Anda di halaman 1dari 6

2016 5th International Conference on Informatics, Electronics and Vision (ICIEV) 402

Design of a Power System Stabilizer Using Adaptive-


Neuro Fuzzy Logic for a Multi-Machine System
Having Dynamic Loads
Moudud Ahmed Naruttam Kumar Roy
Department of EEE Department of EEE
Bangabandhu Sheikh Mujibur Rahman Science and Khulna University of Engineering and Technology
Technology University Khulna, Bangladesh
Gopalganj, Bangladesh e-mail: nkroy@eee.kuet.ac.bd
e-mail: moudud.eee@gmail.com

Abstract—This paper attempts to show how a fuzzy logic controller should generate a stabilizing signal to the high gain excitation
(FLC) is used for the design of a power system stabilizer (PSS) for a system of generators in order to create positive damping torque
multi-machine interconnected power system. It has also been which is in phase with rotor speed deviation [5]. The large
addressed how a fuzzy logic-PSS has been optimized by using an system comprises many interconnected machines. The fine-
adaptive neuro-fuzzy inference system (ANFIS) based technique. tuning of PSS parameters is a difficult job. Furthermore,
The design objective of the PSS is to supply an external damping ongoing variations in power system operating conditions
torque to an automatic voltage regulator (AVR) of a synchronous render this job more difficult [6].
generator without affecting the synchronizing torque. The PSS is
designed based on the use of FLC by Mamdani and Takagi-Sugeno The conventional PSSs (referred to as CPSS) parameters are
(T-S) methods for comparison purposes. The designed controllers derived from a linearized model of the power system [7].
are tested on a test system having a dynamic load. The simulation Furthermore, CPSSs work well with those specific operating
results by MATLAB/SIMULINK demonstrate the effectiveness of points of the system for which they are modeled. Because of
the ANFIS-PSS over FLC-PSS to control the low frequency nonlinear behavior of power systems, configurations and
electromechanical oscillation of interconnected power systems parameters are changed from time to time, so they are not able
under severe disturbances with dynamic loads. to give acceptable results over wider ranges of operating
circumstances. In recent times, with the revolutionary
Keywords—Dynamic Load; Power System Stabilizer; Fuzzy
advancement in digital computers, it is easy to implement
Logic Controller; Adaptive Neuro-Fuzzy Inference System.
novel controllers utilizing intelligent control approaches.
Fuzzy logic PSSs (FLPSSs) have been employed to improve
I. INTRODUCTION the performance of conventional PSSs. It is generally a rule
based system. Complex and nonlinear problems are made
Power systems operating conditions are essentially
easier to solve by FLC without knowing the accurate
nonlinear [1] and load level functions in a complex manner.
mathematical model of the systems [8]. In view of the fact that
Actually the system goes over a broad range of conditions,
the FLPSS shows good performance, there is no orderly design
causing low frequency oscillations (LFOs) [2]. The LFOs of
procedure for the fine-tuning of the parameters of it. Moreover
small magnitudes due to insufficient damping caused by
during modeling, it cannot be guessed which types of
undesirable operating conditions sustain linger and shrink
membership functions (MFs) should be employed from
power transfer capability [2]. The generator terminal voltage
analyzing the data [8]. Therefore, the design of FLPSS
and reactive power flow is maintained by an excitation system
becomes a tough job. On the other hand, an adaptive neuro-
equipped with an automatic voltage regulator (AVR). One of
fuzzy inference system (ANFIS) based technique is a
the dominant objectives of AVR is to keep the terminal voltage
promising method which adjusts membership functions and
of the generator at specified level. Thus, an AVR enhances a
rules adaptively to improve a system’s performance [9].
steady-state stability of power systems; still transient stability
remains a worrying issue. In transient stability when a fault This paper introduces a PSS for a multi-machine
occurs and stays for a limited period of time, it results in a interconnected power system comprising a dynamic load
considerable shrinkage of terminal voltage and machine power grounded on adaptive-neuro fuzzy technique. There is a rich
transfer capacity [3]. Modern generators with high gain exciters literature on the PSS design considering static loads in the
also create oscillatory instability. Power system stabilizer system [5]-[9]. As loads play an important role in power
(PSS) units are often used to improve the damping of local and system stability study [10], [11], the incorporation of load
inter-area modes of electromechanical oscillations. In dynamics during a PSS design is essential. Therefore, the aim
upgrading the dynamic stability of the power system, a PSS of the proposed research is to investigate the performance of
design is inevitably needed [4]. The purpose of a PSS is to ANFIS-PSS in the presence of a dynamic load in the system.
extend stability limit via modulation of the generator’s
excitation [5]. To make satisfactory damping, the stabilizer

978-1-5090-1269-5/16/$31.00 ©2016 IEEE


403

This paper is arranged in six sections. In section II, a TABLE I. LINGUISTIC VARIABLES OF FLC
concise review of the fuzzy logic configuration, MF and fuzzy
logic process is discussed. In Section III, the ANFIS design
procedure is discussed. Consequently, in Section IV,
description of the test system is mentioned. In Section V,
simulation results are shown. Section VI brings a conclusion of
this work.

II. FUZZY LOGIC PSS

A. Structure of FLC
TABLE II. RULE TABLE OF FLC [13]
The basic structure of FLC is shown in Fig. 1. The system
input received by the FLC is turned into fuzzy form. On the
basis of fuzzy rules, input, and MFs, the controller obtains the
output in fuzzy form and finally, turns the output in nonfuzzy
form [12].

Figure 1. FLC structure [12].

The design steps of FLC are as follows.


 Define input and output variables.
 Choice of linguistic variables.
 Selection of fuzzy rule base.
 Defuzzification action.
B. Fuzzy PSS Design in Mamdani Form Figure 2. Input MF (∆wr).

As a first line of FLPSS design input signals to the controller


are selected which characterize the dynamic performance of the
system. In this paper, a Mamdani form of FLC is used which
has two inputs and one output component. Inputs to the FLC
are the generator rotor angular speed deviation (∆wr) and
change of speed deviation (∆wr´) and the output is voltage
(∆VPSS). Seven linguistic fuzzy subset for each of the input-
output variables are used [12], [13]. The linguistic variables are
given in Table I. Two inputs, ∆wr and ∆wr´, generate 49 rules
that define the relation between input and output of FLC. Thus,
the stabilizing signals have been computed [13]. The rule table
of FLC is shown in Table II. Given that the exciter needed a
non-fuzzy signal, the centroid defuzzification approach is used
to obtain output. The MFs for inputs and output are shown in
Figures 2 to 4.

Figure 3. Input MF (∆wr´).


404

Figure 4. Output MF (∆VPSS).

III. ADAPTIVE-NEURO FUZZY CONTROLLER


An ANFIS editor GUI is available in Fuzzy Logic Toolbox
in MATLAB. This toolbar receives input/output data set as
input. Thus, it creates a fuzzy inference system (FIS). FIS’s
MF parameters tend to be adjusted with the use of either only
a backpropagation algorithm, or in a permutation with a least
squares method [14]. The process agrees to the fuzzy systems Figure 5. ANFIS-PSS design procedure.
towards learning from the data they model. At this point of the
system, the input-output data pair is generated using the
FLPSS structure. Afterwards, a fuzzy inference system is
created by using the ANFIS model in MATLAB. Inputs are
used, namely Δwr and Δwr´, and output is ∆VPSS [13]. In this
paper, a Takagi-Sugeno (T-S) type FIS model is used. The
inputs are Gaussian membership type. Seven MFs are required
to cover the full range of respective inputs. In this way, 49
rules appear for the output function which demonstrates a
linear relation to the inputs. The ANFIS-PSS design follows
the steps below.

 Generate input-output data of the system using


FLPSS.
 Then create a suitable ANFIS structure for a given
challenge.
 Select number of training epoch.
 Using the test patterns train ANFIS until it performs
in the expected level.
 Then check the model validation.
Figure 5 shows the flowchart of ANFIS-PSS design
procedure. Figures 6 and 7 show the training error and ANFIS
structure, respectively.

Figure 6. Training error of ANFIS.


405

V. SIMULATION RESULTS
The multi-machine two area complex power system’s
dynamic performance has been analyzed with FLC- and
ANFIS-based PSS. The generators are modeled by a sixth
order state-space model and the dynamic load is presented by
an exponential model [16] with time constants for controlling
the dynamics of active and reactive power. For induction motor
loads, the time constants are in the range up to one second [16].
A symmetrical three phase to ground fault is applied at the
middle of the transmission line, which is cleared by tipping the
fault line to test the robustness of FLPSS and ANFISPSS. The
simulation results can be divided into the following two parts
depending on which controller is used.
(a) for the FLC and
(b) for the ANFIS controller.
Figure 9 shows the comparative analysis of rotor speed
deviation of generator 2 at A-1 for FLPSS and ANFIS-PSS
when symmetrical three-phase fault is occured. The rotor speed
deviation is significantly large for FLPSS with respect to
ANFIS-PSS. Figure 10 shows the performace of the terminal
Figure 7. Architecture of T-S type ANFIS controller. voltage (vt) of generator 1 and generator 3 between FLPSS and
ANFIS-PSS under pre-fault, faulted and post-fault condition.
The active power flow from A-1 to A-2 for FLPSS and ANFIS
PSS is shown in Fig. 11. Figure 12 shows the voltage at the
IV. PROBLEM DESCRIPTION terminal of dynamic load (bus7) for FLPSS and ANFIS-PSS. It
The two area multi-machine power system used in this can be observed that the voltage at the bus 7, where dynamic
paper is shown in Fig. 8 [15]. The test system is modified by load is connected, is reached to the steady-state value quickly
connecting a dynamic load at bus 7. Two symmetrical areas are for ANFIS-PSS compared to FLPSS. In both cases, the system
connected through two 230 kV lines of 220 km long is stable. However, in case of FLPSS rotor speed change,
transmission lines. Each area has two similar round rotor maximum overshoot and settling time are significant, which
generators rated at 20kV/900MVA. The inertia constants of has improved in ANFISPSS. Therefore, it can be concluded
generators in area 1 (A-1) and area 2 (A-2) are 6.5s and 6.17s, that an ANFIS-PSS can be effectively used for the system
respectively. A dynamic load rated at 967 MW and 100MVAR having dynamic loads.
is connected in A-1. Similarly, another static load rated at
1767MW, 100MVAR and -187MVAR is connected in A-2.
Load voltage is improved by inserting a capacitor of
200MVAR and 350MVAR in A-1 and A-2, respectively. In the 1.010

next section, dynamic performance of the system is observed.


1.008

1.006
Speed deviation(pu)

G2(FLPSS)
1.004 G2(ANFIS)

1.002

1.000

0.998

0.996

0 1 2 3 4 5
Time(sec)

Figure 9. Comparative analysis of rotor speed deviation of G2 at A-1


for FLPSS and ANFIS-PSS (symmetrical three-phase fault).

Figure 8. Two area network with a dynamic load.


406

2.0 VI. CONCLUSION


1.8 In this research work, fuzzy logic and adaptive-neuro fuzzy
G1(FLPSS) techniques for the design of PSS are approached. The two
1.6
G3(FLPSS)
G1(ANFIS) different types of stabilizers have been tested on a two area
1.4
G3(ANFIS) complex multimachine system having a dynamic load under
Voltage(pu)

1.2 fault conditions. Simulation results illustrate the effectiveness


1.0 of the proposed adaptive neuro-fuzzy based PSS specially for
dynamic loads. The settling time and overshoot is minimum
0.8
for ANFIS-PSS compared to FLPSS. The ANFIS-PSS shows
0.6 excellent damping of low frequency electromechancal
0.4 oscillation under disturbances.
0.2

0 1 2 3 4 5 APPENDIX
Time(sec) Synchronous generator parameters: G1 to G4:
Nominal power=900e6VA, line-to-line voltage=20e3V, frequency =60Hz, Xd
Figure 10. Comparative performance of terminal voltage (vt) of G1 and =1.8pu, Xd'=0.3pu, Xd''=0.25pu, Xq =1.7pu, Xq'=0.55pu, Xq''=0.25pu, Xl
G3 for FLPSS and ANFIS-PSS (symmetrical three-phase fault). =0.2pu, Tdo'=8s, Tdo''=0.03s, Tqo'=0.4s, Tqo''=0.05s, Stator resistance (Rs)
=0.0025pu, Inertia coefficient, H(s) =6.5s for G1 & G2, H(s) =6.175s for G3
1800
& G4, friction factor (F) =0, pole pairs=4.
Dynamic load parameters:
1600 Nominal L-L voltage, Vn(Vrms)=230e3, frequency, fn(Hz)=60Hz, Active
power, (Po)=967We6, Reactive power(Qo)=100e6Var, Initial positive-
1400 sequence voltage (Vo)=0.994pu, exponents, np=1.3, nq=2, time constants,
Tp1=0.1, Tp2=0.1, Tq1=0.1, Tq2=0.1, Vmin=0.7pu.
1200
Power(pu)

1000

800 A1 to A2(FLPSS)
A1 to A2(ANFIS) REFERENCES
600
[1] M. A. Mahmud, H. R. Pota, M. J. Hossain, and N. K. Roy “Robust
400 Nonlinear Controller Design for Three-Phase Grid-Connected
Photovoltaic Systems Under Structured Uncertainties," IEEE Trans. on
200 Power Delivery, vol. 29, no. 3, pp. 1221-1230, 2014.
0
[2] S. M. Radaideh, I. M. Nejdawi, and, M. H. Mushtaha, “Design of
0 1 2 3 4 5 Power System Stabilizers using Two Level Fuzzy and Adaptive Neuro-
Time(sec)
Fuzzy Inference Systems,” International Journal of Electrical Power &
Energy Systems, vol. 35, no. 1,pp. 47-56, 2012.
Figure 11. Active power flow from A-1 to A-2 for FLPSS and [3] G. Y. R. Vikhram and S. Latha, “Design of Power System Stabilizer
ANFIS PSS (symmetrical three-phase fault). for Power System Damping Improvement using Optimization based
Linear Control Design” IEEE International Conference on Power
3.5 Electronics Drives and Energy Systems, vol. 8, pp. 2231-2307,
Bengaluru, India, 2012.
3.0 [4] A. S. Venugopal, G. Radman, and M. Abdelrahman, “An Adaptive
Neuro Fuzzy Power System Stabilizer for Damping Inter-Area
Oscillations in Power Systems,” Proceedings of the Thirty-Sixth
2.5
Southeastern Symposium on System Theory, pp. 41-44, 2004.
Bus Voltage(pu)

Volt. at bus 7(FLPSS)


[5] R. Gupta, D. K. Sambariya, and R. Gunjan, “Fuzzy Logic based Robust
2.0 Volt. at bus 7(ANFIS)
Power System Stabilizer for Multi-Machine Power System,” IEEE
International Conference on Industrial Technology, pp. 1037-1042,
1.5 Mumbai, India, 2006.
[6] H. Ni, G. T. Heydt, and L. Mili, “Power System Stability Agents Using
1.0 Robust Wide Area Control,” IEEE Power Engineering Review, vol. 17,
no. 4, pp. 58, 2002.
0.5 [7] R. You, H. J. Eghbali, and M. Hashem Nehrir, “An Online Adaptive
Neuro-Fuzzy Power System Stabilizer for Multimachine Systems,”
IEEE Trans. on Power Systems, vol. 18, no. 1, pp. 128-135, 2003.
0.0
0 1 2 3 4 5 [8] J. F. Ardanuy and P. J. Zufiria, “Adaptive Power System Stabilizer
Using ANFIS and Genetic Algorithms,” IEEE Conference on Decision
Time(sec)
and Control, pp. 8028-8033, Spain, 2005.
Figure 12. Voltage across dynamic load terminal (bus 7) for FLPSS and [9] P. Mitra, S. Chowdhury, S. P. Chowdhury, S. K. Pal, Y. H. Song, and G.
ANFIS-PSS (symmetrical three-phase fault). A. Taylor, “Damping Local and Inter-Area Oscillations with Adaptive
Neuro-Fuzzy Power System Stabilizer,” 41st IEEE International
Universities Power Engineering Conference, vol. 2. pp. 457-461, UK,
2006.
407

[10] N. K. Roy, H. R. Pota, M. A. Mahmud and M. J. Hossain, “D- [13] M. F. Othman, M. Mahfouf , and D. A. Linkens, “Designing Power
STATCOM Control in Distribution Networks with Composite Loads to System Stabilizer for Multimachine Power System Using Neuro-Fuzzy
Ensure Grid Code Compatible Performance of Photovoltaic Generators,” Algorithm,” jurnal teknologi, vol. 35, no. 1, pp. 55–64, 2001.
8th IEEE Conference on Industrial Electronics and Applications, [14] Ak. S.R. Swain, and A.K. Swain, and A. Mohapatra, “Design of Power
Melbourne, 2013. System Stabilizer” B. Tech. thesis, National Institute of Technology,
[11] N. K. Roy, H. R. Pota, and T. F. Orchi, “Small-signal Stability Rourkela, May, 2012.
Assessment of Active Distribution Networks with Dynamic Loads,” [15] P. Kundur, Power System Stability and Control. New York: McGraw-
Universities Power Engineering Conference, Bali, 2012. Hill, 1994.
[12] N. Gupta, “Fuzzy Logic Based Power System Stabilizer,” Masters [16] Jovica V. Milanović and I. A. Hiskens,“Effects of Dynamic Load Model
thesis, Thapar University, Patiala, 2008. Parameters on Damping of Oscillations in Power Systems,” Electric
Power Systems Research, vol. 33, pp. 53-61, 1995.