AGC Control Predictivo China

8/13/2019 AGC Control Predictivo China

1/5

Automatic Generation Controller Design in Deregulated and NetworkedEnvironment Using Predictive Control Strategy

J. H. Zhang, J. H. Hao, G. L. Hou

Department of Automation, North China Electric Power University, Beijing, 102206, China(e-mail: [email protected]).

Abstract: In this paper, Generalized Predictive Control (GPC) algorithm is applied to design automaticgeneration control (AGC) systems in deregulated and networked environment. The proposed AGCapproach can be used to deal with the effects caused by power market and communication networks.Finally, the developed scheme is implemented in a two-area AGC system, and the simulation resultsshow the effectiveness of the proposed scheme.

1. INTRODUCTION

Traditional automatic generation control (AGC) systems arefacing new challenges in deregulated and networkedenvironment. The traditional AGC systems have to bereformulated due to the effects caused by the open market for

price based operation and the communication network.

AGC systems after deregulation have been investigated basedon optimization or robust control theory. The concept ofDISCO participation matrix (DPM) and area participationfactor (APF) were introduced to simulate bilateral contract,

moreover, trajectory sensitivities were used to obtain optimal parameters of AGC systems using gradient Newton algorithm(Donde et al. , 2001). The genetic algorithm was used tooptimize integral gains and bias factors (Demiroren et al.,2006). AGC of a hydro-thermal system with generation rateconstraint (GRC) after deregulation is investigated. In

particular, the sensitivity of the optimal controller gains toDPM and APF was given in Paridal and Nandal (2005).Robust control algorithms, such as m -synthesis, H and

mixed2 / H H , were utilized in AGC in a restructured systems

(Bevrani 2003; Feliachi 1997, Bevrani 2004). Multi-stagefuzzy PID controller was designed for AGC systems( Shayeghi , 2006).

Recently, modelling and synthesis of conventional AGCsystems in communication networked environment also have

been published in (Nobile 2000; He 2005; Bevrani 2005).Communication networks inevitably introduced time delaysto AGC systems. There are two time delays induced by twomain communication links, which includes the delay betweenRTU and the control centre, and the delay between thecontrol centre to the individual units. In Nobile (2000), theeffect of signal delay uncertainties in open communicationnetworks on LFC was investigated by constant, random andexponential delays, respectively. It was also pointed out thatthe induced delays might cause deterioration of the AGCsystem performance. Considering the delay between the

power plant automation system and the governor in He

(2005), a robust H controller was designed for conventionalAGC systems and solved by linear matrix inequalities. InBevrani (2005), a PI-based LFC with two kinds ofcommunication delays was designed via a mixed

2 / H H

control techniques and an iterative LMI algorithm.

At present, little research has been carried out on designingcontroller for AGC systems in deregulated and networkedenvironment. In Bhowmik et al. (2004), the effect of signaldelays on load following is summarized, communicationmodels for third parity LFC was investigated by queuingtheory. In Yu and Tomsovic (2004) , the induced time delays

in deregulated and networked AGC systems were simplified.Assumed that the control centre waits to receive the tele-metered signals from remote terminal unit (RTU), and it wasfurther assumed that individual units may communicatedirectly bypassing the control center in case of bilateralcontracts, therefore, two delays were aggregated into a singledelay from control centre. The robust controller of AGCsystems with one communication single delay was presented

based on LMI.

Robust control algorithm can guarantee the stability of AGCsystems in deregulated and networked environment as long astime delays are bounded. However, how to compensate for

the time delay and data dropout has not been considered,moreover, it simply treat networked AGC system as a systemwith time delays, which ignore some features, e.g., randomdelay and data transmission in packet. As such, following theapproaches presented in (Tang 2006; Liu 2005), predictivecontrol algorithm is applied to design controller for AGCsystems in deregulated and networked environment in orderto compensate time delay and take advantage of somefeatures caused by communication transmission. In fact,

predictive control strategy has been utilized in conventionalLFC in (Rerkpreedapong 2003; Atic 2003), because it candeal with the generating rate constraint and incorporateeconomic objectives as a part of control requirements.

The main objective of this work is to investigate predictivecontrol algorithm for AGC systems in deregulated and

Proceedings of the 17th World CongressThe International Federation of Automatic ControlSeoul, Korea, July 6-11, 2008

978-1-1234-7890-2/08/$20.00 2008 IFAC 9410 10.3182/20080706-5-KR-1001.0601


2/5

networked environment. The rest of this paper is organized asfollows. Section 2 designs the predictive controller fornetworked AGC systems after deregulation. Section 3 showsthe application of the proposed controller, and Section 4concludes this paper.

2. DESIGN OF CONTROLLERTake a two-area AGC system in deregulated and networkedenvironment as an example, whose block diagram is shownas Fig. 1.

1 B

ACE1

controller

apf1

apf2

12a

apf3

apf4

ACE2

2 B

PULoad ofDISCO1

PULoad ofDISCO2

PULoad ofDISCO4PULoadof

DISCO3

1c P D

2c P D

11/ R

21 / R

1 f D

2 f D

41/ R31/ R

1, L LOC P D

12 /T s

3

1

1 G sT + 3

1

1 T sT +

2, L LOC P D

4

1

1G

sT + 41

1 T sT +

2

21 P

P

K sT +

12 p

12a

1 2,t ie a ct ua l P -D

Demand ofDISCOs in I t o

GENCOs in II

Demand ofDISCOs in II toGENCOs in I

2

21

P

P

K

sT +

1

2 p

1

1

1 G sT + 1

11 T sT +

2

11 G sT + 2

11 T sT +

1 2,t ie er ro r P -D

Networkcommunication

delay

controller Network

communicationdelay

+

+

-

+

+-

+

+ +

-+

+

+-

++

-++

+

-

++

+

-

-

-+

+

+

+

+ +

+ +

++

+ +

++

+ +

+ +

+ +

11cpf

21cpf

31cpf

41cpf

12cpf

22cpf

32cpf

42cp f

13cpf

23cpf

33cpf

43cpf

14cpf

24cpf

34cp f

44cpf

+Governor Turbine

Powe r System

1wD

2wD

buffer

buffer

Fig. 1 Two-area AGC system in deregulated and networkedenvironment

In deregulated environment, GENCOs sell power to variousDISCOs at competitive prices. A DISCOs may makecontracts with any available GENCOs in their own or otherareas. Thus, it leads to various combinations of the possiblecontracted scenarios between DISCOs and GENCOs. Therestructured AGC system is modeled based on DISCO

participation matrix (DPM) presented in (Donde et al. , 2001),the element of DPM named cpf (contract participation factor)represents the participation of a DISCO in a contract with aGENCO. Likewise, ACE participation factors are employedto the distribution relation between ACE signal and eachGENCO.

The time delays induced by communication channels in AGCsystems can be classified into forward-link delay andfeedback-link delay. The forward-link delay is between thecontroller and the governor due to transmitting control law tothe individual units; the feedback-link delay is between RTUand the control centre due to transmitting tele-meteredfrequency and power tie-line flow signals. In this work, thefeedback-link delay is ignored as shown in Fig. 1. In fact, it isapproximately 80-200 milliseconds if the tele-metered signalsare transmitted via dedicated channel in Synchronous DigitalHierarchy (SDH) or IP switch mode. Or, the feedback-linkand forward-link delays can be regarded as a single delay

under some reasonable assumptions (Yu and Tomsovic,2004).

The following assumptions can reasonably be made whenstudying networked AGC systems:

A1) The random delay t induced by network is bounded.A2) The number of consecutive data dropouts is also

bounded.

A3) The data are transmitted through network with a time-stamp so that the delay information can be extracted at thegovernor node.A4) The bandwidth of the communication network is notlimited.For a two-area AGC system, it can be modelled as thefollowing state space equations

1 1 2( ) ( ) ( ) ( )

( ) ( )

X t A X t B u t B V t

y t CX t

= + + =

&

(1)

where1 2 3 4 1 2 3 41 2

[ ]T tie T T T T V V V V w w P P P P P P P P P X D D D D D D D D D D D= represents the state of the system; 1 2[ ]

T c c P P u D D= is the

manipulated variables or command inputs prior tocommunication channels. 1 2 3 4[ ]

T L L L L P P P P V D D D D= is the

vector of power demands of the DISCOs. 1 A , 1 B and 2 B are appropriate dimension matrixes. The parameters of the

power system are given in Appendix A.

The above AGC model can be transformed to the followingCARIMA model

1 1 1 1( ) ( ) ( ) ( 1) ( ) ( ) ( ) ( )

1 A z y t B z u t D z v t C z e t

- - - -= - + +

D (2)where ( ) nt R , ( 1) mu t R- are the output and controlsequence of AGC system. ( ) l v t R is a vector of measureddisturbance and ( )e t is a zero mean white noise, the operatorD is defined as 11 -D = - , 1( ) A z - and 1( )C z - are n n monic polynomial matrices , 1( ) B z - is an n m

polynomial matrix and 1( ) D z - is an n l polynomialmatrix.

1 1 21 2( ) ...

nn n na

a A z I A z A z A z -- - -= + + + + 1 1 2

0 1 2( ) ...n

nbb B z B B z B z B z

-- - -= + + + + 1 1 2

1 2( ) ...n

n n nccC z I C z C z C z -- - -= + + + +

1 1 20 1 2( ) ...

n

nd d D z D D z D z D z

-- - -= + + + + For simplicity, in the following the C polynomial matrix ischosen to be n n I . Let us consider the following finitehorizon quadratic criterion:

1 1

22

( , ) ( ) ( ) ( 1)

N N u

j j R Q

J N Nu y t j t w t j u t j= =

= + | - + + D + - (3)

17th IFAC World Congress (IFAC'08)Seoul, Korea, July 6-11, 2008

9411


3/5


4/5

In the simulation, the sampling period is T=0.3s, the randomdelays are imposed between the GPC controller and governorall the time, whose upper bound is 0.9s. Fig.2 shows therandom network delay in this simulation. Let the predictivehorizon 3 N = , the control horizon 3 Nu = ,

0.001 nN nN R I = , 0.25 I mM mM Q = . Each of the DISCOsin the two areas demands 0.01 p.u.MW power at 9s, Figs. 3, 4and 5 depict the excursions of area frequencies, tie-line

power flow and GENCOs outputs respectively.

Fig.2 random network delay

(a) Frequency deviation of area1

(b) Frequency deviation of area2Fig.3 Frequency deviation

In Fig.3,it can be seen that the frequency deviation in eacharea approaches zero at 80s around the maximum value is0.014. Fig.4 shows the actual tie line power, and it reaches -

0.005 p.u.MW, which is the scheduled power on the tie linein the steady state. Fig.5 shows actual generated powers ofthe GENCOs. The trajectories settle respective desiredgenerations in the steady state. The simulation results showthe validity of the proposed method.

4. CONCLUSIONS

This paper has studied the design, simulation andimplementation of the networked AGC systems afterderegulation. A GPC controller was used to the networkedand deregulated AGC systems. All control predictions togovernor are transmitted in single package mode. In addition,the data are transmitted with time " stamp, consequently, the

time delay and dropouts can be compensated. The simulationresults demonstrate effectiveness of the presented approach.

Fig.4 Deviation of tie line power flow

(a) GENCO1 power change

(b) GENCO2 power change

(c) GENCO3 power change

(d) GENCO4 power changeFig.5 GENCO power change

REFERENCES

Atic N., Rerkpreedapong D., Hasanovic A. and Feliachi A.(2003). NERC compliant decentralized load frequencycontrol design using model predictive control, Proc. of


9413


5/5

IEEE Power Engineering Society General Meeting . 2: 13-17.

Bevrani A., Mitani Y., Tsuji (2003). Robust load frequencyregulation in a new distributed generation environment,

Proc. of IEEE-PES General Meeting, Toronto.

Canada, 548-553.

Bevrani A., Mitani Y., Tsuji (2004). Robust decentralizedAGC in a restructured power system. EnergyConversion and management , 45:2297-2312.

Bevrani H, Hiyama T (2005). A control strategy for LFCdesign with communication delays. Proc. of the 7 th

International Power Engineering Conference(IPEC) , 1087-1092.

Bhowmik S., Tomsovic K., A. Bose (2004). Communicationmodels for third party load frequency control. IEEETrans.Power System , 19(1):543 " 548.

Donde V, Pai M and Hiskens I (2001). Simulation and

optimization in an agc system after deregulation, IEEETrans. Power Systems , 16(3):481-489.

Demiroren A, Zeynelgil H.L (2006). GA application tooptimization of AGC in three-area power system afterderegulation, Electrical Power and EnergySystems , 29(3):1-11.

Feliachi A (1997). Reduced H load-frequency controllerin a deregulated electric power system environment.

Proceeding of the Conferrence on Decisionand Control, USA, 3100-3104.

He F, Duan X, Su S et. al.(2005). Research on AGC robust

control over communication network. Proc. of IAS2005 , 2068-2074.Liu G. P., Rees D. and Chai S. C (2005). Design and practical

implementation of networked predictive control systems. Proc. of Networking, Sensing and Control2005 . 3:336-341.

Nobile E., Bhowmik S., Tomsovic K., Bose A. (2000).Impact of signal delay uncertainties in opencommunication networks on load frequency control.

Proc. of Euro Conference on Risk Management in Power System Planning andOperation in a Market Environment (RIMAPS'

2000), Sep 25-28 ,2000, Funchal, Portugal. RUR-17.

Paridal M, Nandal J (2005). Automatic generation control ofa hydro-thermal system in deregulated environment.

Electrical Machines and Systems, 2:942-947Rerkpreedapong D., Atic N. and Feliachi A. (2003) Economy

oriented model predictive load frequency control, Proc.of Large Engineering Systems Conference on

Power Engineering 2003 . 12-16.Shayeghi H., Shayanfar H.A, Jalili A (2006). Multi-stage

fuzzy PID power system automatic generation controllerin deregulated environments. Energy Conversionand managemen t , 47:2829-2845.

Tang P.L. and Silva C. W (2006). Compensation for

transmission delays in an Ethernet-based controlnetwork using variable-Horizon predictive control.

IEEE Trans. On Control System Technology, 14(4):707-718.

Yu X, Tomsovic, K (2004). Application of linear matrixinequalities for load frequency control with

communication delays. IEEE Trans. PowerSystems. 19:1508-1515.

Appendix A.

1,2 20.0 P T s= , 1,2 120 / . . P K Hz p u MW = , 1,2,3,4 0.3T T s= ,

1,2,3,4 0.08GT s= , 60 f Hz = , 1,2 2.4 / . . R Hz p u MW = ,

12 0.086 . . /T p u MW Radian= , 1,238.33 10 . . / D p u MW Hz -= .

Acknowledgements

This work is supported by National Natural ScienceFoundation of China under grant (No. 60674051) and Beijing

Natural Science Foundation under grant (No. 4072022).These are gratefully acknowledged.


9414

Documents

AGC Control Predictivo China