International Journal of Electrical Energy, Vol. 3, No.

3, September 2015

Load Frequency Control in Single Area System

Using Model Predictive Control and Linear
Quadratic Gaussian Techniques
Tarek Hassan Mohamed, Gaber Shabib, Esam H. Abdelhameed, and Mohamed Khamies
Faculty of Energy Engineering, Aswan, Egypt
Email: {tarekhie, gabershabib, mohamedahmedmak}@yahoo.com

Yaser Qudaih
Kyushu Institute of Technology, Japan
Email: [email protected]

Abstract—This paper proposes a design and simulation of Optimal control techniques based on feedback
Linear Quadratic Gaussian method (LQG) and model controllers have been proposed to achieve better
predictive controller (MPC) for load frequency control in performance [3], [4]. Other approach based on adaptive
single area power system to improve system performance. neural networks has been employed to achieve better
The proposed controller has been designed such that the
dynamic response [5], [6]. Robust adaptive control
effect of the uncertainty due to governor and turbine
parameters variation and load disturbance is reduced. To schemes have been developed to deal with changes in
account for the modeling uncertainties, the proposed system parameters such as Riccati equation approach [7],
controller estimates the system interface variables and uses H∞-control [8], μ-synthesis approach [9], robust pole
these estimates and optimizes a given performance index assignment approach [10]. Fuzzy logic controllers have
and allocate generating unit output. Digital simulations for a been used in many reports for LFC design in a two area
single control area are provided to validate the effectiveness power system [11], [12]. The MPC controller appears to
of the proposed scheme. The performance of the proposed be an efficient strategy to control many applications in
controller is compared with (MPC) controller. The results industry [13], [14]. In [15] fast response and robustness
show that with proposed (MPC+LQG) technique the system
against parameter uncertainties and load changes can be
performance has good robustness in face of uncertainties
due to governor and turbine parameter variation and load obtained using MPC controller. Despite the fact that
disturbance and carried out the superiority of the proposed controller succeeded in its target, but the door is still
(MPC+LQG) technique. opening to more techniques to improve the system
frequency in face of system power fluctuation and load
Index Terms—linear quadratic Gaussian, load frequency disturbance. This paper proposes a new controller for
control, model predictive controller, Kalman estimator LFC in a single area power system. The proposed control
technique produces its optimal output derived from a
quadratic cost function minimization based on the
I. INTRODUCTION dynamic model of the single area power system.
Load frequency controller (LFC) is very important The technique estimates the optimal control signal
issue in power system operation and control for sufficient while respecting the given constrains over the output
and reliable electric power. LFC acts as a balance frequency deviation and the load change.
between generated power and the power demanded while The effects of the physical constraints such as
keeping the net interchanged tie-line power within the generation rate constraint (GRC) and speed governor
standard acceptable limit. Considerable efforts have been dead band are considered [16]. The power system with
made to design LFC controllers with better performance the proposed (MPC+LQG) technique has been tested
to cope with system parameter changes using various through the effect of uncertainties due to governor and
methodologies. turbine parameters variation and load disturbance using
Various control strategies have been proposed and computer simulation. A comparison has been made
investigated by several researchers for LFC design of between the (MPC+LQG) and the (MPC) controller
power systems. Many classical approaches have been confirming the superiority of the proposed (MPC+LQG)
used to provide supplementary control which will drag technique.
the frequency to normal operating value within very short This paper is organized as follows Section 2 describes
time, such methods include use of proportional integral the system dynamics and implementation scheme of
controllers [1], [2]. single area power system with MPC technique. Section 3
describes the general consideration about LQG technique.
Manuscript received May 5, 2015; revised August 3, 2015. Section 4 describes the implementation scheme of single

©2015 International Journal of Electrical Energy 141

doi: 10.18178/ijoee.3.3.141-144
 International Journal of Electrical Energy, Vol. 3, No. 3, September 2015

area power system with proposed (MPC+LQG) controller.

Simulation results and general remarks are presented in
Section 5. Finally, the paper conclusion is in Section 6.

II. SYSTEM DYNAMICS

In this section, a simplified frequency response model
for a single area power system with an aggregated
generator unit is described in [16].
The overall generator-load dynamic relationship
between the supply error (∆Pd − ∆Pl ) and the frequency
deviation (∆f) can be expressed as:
Figure 2. The block diagram of single area power system including
1 1 D proposed MPC controller.
s∆f = ( ) ∆Pd – ( ) ∆Pl − ( ) ∆f (1)
M M M

The dynamic of the turbine can be expressed as: III. LINEAR QUADRATIC GAUSSIAN
1 1 Load frequency control for a single area power system
s∆Pd = ( ) . ∆PG − ( ) . ∆Pd (2)
Tt Tt has been developed based on both of MPC and LQG
The dynamic of the governor can be expressed as: techniques. The name LQG arises from the use of a linear
model, an integral cost function, and Gaussian white
1 1 1
s∆PG = ( ) . ∆Pc − ( ) ∆f − ( ) . ∆PG (3) noise processes to model disturbance and noise signals.
Tg R.Tg Tg
The LQG controller consists of an optimal state
Equations (1), (2), (3) can be combined in following feedback gain “k” and the Kalman estimator.
state space model: The optimal feedback gain is calculated such that the
D 1 feedback controls law u = -kx minimizes the performance
− 0 0 −
1
index:
M M
s∆f ∆f M ∞
0 −
1 1 ∆P H = ∫0 (X T QX + uT Ru)dt (6)
[s∆Pd ] = [∆Pd ] + [ 0 0 ] [ c]
Tt Tt
1 ∆Pl
s∆PG 1 1 ∆PG 0 where Q and R are positive definite or semi definite
− 0 − Tg
[ R.Tg Tg ] Hermitical or real symmetric matrices [17].
(4) The optimal state feedback u = -kx could not be
∆f implemented without full state measurement.
y = [1 0 0 ] [ ∆Pd] (5) In our case, the states are chosen to be the frequency
∆PG deviation ∆f , mechanical power change ∆Pmi , the
where: governor output change ∆Pgi , and the area tie-line power
S: differential operator.
change ∆Ptie,i . The frequency deviation ∆f, the area tie-
Δ𝑃𝑔 : the governor output change.
line power change ∆Ptie,i and the supplementary control
Δ𝑃𝑚 : the mechanical power change.
action ∆Pci are chosen to be the only measured signals
Δf: the frequency deviation.
which are fed to the Kalman estimator. The Kalman filter
Δ𝑃l : the load change. estimator is used to drive the state estimation:
Δ𝑃𝑐 : supplementary control action.
y: the system output. x̂i = [∆f̂i ∆p̂mi ∆p̂gi ∆p̂tie,i ] (7)
H: equivalent inertia constant.
Such that u = -kx remains optimal for the output
D: equivalent damping coefficient.
feedback problem.
R: speed droop characteristic.
The state estimation is generated from:
Tg and Tt: are governor and turbine time constants.
The block diagram of the past equation is shown in Fig. (ẋ̂ ) = (A − Bk − LC)x̂ + Ly (8)
1.
where L is the Kalman gain which is determined by
knowing the system noise and measurement covariance
Qn and Rn.
However, the accuracy of the filter’s performance
depends heavily upon the accuracy of this covariance. On
the other hand the matrices A and B containing the
machine parameters are not required to be very accurate
Figure 1. The block diagram of uncontrolled single area. due to the inherent feedback nature of the system.
Fortunately, the Kalman filter performs best for linear
Fig. 2 shows the block diagram of simplified systems. The optimal state feedback gains and the
frequency response model for single area power system Kalman state space model have been calculated off-line
with aggregated unit including the proposed MPC which results in great saving in computational burden.
controller. Fig. 3 shows the block diagram of kalman estimator.

©2015 International Journal of Electrical Energy 142

 International Journal of Electrical Energy, Vol. 3, No. 3, September 2015

governor is specified as 0.05%. The parameters of the

controller are set as follows:
 Prediction horizon = 10,
 Control horizon = 2,
 Weights on manipulated variables = 0.8,
 Weights on manipulated variable rates = 0.1,
 Weights on the output signals = 0.1,
 Sampling interval = 0.0003 sec.
For LQG: k = [0.1161 0.4801 0.0574],
Figure 3. The block diagram of kalman estimator.
15.8 0 0
Q=[ 0 0.000233 0 ], r = [150].
IV. SYSTEM CONFIGURATION
0 0 . 00005
The block diagram of a simplified frequency response The system performance with the proposed
model for single area power system with aggregated unit (MPC+LQG) controller at nominal parameter is tested
including the proposed (MPC+LQG) controller is shown and compared with the system performance with MPC
in Fig. 4. controller.

TABLE I. PARAMETERS AND DATA OF PRACTICAL SINGLE AREA

POWER SYSTEM

D(PU/HZ) H(PU S) R(HZ/PU) Tg (S) Tt (S)

0.015 0.08335 3.00 0.08 0.4

A. First Case
The system is tested with load change (the ΔPL
assumed to be 0.02 Pu at t = 3 sec). Fig. 5 shows the
simulation results in this case. The results are the
governor valve position ΔPm of both proposed
(MPC+LQG) controller and MPC controller, the
frequency deviations and the governor’s controlled input
signals of both two systems following a step load change.
Figure 4. The block diagram of single area power system including the It has been noticed that with the proposed (MPC+LQG)
proposed (MPC+LQG) controller.
controller the system is more stable and fast comparing
with the system with MPC controller.
V. RESULTS AND DISCUSSION
B. Second Case
Computer simulations have been carried out in order to
validate the effectiveness of the proposed scheme. The The robustness of the proposed (MPC+LQG)
Matlab/Simulink software package has been used for this controller against parameter uncertainty is validated.
purpose. A practical single area power system having the Both the governor and turbine time constants are
following nominal parameters [1] listed in Table I. The increased to Tg = 0.12 sec and Tt = 0.975 sec respectively.
simulation studies are carried out for the proposed Fig. 6 shows the simulation result of this case, it has
controller with generation rate constraint (GRC) of 10% been indicated that a desirable performance response has
Pu. per minute. The maximum value of dead band for been achieved using (MPC+LQG) controller.

Figure 5. Power system response of first case: (a) the governor valve position (b) the frequency deviation (c) supplementary control action.

©2015 International Journal of Electrical Energy 143

 International Journal of Electrical Energy, Vol. 3, No. 3, September 2015

Figure 6. Power system response of second case: (a) the governor valve position (b) the frequency deviation (c) supplementary control action.

VI. CONCLUSION [12] A. P. Birch, A. T. Sapeluk, and C. S. Ozveren, “An enhanced

neural network load frequency control technique,” in Proc.
This paper investigates robust load frequency control International Conference on Control, Mar. 1994, pp. 409-415.
[13] J. Thomas, D. Dumur, J. Buisson, and H. Gueguen, “Model
of a single area power system based on the (MPC+LQG) predictive control for hybrid systems under a state partition based
control technique. Digital simulations have been carried MLD approach (SPMLD),” in Proc. International Conference on
out in order to validate the effectiveness of the proposed Informatics in Control, Automation and Robotics, Setúbal, 2004,
scheme. Simulation results show that the fast response, pp. 78-85.
[14] J. Richalet, A. Rault, J. L. Testud, and J. Japon, “Model predictive
robustness against parameter uncertainties and load heuristic control application to industrial processes,” Automatica,
changes can be considered as some advantages of the vol. 14, no. 5, pp. 413-428, 1978.
proposed (MPC+LQG) controller. In addition, a [15] T. H Mohamed, H. Bevrani, A. A. Hassan, and T. Hiyama,
performance comparison between the proposed controller “Model predictive based load frequency control design,” in Proc.
16th International Conference of Electrical Engineering, Busan,
and a MPC control scheme is carried out. It is shown that Korea, Jul. 2010.
the (MPC+LQG) controller response is much better than [16] H. Bevrani, Robust Power System Control, New York: Springer,
that of MPC controller response and able to deal with 2009, pp. 15-61.
both of parameter uncertainty and load changes more [17] A. A. Hassan, Y. S. Mohamed, and T. H. Mohamed, “Robust
control of a field oriented linear induction motor drive,” in Proc.
efficiently. 11th International Middle East Power Systems Conference, Cairo
University, Egypt, Dec. 19-21, 2010.
REFERENCES
T. H. Mohamed was born in Libia in 1975.
[1] P. Kundur, Power System Stability and Control, New York: He received his B.Sc. degree in Automatic
McGraw – Hill Book Company, 1994. control from Minofia University, Egypt in
[2] M. Aldeen and J. F. Marah, “Decentralized proportional-plus- 1997, and his M.Sc. and PhD from Minia
integral design method for interconnected power systems,” IEE University, Egypt, in 2006 and 2012
Proceedings-C, vol. 138, no. 4, pp. 263-274, 1991. respectively. He is now a lecturer at the
[3] K. Yamashita and H. Miyagi, “Multivariable self-tuning regulator Faculty of Energy Engineering, Aswan
for load frequency control system with interaction of voltage on University, Egypt. His area of interest
load demand,” IEEE Proceedings-D, vol. 138, no. 2, Mar. 1999. includes control of electrical machines, and
[4] C. I. Pan and L. M. Liaw, “An adaptive controller for power power systems.
system load-frequency control,” IEEE Transactions on Power
Systems, vol. 4, no. 1, pp. 122-128, Feb. 1989. Yaser Soliman Qudaih Graduated from the
[5] M. Dyukanovic, et al., “Two-Area load frequency control with University of Engineering and Technology,
neural networks,” in Proc. North American Power Symposium, Lahore, Pakistan in 1996 as an electrical
1993, pp. 161-169. engineer. He Completed his M.Sc. and PhD
[6] A. P. Brich, et al., “Neural network assisted load frequency from Kumamoto University, Japan in
control,” in Proc. 28th University Power Engineering Conf., 1993, Electrical Engineering. He is currently a
pp. 518-521. researcher (Project Assistant Professor) at the
[7] Y. Wang, R. Zhou, and C. Wen, “Robust load-frequency Department of Electrical Engineering and
controller design for power systems,” IEE Proceedings-C, vol. Electronics, Kyushu Institute of Technology
140, pp. 11-16, 1993. (KIT), Japan. His area of interest including
[8] D. Rerkpreedapong, A. Hasanovic, and A. Feliachi, “Robust load power system is renewable energy and Smart Grid Applications. He is a
frequency control using genetic algorithms and linear matrix member of the Institute of Electrical Engineers of Japan and IEEE.
inequalities,” IEEE Trans. on Power Systems, vol. 18, pp. 855-861,
2003. Mohamed Ahmed Khamies was born in
[9] Z. Q. Wang and M. Sznaier, “Robust control design for load Sohag in 1990. He received Bachelor degree
frequency control using μ-synthesis,” in Proc. Southco/94, in Electrical Engineering from Aswan
Conference Record, Orlando, FL, USA, 1994, pp. 186-190. University, Egypt in 2011. He is now electric
[10] M. Azzam, “Robust automatic generation control,” Energy engineer in ministry of electricity and
Conversion and Management, vol. 40, pp. 1413-1421, 1999. renewable energy, Egypt. His area of interest
[11] E. Cam and I. Kocaarslan, “Load frequency control in two area including power system is renewable energy
power systems using fuzzy logic controller,” Energy Conversion and control of electric machines and power
Management, vol. 46, pp. 233-243, 2005. system.

©2015 International Journal of Electrical Energy 144