Vous êtes sur la page 1sur 10

1026

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 2, APRIL 2010

A New Approach for Allocation and Sizing of Multiple Active Power-Line Conditioners
Iman Ziari, Student Member, IEEE, and Alireza Jalilian
AbstractIn this paper, a particle swarm optimization (PSO) algorithm is developed for allocation and sizing of multiple active power-line conditioners (APLCs) in power systems. The utilized objective function comprises four factors as total harmonic distortion, total size of active power-line conditioners (APLCs), harmonic transmission-line loss, and motor load loss. To evaluate the capability of the proposed method, the IEEE 18-bus test system is employed and investigated in three different cases. These cases are based on the assumption of continuous/discrete and limited/unlimited size for APLCs, requiring the optimization method to solve discrete/nondiscrete nonlinear problems. Therefore, in addition to PSO, an integer nonlinear optimizer (discrete PSO called DPSO) algorithm is also developed. Simulation results are compared with results obtained by genetic algorithm as well as with nonlinear programming (discrete nonlinear programming). It is demonstrated that analytical methods enjoy higher accuracy in smooth objective functions while they are not accurate enough in practical situations where nonsmooth objective functions are involved. If improper initial solutions are provided, achieving optimized results cannot be guaranteed by using analytical approaches. It is also shown that heuristic algorithms due to their randomness characteristics are more useful in practical cases where a number of local minima are present. Simulation results conrmed the capability and effectiveness of the proposed PSO-based algorithm in the allocation and sizing of multiple APLCs in a test power system compared with analytical methods and other heuristic algorithms. Index TermsActive power-line conditioner (APLC), genetic algorithm (GA), harmonics, nonlinear programming, particle swarm optimization (PSO).

I. INTRODUCTION

N modern electrical distribution systems, there has been a remarkable growth in the use of nonlinear loads, such as rectiers, converters, adjustable speed drives, arc furnaces, computer power supplies, etc. Nonlinear loads act as current sources injecting harmonic currents into the power systems. These power-electronic-based loads have caused severely distorted voltage waveforms at the point of common coupling (PCC). Other linear loads connected at the same PCC will receive a distorted supply voltage, which may lead to various

Manuscript received September 30, 2008; revised May 08, 2009. First published December 28, 2009; current version published March 24, 2010. Paper no. TPWRD-00750-2008. I. Ziari is with the Department of Electrical Engineering, Iran University of Science and Technology (IUST), Narmak, Tehran 1684613114, Iran (e-mail: iman_ziari@ee.iust.ac.ir). A. Jalilian is with the Department of Electrical Engineering, Centre of Excellence for Power System Automation and Operation, Iran University of Science and Technology (IUST), Tehran 1684613114, Iran (e-mail: jalilian@iust.ac.ir). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TPWRD.2009.2036180

unwanted effects. The overheating of motors, transformers, cables, maloperation of some protection devices, and resonance with capacitors are some of these effects [1]. Conventionally, passive L-C lters are used to reduce harmonics; nevertheless, they have demerits of xed compensation, large size, and resonance. In order to overcome these problems, active power lters (APFs), which are also called APLCs, have been developed [2]. The basic compensation principles of APLCs were proposed in the 1970s. APLCs provide injected equal-but-opposite currents to the PCC that completely eliminate the nonsinusoidal requirements of the nonlinear loads. However, in practice, a more reasonable requirement is to reduce harmonic distortion to a minimum acceptable level for a given condition [3][5]. The rst APLC with power rating of 800 kVA was put into practical use for harmonic compensation in 1982. Later, in 1986, a combined system of an APLC of 900 kVA and a passive lter of 6600 kVA was practically installed with steel mill drives [6]. Currently, APLCs have attracted a lot of attention owing to their capabilities, particularly in the elimination of harmonics [7][9]. Despite a large number of benets that APLCs enjoy, their huge installation and operation costs prevent electrical engineers from employing these protable instruments greatly without any restriction at all buses in power systems. Owing to this fact, a variety of solution techniques has been utilized to solve the APLCs allocation and sizing (AAS) problem. These techniques are mainly categorized into two groups: 1) allocation of only one APLC and 2) allocation of multiple APLCs. References [10][12] introduced the initial steps toward solving the AAS problem; nonetheless, using only one APLC may not guarantee satisfaction of the harmonic limits at all buses, in general, and when many nonlinear loads are available in a power system in particular. References [13][15] are based on allocating and sizing of multiple APLCs using an analytical optimization algorithm called nonlinear mixed integer programming. Although the analytical optimization algorithms do not suffer from a time-consuming problem, the local minimum is their main drawback. The need for the initial solution and difculty in differentiation from various types of nonlinear objective functions is among other disadvantages of these methods. It should be noticed that an incorrect selection of initial values leads to inaccurate results when these methods are applied. As the essential part of the objective function, THD level and APLCs size have been considered in [11][14], [16]. Harmonic transmission-line loss (HTLL) and motor load loss (MLL), due to harmonics and telephone inuence factor (TIF), are included in the objective function in other studies [10], [15].

0885-8977/$26.00 2010 IEEE

ZIARI AND JALILIAN: NEW APPROACH FOR ALLOCATION AND SIZING OF MULTIPLE APLCS

1027

In [16], a genetic-based algorithm (GA) is presented to solve the AAS problem. Compared to the analytical algorithm, GA as a heuristic algorithm enjoys the luxury of simple concept and easy implementation. It also does not need an initial solution and differentiation from the nonlinear objective functions. Nevertheless, time consuming, local minimum, and the need to adjust several parameters and lack of memory are still among substantial imperfections. Particle swarm optimization (PSO), as another heuristicbased approach, enjoys all plus characteristics of heuristic algorithms [17]. Compared to the other heuristic algorithms, such as GA, the PSO method, to some extent, is not time consuming, does not suffer severely from local minimum, needs to adjust only a few parameters, and has memory. These advantages lead PSO to be applied to different areas of electric power systems, such as economic dispatch, reactive power control, power losses reduction, optimal power ow, power system control design, and unit commitment [17][20]. In this paper, a PSO-based algorithm is proposed for allocation and sizing of multiple APLCs in a standard IEEE test system. The objective function consists of four different factors as: 1) THD, 2) total APLC current, 3) MLL, and 4) HTLL. Furthermore, APLCs with continuous/discrete and limited/unlimited sizes are considered in this investigation. Simulation results are nally compared with the results obtained by another heuristic method (GA in this case) and a conventional analytical optimization technique to validate the capability of the proposed method. II. PROBLEM FORMULATION In this section, rst of all, the APLC model assumed in this paper is presented; then, the AAS problem and the relative objective function and constraints are formulated. A. APLC Model The common model for APLC employed in almost all references in the eld of AAS [10][16] is a current source injecting harmonics to the PCC in a power system. In order to assume the APLC current with certain amplitude and phase angle, the phasor model of (1) is used (1) where APLC current at bus for harmonic order ;

load losses (MLL), and to minimize the cost of APLCs. The constraints also maintain individual and total harmonic distortions within a standard level. The mentioned parts will be lumped into the objective function which is expressed as follows: (2) where weight coefcient for THD; weight coefcient for MLL; weight coefcient for HTLL; weight coefcient for total APLCs current. As seen, the given objective function is composed of four parts: ; 2) ; 3) ; and 4) . 1) , , and are the summation of THDs, MLLs and APLCs current magnitude at all buses, respectively. is the summation of transmission line losses at all harmonic orders. These factors are formulated as (3)

(4) where number of buses; maximum considered harmonic order; voltage at bus for harmonic order ; THD at bus . can also be formulated as follows [10]: (5)

(6) where is the MLL at bus . The formulation indicates that lower order harmonics affect electrical motors more than higher order harmonics. can be obtained as [15], as On the other hand, and shown in (7) at the bottom of the next page, where are line resistance and impedance between buses and for harmonic order h, respectively. nally can be determined by using (8) as (8) where is the amplitude of APLC current at bus for harmonic order . The rest of the mathematical formulation will be described in Section III.

real part of APLC current at bus for harmonic order ; imaginary part of APLC current at bus for harmonic order . The indices and represent the real and imaginary parts of the APLC current, respectively. B. Objective Function The most important objectives of placement and sizing of APLCs in a power system are to reduce total harmonic distortion (THD), harmonic transmission line losses (HTLL), motor

1028

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 2, APRIL 2010

III. IMPLEMENTATION OF PSO FOR AAS PROBLEM A. Overview of PSO PSO is a population-based and self-adaptive technique introduced originally by Kennedy and Eberhart in 1995 [19]. This stochastic-based algorithm handles a population of individuals in parallel to probe capable areas of a multidimensional space where the optimal solution is searched. The individuals are called particles and the population is called a swarm. Each particle in the swarm moves toward the optimal point with adaptive velocity [20], [21]. Mathematically, the position of particle in an -dimensional . The vevector is represented as locity of this particle is also an -dimensional vector as . Alternatively, the best position related to the lowest value of the objective function for each particle is and represented as the global best position among all particles or best pbest is de. During noted as the iteration procedure, the velocity and position of particles are updated [22]. B. Implementation of the PSO Algorithm It should be noticed that the APLC current at each bus for each harmonic is regarded as the position of a particle during the optimization process. The proposed PSO-based algorithm as an APLC location and size optimizer is illustrated in Fig. 1. The description along with the required comments on the proposed algorithm steps will be given. The coding is written in Matlab 7.4 programming language and executed by using a Pentium IV, 1.8 GHz, with 2-GB random-access memory (RAM) processor. 1) Step 1: (Input System Data and Initialize): In this step, the power system conguration and data, nonlinear loads loca, and constraints, such as maximum-allowed tion, current individual harmonic distortion and THD, and APLCs size are specied. The number of population members and iterations are set. (consisting of real and imaginary The population of particles parts of APLC current at all buses for all harmonic orders) as in the search space are initialized in well as their velocity this step. Vectors X and V are described as shown in the equation at the bottom of the next page. The PSO weight factors are also set in this step. 2) Step 2: (Calculate the Objective Function): For this purpose, APLCs current and bus voltages should be available. As mentioned, APLCs current at each bus for each harmonic is considered as the position of a particle which is initialized in Step 1.

Fig. 1. Algorithm of proposed PSO-based approach for the AAS problem.

The bus voltages can be attained as (9) and are the bus voltages and injecting current where vectors for harmonic order , respectively. The corresponding bus admittance matrix for harmonic order is also presented . Considering an APLC located at each bus, the bus inby jecting current is implied as (10) and are the APLC and nonlinear load current where vectors for harmonic order h, respectively. After calculation of the objective function by using (2), the best position related to each member of particles can be evaluated as described in the next step. 3) Step 3: (Calculate Pbest): The objective function related to each particle in the population in the current iteration is compared with it in the previous iteration and the position of the par-

(7)

ZIARI AND JALILIAN: NEW APPROACH FOR ALLOCATION AND SIZING OF MULTIPLE APLCS

1029

ticle enjoying a lower objective function as Pbest for the current iteration is recorded if if (11)

where is the number of iterations, and is the objective function evaluated for particle . 4) Step 4: (Calculate Gbest): In this step, the best objective function associated with the Pbests among all particles in the current iteration is compared with it in the previous iteration and the lower value is chosen as the current overall Gbest if if (12)

5) Step 5: (Update Velocity): After calculation of Pbest and Gbest, the velocity of particles for the next iteration should be modied by using
Fig. 2. Single-line diagram of the 18-bus IEEE distorted system [6], [10][16].

(13) where velocity of particle j at iteration k; inertia weight factor; acceleration coefcients; position of particle j at iteration k; best position of particle j at iteration k; best position among all particles at iteration k. In the velocity updating process, as inertia weight, and and as acceleration coefcients should be determined in advance. The acceleration coefcients are two different random and are dened as follows: values in the interval (14) where initial inertia weight factor; nal inertia weight factor; current iteration number; maximum iteration number. 6) Step 6: (Update Position): The position of each particle is modied as at the next iteration (15) 7) Step 7: (Check Convergence Criterion): If or , the program is terminated and the results are printed; otherwise, the programs goes to Step 2. IV. SIMULATION RESULTS To assess the applicability of the proposed approach, the 18-bus IEEE power system is considered. The system contains several linear loads and three nonlinear loads located at buses 7, 24, and 25, where the 5th, 7th, 11th, 13th, 17th, 19th, 23rd, and 25th harmonics are produced. The single-line diagram for the testing system is depicted in Fig. 2. The detailed specication of this system is given in [10]. In this power system, three identical harmonic current sources are employed as nonlinear loads. The harmonic contents of the employed harmonic current sources (the nonlinear loads) are shown in Fig. 3. Voltage distortions for all harmonic orders and at all buses as well as THD and MLL at all buses can be calculated by using the admittance matrix for all harmonic orders and

1030

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 2, APRIL 2010

TABLE II SITUATION OF BUSES REGARDING THE AVAILABILITY OF APLC

Fig. 3. Harmonic contents of the nonlinear loads in the test system.

TABLE I VOLTAGE CONDITIONS AT DIFFERENT BUSES IN THE ABSENCE OF APLCS

TABLE III TOTAL APLCS CURRENT CALCULATED IN CASE 1

the harmonic contents of nonlinear loads shown in Fig. 3 [using (9) and (10)]. It should be noted that because no APLC is inin (10) is considered a zero matrix. stalled, APLCs current Table I illustrates the obtained simulation results in no APLC state. It can be found from Table I that the system is badly distorted by nonlinear loads. As shown in Table I, the resulting THD at all buses is, on average, 10.25%, which represents a rather unallowable harmonic distortion level in reference to the IEEE standard limit of 5%. Using (5) and (7), the average MLL and HTLL in no APLCs state were calculated to be 0.0035 and 0.0394 p.u., respectively. The proposed AAS problem is evaluated in three different cases. In case 1, no constraints are applied to the APLCs size. The rms value of the APLCs current is limited to 0.05 p.u. in the second case. At last, the employed APLCs size is assumed to be integer multiples of 0.01 p.u.; plus, the size of APLCs is limited to 0.07 p.u. in case 3. Furthermore, the investigations are performed in six states as represented in Table II. In Table II, the signs and indicate the availability and unavailability of APLC in the relative bus, respectively. Therefore, state 1 considers only buses 7, 24, and 25 as the only candidate buses for installing APLCs. Buses 7, 23, 24, 25, and 26 are also the only candidate buses in state 2. The rest of the states can be simply recognized by using the signs marked by in Table II. As a well-established and acceptable optimization method, the genetic algorithm (GA) is used to evaluate and compare the

results. Two GA algorithms as GA1 and GA2 are utilized to solve AAS problem in this application. Population size and generation number are set in GA1 so that the computation time of GA-based optimization process would be equal to the PSO one Population size Generation number . In GA2, the population size and generation number are selected exactly the same as those used in the proposed PSO Population size Generation number . The implementation time of GA2 to solve the AAS problem of the test power system is about four times that of PSO and GA1. The GAs output is presented in two last rows of the relative tables. Furthermore, to illustrate the priority of PSO over analytical methods in problems with many local minimums, a comparison with the nonlinear programming (NLP and DNLP)-based algorithm as an analytical method is implemented. This program is written in GAMS software [23] and the simulation results are given in the last row of the following tables under the NLP heading. A. Case 1 In this case, no current constraints are applied to the APLCs size. Tables IIIVII demonstrate the size of APLCs, THD, MLL, and HTLL indices obtained from the simulation of the system in six states corresponding to 3, 5, 7, 10, 13, and 16 APLCs, respectively. Among the aforementioned states, the rst and the sixth states correspond to the least and the most variables for the PSO algoand variables), rerithm (i.e., spectively. Numbers 2 and 8 indicate the real and imaginary parts of the APLCs current and harmonic orders, respectively.

ZIARI AND JALILIAN: NEW APPROACH FOR ALLOCATION AND SIZING OF MULTIPLE APLCS

1031

TABLE IV AVERAGE THD CALCULATED IN CASE 1

TABLE V AVERAGE MLL CALCULATED IN CASE 1

Fig. 4. Convergence characteristics for the states in case 1.

TABLE VI AVERAGE HTLL CALCULATED IN CASE 1

TABLE VII OBJECTIVE FUNCTION CALCULATED IN CASE 1

Numbers 16 and 3, on the other hand, are the number of candidate buses for the installation of APLCs. The population number is about twice and ten times more than the variables in state 6 and state 1, respectively. Therefore, the number of variables exactly impresses the accuracy of the results as shown in the following tables and Fig. 4. States 5 and 6 suffer from the highest OF values compared to states 1 and 2 which enjoy the lowest OF values. As illustrated in Table IV, the maximum average THD among the states, 0.00639 (in state 6), is about 16 times lower than the average THD before installation of the APLCs, which is 0.1025. MLL and HTLL indices also decrease from 0.0394 to 0.00181 and from 0.0035 to 0.000254, respectively, indicating the effectiveness of the PSO algorithm. The best output in all states is that the current of installed APLCs at candidate buses 7, 24, and 25 would be equal to the nonlinear loads current located in these buses and zero at the rest of the candidate buses. The small discrepancy observed among the total APLCs current in different states given in Table III and the nonlinear load currents given in Fig. 3 is due to two points: 1) The nature of PSO which converges to some extent prematurely and 2) THD which is not the only objective function in this study. This fact is vividly perceived in three APLC states.

0 in this state means that the APLCs current is equal to nonlinear loads current, 0.29154, which imposes a higher cost than the APLCs current obtained, which is 0.29066. In order to improve the accuracy of results, two different ways can be employed: 1) increasing the number of initial random inputs and 2) increasing the number of iterations. However, they aggravate the time-consuming problem as a remarkable difculty. Compared to PSO, the analytical method is expected to represent more accurate results since no constraint is applied to the size of APLCs as optimizing variables in case 1. Therefore, the objective function seems to be smooth. As a result, analytical methods can be more useful in this case so that the aforementioned tables clarify the convergence in OFs with smaller values. The trend of the objective function with respect to the number of iterations is shown in Fig. 4 during PSO implementation. This is to test the convergence of the objective function to determine the quickness of PSO in terms of iteration numbers. As observed, the best and the worst results belong to states 1 and 6, respectively, which is related to the number of variables. It is clear that the higher aggregation of APLCs should be around the nonlinear loads which can be seen in Table V. Quick convergence is a remarkable benet of the PSO-based AAS compared with other heuristic methods (Fig. 4). The convergence is accomplished in about 100 and 200 iterations in states 1 and 6, respectively. This fact is due to the number of particles: 500 in this test, which is about twice the number of variables in state 6 and 10 times in state 1. In comparison with the PSO algorithm, GA outcomes suffer from lower accuracy. This fact is more clearly recognized in higher states and more variables. As represented, THD, HTLL, MLL, and total size of the APLCs in the PSO-based approach are severely more efcient than GA1; however, the length of implementation in both methods is the same. For instance, the total APLCs size obtained by GA1 is almost 38% more than that obtained by GA1 for state 6. It is undeniable that the case with no current constraints for APLCs is impractical because the DC source in the APLC is implemented by capacitors and inductors which are discrete devices. However, this case is assessed only for showing the capability of the proposed algorithm.

1032

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 2, APRIL 2010

TABLE VIII TOTAL APLCS CURRENT CALCULATED IN CASE 2

TABLE IX AVERAGE THD CALCULATED IN CASE 2

Fig. 5. Convergence characteristics for states in case 2.

B. Case 2 The rms value of the APLCs current is restricted to 0.05 p.u. in this case. Tables VIII and IX illustrate the total size of APLCs and THD for different states, respectively. Despite the previous case, the obtained outcomes are relatively practically feasible owing to the assumption of discrete APLCs; however, considering 16 buses as candidate buses of APLCs is not commercially feasible. The proposed heuristic strategy is completely the same as the algorithm presented in Section III-B, but a change in vector as shown in (16) and (17) should be done at the beginning of step is 2 as shown in (17) at the bottom of the page, where the maximum current magnitude of APLCs, which is set to 0.05 p.u. in this test. As mentioned before, there are only three buses as candidates for installing APLCs in state 1 and the maximum current of each APLC is also limited to 0.05 p.u. Therefore, the total current of compensators is 0.15 p.u., which is much less than the total current of nonlinear loads, 0.2916. This fact causes voltage THD at buses 20, 21, 22, 23, 24, 25, and 26 in state 1 to not be in the standard limit of 5%, as demonstrated in Table VII. Although, states 4, 5, and 6 in this case show acceptable results in the average THD and total APLCs current, they are not comparable with state 2. State 2 devotes the most reasonable results so that it gets benet from almost the lowest APLCs current0.24 p.u.while the voltage THD standard range is also satised at 2.23%.

As a result, in addition to three APLCs state, the application of 16 APLCs is also not recommended in this case, because the APLCs current in this state is the highest one, more than 0.36 p.u. In comparison with the no-current constraint state, the results are converted to the feasible outcomes Although the GA2-based average THDs and MLLs as well as the total APLCs current or the total APLCs size are fairly close to the PSO-based results in states 1 and 2, they are almost severely higher in other states. As illustrated in the aforementioned tables, again in this case, when the objective function is still smooth and only a constraint is applied to the variables, nonlinear programming as an analytical method shows better results compared with PSO and GA; however, this case also suffers from inapplicability. Fig. 5 depicts a comparison among the trend of objective functions yielded from the aforementioned six states in terms of the iteration number. As seen in Fig. 5, the third and fourth states present almost similar behavior in convergence and show the superiority over other states because they converge in the lowest objective function values. The quickest convergence characteristics belong to states 1 and 2, but in the highest objective function value with respect to other states. C. Case 3 The APLCs size in this case study is assumed as integer multiples of 0.01 p.u. and is limited to 0.07 p.u. As a practical and feasible case, this case is described in detail. All of the aforementioned states are assessed in this case, and the results are nally compared.

if if if if (17) (16)

ZIARI AND JALILIAN: NEW APPROACH FOR ALLOCATION AND SIZING OF MULTIPLE APLCS

1033

TABLE X APLCS CURRENT INJECTING INTO BUSES FOR DIFFERENT STATES IN CASE 3

TABLE XI THD OF BUSES FOR DIFFERENT STATES IN CASE 3

For including the mentioned limits to the algorithm presented in this subsection, some changes should be applied to the vector at the beginning of step 2. The following procedure is executed in order to divide APLCs current into 0.01-p.u. steps: First, the APLC current is multiplied by 100 to make it a digit between 0 and 10. Then, the integer part of the obtained digit is kept, and divided by 100 to return it into the original value, between 0 and 0.10. Therefore (18)

TABLE XII MLL OF BUSES FOR DIFFERENT STATES IN CASE 3

(19)

where operator int shows rounding the variable up to the nearest integer. After that, for limiting the APLCs current to 0.07 p.u., (16) and (17) are used and MaxIF is set to 0.07. The constraints applied to the APLCs size make case 3 a practical and feasible case and make the optimizing problem a discrete/nonlinear programming with a nonsmooth objective function. As mentioned before, the main drawback of analytical methods is local minimum. This problem clearly occurs in the nonsmooth objective function in case 3. Table X demonstrates the simulation results indicating the size of APLCs installed in the candidate buses in all six states. To compare with the previous cases, THD and MLL are also represented in Tables XI and XII, respectively. As demonstrated in Table XI, the maximum THD in this case belongs to bus 24, which is 0.0353, 0.0161, 0.0143, 0.0167, 0.0143, and 0.0208 in states 1 to 6, respectively. As a result, it is expected that the maximum current to be injected by APLC in bus 24 (0.07 p.u. in all states is illustrated in Table X).

As seen, the obtained results in this case are more acceptable than previous ones owing to the fact that applying restriction and discreteness on APLCs size makes the problem more feasible and practical. As an illustration, only seven buses are selected to employ APLCs in states 6, while 16 buses are candidate. It is observed that the APLCs with very low current are eliminated when the discreteness criterion is applied compared with case 1 results. Therefore, the concept of allocation is vividly demonstrated, because only a few buses are selected for installing APLCs, while more choices are available.

1034

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 2, APRIL 2010

TABLE XIII AVERAGE HTLL CALCULATED IN CASE 3

TABLE XIV OBJECTIVE FUNCTION CALCULATED IN CASE 3

Fig. 6. Convergence characteristics for states in case 3.

V. CONCLUSION This restriction and discreteness also cause total APLCs current in all states in this case to be much less than these in case 1 and fairly less than these in case 2 except with state 2 (0.27 in case 3 compared with 0.24 p.u. in case 2). The reduction of step size of APLCs (0.01 p.u.) leads to making the problem closer to the continuous size; therefore, more buses will be selected as candidate buses for them. Since the total nonlinear loads current is constant, the total APLCs size is known to some extent. Accordingly, the reduction of size of APLCs may lead to employ more APLCs (case 1). It has been simply gured out that the aggregation of APLCs is around the nonlinear loads such as cases 1 and 2 (Tables V and VI). For instance, the highest APLCs current are allocated at buses 7 (0.05), 24 (0.07), and 25 (0.07) in state 6 as shown in Table X. The results in seven APLCs state are remarkably more efcient than 10, 13, and 16 APLCs states as represented in Table XIV. It should be noticed that if THD is considered only as a constraint instead of a part of the objective function, 3APLCs state is the best because only 0.2100-p.u. APLC current is needed while THD is 0.0204 (within the standard limit). The outcomes vividly show that in integer nonlinear optimization, DPSO leads to more operational results with respect to GA and discrete nonlinear programming (DNLP). This fact is illustrated specically in Table XIV so that a discrepancy of about 70% and 200% in DPSO-based OF as the main criteria for comparison exists with GA and DNLP in state 6 with a higher number of variables, respectively. This difference is severely distinguished in APLCs current so that the calculated total APLCs injected current in DNLP is eight times more than DPSO. Fig. 6 depicts the convergence characteristics of the objective function evaluated in case 3 for all six states with respect to the iteration number. Among the mentioned states, the 3 APLCs state obtains a benet from quick convergence so that it converges in only 30 iterations, but in the highest objective function value than others, the second and third states are in the second and third priorities, respectively. The best results are associated with the third state which converges in the lowest objective function. A PSO-based approach has been proposed in this paper for optimal allocation and sizing of multiple APLCs in power systems. In this optimization procedure, a comprehensive objective function is evaluated. The objective function is composed of four parts: 1) THD; 2) total APLCs current; 3) MLL; and 4) HTLL. The proposed algorithm has been successfully implemented to an 18-bus IEEE power system in which three nonlinear loads are located. The results were achieved in three different cases. Case 1 in which the size of APLCs is continuous and unlimited, case 2 in which the size of APLCs is continuous but limited to 0.05 p.u.; and case 3 in which the size is discrete by 0.01-p.u. steps and limited to 0.07 p.u. These cases were assessed in six different states depending on the number of candidate buses to install APLCs. To determine the capability of the proposed method, two GA-based algorithms (GA1 with the same computation time as PSO and GA2 with the same population size and iteration number as PSO, but with longer computation time) are developed. An analytical optimization approach based on nonlinear programming (NLP) is also implemented for comparison purposes. As expected, when no local minimum exists in the objective function, analytical methods provide the most accurate results as observed in cases 1 and 2. It is also demonstrated that when no constraint is applied to the problem (cases 1 and 2); PSO enjoys more accurate results compared to GA, but less accurate results compared with NLP. However, the main shortcoming of the analytical methods is the local minimum being highlighted when the problem is a nonsmooth function. The realistic AAS problem is a practical case with a nonsmooth function and a number of local minimums. In this type of function, analytical methods can optimize properly only when perfect initial solutions are provided. This, however, is not an easy task, and hence, to solve this problem, heuristic algorithms are highly recommended. In this paper, a discrete PSO algorithm (DPSO) is developed to overcome a discrete nonlinear problem, including discreteness and limited size for APLCs (case 3). It is demonstrated that the obtained results using DPSO are more accurate compared with the results obtained by the discrete nonlinear programming (DNLP) method and other heuristic approaches, such as GA.

ZIARI AND JALILIAN: NEW APPROACH FOR ALLOCATION AND SIZING OF MULTIPLE APLCS

1035

In terms of computation time, the PSO is more effective than GA as a well-known heuristic method so that a remarkable difference is observed between the PSO-based results and GA1 results. This difference is still remarkable when the GA2 algorithm is used for comparison. The results clearly state the mentioned plus properties, such as the applicability and capability of using the PSO for allocation and sizing of multiple APLCs in power systems.

REFERENCES
[1] J. Arrillaga and B. C. Smith, Power System Harmonic Analysis. New York: Wiley, 1997. [2] B. Singh and K. Al-Haddad, A review of active lters for power quality improvement, IEEE Trans. Ind. Electron., vol. 46, no. 5, pp. 960971, Oct. 1999. [3] H. Sasaki and T. Machida, A new method to eliminate AC harmonic currents by magnetic compensation-considerations on basic design, IEEE Trans. Power App. Syst., vol. PAS-90, no. 5, pp. 20092019, Sep. 1971. [4] G. H. Choe and M. H. Park, A new injection method for AC harmonic elimination by active power lter, IEEE Trans. Power Electron., vol. 35, pp. 141147, Feb. 1988. [5] G. H. Choe and M. H. Park, Analysis and control of active power lter with optimized injection, IEEE Trans. Power Electron., vol. 4, no. 4, pp. 427433, Oct. 1989. [6] H. Akagi, Trends in active power line conditioners, IEEE Trans. Power Electron., vol. 9, no. 3, pp. 263268, May 1994. [7] S. Mishra and C. N. Bhende, Bacterial foraging technique-based optimized active power lter for load compensation, IEEE Trans. Power Del., vol. 22, no. 1, pp. 457465, Jan. 2007. [8] I. Ziari and A. Jalilian, A new control strategy for an active power lter to compensate harmonics and reactive power, presented at the ICHQP, Cascais, Portugal, 2006. [9] I. Ziari, A. Kazemi, and A. Jalilian, Using active power lter based on new control strategy to compensate power quality, presented at the UPEC, Newcastle-upon-Tyne, U.K., 2006. [10] W. M. Grady and M. J. Samotyj, The application of network objective functions for actively minimizing the impact of voltage harmonics in power systems, IEEE Trans. Power Del., vol. 7, no. 3, pp. 13791386, Jul. 1992. [11] W. K. Chang and W. M. Grady, Meeting IEEE-519 harmonic voltage and voltage distortion constraints with an active power line conditioner, IEEE Trans. Power Del., vol. 9, no. 3, pp. 15311537, Jul. 1994. [12] W. M. Grady and M. J. Samotyj, Minimizing network harmonic voltage distortion with an active power line conditioner, IEEE Trans. Power Del., vol. 6, no. 4, pp. 16901697, Oct. 1991. [13] Y.-Y. Hong, Y.-L. Hsu, and Y.-T. Chen, Active power line conditioner planning using an enhanced optimal harmonic power ow method, Elsevier Trans. Elect. Power Energy Syst., pp. 181188, Feb. 1999.

[14] W. K. Chang and W. M. Grady, Minimizing harmonic voltage distortion with multiple current-constrained active power line conditioners, IEEE Trans. Power Del., vol. 12, no. 2, pp. 837843, Apr. 1997. [15] W. K. Chang and W. M. Grady, Controlling harmonic voltage and voltage distortion in a power system with multiple active power line conditioners, IEEE Trans. Power Del., vol. 10, no. 3, pp. 16701676, Jul. 1995. [16] R. Keypour and H. Sei, Genetic based algorithm for active power lter allocation and sizing, Elsevier Trans. Electrical Power Syst. Res., pp. 4149, 2004. [17] M. R. Alrashidi and M. E. El-Hawary, A survey of particle swarm optimization applications in electric power systems, IEEE Trans. Evol. Comput., to be published. [18] H. Bai and B. Zhao, A survey on application of swarm intelligence computation to electric power system, Proc. 6th World Congr. Intelligence Control and Automation, vol. 2, pp. 75877591, Jun. 2006. [19] J. Kennedy and R. Eberhart, Particle swarm optimization, in Proc. IEEE Int. Conf. Neural Networks, pp. 19421948. [20] J. Park and K. Lee, A particle swarm optimization for economic dispatch with non-smooth cost functions, IEEE Trans. Power Syst., vol. 20, no. 1, pp. 3442, Feb. 2005. [21] A. I. Selvakumar and K. Thanushkodi, A new particle swarm optimization solution to nonconvex economic dispatch problems, IEEE Trans. Power Syst., vol. 22, no. 2, pp. 4251, Feb. 2007. [22] K. E. Parsopoulos and M. N. Vrahatis, On the computation of all global minimizers through particle swarm optimization, IEEE Trans. Evol. Comput., vol. 8, no. 3, pp. 211224, Jun. 2004. [23] D. Chattopadhyay, Application of general algebraic modelling system to power system optimization, IEEE Trans. Power Syst., vol. 14, no. 1, pp. 1522, Feb. 1999. Iman Ziari (S09) was born in Abadan, Iran, on September 11, 1978. He received the B.Sc. degree from Sahand University of Technology (SUT), Tabriz, Iran, and the M.Sc. degree in electrical engineering from Iran University of Science and Technology (IUST), Tehran, Iran, where he is currently pursuing the Ph.D. degree. Currently, he is a Lecturer in the Electrical Engineering Department of Islamic Azad University (IAU)-Damavand branch, Tehran. His research topics include power system harmonics.

Alireza Jalilian was born in Yazd, Iran, in 1961. He received the B.Sc. degree in electrical engineering from Mazandaran University, Babol, Iran, in 1989, and the M.Sc. and Ph.D. degrees from the University of Wollongong, Wollongong, New South Wales, Australia, in 1992 and 1997, respectively. He joined the Power Group of the Department of Electrical Engineering at the Iran University of Science and Technology in 1998, where he is Assistant Professor. His research interests are power-quality causes and effects as well as mitigations.

Vous aimerez peut-être aussi