has
d = h2 . (20) solution, they may be two or one positive real roots. Generally,
the maximum positive real roots are chosen to ensure that the
Equation (19) can be simplified as follows:
ZVS can be achieved. Then, the switching frequency can be
ax3 + bx2 + cx + d = 0. (21) obtained

By using the Cardano formula, the equation can be solved. ωr Δ>0
ωs = g (R, vr n ) =  . (26)
First, calculate sign of the discriminants of the equation ωr max(x1 , x2 , x3 ) Δ ≤ 0
Δ = q 2 /4 + p3 /27 The ac subsystem is fed with the switching frequency. This
2 scheme uses the function of the output feedback to linearize
3ac − b
p= the resonant converter model. The function can be implemented
3a2
using the output voltage and current, as shown in Fig. 8.
2b3 − 9abc + 27a2 d By combining the nonlinear ac subsystem model, the linear
q= . (22)
27a3 dc subsystem model, and the nonlinear function, as shown in
Then, the solutions of the equation can be obtained. Fig. 8, the transfer function of the modulation voltage vr n and

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
892 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 1, JANUARY 2018

Fig. 9. Closed-loop control structure of an LLC resonant converter.

Fig. 8. Output feedback linearization control of an LLC resonant converter. TABLE I


PARAMETERS OF THE LLC RESONANT CONVERTER

output voltage vo is linear. The switching frequency ωs satisfies Parameter Symbol Value
(18). In order to build the transfer function from the modulation
voltage to the rectifier voltage, some changes of the equations Input voltage (V) vi 240
Output voltage (V) vo 24
are expressed. Define x̂ is the virtual state variables, which is Maximum output current (A) io 8
[îr s , îr c , îm s , îm c , v̂cs , v̂cc ]. The virtual state variables are the Resonant capacitor (nF) Cr 23.5 nF
variables of the inverse function ωs = g(vr n , R). They can sat- Resonant inductor (μH) Lr 86 μH
Magnetic inductor (μH) Lm 266.5 μH
isfy (18). Then, the inverse function can be expressed as follows. Transformer ratio n 10:1:1
Since the state variables are in a steady state, the gradients of Output capacitor (μF) C 3960
these variables are zero Switching frequency (kHz) f 0–300

0 = f (x̂, ωs )
 2  2 maximum resonant current. The output voltage is added to the
vr n = 2nR îr s − îm s + îr c − îm c /π. (27) output value of the controller for better dynamic performance.
Then, the inner loop transfer function can be derived
Define the state error between the virtual and real variables
as follows: Gi = kpi / (Ls s + kpi ) . (32)
Δx = x − x̂. (28) A PI controller is applied in the outer loop feedback to regulate
The state equation of error can be achieved from (4) and (27) the output voltage. The closed-loop transfer function of an LLC
resonant converter can be achieved
Δẋ = f (Δx, ωs ) . (29)
vo Gi (kpv s + kiv )
G= = 2
Since the dynamic of resonant network is much faster than the vref Cs + Gi (kpv s + kiv )
output voltage and the controller, the state error reaches at steady
kpi kpv s + kpi kiv
state during every control period. The transfer function from the = . (33)
modulation voltage to the rectifier voltage can be achieved as Ls Cs3 + kpi Cs2 + kpi kpv s + kpi kiv
follows: where kpv and kiv are the integral and proportional gain of the
out loop, respectively. Using the method of zero-poles assign-
vn = v r n . (30)
ment, the values of kpv and kiv can be achieved. Then, the
Combining (14) with (30), the linearized model of an LLC fluctuation transfer function can be achieved
resonant converter in Fig. 8 can be expressed as follows:
vo = Δvn s/ Ls Cs3 + kpi Cs2 + kpi kpv s + kpi kiv . (34)
C v̇o = iB r − vo /R
The steady value of the fluctuation transfer function is zero.
Ls i̇B r = vr n − vo. (31) Thus, the influence of parameter fluctuations can be eliminated.
Such a simplified two-order model cannot replace the derived
V. SIMULATION AND EXPERIMENTAL RESULTS
model, which provides detailed dynamics of an LLC resonant
converter. But it is useful for simplifying the controller design. In order to verify the effectiveness of the proposed model and
The LLC resonant converter has been linearized using the load control method, a simulation model of full-bridge LLC resonant
feedback linearization. Then, we can design a conventional PI converter has been built in MATLAB. The specifications of the
controller to achieve the modulation voltage. The LLC resonant parameters used for the simulation are given in Table I.
converter is designed as a voltage source. The output voltage In order to verify the accuracy of the simplified model, the
should be regulated. Then, the output voltage is used for the bode plots of the full order and reduced order models are given
feedback in the control scheme. The control structure is shown in Fig. 10.
in Fig. 9. In Fig. 10, the blue and red lines represent full order and
The rms value of resonant current is measured as the inner simplified model, respectively. It can be seen that in low and
loop feedback. Δvn is the disturbance between the modulation medium frequency range, two models have similar frequency
voltage and the rectifier voltage, which is caused by the mea- characteristics. The simplified model is accurate enough to guide
surement error of the resonant parameters. The proportional gain the controller design. However, in the high frequency range (near
kpi is designed to damp the oscillatory response and limit the the switching frequency), the simplified model shows its merits

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
FANG et al.: CONTROL OF AN LLC RESONANT CONVERTER USING LOAD FEEDBACK LINEARIZATION 893

Fig. 10. Bode plots of full order and reduced simplified model.

Fig. 12. Loop transfer function bode plot comparison of the traditional and
proposed control scheme. (a) At full load. (b) At half load.

Fig. 11. Bode plots of simulation mode, full order, and simplified model. proposed control scheme. In addition, comparisons between the
linearized and traditional control scheme, both of which have
the same PI parameters, are also implemented by simulation and
noticeably in which the simplified model excludes poles caused Fig. 12 shows the related frequency response results.
by the resonant components, which has no effects on the control The blue lines are the traditional control scheme and red
performance of an LLC resonant converter. lines are the proposed scheme. Fig. 12(a) shows the frequency
Fig. 11 shows the frequency response of the simulated LLC response plots at full load and Fig. 12(b) shows the plots at
resonant converter model, full-order model, and simplified half load. From both figures, it can be found that the proposed
model. The frequency response curves of the three models are control scheme has lower resonant peak value and larger phase
quite similar as they obtained similar poles and zeros. Moreover, margin. The proposed control scheme can effectively improve
the steady-state gains of theoretical modes (full-order model and the performance of the LLC resonant converter.
simplified model) are lower than that of simulated model due to A 200 W laboratory prototype of a full-bridge LLC resonant
the ignorance of harmonics. Fig. 11 proves that the simplified converter has also been built. The specifications of the param-
model can describe the frequency response of an LLC resonant eters used for the prototype are given in Table I. Fig. 13 shows
converter efficiency. the proposed control scheme of an LLC resonant converter. The
The feedback linearization control scheme and traditional linearization function is applied in this paper based on the typ-
control scheme are also applied in the simulation. Fig. 12 shows ical PI controller to linearize the converter model and improve
the frequency response of the traditional control scheme and the the performance.

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
894 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 1, JANUARY 2018

Fig. 13. Closed-loop output voltage control of an LLC converter using pro-
posed linearized control.

Fig. 14. Experimental setup of a 200 W LLC resonant converter.

Fig. 16. Experimental results of LLC resonant converter against load step
from no load to full load. (a) Case 1 with proposed strategy. (b) Case 2 with PI
controller. (c) Case 3 with PI controller in switching frequency.

load condition. Then, the burst control is applied to regulate


the output voltage under no load condition. Fig. 14 shows the
Fig. 15. Time signal waveform of interrupt function.
experimental setup of this LLC resonant converter.
In order to accelerate the computer speed, IQ math library
is used in the program code. A General-purpose input/output
Texas Instruments TMS320F28335 control card, which con- (GPIO) signal waveform is shown in Fig. 15 to measure the
tains a 32-b floating point DSP microcontroller, is used to imple- interruption time and computer time. The GPIO port is set at the
ment the control scheme, produces the pulse frequency modula- beginning of interrupt function and cleared at end. The computer
tion for the power MOSFETs. The sample and control frequency time of interruption function is only 6 μs. And the control period
is 10 kHz, which is lower than the switching frequency. The is 100 μs. The interruption percentage is only 6%. If we want
sample and calculation time can be ensured. Due to the influ- to maximize the bandwidth, the interrupt percentage usually
ence of the interturn distributed capacitor in the transformer, the should be smaller than 60%. Then, the minimum interrupt period
output voltage is higher than simulation results under the light is 10 μs. The maximum bandwidth we can get is 100 kHz.

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
FANG et al.: CONTROL OF AN LLC RESONANT CONVERTER USING LOAD FEEDBACK LINEARIZATION 895

Fig. 17. Experimental results of LLC resonant converter against load step Fig. 18. Experimental results of LLC resonant converter against load step
from half load to full load. (a) Case 1 with proposed strategy. (b) Case 2 with from full load to half load. (a) Case 1 with proposed strategy. (b) Case 2 with
PI controller. (c) Case 3 with PI controller in switching frequency. PI controller. (c) Case 3 with PI controller in switching frequency.

frequency to the expectable value. At case 1, the settling time


In order to verify the effectiveness of the proposed feedback is only 20.4 ms and voltage drop is 6 V, as shown in Fig. 16(a).
linearized control strategy, the experiments of the proposed strat- Due to the complex computing, the control frequency in case 1
egy with control frequency of 10 kHz (Case 1), the conventional is only 10 kHz to ensure the enough computing time in every
PI controller with control frequency of 10 kHz (Case 2), and the control period. At case 2, the setting time is 38.8 ms and the
PI controller sampling at every switching period (Case 3) are voltage drop is 6 V. The conventional PI controller’s control
expressed. Fig 16 shows the experimental results against the frequency is still 10 kHz. Then, the output voltage can reach at
step load from no load to full load in the three cases. Due to the steady state at the ending time of every control period. At case
influence of junction and winding capacitor, the output voltage 3, the setting time is 41.2 ms and voltage drop is 6.4 V. At this
is higher than the reference voltage even at maximum frequency. case, the switching frequency is regulated at every switching
Then, the converter operates at burst mode to maintain the ref- period. Since the output voltage may not reach steady state at
erence voltage under the light load condition. When the loads every switching period, some oscillations occur. The dynamic
step up, the switching frequency is regulated from the maximum response is worse than case 2.

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
896 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 1, JANUARY 2018

TABLE II
DYNAMIC RESPONSE COMPARE WITH THREE CASES UNDER
DIFFERENT CONDITIONS

Setting time Case 1 Case 2 Case 3

Load step from no to full 20.4 ms 38.8 ms 41.2 ms


Load step from half to full 29 μs 44 μs 22.4 μs
Load step from full to half 28 μs 40 μs 19.2 μs
Load step from full to no 2.24 s 3.16 s 2.64 s

Fig. 20. Waveforms of ripple current and voltage.

high control frequency near switching frequency can make the


converter unstable.
Figs. 16 and 17 shows the experimental results of the LLC
resonant converter against load step up. The results of converter
against load step down are shown in Figs. 18 and 19. The dy-
namic response of the LLC resonant converter from full load to
half is shown in Fig. 18. At case 1, the setting time is 28 μs,
similar with load step up. At case 2, the setting time is 40 μs, as
shown in Fig. 18(b). At case 3, the setting time is 19.2 ms due
to the oscillation.
When the load step down from full to no load, output voltage
will have a large overshoot. The converter operates at burst
mode at beginning. Due to the higher voltage, all the switches
stop working. The output voltage reduces to the reference value
Fig. 19. Experimental results of LLC resonant converter against load step
using the dead load. The setting time is very long, as shown
from full load to no load. (a) Case 1 with proposed strategy. (b) Case 2 with PI in Fig. 19. At case 1, the setting time is 2.24 s and voltage
controller. (c) Case 3 with PI controller in switching frequency. overshoot is 4 V. At case 2, the setting time is 3.16 s and the
overshoot is 4 V. The proposed strategy has a better dynamic
Fig 17 shows the experimental results against step load from response. At case 3, the setting time is 2.64 s and the overshoot
half load to full load in the three cases. At half load condition, is 4.4 V. The oscillation is eliminated since the converter works
the switching frequency is near the resonant frequency to regu- at the burst mode. The dynamic performance is improved due
late the output voltage. The influence of junction and winding to the high control frequency.
capacitor is eliminated. The dynamic response is faster than the Table II gives the setting time of the three cases under different
condition that load step up from no load to full load. At case 1, conditions. The LLC resonant converter works at the burst mode
the settling time is only 29 μs and the voltage drop is very small, with maximum switching frequency under on load condition.
near zero, as shown in Fig. 18(a). At case 2, the setting time is And the converter works near the resonant frequency at half
44 μs. The output voltage nearly does not change. At case 3, the load. Therefore, the setting time against load step up from no
setting time is 22.5 ms. The voltage drop is 2.4 V and the voltage to full load is longer than the setting time from half to full
overshoot is 1.2 V. Since the output voltage may not reach at load. At case 3, the higher control frequency, which is equal
steady state at every switching period, the oscillations worsen with the switching frequency, causes some oscillations during
the dynamic characteristics of an LLC resonant converter. The the loads’ variation period. The dynamic response of case 3 is

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
FANG et al.: CONTROL OF AN LLC RESONANT CONVERTER USING LOAD FEEDBACK LINEARIZATION 897

worse than the case 2. Since the nonlinear characteristics are The phase difference is decided by the difference between
eliminated at case 1, the proposed control strategy has a best resonant and magnetic current
dynamic performance.
ir − im = ip sin (ωs t + ϕ)
Fig. 20 shows the waveforms of ripple current and ripple
voltage of the output capacitor. The yellow line is the output = (ir s − im s ) sin ωs t + (ir c − im c ) cos ωs t. (37)
voltage and it has been regulated to 24 V. The blue line is the
Then, we have
ripple voltage and due to large output capacitor value, the peak-
peak value of ripple voltage is only 150 mV. Green line is the cos ϕ = (ir c − im c ) /ip ; sin ϕ = (ir s − im s ) /ip . (38)
rectifier diode current and its peak-peak value is about 14 A.
Therefore, the magnetic voltage can be derived as follows:
Thus, large output capacitors are needed to absorb the large
ripple currents. vm = [4nvo (ir s − im s ) / (πip )] sin ωs t
+ [4nvo (ir c − im c ) / (πip )] cos ωs t. (39)
VI. CONCLUSION
This paper proposes a load feedback linearized control strat- REFERENCES
egy of an LLC resonant converter to improve the performance. [1] G. Ivensky, S. Bronshtein, and A. Abramovitz, “Approximate analysis of
The nonlinear characteristics of the LLC resonant converter resonant LLC DC-DC converter,” IEEE Trans. Power Electron., vol. 26,
caused by the variable frequency control are eliminated. The no. 11, pp. 3274–3284, Nov. 2011.
[2] X. Mathu, C. O. Na, W. Ningning, S. Kulkarni, and S. Roy, “Review of
dynamic response is improved. The settling time of the load integrated magnetics for power supply on chip (PwrSoC),” IEEE Trans.
step up from no to full load is reduced from 38.8 to 20.4 ms Power Electron., vol. 27, no. 11, pp. 4799–4816, Nov. 2012.
compared with the conventional PI controller. The main contri- [3] K. Bong-Chul, P. Ki-Bum, K. Chong-Eun, L. Byoung-Hee, and M.
Gun-Woo, “LLC resonant converter with adaptive link-voltage variation
butions of this paper are as follows. for a high-power-density adapter,” IEEE Trans. Power Electron., vol. 25,
1) Propose a load feedback linearization control strategy, no. 9, pp. 2248–2252, Sep. 2010.
which can effectively improve the dynamic performance [4] L. Huber, G. Liu, and M. M. Jovanovic, “Design-oriented analysis and
performance evaluation of buck PFC front end,” IEEE Trans. Power Elec-
of the LLC resonant converter. Using the loads feedback, tron., vol. 25, no. 1, pp. 85–94, Jan. 2010.
the nonlinear characteristics are eliminated. The output [5] J. Cao and A. Emadi, “A new battery/ultra capacitor hybrid energy storage
characteristics can be regulated as expected. system for electric, hybrid, and plug-in hybrid electric vehicles,” IEEE
Trans. Power Electron., vol. 27, no. 1, pp. 122–132, Jan. 2012.
2) Build a simple two-order model of an LLC resonant con- [6] L. Wuhua and H. Xiangning, “Review of nonisolated high-step-up DC/DC
verter. Although ignoring the dynamic influence of the converters in photovoltaic grid-connected applications,” IEEE Trans. Ind.
resonant network, the complex seven-order model of con- Electron., vol. 58, no. 4, pp. 1239–1250, Apr. 2011.
[7] Q. Zhijun, O. Abdel-Rahman, H. Al-Atrash, and I. Batarseh, “Model-
verter based on FHA is simplified as a two-order model. ing and control of three-port DC/DC converter interface for satellite ap-
The LLC resonant converter can be equal as a controlled plications,” IEEE Trans. Power Electron., vol. 25, no. 3, pp. 637–649,
voltage source with a LC filter. This model is helpful to Mar. 2010.
[8] A. Kwasinski and C. N. Onwuchekwa, “Dynamic behavior and stabiliza-
guide the design of the resonant converter’s controller. tion of DC microgrids with instantaneous constant-power loads,” IEEE
Experimental results verify the effectiveness of the proposed Trans. Power Electron., vol. 26, no. 3, pp. 822–834, Mar. 2011.
control strategy. And the proposed controller and modeling [9] M. Saeedifard and R. Iravani, “Dynamic performance of a modular mul-
tilevel back-to-back HVDC system,” IEEE Trans. Power Del., vol. 25,
method can also be used for other resonant converters. no. 4, pp. 2903–2912, Apr. 2010.
[10] L. Xiaodong and A. K. S. Bhat, “Analysis and design of high-frequency
isolated dual-bridge series resonant DC/DC converter,” IEEE Trans. Power
APPENDIX Electron., vol. 25, no. 4, pp. 850–862, Apr. 2010.
[11] Q. Ying, C. Xiyou, Z. Chongquan, and Q. Chen, “Uniform models of
vm is the magnetic voltage of the LLC resonant converter. PWM DC-DC converters for discontinuous conduction mode considering
When the resonant current is larger than the magnetic current, parasitics,” IEEE Trans. Ind. Electron., vol. 61, no. 11, pp. 6071–6080,
the magnetic voltage is equal to the output voltage, provided Nov. 2014.
[12] B. Yang, F. C. Lee, A. J. Zhang, and H. Guisong, “LLC resonant converter
that the turns ratio is 1; otherwise, the negative value of output for front end DC/DC conversion,” in Proc. IEEE Appl. Power Electron.
voltage instead is Conf. Expo., Mar. 2002, vol. 2, pp. 1108–1112.
[13] L. Bing, L. Wenduo, L. Yan, F. C. Lee, and J. D. Van Wyk, “Optimal
vm = nvo · sign (ir − im ) design methodology for LLC resonant converter,” in Proc. IEEE Appl.
 Power Electron. Conf. Expo., Mar. 2006, pp. 533–538.
1 ir > im [14] M. P. Foster, C. R. Gould, A. J. Gilbert, D. A. Stone, and C. M. Bingham,
sign (ir − im ) = . (35) “Analysis of CLL voltage-output resonant converters using describing
−1 ir < im functions,” IEEE Trans. Power Electron., vol. 23, no. 4, pp. 1772–1781,
Apr. 2008.
Since the resonant and magnetic current are both sinusoidal [15] F. Zhijian, C. Tao, D. Shanxu, and C. Changsong, “Optimal design method-
ology for LLC resonant converter in battery charging applications based on
signals, the magnetic voltage is a square wave signal, which is time-weighted average efficiency,” IEEE Trans. Power Electron., vol. 30,
half positive and half negative values and the waveform is similar no. 10, pp. 5469–5483, Oct. 2015.
with that of input voltage vab but with a difference in amplitude [16] G. C. Verghese, M. E. Elbuluk, and J. G. Kassakian, “A general approach to
sampled-data modeling for power electronic circuits,” IEEE Trans. Power
and phase. Then, the fundament component of magnetic voltage Electron., vol. PE-1, no. 2, pp. 76–89, Feb. 1986.
can be written as follows: [17] B. Yang, F. C. Lee, and M. Jovanovic, “Small-signal analysis for
LLC resonant converter,” Center Power Electron. Syst. Semin., 2003,
vm = (4nvo /π) sin (ωs t + ϕ) . (36) pp. 144–149.

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.
898 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 1, JANUARY 2018

[18] C. Chien-Hsuan, C. En-Chih, C. Chun-An, C. Hung-Liang, and L. Sheng- Zhijian Fang received the B.S. and Ph.D. degrees
Chang, “Small signal modeling of LLC resonant converters based on in electrical engineering and automation from
extended describing function,” in Proc. Int. Symp. Comput., Consum. Huazhong University of Science and Technology,
Control, Jun. 2012, pp. 365–368. Wuhan, China, in 2010 and 2015, respectively.
[19] J. Stahl, H. Steuer, and T. Duerbaum, “Discrete modeling of resonant Since 2015, he has been a Postdoctoral
converters steady state and small signal description,” in Proc. IEEE Energy Research Fellow in the School of Electrical Engineer-
Convers. Congr. Expo., Sep. 2012, pp. 1578–1584. ing, Wuhan University, Wuhan, China. From Febru-
[20] B. Cheng, F. Musavi, and W. G. Dunford, “Novel small signal modeling ary 2016 to January 2017, he was a Postdoctoral
and control of an LLC resonant converter,” in Proc. IEEE Appl. Power Research Fellow with Department of Electrical and
Electron. Conf. Expo., Mar. 2014, pp. 2828–2834. Computer Engineering, Ryerson University, Toronto,
[21] J. Jinhaeng, J. Minjae, C. Seokjae, C. Youngho, and C. Byungcho, “Current ON, Canada. His research interests include high per-
mode control for LLC series resonant dc-to-dc converters,” in Proc. IEEE formance dc/dc converter, battery charger, and renewable energy applications.
Appl. Power Electron. Conf. Expo., Mar. 2011, pp. 21–27.
[22] J. Jinhaeng, P. S. Kumar, K. Dongyun, and C. Byungcho, “Average current-
mode control for LLC series resonant dc-to-dc converters,” in Proc. Power
Electron. Motion Control Conf., Jun. 2012, vol. 2, pp. 923–930. Junhua Wang was born in Shandong, China, in 1981.
[23] Z. Sheng, L. Haoze, L. Wuhua, H. Xiangning, and X. Changliang, “The- He received the Ph.D. degree in Electrical Engineer-
oretical evaluation of stability improvement brought by resonant current ing from Hong Kong Polytechnic University, Hong
loop for paralleled LLC converters,” IEEE Trans. Ind. Electron., vol. 62, Kong, China, in 2012.
no. 7, pp. 4170–4180, Jul. 2015. He joined the Carnegie Mellon University as a
[24] X. Fan, Z. Bo, and Q. Dongyuan, “Stabilizing the nonlinear dynamic Postdoctoral Researcher in 2012 and then worked
behavior of LLC resonance full-bridge DC-DC converter under volt- as a Research Fellow in the GATE Center for Elec-
age mode control,” in Proc. Electron. Appl. Conf. Expo., Nov. 2014, tric Drive Transportation, Dearborn, MI, USA. He
pp. 1093–1097. is currently a Professor in the School of Electrical
[25] O. Dranga, B. Buti, and I. Nagy, “Stability analysis of a feedback- Engineering, Wuhan University, Wuhan, China. His
controlled resonant DC-DC converter,” IEEE Trans. Ind. Electron., vol. 50, research interests include wireless transmission tech-
no. 1, pp. 141–152, Jan. 2003. nology based on magnetic resonance, applied electromagnetics, system equip-
[26] K. Mandal, S. Banerjee, and C. Chakraborty, “Determination of stable ment for power transmission and distribution.
region of controller parameters for series-parallel resonant converter with
capacitive output filter,” in Proc. IEEE Int. Symp. Ind. Electron., Jun. 2011,
pp. 229–232.
[27] J. Jinhaeng, J. Minjae, C. Byungcho, and K. Heung-Geun, “Dynamic
analysis and control design of optocoupler-isolated LLC series resonant Shanxu Duan received the B.S., M.S., and Ph.D. de-
converters with wide input and load variations,” in Proc. IEEE Energy grees in electrical engineering from Huazhong Uni-
Convers. Congr. Expo., Sep. 2009, pp. 758–765. versity of Science and Technology, Wuhan, China, in
[28] F. Kurokawa and K. Murata, “A new quick response digital modified P-I- 1991, 1994, and 1999, respectively.
D control LLC resonant converter for DC power supply system,” in Proc. Since 1991, he has been a Faculty Member in
IEEE Power Electron. Drive Syst., Dec. 2011, pp. 35–39. the School of Electrical and Electronics Engineer-
[29] L. Chun-Liang, C. Yi-Hsun, L. Yi-Feng, W. Shun-Chung, and L. Yi-hua, ing, Huazhong University of Science and Technol-
“Design and implementation of a digitally-controlled LLC resonant con- ogy, Wuhan, China, where he is currently a Professor.
verter for battery charging applications,” in Proc. IEEE Power Electron. His research interests include stabilization, nonlinear
Drive Syst., Apr. 2013, pp. 804–808. control with application to power electronic circuits
[30] M. I. Shahzad, S. Iqbal, and S. Taib, “LLC series resonant converter with and systems, fully digitalized control techniques for
PID controller for battery charging application,” in Proc. IEEE Energy power electronics apparatus and systems, and optimal control theory and cor-
Convers., Oct. 2014, pp. 84–89. responding application techniques for high frequency pulse width modulation
[31] C. Buccella, C. Cecati, H. Latafat, P. Pepe, and K. Razi, “Observer-based power converters.
control of LLC DC/DC resonant converter using extended describing Dr. Duan is a Senior Member of the Chinese Society of Electrical Engi-
functions,” IEEE Trans. Power Electron., vol. 30, no. 10, pp. 5881–5891, neering and a council member of the Chinese Power Electronics Society. He
Oct. 2015. was selected as one of the New Century Excellent Talents by the Ministry of
[32] M. Hao, L. Qinwei, and G. Jin, “A sliding-mode control scheme for LLC Education of China in 2007. He received the honor of “Delta Scholar” in May
resonant DC/DC converter with fast transient response,” in Proc. IEEE 2009.
Energy Convers., Oct. 2012, pp. 162–167.
[33] H. Zhiyuan, W. Laili, L. Yan-Fei, and P. C. Sen, “Bang-Bang charge
control for LLC resonant converters,” in Proc. IEEE Energy Convers.
Congr. Expo., Sep. 2013, pp. 140–146. Kaipei Liu was born in Hubei, China, in 1962. He
[34] T. Nishimura et al., “Robust digital control for an LLC current-resonant received the Ph.D. degree in computer application
DC-DC converter,” in Proc. Elect. Eng./Electron., Comput., Telecommun. technology from Wuhan University, Wuhan, China,
Inf. Technol., May 2012, pp.1–4. in 2001.
[35] C. L. Chia and E. K. K. Sng, “A novel robust control method for the He is currently a Professor in the School of Elec-
series–parallel resonant converter,” IEEE Trans. Power Electron., vol. 24, trical Engineering, Wuhan University, where he has
no. 8, pp. 1896–1904, Aug. 2009. been researching power electronics technology.
[36] F. Weiyi, F. C. Lee, and P. Mattavelli, “Optimal trajectory control of burst
mode for LLC resonant converter,” IEEE Trans. Power Electron., vol. 28,
no. 1, pp. 457–466, Jan. 2013.
[37] F. Weiyi, F. C. Lee, and P. Mattavelli, “Simplified optimal trajectory con-
trol (SOTC) for LLC resonant converters,” IEEE Trans. Power Electron.,
vol. 28, no. 5, pp. 2415–2426, May 2013.
[38] F. Weiyi and F. C. Lee, “Optimal trajectory control of LLC resonant Tao Cai was born in Hubei Province, China, in 1974.
converters for soft start-up,” IEEE Trans. Power Electron., vol. 29, no. 3, He received the Ph.D. degree in control science and
pp. 1461–1468, Mar. 2014. engineering from Huazhong University of Science
[39] F. Weiyi, F. C. Lee, and P. Mattavelli, “Optimal trajectory control of and Technology, Wuhan, China, in 2004.
LLC resonant converters for LED PWM dimming,” IEEE Trans. Power His research interests include advanced signal pro-
Electron., vol. 29, no. 2, pp. 979–987, Feb. 2014. cessing and energy management of renewable power
[40] B. Yang, “Topology investigation of front end DC/DC converter for dis- generation. He has authored about 20 technical pa-
tributed power system,” Ph.D. dissertation, , Virginia Tech, Blacksburg, pers in journal and conferences.
VA, USA, Sep. 2003.

Authorized licensed use limited to: National Taiwan Univ of Science and Technology. Downloaded on September 24,2024 at 01:31:57 UTC from IEEE Xplore. Restrictions apply.

You might also like