Load Frequency Control Using PID Tuned

ANN Controller in Power System

V.Shanmuga sundaram , Dr.T.Jayabarathi
Lecturer/EEE , Professor,
Sona college of technology School of Electrical Engineering,
Salem-5 ,Tamilnadu,India VIT University,Vellore
[email protected] Tamilnadu,India.

Abstract: The main objective of Load Frequency which they are connected. The net power flow on tie-
Control(LFC) is to regulate the power output of the lines is scheduled on a priori contract basis. It is
electric generator within an area in response to changes therefore important to have some degree of control
in system frequency and tie-line loading .Thus the LFC over the net power flow on the tie-lines. Load
helps in maintaining the scheduled system frequency
Frequency Control (LFC) allows individual utilities
and tie-line power interchange with the other areas
within the prescribed limits. Most LFCs are primarily to interchange power to aid in overall security while
composed of an integral controller. The integrator gain allowing the power to be generated most
is set to a level that the compromises between fast economically.
transient recovery and low overshoot in the dynamic The variation in Load frequency is an
response of the overall system. This type of controller is index for normal operation of the power systems.
slow and does not allow the controller designer to take When the load perturbation takes place, it will affect
in to account possible changes in operating condition the frequency of other areas also. In order to control
and non-linearities in the generator unit. Moreover, it frequency of the power systems various controllers
lacks in roubustness.Therefore the simple neural
are used in different areas, but due to the non-
networks can alleviate this difficulty. The ANN is
applied to self tune the parameters of PID controller. linearity in system components and alternators, these
Multi area system, have been considered for simulation conventional feedback controllers could not control
of the proposed self tuning ANN based PID controller the frequency quickly and efficiently.
.The performance of the PID type controller with fixed
gain, Conventional integral controller, and ANN based 1.1 PROBLEM DEFINITION
PID controller have been compared through MATLAB
Simulation results. Comparison of performance In order to keep the power system in
responses of integral controller & PID controller show normal operating state, a number of controllers are
that the neural-network controller has quite satisfactory
used in practice. As the demand deviates from its
generalization capability, feasibility and reliability, as
well as accuracy in multi area system. The qualitative normal operating value the system state changes.
and quantitative comparison have been carried out for Different types of controllers based on classical linear
Integral,PID and ANN controllers. The superiority of control theory have been developed in the past.
the performance of ANN over integral and PID Because of the inherent non-linearities in system
controller is highlighted. components and synchronous machines, neural
Keywords: Power system, Neural network, Back network techniques are considered to build non-linear
propagation algorithm, Controllers. ANN controller with high degree of performance.
Most load frequency controllers are
1. INTRODUCTION primarily composed of an integral controller. The
In electric power generation, system integrator gain is set to a level that compromise
disturbances caused by load fluctuations result in between fast transient recovery and low overshoot in
changes to the desired frequency value. Load the dynamic response of the overall system. This type
Frequency Control (LFC) is a very important issue in of controller is slow and does not allow the controller
power system operation and control for supplying designer to take into account possible non-linearities
sufficient and both good quality and reliable power. in the generator unit.
Power networks consist of a number of
utilities interconnected together and power is
exchanged between the utilities over the tie-lines by

978-1-61284-379-7/11$26.00 2011
c IEEE 269
1.2 OBJECTIVES OF LFC: equation to obtain deviations from nominal flow as a
function of deviations in phase angle from nominal.
In order to regulate the power output of
the electric generator within a prescribed area in 1
response to changes in system frequency, tie line Ptieflow + ΔPtieflow = [(β 1 + Δβ 1 ) − ( β 2 + Δβ 2 )]
loading so as to maintain the scheduled system X tie
frequency and interchange with the other areas within 1 1
the prescribed limits. = ( β1 − β 2 ) + (Δβ 1 − Δβ 2 )
X tie X tie
then
1.2.1Multi Area System Modeling 1
ΔPtieflow = (Δβ 1 − Δβ 2 )
X tie
In multi area system load change in one
area will affect the generation in all other
interconnected areas. Tie line power flow should also where ∆β1 and ∆β2 are equivalent to ∆δ1 and
be taken into account other than change in frequency. ∆δ2 .
We will discuss two area and three area systems in
the following session.
Then equation (1.16) can be expressed as,
1.2.2 Two Area System
T
ΔPtieflow = ( Δω1 − Δω 2 )
In two-area system, two single area s
systems are interconnected via the tie line.
Interconnection established increases the overall where T is “tie-line stiffness” coefficient and ,
system reliability. Even if some generating units in
one area fail, the generating units in the other area T = (2×3.14×50)× 1 / Xtie (for a 50-Hz system).
can compensate to meet the load demand.
Suppose now that we have an
1.2.3Tie Line Model interconnected power system broken into two areas
each having one generator. The areas are connected
The power flowing across a by a single transmission line. The power flow over
transmission line can be modeled using the DC load the transmission line will appear as a positive load to
flow method as one area and an equal but negative load to the other,
or vice versa, depending on the direction of flow. The
1 direction of flow will be dictated by the relative
Ptieflow = (β1 − β 2 ) phase angle between the areas, which is determined
X tie
by the relative speed -deviations in the areas.
This tie flow is a steady-state quantity. For
purposes of analysis here, we will perturb the above
A block diagram representing this interconnection can be drawn as in figure .

270 2011 1st International Conference on Electrical Energy Systems

 1 / R1

∆Pvalve

1 1 1
∑
1 + sTG1 1 + sTCH 1 M 1 s + D1
Load ∆ω1
reference ∆PL1 ∆Ptie

T/s
∆PL2
∆ω2
1 1 1
∑
1 + sTG 2 1 + sTCH 2 M 2 s + D2
Load
reference
∆Pvalve

1 / R2

Figure 1. Block diagram of interconnected areas

assumes that mechanical powers are constant, the

It should be noted that the tie power rotating masses and tie line exhibit damped
flow was defined as going from area 1 to area 2; oscillatory characteristics are known as
therefore, the flow appears as a load to area 1 and a synchronizing oscillations.
power source (negative load) to area 2. If one
It is quite important to analyze the ⎛ 1 ⎞
steady-state frequency deviation, tie-flow deviation − ΔPL1 ⎜⎜ + D2 ⎟⎟
and generator outputs for an interconnected area after
ΔPtie = ⎝ R2 ⎠
a load change occurs. Let there be a load change ∆PL 1 1
in area 1. In the steady state after all synchronizing + + D 1 + D2
oscillations have damped out, the frequency will be R1 R2
constant and equal to the same value on both areas.

Then
∆ω1 = ∆ω2 = ∆ω and The new tie flow is determined by the
Finally we have, net change in load and generation in each area. We
− ΔPL1 do not need to know the tie stiffness to determine this
Δω = new tie flow, although the tie stiffness will determine
1 1
+ + D 1 + D2 how much difference in phase angle across the tie
R1 R2 will result from the new tie flow.[6]
2. BACK PROPAGATION ALGORITHM
Almost any function can be approximated
using the multilayer network if we have sufficient
number of neurons in the hidden layer. Infact it has
been shown that two layer networks, with sigmoid

2011 1st International Conference on Electrical Energy Systems 271

transfer functions in the hidden layer and linear NNTOOL method provides the facility
transfer functions in the output layer can approximate to train through one of the methods Say conjugate
virtually any function of interest to any degree of gradient method, Levenberg-Marquardt method for
accuracy , provided sufficiently many hidden units back propagation. It is superior to approximate
are available. steepest descent method.
Hence at first training is carried out
For multilayer networks output of one using the nntool method. In the neural network we
layer becomes input to the following layer. have employed TANSIG as transfer function in the
Step-1 hidden layer and PURELIN in the output layer. Then
the obtained weights and biases are chosen as the
initial weights and biases.
Propagation of input forward through the Now the training is carried out using
network the first method (using approximate steepest descent
a0 = p , 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 m +1 = f m +1
( W m +1 a m + b m +1 ) for m = 0 , 2 , ... , M − 1,
2.2 Design of ANN Controller:
The range over which error signal is in
a = am 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
Step -2 taken as the input.
This input – target pair is fed and new
Propagation of sensitivities backward through neural network is formed using “nntool” in the
the network MATLAB Simulink software. Weights and Biases
obtained are fed to back propagation algorithm using
sM =−2F' M (nM)*(t −a) , approx. steepest descent method. Thus the neural
network is well trained. Updated weights and biases
are given to a fresh neural network. Now the neural
sm = F'm(nm)*(Wm+1 )T sm+1, form
=M−1,...,2,1. network is ready for operation.

In ANN controller is trained with the set of

Step -3 data to determine the PID controller parameters:
Proportional constant, KP
Updation of weights and biases using Integral constant, KI
approximate steepest descent rule Derivative constant, KD.
Normal PID controller has fixed
constants for proportional, integral and derivative
term .But during transient state controller effect can
W m (k + 1) = W m (k ) − α s m (a m−1 ) T , be made to perform better by variation of the
constants. This is accomplished by ANN controller.

b m (k + 1) = b m (k ) − α s m .

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).

272 2011 1st International Conference on Electrical Energy Systems

 3. DYNAMIC SIMULATION OF INTEGRAL, Peak undershoot is reduced, Settling time is reduced.
PID & ANN CONTROLLERS The superiority of ANN controller is established in
the cases of two area systems. From the Qualitative
3.1Two Area LFC-Area 1:
PERFORMANCE INDICES
CHANGE IN IAE ISE ITAE
FREQUENCY IN
Hz
INTEGRAL
Δ f1 0.1793 0.0072 0.5572

Δ f2 0.2932 0.0142 1.2304

PID
Δ f1 0.0487 8.0973e- 0.0799
004
Figure 3.1 Comparison of Integral, PID, and ANN Δ f2 0.0577 0.0014 0.0885
Controller Performance for Two Area System-
Change in frequency in Area 1 . ANN
Δ f1 0.0443 8.2026e- 0.0602
004
3.2 Two Area LFC-Area 2: Δ f2 0.0513 0.0014 0.0649

and Quantitative comparison of the results we can

conclude that the ANN controller yields better
results. ANN controller gives minimum
IAE/ISE/ITAE compared to the conventional integral
and PID controllers. The simulation studies were also
done for change in the operating conditions like
change in governor time and turbine time constants.
The Qualitative comparisons of all the controllers
show that the ANN results in robust performance.
Hence ANN controller has large potential to be used
as a control strategy for the Load Frequency Control.
5.REFERENCES
Figure 3.2 Comparison of Integral ,PID and ANN [1].Aanstad, O. J. and Lokay, H. E. (1970) Fast valve
Controller Performance for Two Area System- control can improve turbine generator response to
Change in frequency in Area 2. transient disturbances. Westinghouse Engineer, July,
pp. 114-119.
3.3 Quantitative Comparison of [2] Dr.Chaturvedi D.K Prof.Satsangi P.S & Prof.kalra
Integral,PID and ANN Controllers-Two Area P.K, Application of Generalized Neural Network to
LFC . Load Frequency Control Problem.
[3].Djukanovic, M., Sobajic, D. J. and Pao, Y. H.
(1992) Neural net based determination of generator-
With reference to the results obtained shedding requirements in electric power systems.
as shown in Table3.3, we can conclude that the IEEE Proceedings--C 139, 5 427-436.
performance indices IAE/ISE/ITAE are minimum [4] Elgerd OI. Electric energy systems theory: An
both Area 1 and Area 2 .When ANN controller introduction. McGraw-Hill;
compared to that of Integral and PID controllers. [5] Kundur.P (1994) Power System Stability and
Control New York McGraw-
4.CONCLUSION Hill
From the simulation results obtained [6] Hadi Saadat Power System Analysis Tata
for load disturbances for ANN controller, PID Mcgraw-Hill Publishing Company Limited –New
controller, Conventional integral controller we can Delhi.
conclude that ANN controller is faster than the other,

2011 1st International Conference on Electrical Energy Systems 273

[7].Hertz, J., Krogh, A. and Palmer, R. G. (1991)
Introduction to the Theory of Neural Computation.
Addison-Wesley, Reading, MA.

[8] Ha QP. A fuzzy sliding mode controller for power

system load-frequency control. In: IEEE Second
Conf -Knowledge-Based Intelligent Systems, 1998
[8] Jaleeli N, VanSlyck LS, Ewart DN, Fink
LH,Hoffmann AG. Understanding automatic
generation control. IEEE Trans Power Systems
1992;7(3):1106–12.
[9] Kanniah J, Tripathy SC, Malik OP, Hope GS.
Microprocessor-based adaptive load-frequency
control. IEE Proc C1984; 131(4):121–8.
[10] Pan CT, Liaw CM. An adaptive controller for
power system load-frequency control. IEEE Trans
Power Systems 1989;4(1):122–8.
[11] 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.
[12] Talaq J, Al-Basri F. Adaptive fuzzy gain
scheduling for load frequency control. IEEE Trans
Power Systems 1999;14(1):145–50.
[13] Vajk I. Adaptive load-frequency control of the
Hungarian power system. Automatica 1985;
21(2):129–37.
[14] Yamashita K, Miyagi H. Multivariable self-
tuning regulator for load frequency control system
with interaction of voltage on load demand. IEEE
Proc D 1991; 138(2).

274 2011 1st International Conference on Electrical Energy Systems