An Investigation of ANN based PID
Controllers using Three- Area Load
Frequency Control in Interconnected
Power System
V.Shanmuga Sundaram 1 , T.Jayabarathi 2
1
Lecturer, Sona College of Technology,Salem-5,Tamilnadu,India
2
Professor / Director, School of Electrical Engineering, VIT University,vellore.Tamilnadu,India
Abstract: The LFC problem, which is the major requirement in parallel operation of several interconnected systems, is one of very important subjects in power system studies. In this study, the power systems with three areas connected through lines are considered. The perturbation of frequencies at the areas and resulting tie-line power flows arise due to unpredictable load variations that cause mismatch between the generated and demanded powers. The objective of LFC is to minimize the transient deviations and to provide zero steady state errors of these variables in a very short time. Variation in load frequency is an index for normal operation of power systems. When load Perturbation takes place anywhere in any area of the system, it will affect the frequency at other areas also. To control load frequency of power systems various controllers are used in different areas, but due to non-linearity's in the system components and alternators, these controllers cannot control the frequency quickly and efficiently. The simple neural networks can alleviate this difficulty. This paper deals with various controllers like proportional integral (PI), Proportional Integral Derivative (PID) and ANN (Artificial neural network) tuned PID controller for three area load frequency control.The performance of the PID type controller with fixed gain, Conventional integral controller (PI) and ANN based PID (ANN-PID) controller have been compared through MATLAB Simulation results. Comparison of performance responses of integral controller & PID controller show that the ANN- PID controller has quite satisfactory generalization capability, feasibility and reliability, as well as accuracy in three area system. The qualitative and quantitative comparison have been carried out for Integral,PID and ANN- PID controllers. The superiority of the performance of ANN over integral and PID controller is highlighted.
Keywords: Power system, Neural network, Back propagation algorithm, PI ,PID and ANN -PID controllers
I.INTRODUCTION
In electric power generation, system disturbances caused by load fluctuations result in changes to the desired frequency value. Load Frequency Control (LFC) is a very important issue in power system operation and control for supplying sufficient and both good quality and reliable power.
Power networks consist of a number of utilities interconnected together and power is exchanged between the utilities over the tie-lines by which they are connected. The net power flow on tie-lines is scheduled on a priori contract basis. It is therefore important to have some degree of control over the net power flow on the tie-lines. Load Frequency Control (LFC) allows individual utilities to interchange power to aid in overall security while allowing the power to be generated most economically.
The variation in Load frequency is an index for normal operation of the power systems. When the load perturbation takes place, it will affect the frequency of other areas also. In order to control frequency of the power systems various controllers are used in different areas, but due to the non-linearity in system components and alternators, these conventional feedback controllers could not control the frequency quickly and efficiently.
A.Problem Formulation
gain is set to a level that compromise between fast transient recovery and low overshoot in the dynamic response of the overall system. This type of controller is slow and does not allow the controller designer to take into account possible non-linearities in the generator unit.
B. Objectives of LFC
In order to regulate the power output of the electric generator within a prescribed area in response to changes in system frequency, tie line loading so as to maintain the scheduled system frequency and interchange with the other areas within the prescribed limits.
D. Three Area System Modeling
In three area system, generation and load demand of one domain is dealt. Any load change within the area has to be met by generators in that area alone through suitable governor action. Thus we can maintain the constant frequency operation irrespective of load change.
E.Generator Model
A single rotating machine is assumed to have a steady speed of ω and phase angle δ0. Due to various electrical or mechanical disturbances, the machine will be subjected to differences in mechanical and electrical torque, causing it to accelerate or decelerate. We are mainly interested in the deviations of speed, ∆ω, & and deviations in phase angle ∆δ, from nominal.
F. Load Model
The load on a power system comprises of a variety of electrical devices. Some of them are purely resistive. Some are motor loads with variable power frequency characteristics, and others exhibit quite different characteristics. Since motor loads are a dominant part of the electrical load, there is a need to model the effect of a change in frequency on the net load drawn by the system. The relationship between the change in load due to the change in frequency is given by
∆PL (freq) = D ∆ω (or) D = ∆PL (freq) / ∆ω
The net change in Pelec is
Pelec = ∆PL + D ∆ω
G. Prime-Mover Model
The prime mover driving a generator unit may be a steam turbine or a hydro turbine. The models for the prime mover must take account of the steam supply and boiler control system characteristics in the case of a steam turbine, or the penstock characteristics for a hydro turbine. Here only the simplest prime-mover model, the non reheat turbine, is considered. The model for a non reheat turbine shown in figure 1. Relates the position of the valve that controls emission of steam into the turbine to the power output of the turbine.
H.Governor Model
Suppose a generating unit is operated with fixed mechanical power output from the turbine, the result of any load change would be a speed change sufficient to cause the frequency-sensitive load to exactly compensate for the load change. This condition would allow system frequency to drift far outside acceptable limits. This is overcome by adding mechanism that senses the machine speed, and adjusts the input valve to change the mechanical power output to compensate for load changes and to restore frequency to nominal value. The earliest such mechanism used rotating “fly balls” to sense speed and to provide mechanical motion in response to speed changes. Modern governors use electronic means to sense speed changes and often use a combination of electronic, mechanic and hydraulic means to effect the required valve position changes.
II.BACK PROPAGATION ALGORITHM
Almost any function can be approximated using the multilayer network if we have sufficient number of neurons in the hidden layer. In fact it has been shown that two layer networks, with sigmoid transfer functions in the hidden layer and linear transfer functions in the output layer can approximate virtually any function of interest to any degree of accuracy , provided sufficiently many hidden units are available.
Basic steps of using MATLAB toolbox, nntool .
In this thesis training is carried out using the two methods. However the updation of weights and biases from the initial set values, sensitivities, trained data, coordination between the target and trained data can’t be explicitly in the second method (using nntool).
NNTOOL method provides the facility to train through one of the methods Say conjugate gradient method, Levenberg-Marquardt method for back propagation. It is superior to approximate steepest descent method.Hence at first training is carried out using the nntool method. In the neural network we have employed TANSIG as transfer function in the hidden layer and PURELIN in the output layer. Then the obtained weights and biases are chosen as the initial weights and biases.
Now the training is carried out using the first method (using approximate steepest descent method). All results such as the updating of weights and biases from the initial set values, sensitivities, trained data, coordination between the target and trained data can be obtained.
A. Design of ANN Controller:
The range over which error signal is in transient state, is observed. Responding values of the proportional, integral and derivative constants are set. This set is kept as target. Range of error signal is taken as the input.This input – target pair is fed and new neural network is formed using “nntool” in the MATLAB Simulink software. Weights and Biases obtained are fed to back propagation algorithm using approx. steepest
1
2
P
net3
∆
P 12
∆
P
13∆
P
23Area1
Area3
Area2
GsT
+
1
1
CH
sT
+
1
1
D
Ms
+
1
1 / R
Load
reference
set point
Governor prime mover
∆
P
Lrotating mass
& load
∆
P
valvenetwork. Now the neural network is ready for operation.
In ANN controller is trained with the set of data to determine the PID controller parameters: Proportional constant, KP
Integral constant, KI Derivative constant, KD.
Normal PID controller has fixed constants for proportional, integral and derivative term .But during transient state controller effect can be made to perform better by variation of the constants. This is accomplished by ANN controller.
III. DYNAMIC SIMULATION OF INTEGRAL, PID & ANN -PID CONTROLLERS
Figure1. Comparison of Integral, PID, and ANN Controller Performance for Three System-Change in frequency in Area 1
Figure3. Comparison of Integral, PID, and ANN Controller Performance for Three System-Change in frequency in Area 3
From the response of the Fiure 1,2 &3 of Area 1 to Area 3 we can conclude that the ANN-PID Controller yields better Performance than the Conventional Integral and PID Controller since the ANN-PID gives less undershoot and less settling time of all the areas.
IV.Quantitaive Comparison
Tabe-1
Controller
Integral PID ANN-PIDPEAK UNDER SHOOT (Hz)
Area 1 -0.065 -0.042 -0.04
Area 2 -0.095 -0.028 -0.025
Area 3 -0.008 -0.030 -0.028
Area 1 -0.065 -0.042 -0.04
SETTLING TIME IN(secs)
Area 1 40.0 34.0 28.0
Area 2 30.0 25.0 20.0
Area 3 36.0 32.0 28.0
PERFORMANCE INDICES CHANGE IN FREQUENCY IN
Hz
IAE ISE ITAE
INTEGRAL
(Area1)
Δ
f1 0.3233 0.0074 3.6350(Area2)
Δ
f2 0.2961 0.0130 1.3146(Area3)
Δ
f3 0.3408 0.0051 4.1500PID
(Area1)
Δ
f1 0.1003 8.2525e-004 0.8964(Area2)
Δ
f2 0.0527 0.0015 0.0601(Area3)
Δ
f30.1232 0.0015 0.9305
ANN
(Area1)
Δ
f1 0.1020 8.1251e-004 0.8939(Area2)
Δ
f2 0.0576 0.0014 0.0845(Area3)
Δ
f30.1224 0.0015 0.9139
With reference to the results obtained as shown in Table 1, we can conclude that the performance indices IAE/ISE/ITAE are minimum .When ANN-PID controller compared to that of Integral and PID controllers.
IV.CONCLUSION
From the simulation results obtained for load disturbances for ANN- PID controller, PID controller, Conventional integral controller we can conclude that ANN based PID controller is faster than the other, Peak undershoot is reduced, Settling time is reduced. The superiority of ANN controller is established in the cases of Three area systems. From the Qualitative and Quantitative comparison of the results we can conclude that the ANN-PID controller yields better results. ANN controller gives minimum IAE/ISE/ITAE compared to the conventional integral and PID controllers. Hence ANN-PID controller has large potential to be used as a control strategy for the Load Frequency Control.
V.REFERENCES
[1] Aanstad, O. J. and Lokay, H. E. (1970) Fast valve control can improve turbine generator response to transient disturbances.
Westinghouse Engineer, July, pp. 114-119.
[2] Dr.Chaturvedi D.K Prof.Satsangi P.S & Prof.kalra P.K, Application of Generalized Neural Network to Load Frequency Control Problem.
[3] V.Shanmuga sundaram and T.Jayabarathi A novel approach of load frequency control in multi area power system ,International journal of engineering science and technology Vol.3 no 3. March 2011.
[4] V.Shanmuga sundaram and T.Jayabarathi An investigation of PID tuned ANN controllers using load frequency control in single area power system ,International journal of recent trends in engineering and technology Vol.5. no 2. March 2011
[5] Djukanovic, M., Sobajic, D. J. and Pao, Y. H. (1992) Neural net based determination of generator-shedding requirements in electric power systems. IEEE Proceedings--C 139, 5 427-436.
[6] Elgerd OI. Electric energy systems theory: An introduction. McGraw-Hill; [7] Kundur.P (1994) Power System Stability and Control New York McGraw-Hill
[8] Hadi Saadat Power System Analysis Tata Mcgraw-Hill Publishing Company Limited –New Delhi.
[9] Hertz, J., Krogh, A. and Palmer, R. G. (1991) Introduction to the Theory of Neural Computation. Addison-Wesley, Reading, MA [10] Shayeghi H., Shayanfar.H.A, Jalili.A Load frequency control strategies:A state-of the art survey for the researcher. Energy
Conservation and Management.2009 :20(1):346-357.
[11] Ibraheem I , Kumar P and Kothari DP Recent philosophies of automatic generation control strategies in power systems IEEE Transactions on Power Syatems 2005:20(1):375-382
[12] Pan CT, Liaw CM. An adaptive controller for power system load-frequency control. IEEE Trans Power Systems 1989;4(1):122–8. [13] Reddoch P, Julich TT, Tacker E. Model and performance functional for load frequency control in interconnected power systems. In:
IEEE Conf Decision and Control, 1971.
APPENDIX
Plant parameters and contants
Parameters Area 1 Area 2 Area3
Turbine time constant 0.5 s 0.6s 0.5 s
Generator Time constant 0.2s 0.3s 0.2s
Generator Angular Momentum
10 MJrad/s 8 MJrad/s 10 MJrad/s
Governor Speed Regulation
0.05 pu 0.065 pu 0.05 pu
Load change for
Frequency change of 1% D = ∆p / ∆f
0.6% 0.6
0.9% 0.9
0.6% 0.6
Rated output 250 MW 250 MW 250 MW
Sudden Load Variation 250 MW 250/250=1p.u
250 MW 250/250=1p
250 MW 250/250=1p.u
BIOGRAPHY
V.Shanmuga Sundaram. received the M.E. degree in Power Systems Engineering from Anna University, Chennai, Tamilnadu, India, in 2005. He is a research Scholar of VIT University, Vellore ,Tamilnadu India and he has published 4 international journals and presented a paper 8 international national conferences. Currently, he is an Lecturer in Sona College of Technology Salem, His interests are in power system control design , Deregulated power systems and Intelligent control.