Parallel Load Sharing Rev0
Parallel Load Sharing Rev0
Loadsharing for
Compressor Networks
Compressor Networks
individual machines
• A “network view” of the application is
essential to achieve good surge protection
and good performance control of the
network
Equidistant Loadsharing
2011 Compressor Controls Corporation
Dev1 = Dev2
Q1 = Q2
2013 Compressor Controls Corporation
N1 = N2
2 2
qr,1 qr,2
DEV1 DEV2
• Machines operate at same Rc
Notes:
• The DEV is a dimensionless number representing the relative
distancesystem
• Maximum between the operating
turndown withoutpoint and or
recycle theblow-off
SCL
• Machines are kept at the same DEV which may correspond to
• Minimizes
different the risk of
speeds andsurge
flow since
rates all machines’ (recycle valves)
• absorb partwill
Recycle of only
the disturbance
start when all machines are on their SCL
• • Automatically
Since DEV isadapts to different
dimensionless, alltypes
sortsof
ofcompressors
machines can be
• CCC mixed: small,algorithm
patented big, centrifugals, turbine & motor drives
Dynamic simulation:
Variable speed drive - parallel compressors
VSDS
RSP
Compressor 1
out
UIC Serial LSIC Serial
1 network 1 network
Load
2013 Compressor Controls Corporation
MPIC HIC
1 1
VSDS
RSP
Suction
Compressor 2 Process
out
UIC Serial LSIC
2 network 2
MPIC
(P/F)
1
Process Pd
Process Ps PV
L/S data flow
Process Mass Flow PID
SP
DEV avg
OUT
LSIC 1 LSIC 2
(P/F) (P/F)
4 4
PV PV
SP
SP
5
2
OUT OUT
2013 Compressor Controls Corporation
3 To Speed 3
To Speed
Governor 1 Governor 2
UIC 1 UIC 2
(A/S) (A/S)
DEV DEV
6
OUT OUT
To A/S 1 To A/S 2
valve valve
How does L/S work?
1. Master is reading/controlling the Process Variable.
2. Based on the deviation (error) between the process PV and the
process SP, the Master initiates a PID response to the P/F
controllers to increase/decrease speed.
3. The A/S controllers continuously calculate the DEV and send it
to the P/F controllers as their PV. Suppose DEV1=0.3 &
DEV2=0.2.
4. The P/F controllers send the DEV to the Master which calculates
the average deviation. (Avg DEV= 0.25 in our case).
5. The average deviation is sent back to the P/F controllers as a
new SP. The P/F controllers compare the new SP (new DEV)
2013 Compressor Controls Corporation
with their PV (old DEV) and adjust their DEV slowly. If the SP is
lower than PV, the P/F controller lowers the speed (at the speed
governor output) since it is running at a higher flow rate [here
we need to lower the DEV move to the left less speed]. In
our case, P/F1 (having SP=0.25 and PV=0.3) will decrease its
speed and P/F2 (having SP=0.25 and PV=0.2) will increase its
speed to run on higher DEV. This process is continuous.
• If disturbance happens and we have DEV < or = 0, Master will
help the A/S controllers by sending its PID output directly to
the A/S controllers to increase its output (whichever needs
help). So both PI from A/S and PID from Master help in
regulating the valve throughput.
• POC can be enabled in Master to improve performance.
Parallel Load-sharing
out
UIC Serial UIC Serial LSIC Serial
1A network 2A network A network
Suction
MPIC
Header VSDS RSP 1
Train B
2013 Compressor Controls Corporation
Section 1 Section 2
out Process
UIC Serial UIC Serial LSIC
1B network 1B network B
• Select per train -- in the loadsharing controller (Master) - the section closest to the SCL (higher S
(lower DEV)). From UIC1A & UIC2A, minimum DEV is selected per train and sent back via LSIC-A to
Master. From UIC1B & UIC2B, minimum DEV is selected and sent back via LSIC-B to Master. Master
calculates new avg DEV for the network.
• By selecting the section closest to the SCL it is guaranteed that the other section on the same train
is not in recycle
The Primary Response
• Master controller (MPIC) controls
the main Process Variable (PV) via
its PID control block
• The output of the MPIC PID goes to
the primary response block in the
loadsharing controller
MPIC
PV PID
1
• In the primary response block, the
controller checks if the machine is
close to the SCL:
– Yes: don’t reduce compressor
2013 Compressor Controls Corporation
throughput
– No: reduce compressor
throughput as necessary
• Apply loadsharing gain M0
• The primary loop gain of MPIC goes
directly to the performance control
element via the primary response Primary
block Response
To performance
control element
The Primary Response
• In order to check if the machine is
close to the SCL, the LSIC primary
response block needs the A/S DEV
•
•
The DEV is reported by the UIC
When the machine is close to the
MPIC
SCL, MPIC output changes will be PV
•
redirected from the LSIC to the UIC
to help in opening the A/S valve
When DEV < or = 0, further
1 PID
To antisurge
valve
Loadsharing Response
DEV
Primary
Response
Loadsharing Response
DEV
Primary
Response
To performance
control element +
Pressure Override Control (POC)
• Performance control loop may be too slow
Benefits:
to react to a large disturbance
• The pressure excursion may be dangerously PID
• large
Fastandresponse
threatenduring fast/ machinery
a process upsets trip PV
PV
• • Pressure
Avoid process
Overridetrips due(POC)
Control to lack
canofbe PI
response
used to openintheperformance control
antisurge valve when a (One-Sided)
SP
elementslimit is exceeded
configured
• • When
Allowsthe closer
operating point drops
operation below the
to process
POC setpoint,
limits withoutnormal
takinga/srisks
control is resumed
• MPIC will ultimately stabilize the operating
2011 Compressor Controls Corporation
POC-SP
Antisurge
PIC-SP Response
2010
To antisurge
2
valve
qr
Loadsharing Overview
(Simplified)
PV
POC PV
SP PID
PI
(One Sided)
POC SP
DEV
fA Mode
PV SP
+ +
Loop Loop
Decoupling Decoupling
LSIC
To antisurge valve
To performance
control element