Genetic Algorithm Optimised Fuzzy Control of ... - IEEE Xplore

Australasian Universities Power Engineering Conference, AUPEC 2014, Curtin University, Perth, Australia, 28 September – 1 October 2014

1

Genetic Algorithm Optimised Fuzzy Control of DSTATCOM for Improving Power Quality Juan Shi

Amin Noshadi

College of Engineering and Science Victoria University Melbourne, Australia [email protected]


Akhtar Kalam

Peng Shi



Abstract—Proportional and Integral (PI) controllers are often used to control the operation of the Distribution Static Compensator (DSTATCOM) for the distribution power system. However, since the power system is highly nonlinear and subject to various disturbances, the PI controlled DSTATCOM cannot provide optimal performance for different operating points. More robust controllers such as the ones based on fuzzy logic approach are required for the DSTATCOM to provide adequate dynamic voltage control and to improve power quality and stability of the distribution power system. This paper presents the design of Fuzzy Logic Controllers (FLC) of a DSTATCOM for improving power quality for a distribution power system. The scaling factors of the designed FLC are optimised using Genetic Algorithms (GA). The GA optimised FLC for the DSTATCOM used for improving power quality and dynamic performance of a distribution power system is compared with a DSTATCOM controlled by PI controllers. The performances of the DSTATCOM controllers are evaluated during grid-side voltage and large load variations of the distribution power system. Simulation results show that the DSTATCOM with the GA optimised FLC can improve the power quality and dynamic performance of distribution power system and provide better system dynamic performance than the DSTATCOM with the conventional PI controllers. Index Terms--Distribution Static Compensator (DSTATCOM), Distribution power system, Fuzzy Logic Controller (FLC), Genetic Algorithm (GA), Proportional and Integral (PI) Control.

I.

INTRODUCTION

Distribution power system has poor power quality and dynamic performance due to insufficient reactive power support during disturbances. Low voltage poor power quality can be caused by the demand in reactive power as it loads up the supply system unnecessary. This can also be due to harmonic pollution and load imbalance as these

cause extra stress on the networks and excessive voltage imbalance causing stress on other loads connected to the same network [1]. Flexible AC Transmission Systems (FACTS) devices such as Static Synchronous Compensator (STATCOM) can address the power quality issues related to transmission lines while DSTATCOM can improve the power quality and dynamic performance in a distribution network [2]. A DSTATCOM is a shunt connected bidirectional converter based device which can provide adequate level of reactive power to improve the quality of electrical power featured as the voltage at the point of common coupling (PCC) in distribution network [3-5]. The control scheme is crucial to the operation of the DSTATCOM to produce the desired performance and to improve the power quality of the power system. Various control structure and algorithms for DSTATCOM converter such as phase shift control with PI controller, carrier based PWM control with PI controller, and carrier less hysteresis control with PI controller have been proposed to address the power quality issues [2]. In the work by Singh and Solanki [6], instantaneous reactive power theory, a synchronous reference frame theory, and an Adaline-based algorithm have been compared for extracting reference current signals. Kora [7] presented FLC for three-phase DSTATCOM to compensate for AC and DC loads. However, the FLC is only proposed for controlling the dclink voltage based on the energy of a dc-link capacitor. In this paper, carrier based PWM control with FLCs are designed for controlling the DC voltage, AC voltage, and current regulators. In reference [8], a Mamdani type of fuzzy controller was employed whereas a Sugeno type of FLC is designed in this paper. Previous research results by the authors show that a Sugeno type of FLC for STATCOM has many advantages over the Mamdani type [9]. A Sugeno type of FLC can overcome the computational problem a complex system such as distribution power system encounters in both software


simulation and real time implementation. A Sugeno type FLC can be analysed and implemented more effectively than the Mamdani type. In addition, it is more convenient in mathematical analysis and in system analysis for a DSTATCOM equipped with a Sugeno type FLC as the membership functions for the output are singletons. After designing the fuzzy controller, fine tuning can be made in order to improve the performance of the controller. Tuning can be made either to the membership functions or to the scaling factors (note that it is common to use normalised inputs and outputs for fuzzy controller and hence scaling factors are required to normalise these inputs and outputs). However, as the rule-base conveys a general control policy, it is preferred to keep the rule-based unchanged and the tuning exercise is focused on the scaling factors. Genetic Algorithm (GA) has been utilised to optimise the FLC scaling factors in order to achieve the optimum trade-off between the desired objective functions in this paper. This paper presents the design of GA optimised FLCs for a 3MVA DSTATCOM for improving the power quality and transient behaviour of a distribution network. Comparison study of PI controlled and fuzzy logic controlled DSTATCOM for improving the power quality and dynamic performance of a distribution power system is simulated using SimPowerSystem in MATLAB/Simulink environment. The performances of the DSTATCOM controllers are evaluated during grid-side voltage and large load variations.

2

In Fig. 3, PLL represents the phase-locked loop used to synchronise on the positive sequence component of the three phase (3Φ) primary voltage V1. The output of the PLL is θ =ωt and it is used to compute the direct-axis and quadrature-axis components of the AC (3Φ) voltages (Vd and Vq) and currents (Id and Iq). The DC measurement system in Fig. 3 provides the measurement of the DC voltage Vdc. The AC voltage measurement and current measurement systems in Fig. 3 measure the d and q components of AC positive-sequence voltage and currents to be controlled, respectively.

Figure 2. System configuration of a distribution network.

II. THE OPERATING PRINCIPLE OF LOAD COMPENSATION DSTATCOM AND SYSTEM CONFIGURATION Fig. 1 shows a schematic diagram for load compensation using an ideal shunt compensator like a DSTATCOM by inject current ic at the PCC to cancel the reactive, non-linear and unbalanced load current il [2], [7]. Fig. 2 shows a configuration of a network where DSTATCOM is used as a shunt compensator to regulate voltage on a 25-kV distribution power system [10]. 21 km and 2 km feeders are used to transmit power to loads at buses B2 and B3, respectively. A variable load producing continuously changing currents and voltage flicker is connected to bus B3 through a 25kV/600V transformer. The DSTATCOM uses Voltage Source Converter (VSC) to regulate voltage at PCC by absorbing or generating reactive power using power electronics to regulate three phase sinusoidal voltage at its terminal. The VSC uses forcedcommutated power electronic devices (GTOs, IGBTs or IGCTs) to synthesise the voltage on the secondary side of the coupling transformer from a DC voltage source [10], [11]. A DSTATCOM with VSC using PWM inverters has been used in this study. Fig. 3 depicts a single-line diagram of a DSTATCOM and the control system block diagram of DSTATCOM.

Figure 1. Schematic diagram of load compensation [2].

Figure 3. A single-line diagram of a DSTATCOM and its control system block diagram.

III.

FUZZY LOGIC CONTROLLER DESIGN FOR THE DSTATCOM

A fuzzy logic controller (FLC) consists of four elements. These are a fuzzification interface, a rule base, an inference mechanism, and a defuzzification interface [12]. A FLC has to be designed for the DC voltage regulator, the AC voltage regulator, and the current regulator. The design of the FLC for DC voltage regulator is described in detail first. The design of the fuzzy controllers for the AC and current regulators follows similar procedure. The PID-type FLC designed for DC voltage regulator has two inputs and one output. The error e (e = Vdcref -Vdc) and the rate change of error are the inputs and the output of the FLC is ΔId. ΔId is integrated to produce Idref. In order to reduce the number of rules that is used in regular three-input PID-type FLC, two Sugeno-type FLC shown in Fig. 4 are used in this


3

paper. In fact, the PID action is separaated into a PI-type FLC part and a PD-type FLC part. The outputs from these two FLCs are then added to produce a PIID-type FLC. As it can be seen from Fig. 4, four parameterrs need to be tuned using optimisation methods. Namely, GE E is the input error scaling factor, GCE is the input rate of o change of error scaling factor, GCU is the PI-type FL LC output scaling factor, and GU is the PD-type FLC ouutput scaling factor respectively.

Figure 5. Membership functiions of the input error (e).

Figure 4. Block diagram of the fuzzy logic based DC voltage regulator.

The linguistic variables for error e, the t rate of change of error and the controller output ΔI Δ d will take on the following linguistic values shown in Tablle I: TABLE I.

“NL” “NM” “NS” “ZO” “PS” “PM” “PL”

FUZZY LINGUISTIC VARIABLES

Negativee Large Negativee Medium Negativee Small Zero Positive Small Positive Medium Positive Large

Figure 6. Membership functions of thhe input the rate of change of error e t.

The above linguistic quantification haas been used in this paper to specify a set of rules or a rule-bbase. The rules are formulated from practical experience. For F the FLC with two inputs and seven linguistic values foor each input, there are 72 = 49 possible rules with all coombination for the inputs. The tabular representation of thhe FLC rule base (with 49 rules) for the DC voltage reguulator is shown in Table II. TABLE II.

THE RULE TABLE WITH H 49 RULES

Figure 7. Membership functionns of the output of the FLC.

Figs. 8 and 9 illustrate the block diagram of the PIDtype fuzzy controllers as the AC voltage regulator and current regulator which has sim milar structure of the FLC DC voltage regulator. Again GE, GCE, G GU and GCU are the scaling factors for the inputs andd output respectively.

The size of inputs and output membbership functions is chosen to be seven. The membership funnctions of the input variables for both the PI-type FLC and thhe PD-type FLC to be employed are of the triangular type annd they are defined as shown in Figs. 5 and 6. The membersship function of the output variable for both the PI-type FLC C and the PD-type FLC are singletons as depicted in Fig. 7.

Figure 8. Block diagram of the PID-type fuzzy AC voltage regulator.


V.

Figure 9. Block diagram of the PID-type fuzzy current regulator.

IV.

TUNING OF THE SCALING FACTORS OF THE FLC USING GENETIC ALGORITHM (GA)

The Genetic Algorithm (GA) is a global search technique based on the principles of genetics and natural selection developed by Holland [13]. In GA, search parameters are represented as genes and these parameters evolve under specified rules in order to find the optimum solution of the predefined cost function. GA is easy to implement and it is a robust and reliable method. A set of population or candidates for the solution is first initialised. The cost function is then evaluated in each generation using the population. The cost function test of these chromosomes determines which candidates are used to mutate and form a new set of population in the next generation. New chromosomes (candidates) are created by various operations such as crossover, mutation and reproduction. The basic structure of the GA consists of: coding, selection, mating (crossover) and mutation. In the first step of optimisation, the chromosomes are converted into code in which the GA can work with. The most common way of coding is binary coding. Each chromosome is represented in a binary form. Next, the fitness value of each chromosome is evaluated and the chromosomes with the best solution are used to produce the next generation using mutation and crossover. |

4

COMPARISON STUDY OF THE PERFORMANCE OF PI AND GA-OPTIMISED FLC FOR THE DSTATCOM

The power system used in the simulation study was shown in Fig. 2 in Section II. The system parameters are shown in Table III. The system dynamic responses to various grid-side voltage and load variations of the distribution network have been investigated. Genetic Algorithm (GA) was employed to find the optimal values of the scaling factors of the designed FLCs for the DC voltage regulator, the AC voltage regulator, and the current regulators of the DSTATCOM. The integral of the absolute magnitude of the error (IAE) between the reference DC voltage Vdcref and the measured DC voltage Vdc was chosen as the fitness function for the GA for optimising the scaling factors of the fuzzy DC voltage regulator. Similar fitness function was used for the fuzzy AC regulator and fuzzy current regulators. The convergences of the objective functions using GA for the FLCs as the DC voltage regulator, AC voltage regulator, and current regulators are depicted in Figs. 11 to 13. Optimal values of GCU, GU, GE, and GCE for each FLCs and shown in Table IV, respectively. The fine-tuned conventional PI controller gains provided in [10] are also provided in Table IV.

|

In recent years, GA has been used in many optimisation problems as it can start from any solution and it gradually converge to the optimal solution by improving the results at each generation. A flowchart of GA process is illustrated in Fig. 10.

Figure 11. Tuning of the fuzzy DC Voltage regulator parameters using GA.

Figure 12. Tuning of the fuzzy AC Voltage regulator parameters using GA.

Figure 10. Flowchart of Genetic Algorithm.


5

the system performance, except slight reduction in the peak overshoot as PD-type FLC only improves system transient performance.

Figure 13. Tuning of the fuzzy current regulator parameters using GA. TABLE III. PARAMETER SOURCE VOLTAGE DISTRIBUTION LINE VOLTAGE FREQUENCY FEEDER RESISTANCE R FEEDER INDUCTANCE L FEEDER CAPACITANCE C FIXED LOAD AT BUS 2 FIXED LOAD VARIABLE LOAD

DC LINK CAPACITOR DC VOLTAGE SET POINT TABLE IV.

SYSTEM PARAMETERS NUMERICAL VALUE 25 KV

25 KV 60 HZ 0.1153Ω/KM 1.048MH/KM 11.33NF/KM 3.007MVA WITH POWER FACTOR =0.998 1MVA WITH POWER FACTOR =1 NOMINAL: 1.8MVA WITH POWER FACTOR= 0.9 MODULATION: 1.2MVA WITH FREQUENCY OF 5HZ 10MF 2400 V

OPTIMAL SCALING FACTORS FOR THE FLCS AND PI CONTROLLER GAINS

PARAMETER DSTATCOM PI DC VOLTAGE REGULATOR GAINS DSTATCOM PI AC VOLTAGE REGULATOR GAINS DSTATCOM PI CURRENT REGULATOR GAINS FLC SCALING FACTORS FOR THE DC VOLTAGE REGULATOR FLC SCALING FACTORS FOR THE AC VOLTAGE REGULATOR FLC SCALING FACTORS FOR THE CURRENT REGULATOR

NUMERICAL VALUES KP=0.001; KI = 0.15 KP =0.55; KI = 2500 KP = 0.8; KI = 200 GE=0.00319; GCE =1.7778; GCU= 0.53559; GU=0.0011 GE=1.2229; GCE=4.1472; GCU= 2450; GU=0.6 GE=0.1223; GCE = 5.7638; GCU=1060; GU=0.25578

A. Dynamic Response of the Fuzzy Controlled DSTATCOM The dynamic response of a distributed power system subject to step changes in source voltage at the infinite bus is observed with the fuzzy controlled DSTATCOM and conventional PI controlled DSTATCOM. Fig. 14 shows the voltage at B3 when the source voltage has been changed by successively increasing the source voltage by 6%, decreasing it by 6% and bringing it back to its initial value at 0.2s, 0.3s, and 0.4s. It can be deduced from the results that the dynamic response of the DSTATCOM can be improved much more by utilising FLCs instead of conventional PI controllers. The advantage of using the FLCs with the proposed structure is that PI-type part of the FLC can be used separately without any additional adjustment required. It can also be inferred from Fig. 14 that the PD-type FLC does not have a significant effect on

Figure 14. Dynamic response of the fuzzy controlled DSTATCOM and conventional PI controlled DSTATCOM.

B. Mitigation of Voltage Flicker The mitigation of voltage flicker with the fuzzy logic controlled and conventional PI controlled DSTATCOM can be observed when the load is varied. Fig. 15 shows the voltage at B3 with the fuzzy controlled DSTATCOM and conventional PI controlled DSTATCOM. It has been observed that voltage at B3 varies between 0.96 pu and 1.04 pu (±4% variation) without DSTATCOM [14]. It is observed in Fig. 15 that the voltage fluctuation at bus B3 is reduced to ±0.7% with the PI controlled DSTATCOM. The voltage fluctuation at bus B3 is further reduced to ±0.6% with both the PI-type and PID-type fuzzy controlled DSTATCOM as indicated in the solid curve and dotted red curve shown in Fig. 15. Since PID-type and PI-type FLCs provide similar result, it is recommended that only PI-type FLC would be adequate for controlling the DSTATCOM in real-time implementation.

Figure 15. Mitigation of voltage flicker with the fuzzy controlled DSTATCOM and conventional PI controlled DSTATCOM.


VI.

CONCLUSIONS

This paper has presented the design of FLCs for a DSTATCOM to improve power quality and dynamic performance of a distribution power system. Genetic Algorithm (GA) was used to tune the scaling factors of the PID-type FLC to achieve optimal performance. The optimised FLC was then applied to control the DSTATCOM in the distribution power system. The results were compared with those of a conventional PI controlled DSTATCOM in the presence of source voltage variation and large load variations. The results show that the system’s performance was dramatically improved by using both the PID-type FLC and the PI-type FLC compared to the PI controller. The simulation results obtained in MATLAB/SimPowerSystems show that the DSTATCOM controlled by the GA optimised FLC provides better system dynamic response and hence improves power quality and transient behaviour for the distribution power system. VII. [1] [2]

[3]

[4] [5]

[6] [7]

[8]

[9]

[10] [11] [12] [13] [14]

REFERENCES

K. Schipman and F. Delince, “The Importance of Good Power Quality,” ABB Power Qual. Prod., Charleroi, Belgium, ABB Review, Apr., 2010. A. Banerji, S. K. Biswas, and B. Singh, “DSTATCOM Control Algorithms: A Review”, International Journal of Power Electronics and Drive System (IJPEDS), Vol.2, No.3, pp. 285-296, ISSN: 20888694, September 2012. G. F. Reed, M. Takeda, F. Ojima, A. P. Sidell, R. E. Chervus and C. K. Nebecker, “Application of a 5MVA, 4.16kV D-STATCOM System for Voltage Flicker Compensation at Seattle Iron & Metals,” IEEE PES SM, pp. 1605-1611, 2000. A. Ghosh and G. Ledwich, “Load Compensating DSTATCOM in Weak AC Systems,” IEEE Trans. on Power Delivery, vol. 18, no.4, October 2003. S. Kincic and A. Chandra, “Distribution Level STATCOM (DSTATCOMs) for Load Voltage Support,” IEEE Proceedings of Power Engineering Conference on Large Engineering Systems, pp 30-37, 2003. B. Singh, and J. Solanki, “A Comparison of Control Algorithms for DSTATCOM”, IEEE Trans. on Industrial Electronics, Vol. 56, NO. 7, pp. 2738-2745, July 2009. K. K. Kora, “A Fuzzy Logic DC-link Voltage Controller for ThreePhase DSTATCOM to Compensate AC and DC Loads”, International Journal of Scientific & Engineering Research, Vol. 2, Issue 10, ISSN 2229-5518, pp. 1-13, October 2011. D. Prasad, T. S. Kumar, B.V. Prasanth, and K.S.G.Sankar, “Fuzzy Logic Control of D-Statcom for Power Quality Improvement”, Int. Journal of Engineering Research and Applications, ISSN: 22489622, Vol. 3, Issue 6, pp.398-403, November-December 2013. J. Shi, A. Kalam, and P. Shi, “Improving Power Quality and Stability of Wind Energy Conversion System with Fuzzy Controlled STATCOM”, paper accepted and will be published in the Australian Journal of Electrical and Electronic Engineering, 2014. The MathWorks, SimPowerSystems User’s Guide, MatlabR2013b. N. G. Hingorani, L. Gyugyi, “Understanding FACTS; Concepts and Technology of Flexible AC Transmission Systems,” IEEE Press book, 2000. K.M. Passino and S. Yurkovich, Fuzzy Control. Addison-Wesley Longman, Inc. 1995. J. H. Holland, "Genetic algorithms." Scientific American 267.1, pp. 66-72, 1992. J. Shi, A. Noshadi, A. Kalam, P. Shi, “Fuzzy Logic Control of DSTATCOM for Power Quality Improvement and Dynamic Performance”, To be published in the proceedings of Saudi Arabia Smart Grid 2014, 16-19 November 2014, Saudi Arabia, 2014.

6