Electric Power Systems Research 74 (2005) 915
Hardware implementation of a fuzzy logicscale power system
A power d in ththe perform powerbeen used to lidateto the system n studibetter perfor eratinexperimentally implemented using MATLAB/Real-Time Windows Target toolbox software on a laboratory set up to model a simple powersystem of 1 KVA machine connected to an infinite bus through a transmission line. Experimental tests and results revealed the effectivenessof FLPSS, especially in vulnerable operating points. 2004 Published by Elsevier B.V.
The appprove the dthe focus o[1,2]. The oof continuobances, sucthe stabilizsurementsover a widcontrol menon-lineartroller appeas it consu
0378-7796/$doi:10.1016/jardware implementation; Fuzzy logic stabilizer; Laboratory scale power system
lication of power system stabilizers (PSS) to im-ynamic performance of a power system has beenf extensive studies for more than three decadesperating point of a power system drifts as a resultus load changes or unpredictable major distur-h as three-phase faults. It is necessary to adapt
er parameters in real-time based on on-line mea-in order to maintain good dynamic performancee range of operating conditions. The fuzzy logicthod has been found to be a good tool to deal withand ill-defined systems. Recently, the fuzzy con-ared to be one of the most convenient techniquesmes less computational time and it is robust .
ding author. Tel.: +966 5052 07454; fax: +966 1452 8148.dresses: email@example.com (S.A. Al-Osaimi),.edu.sa (A. Abdennour), firstname.lastname@example.org).
Also, this controller could easily be constructed utilizing asimple microcomputer paired with A/D and D/A converters. The implementation of fuzzy logic power system stabi-lizers has been introduced in a number of publications .Possible input signals to the fuzzy excitation controller arethe sensed generator speed deviation and acceleration. Thesesignals are first described by some linguistic variables us-ing the membership function in fuzzy set notation before thefuzzy controller can process them. A database, which con-tains all the decision rules, expressed in linguistic variablesis set up to form the basis for the fuzzy logic operation per-formed by the fuzzy excitation controller to reach a desiredoutput . The aim of this paper is to design and implementa fuzzy logic stabilizer to improve the stability of a 1 KVAlaboratory scale model of power system. The studied systemconsists of an alternator connected to an infinite bus througha short transmission line. The stabilizing signal is computedusing the standard fuzzy membership functions, which de-pend on the speed and acceleration state of the generator. The
see front matter 2004 Published by Elsevier B.V..epsr.2004.10.001Saud A. Al-Osaimi a, , Adel Abdennour b,a Saudi Telecom Company, P.O. Box 152071, R
b King Saud Universi
Received 7 May 2004; received in revised form 23 Aug
system stabilizer using the fuzzy logic is designed and implementeance of the fuzzy logic power system stabilizer (FLPSS) and PID
optimize the parameters of the fuzzy and PID stabilizers. To vaand the results of both stabilizers are compared. The simulatio
mance comparable to that of the PIDPSS over a wide range of opstabilizer on a laboratory
ullaziz A. Al-Sulaiman b
1785, Saudi Arabia
4; accepted 3 October 2004
is paper. Simulation studies are performed to evaluatesystem stabilizer (PIDPSS). Genetic algorithms have
the design, different types of disturbances are appliedes show that the fuzzy stabilizer provides a relativelyg conditions. The FLPSS and PIDPSS have also been
10 S.A. Al-Osaimi et al. / Electric Power Systems Research 74 (2005) 915
influence of the proposed stabilizer is demonstrated throughsimulation studies for different operating conditions and dis-turbances. The performance of this fuzzy logic stabilizer isalso evaluated experimentally in the laboratory.
2. Design of fuzzy logic stabilizer
The design process of FLPSS is consisted of the followingsteps : Selection of FLPSS input/output variables; Fuzzification; Rule definition; Rule inference and Defuzzification.
The block diagram of the FLPSS is shown in Fig. 1. TheFLPSS has two inputs and one output. The speed deviation() and its derivative () are considered as the inputsof the FLPSS. The speed deviation and its derivative passthrough two suitable gains, Ke and Ke, respectively, beforethey are fed to the FLPSS. The output signal of the stabilizeris also scathen is sentgains is tonormalized
Fuzzificvariables toinputs of Ffunctions slinguistic vMP, LP), wsmall negaand large pfunctions aTable 1.
Table 1Output memb
For thisables for ea
as shoe. Theed asle 1:e infet mem
ntecedhen them (bee deft into
ficatiot is th
e (ui) denotes the output membership grade for the ith
t Acceleration ()LN MN SN VS SP MP LP
deviation ()VS SP MP LP LP LP LPSN VS SP MP MP LP LPMN SN VS SP SP MP LPMN MN SN VS SP MP MPLN MN SN SN VS SP MPLN LN MN MN SN VS SPLN LN LN LN MN SN VSled by passing through the output gain, Ku, andto the power system. The main objective of these
allow the use of speed and acceleration signals inquantities.
ation is the process of transferring the crisp inputcorresponding fuzzy variables. In this paper, the
LPSS are fuzzified according to the membershiphown in Fig. 2. For each input variable, sevenariables are defined as (LN, MN, SN, VS, SP,hich indicate large negative, medium negative,
tive, very small, small positive, medium positive,ositive, respectively. Also, the output membershipre chosen as singleton functions as indicated in
Fig. 1. Block diagram of a system with FLPSS.
outputhe a0.8, tof th
SpeedLPMPSPVSSNMNLNFig. 2. Membership functions for , .
FLPSS, with two inputs and seven linguistic vari-ch input, there will be a maximum of 49 decisionwn in Table 2. Every entity in the table representsrule relating two inputs and one output can be
the following logic,If () is LP and () is LN, then (UPSS) is VS.rence mechanism is used to compute the FLPSSbership grades. For example, if the two parts of
ent yielded the fuzzy membership values 0.5 ande fuzzy operator simply selects the minimum valuecause of the AND operator), which equals 0.5.uzzifier converts the fuzzy value of the FLPSSa crisp (numerical) value. The input for the de-n process is the aggregated fuzzy output. The finale stabilizer signal UPSS. In this paper, the follow-n is used for defuzzification:
S.A. Al-Osaimi et al. / Electric Power Systems Research 74 (2005) 915 11
Table 3Optimal parameters for fuzzy and PID stabilizers
Type Parameter Value
In this ptune the fupurposes. Trithm is toiterative mtuning algogiven by:
where N istime and G
(i) Initiali(ii) For eac
(iii) Checkto step
(iv) Producand mu
(v) DetermThe FL
a single ouKe, and onparameters
The parasystem mocondition:
A 0.25 pplied. The pin Table 3lizers are s
Fig. 5 stem with th0.25 pu ste
3. Fitness function for tuning the parameters of the fuzzy stabilizer.
4. Fitness function for tuning the parameters of the PID stabilizer.
he system without stabilizer is highly oscillatory whilexistence of FLPSS or PIDPSS assists in damping thislation. Also one can say that the fuzzy and PID stabi-under this operating condition have the same response,
h is expected since they are both designed and optimizedis operating condition. The fuzzy and PID parameters
ept unchanged for all other tests performed in this paper.Ke 1.15Ke 0.38Ku 65.58
Kp 2.60Ki 1.00Kd 2.95
of FLPSS parameters
aper, a genetic algorithm program has been used tozzy stabilizers and PID stabilizer for comparisonhe aim of the proposed parameter-tuning algo-change the stabilizers gains in an intelligent an
anner to achieve a desired system response. Therithm attempts to maximize the fitness function
ction = G 11 + Je
the number of data acquired during simulation, the normalization gain = 100.ole procedure of applying GAs to determine theof PSSs is summarized as [11,12]:
ze population using random selection method;h individual string, compute its fitness function;whether the stopping criterion is met. If yes, go(v), otherwise, continue to step (iv);e new population using reproduction, crossovertation, then go back to step (ii) andine the most fit parameters.
PSS developed in this paper has two inputs andtput, which implies two input parameters Ke ande output parameter Ku. The PIDPSS has threeKp, Ki and Kd.meter-tuning algorithm was applied to the power
del for both stabilizers at the following operating
that tthe eoscillizerswhicfor thare kpower (P) = 0.5pu, power factor = 0.95u step decrease in the mechanical power is ap-arameters of both stabilizers after tuning are listed
and the fitness functions for fuzzy and PID stabi-hown in Figs. 3 and 4, respectively.hows the speed deviation response for the sys-e fuzzy, with PID, and without a stabilizer, underp decrease in torque at P= 0.5 pu. It can be seen Fig. 5. Response to 0.25 pu step decrease in torque at P= 0.5 pu.
12 S.A. Al-Osaimi et al. / Electric Power Systems Research 74 (2005) 915
Fig. 6. Powerin field voltag
Also, it is n+2 pu in all
The behious typesdifferent options.
The systious types oent operatinare to see tthe systemtwo stabiliz
The perfuated for thfactor laggapplied at tat time 3 s.stabilizerspower anglcal.
4.2. Step c
The perfated for thefactor laggiplied at timtime 3 s. Tstabilizersbetter perfo
. Power angle response of FLPSS and PIDPSS for a 0.15 pu loadcted at generator terminals, P= 0.25 pu.
Sudden connection of resistive load
e performances of the PIDPSS and the FLPSS is evalu-for the generator operating at P= 0.25 pu and 0.9 powerangle response of FLPSS and PIDPSS for a 0.05 pu step changee, P= 0.25 pu.
oted here that the field voltage is limited to 2 totests.
avior of the fuzzy and PID stabilizers under var-of disturbances and with the system operating aterating conditions is studied in the following sec-
em dynamic behavior is investigated through var-f disturbances applied to the system under differ-g conditions. The objectives of this investigation
he impact of the fuzzy and the PID stabilizers ondynamics performance as well as to compare theers responses.
nce voltage disturbance test
ormances of the PIDPSS and the FLPSS are eval-e generator operating atP= 0.25 pu and 0.9 powering. A 0.05 pu decrease in reference voltage was
Fig. 7.in torq
Thatedime 1 s and the system returns to the original value
The system response for both the fuzzy and PIDare shown in Fig. 6. This figure shows that thee responses for both stabilizers are almost identi-
hange in input power
ormances of the PIDPSS and the FLPSS are evalu-generator operating atP= 0.25 pu and 0.95 powerng. A 0.25 pu decrease in reference torque was ap-e 1 s and the system returns to original value at
he system response for both the fuzzy and PIDis shown in Fig. 7. In this case, the FLPSS givesrmance.
factor laggiat the genesponse forthe FLPSS
From thand PID stHowever, uappears to
Experimsystem avabased PSSangle response of FLPSS and PIDPSS for 0.25 pu step change0.9 pu.ng. A 0.15 pu resistive load is suddenly connectedrator terminals. The generator power angle re-
both stabilizers is shown in Fig. 8. It is visible thatslightly better the PIDPSS.e above results, One can conclude that the FLPSSabilizers under the light disturbance are similar.nder the large disturbance, the fuzzy stabilizer
provide better performance.
ental studies have been performed on the machineilable in the laboratory to test the fuzzy logic-. Using a personal computer (PC), the proposed
S.A. Al-Osaimi et al. / Electric Power Systems Research 74 (2005) 915 13
Fig. 9. Schematic diagram of the physical model.
FLPSS has been implemented on a physical model of a powersystem. This system consists of a 1 KVA, 380 V, three-phasealternator driven by a 1 KVA separately excited dc machineand is connected to an infinite bus (city of Riyadh utility bus)
through a transmission line. An overall schematic diagram ofthis physical model is shown in Fig. 9.
5.1. Speed measurement
In the proposed FLPSS, generator speed deviation andacceleration are used as the inputs. The speed signal can beobtained from a tacho generator, which is coupled with themachine shaft and generates an analog voltage signal (1 Vper 1000 rpm). For the speed deviation and its accelerationsignals, there is no direct transducer available on the physicalmodel of the power system. However, the speed signal is usedin the software program to generate a signal proportional tospeed deviation and its acceleration.
5.2. Data acquisition system (DAS)
The PC communicates with the outside environmentthrough a data acquisition card. This card is equippedwith sixteen inputs with 12 bit dynamic range and ana-logue interface channels . These channels have a
MATLFig. 10. Real time implementation of the fuzzy logic stabilizer usingFig. 11. Real time implementation of the PID stabilizer using MATLABAB/Real Time Windows Target toolbox.
/Real Time Windows Target toolbox.
14 S.A. Al-Osaimi et al. / Electric Power Systems Research 74 (2005) 915
Fig. 12. Response to 0.45 pu step decrease in torque with FLPPSS and with-out PSS, P= 0.9 pu.
Fig. 13. Response to 0.45 pu step decrease in torque with FLPPSS andPIDPSS, P= 0.9 pu.
Fig. 14. Response to 0.3 pu load connected at generator terminals withFLPPSS and without PSS, P= 0.3 pu.
Fig. 15. RespFLPPSS and
The MAis used to iprogram w
The anaceives the spasses throacceleratiosignals (the controltion limiternel of DASjunction.
For combuilt usingimplement
The permodel undto those of
5.4.1. StepThe dy
for loadingstep decreafor the FLPFigs. 12 an
From thto damp thserved in conse to 0.3 pu load connected at generator terminals withPIDPSS, P= 0.3 pu.
sampling rate with a maximum of 330 kHz.has, also, two output-range analogue interface
S Implementation Using MATLAB
TLAB/Real Time Windows Target toolbox ,mplement the FLPSS, as shown in Fig. 10. Thisorks as follows:logue to digital (A/D) input channel of DAS re-peed signal and samples it at a 1 ms rate. The signalugh a low-pass filter to eliminate noise. Then, then is calculated by using a derivative block. The two, ) are fed to the fuzzy logic controller. Then,
signal is determined and passed through a satura-. Finally, the digital to analog (D/A) output chan-sends the control signal to the exciter-summing
parison, a digital PID power system stabilizer wasthe same program. A block diagram for PIDPSSation is shown in Fig. 11.iments and test results