Security-Constrained Economic Scheduling of Generation Considering Generator Constraints Zwe-Lee Gaing, Member, IEEE Abstract-- This paper proposes an efficient constriction particle swarm optimization (CPSO) with mutation mechanism for solving the steady-state economic dispatch (ED) problem with contingency constraints and operating limits of series FACTS devices in power systems. The objective of security-constrained economic dispatch (SCED) is defined to not only minimize total generation cost but also to enhance transmission security, reduce transmission loss, and improve the bus voltage profile under pre-contingent and post-contingent states. Many non-linear characteristics of the generator, such as ramp rate limits and valve-point loading effects are considered using the proposed method for practical generator operation. The effectiveness of the proposed method is demonstrated for the IEEE 118-bus system with series FACTS devices, and it is compared with the other stochastic optimization methods in terms of solution quality and convergence rate. The experimental results show that the proposed CPSO method was indeed capable of efficiently obtaining higher quality solutions in SCED problems. Rung-Fang Chang Previous efforts in solving traditional ED problems have employed various mathematical programming methods and optimization techniques. In conventional numerical methods for the solution of ED problems, such as gradient-based, linear-programming, interior-point methods, an essential assumption is that the incremental cost curves of the units are monotonically increasing piecewise-linear functions. Unfortunately, this assumption may render these methods infeasible because of its non-linear characteristics in practical systems. These non-linear characteristics of a generator include discontinuous prohibited zones, ramp rate limits, valve-point loading effect, whose cost functions are not smooth or convex [12]-[14]. Recently, many global optimization techniques known as genetic algorithms (GA), simulated annealing (SA), evolutionary programming (EP), and particle swarm optimization (PSO), has been successfully used to solve the variant ED problems [2][12]-[17]. However, the objective function is usually not integrated with the security constraints or operating limits of FACTS devices, thus resulting in the solution perhaps being unsuitable for practical operations. In addition, premature convergence may result in the local optima by obtaining [18]-[22]. In this paper, an efficient constriction PSO (CPSO) method with mutation mechanism for solving the steady-state ED problem with security constraints and operating limits of series FACTS devices is proposed. The objective of SCED is defined to not only minimize total generation cost but also to enhance transmission security, reduce transmission loss, and improve the bus voltage profile under pre-contingency and post-contingency states. Many non-linear characteristics of the generator, such as ramp rate limits and valve-point loading effects are considered using the proposed method for practical generator operation. The effectiveness of the proposed method is demonstrated for the IEEE 118-bus system with series FACTS devices, and it is compared with the other stochastic optimization methods in terms of solution quality and convergence rate. Index terms-- economic dispatch, contingency analysis, ramp-rate limit, flexible ac transmission systems (FACTS), particle swarm optimization I. and INTRODUCTION C urrently, the concept of performing the optimal plan of power system operation with considering system security assessment is positively presented to ensure the system can secure operation without interruption to customer service even though the system suffered the contingency impact [1]-[8]. For the reason, installing the suitable FACTS (Flexible AC Transmission Systems) devices at key locations to increase the power-transfer capability of transmission system and keep power-flow over designed routes has been developed actively [9]-[11]. Therefore, to perform the optimal economic dispatch (ED) scheduling, the control variables should include the power output of generators and the parameters setting of FACTS devices. In addition, the pre-protection strategy of system and the security constraints also should be taken into consideration for enhancing the system security. The constraints included the transmission thermal limit and the bus voltage limit, to expect an economy-security operation model, regardless of whether the system is in a normal operation state or a contingent state. Because the contingency constraints are a fundamental element of economy-security control, therefore, the security and optimality of system operation should be treated simultaneously for a power system economy-security control, thus would add to the complexity of system operation [3]-[8]. II. PROBLEM DESCRIPTION A. Security-Constrained ED The steady-state SCED formulation can be stated as: Min f (u ( 0 ) , x ( 0 ) ) (1) u ,x g ( k ) (u ( k ) , x ( 0 ) ) 0 , s.t. for k 0,1,..., N c (2) h (u , x ) 0 , for k 0,1,..., N c (3) where is x a vector of control variables of pre-contingency, such as generation of generator ( Pg ), while u is the vector (k ) (k ) (0) (o) ----------------------------------------------Z. L. Gaing and R. F. Chang are with the Department of Electrical Engineering of Kao Yuan University, Kaohsiung, Taiwan 821 (e-mail: zlgaing@ms39.hinet.net, Fax: 886-7-6077009). of state variables, such as bus voltage ( V ) and reactive 1 TCSC device can permit to modify the reactance X L of transmission line. The models of TCPS and TCSC are shown in Fig. 1. power output of generator ( Q g ) . Objective function (1) is scalar. Equalities (2) are the conventional power equations. Inequalities (3) are the limits on the control variables x, and the operating limits on the power system. The superscript “ o”represents the pre-contingency (base-case) state being optimized, and superscript “ k” (k>0) represents the post-contingency states for the Nc contingency cases. So, the equality constraints g(o) change to g(k) to reflect the outage equipment. Performing the SCED scheduling, if a vector of control variables x of pre-contingency while satisfying all equality and inequality constraints during whole operation period, regardless of whether the system is in a pre-contingent state or a post-contingent state, is a feasible operation point that can ensure the system to locate in the security region. In addition, if the x within the security region can drive the minimization of objective function f, it is the optimal operation point that is the pursuing goal of this paper. 1 : 1ij Vi jXc where Pgi is the current output power, and P RL+ jXL Vi Vj (b) TCSC Fig. 1 Models of TCPS and TCSC D. Objective Function As mentioned previously, the objective of SCED is to consider simultaneously the pre-protection strategy of system, security constraints and operating limits of FACTS devices. The control variables x ( 0 ) must be solved subject to both the pre-contingency constraints ( u ( 0 ) , g ( 0 ) , h ( 0 ) ) and the post-contingency constraints ( u ( k ) , g ( k ) , h ( k ) ) of the selected contingency events. Hence, the SCED is expressed as a non-convex programming problem. Min f ( x ( 0 ) ) (7) s.t. i) power balance NB Y Pi ( k ) (k ) ij Vi ( k ) V j( k ) cos(i( k ) j( k ) ij( k ) ) 0 , (8) j 1 NB Y Qi( k ) (k ) ij Vi ( k ) V j( k ) sin(i( k ) j( k ) ij( k ) ) 0 , j N B j 1 (9) ii) unit operation constraints max( Pgimin , Pgi( 0 ) DRi ) Pgi( 0 ) min( Pgimax , Pgi( 0 ) URi ) (10) (5) ( 0) gi Vj (a) TCPS B. Operation Constraints of Generator The unit generation output is usually assumed to be able to be adjusted smoothly and instantaneously. Practically, the operating range of all on-line units is restricted by their ramp rate limits for forcing the units operation continually between two adjacent specific operation periods [9]-[10][13]. In addition, the valve-point loading effect in the input-output curve of unit is due to steam valve operation or vibration in a shaft bearing. Hence, the two constraints of the unit operation must be taken into account to achieve true economic operation. ˙Ramp Rate Limit According to [12] and [14], the inequality constraints due to ramp rate limits for unit generation changes are given: 1) as generation increases Pgi Pgi( 0 ) URi (4) 2) as generation decreases Pgi( 0 ) Pgi DRi RL+ jXL is the previous output power. URi is the up ramp limit of the i-th generator (MW/time-period); and DRi is the down ramp limit of the i-th generator (MW/time period). Qgimin Qgi( k ) Qgimax , i N g iii) security constraints V jmin V j( k ) V jmax , j N B (11) , m N l (13) S (k ) Lm S max Lm iv) operating limits of FACTS devices imin i( 0 ) imax , i N TCPS ˙Valve-Point Loading Effect The valve-point loading effect of thermal units should be taken into consideration where the fuel cost function of unit i is as follows. Fi ai bi Pgi ci Pgi2 d i sin(ei ( Pgimin Pgi )) (6) X min cj X ( 0) cj X max cj , j N TCSC (12) (14) (15) where the power flow equations (8)-(9) are used as equality constraints; the active and reactive power generation limits (10)-(11), bus voltage limits (12), thermal capacity limits of transmission lines (13) and operating limits of FACTS devices (14)-(15) are used as inequality constraints. where ai, bi, ci, di, and ei are the cost coefficients of unit i. C. Function of Series FACTS Devices According to their steady-state characteristics of variant FACTS devices, the function of series controllers, such as TCPS ( Thyristor-controlled phase shifter) and TCSC( Thyristor-controlled series capacitor), is mainly used to control the power flow of the lines close to their thermal limits. The TCPS device can to adjust the phase-angle to control the active power flow of transmission line. The III. CONSTRICTION PARTICLE SWARM OPTIMIZATION WITH MUTATION MECHANISM Let xi and vi denote the positions and the corresponding flight speed (velocity) of the particle i in a continuous search space, respectively. In a traditional PSO algorithm, the particles are manipulated according to the following 2 If rand () pm then equations [17]. vi( t 1) w vi(t ) c1r1i ( pbest ( t ) xi( t ) ) c2 r2i ( gbesti( t ) xi(t ) ) ) (16) xi( t 1) xi( t ) vi( t 1) (17) where t : pointer of iterations (generations), w : inertia weight factor, c1 , c2 : acceleration constant, r1i, r2i : uniform random value in the range [0,1], vi(t ) : velocity of particle i at iteration t, x i( t ) pbest xk xk (1 Gaussian()) . IV. : current position of particle i at iteration t, : the previous best position of particle xi at iteration t, gbest (t ) : the best position among all individuals in the population at iteration t, vi( t 1) : new velocity of particle i that is limited to a maximum velocity v max , DEVELOPMENT OF THE PROPOSED METHOD Before employing the CPSO method to solve the SCED problem, two definitions must be made as follows. A. Individual String In this paper, the individual is composed of continuous control variable (as power output of generator) and discrete control variable (as the value of FACTS parameters). The individual x (j0 ) is defined as follows: (t ) i predefined (21) where Gaussian() is a Gaussian distribution function, is set to be 0.1. x (j0 ) [ Pg(10 ) ,..., PgN( 0 ) , 1( 0 ) ,..., N( 0 ) , X c(10 ) ,..., X cN( 0 ) ] j (22) According to the power limits of the on-line units randomly establish the initial population x(0). t=1 i.e. vi( t 1) vimax , xi( t 1 ) : new position of particle i. One of the main drawbacks of the traditional PSO is its premature convergence, especially while handling problems with more local optima and heavier constraints. To overcome this disadvantage, the concept of constriction factor was suggested by [21] to gain both speed up convergence and escape local minima. The constriction PSO (CPSO) algorithm with a new scheme of velocity updating is as (18). The new scheme will replace the (16). And, in (18), the relationship between the parameters is given by (19) and (20). In (18), is the constriction factor. r r vi( t 1) (vi( t ) c1 1i ( pbest ( t ) xi( t ) ) c2 2i ( gbesti( t ) xi( t ) ) ) ri ri (18) where ri r1i r2i , (19) 2 , c1 c2 , 4 . (20) 2 2 4 Perform CPSO operations using (18)- (20), and (17) New offspring population Perform system steady-state contingency analysis using security (N-1) criterion No Fitness Satisfy all constraints (8)-(15)? Yes t=t+1 Evaluate the fitness using (23) Stopping rule is satisfied ? No Yes Select the best individual Because the global best individual attracts all members of the swarm in PSO, it is possible to lead the swarm away from a current location by mutating a single individual if the mutated individual becomes the new global best. This mutation mechanism potentially provides a means both of escaping local optima and speeding up the search. Therefore, to enhance the effectiveness of CPSO, a mutation operator of real-valued GA should be integrated with the CPSO [22]. The mutation process is employed as follows: Let the i-th individual xi [ x1 ,..., xk ,..., x N ]i , and the gene xk be selected for mutation according to the probability pm. The new gene xk can be obtained by (21) and PGj( 0 ) END Fig. 2 Operating procedures of the proposed CPSO-based SCED method B. Fitness Function In this paper, the objective of steady-state SCED is not only to minimize total generation cost but also to enhance transmission security, reduce transmission loss, and improve the bus voltage profile under normal or post-contingent states. The fitness function is as (23). If an individual x (j0 ) is a feasible solution and satisfies xk [ xkmin , xkmax ] . Then the next offspring of the application all constraints, its fitness will be measured by using the fitness function f as in (23). Otherwise, its fitness will be penalized with a very large positive constant (i.e. violates the equality constraints (8)-(9) or the control variable of the mutation operator is xi [ x1 ,..., xk ,..., x N ]i . 3 violates the inequality constraints (10)-(15)). The infeasible individual will not be selected by the proposed scheme for evolution in the next generation, so the proposed method can converge rapidly. FT ( x (j0 ) ) , x (j0 ) Security f ( x (j0 ) ) , x (j0 ) Un sec urity where i i (23) TABLE I SYSTEM STATUS UNDER NORMAL OPERATION AND POST-CONTINGENCY. NB Ng F ( x FT The proposed method was compared with traditional PSO and GA, in terms of solution quality and convergence rate using the same fitness function and individual definition. The software was written in Matlab language and executed on a Pentium IV 1.8 GHz personal computer with 512MB RAM. ( 0) j ) w L Ploss ( x (j0 ) ) ( wnV Vk ( x (j0 ) ) Vref ) k Study Case Load (MW) Case 1 3668.0 (100%) 3668.0 (100%) 4034.8 (110%) (24) Case 2 (25) Case 3 Ng w L Fi ( x (j0 ) ) / PDt i Normal operation (Pre-contingency) Line Line flow (Mva) L17-18 100.64 Post-contingency Line Outage L17-18 Overload L15-17, L15-19 L8-30 124.36 L8-30 L5-8, L23-24 L8-30 121.26 L8-30 L5-8 Ng wVn Fi ( x (j0 ) ) PDn ( x (j0 ) ) / PDt B. Selected Contingency Events From the result of contingency selection, three of the most critical faults are proven such as line 17-18 outage, line 8-30 outage, and a heavy load demand (110%) with line 8-30 outage, respectively. Table I shows that the status of the system under normal operation state (pre-contingency) and post-contingency state. In Case 1, when line 17-18 faulted, two lines were overloaded (line 15-17 and line 15-19). In Case 2, when line 8-30 faulted, two lines were overloaded (line 5-8 and line 23-24). In Case 3, when the heavy load (3668.0*110%=4034.8MW) with line 8-30 outage, the, one line was overloaded (line 5-8). In this paper, we employed these selected contingency events to test the performance of the proposed method. C. Parameters of Algorithm Through many experiments, the results revealed that the appropriate values for c1, c2, and Pm are 1.65, 2.45, and 0.01, respectively. They can yield an optimal evaluation value. Therefore, the following parameters of CPSO are used: • individual length= 41, • population size= 30, • c1=1.65, c2=2.45 • vimax Pgimax / 2 , (26) i wL is a weight factor, the purpose of which is mainly to transfer the transmission loss into a penalty cost. wnV is a weight factor of voltage deviation at bus n, the purpose of which is mainly to transfer the voltage deviation into a penalty cost. PDn is the load demand at bus n. PDt is the total load demand of the system. Vref is a magnitude of reference voltage, in general, Vref = 1.0 p.u. C. CPSO-based SCED Operating procedures of the proposed CPSO-based SCED method is shown in Fig. 2. For a power system with a higher X/R ratio of transmission line, the fast-decoupled load flow (FDLF) method has superior computation efficiency. To enhance the effectiveness of the proposed method, the FDLF method is employed to measure the fitness of the individual for the acceptable solution quality. The maximum number of iterations is set at 30, and the power mismatch accuracy is 0.001 p.u. in the FDLF method. V. NUMERICAL EXAMPLES AND RESULTS To verify the effectiveness of the proposed method, the IEEE 118-bus system with series FACTS devices was tested. The test system mainly contains 36 thermal units, 118 buses and 179 transmission lines. In addition, six series FACTS devices (three TCPS devices and three TCSC devices) are installed on it. The TCPS devices were installed on branches 24-72, 71-72 and 82-83. The TCSC devices were installed on branches 5-8, 23-24 and 92-100. The detailed characteristics of the 36 thermal units with the ramp-rate limit and valve-point loading effect are shown in Table AI. Bus 69 is the reference bus. The system has 41 control variables as the active outputs of 35 PV-bus units and six parameter values of FACTS devices. The limits of the installed TCPS are taken 5 0 50 and the limits of • mutation rate is Pm=0.01, • number of iterations=30. D. Results and Discussions First, to prove the effectiveness of the proposed CPSO-based SCED method is superior to traditional PSO and GA, the IEEE 118-bus system in normal operating state is used to test and verify. Figure 2 shows the convergence tendency of the average over 20 trials. The simulation results are summarized in Table II, satisfying the all constraints of the system. The optimal settings of the control variables that are obtained by the three proposed methods are shown in Table III. the installed TCSC are taken 0.8 X L X c 0.2 X L . The upper and lower limits of the PV-bus and load-bus voltages are 0.9 p.u. and 1.1 p.u., respectively. In a stable operation state, the load demand progressively increases from 3204.1 MW (at am 10:00) to 3668.0 MW (at am 11:00). The previous output power of unit Pgi( 0 ) is also shown in TABLE II COMPARISON BETWEEN THREE METHODS IN NORMAL OPERATING STATE Table AI in Appendix. 4 Method Ave. Fitness Min. Max. (best) (worst) CPSO PSO GA 48670 48680 48757 48373 48421 48555 48919 49028 49035 Ave. CPU time /generation (sec.) 26.31 26.02 27.78 4 loss of 73.99MW, and a summation of bus voltage deviation of 1.73 p.u. This is because the CPSO with mutation mechanism can avoid the premature convergence so that the average fitness obtained by the proposed CPSO method is superior to that obtained by the PSO and GA. In regard to the convergence rate, as seen in Fig. 2 and Table II, the CPSO method is also superior to the PSO and GA. This is also an evidence that the CPSO method can avoid the phenomenon of premature convergence. Figure 3 shows that the bus voltage profile obtained by the proposed CPSO method satisfies the bus voltage constraints (0.9 p.u.~1.1 p.u.) while the system is suffering the contingency impact. 118-bus system x 10 CPSO PSO GA 5 4.98 Fitness 4.96 4.94 4.92 4.9 4.88 4.86 0 5 10 15 20 Number of generations 25 30 TABLE IV COMPARISON OF VARIED CONTINGENCY EVENTS Fitness Ave. CPU time Selected /generation Min. Max. Contingency Ave. (sec.) (best) (worst) Events Fig. 2. Convergence tendency of the evaluation value TABLE III OPTIMAL SETTINGS OF CONTROL VARIABLES IN NORMAL OPERATING STATE USING THREE PROPOSED METHODS Normal Operating State Control variable Pg1 Pg4 Pg8 Pg10 Pg12 Pg24 Pg25 Pg26 Pg27 Pg31 Pg40 Pg42 Pg46 Pg49 Pg54 Pg59 Pg61 Pg65 Pg66 Pg69 Pg70 Pg72 Pg73 Pg80 Pg87 Pg89 Pg90 Pg91 Pg99 Pg100 Pg103 Pg107 Pg111 Pg112 Pg113 Pg116 2472 CPSO 25.00 68.95 282.85 68.95 67.77 253.78 54.25 68.95 68.95 56.29 250.66 68.95 51.12 69.87 197.00 162.51 68.95 223.96 140.00 265.26 17.87 29.12 76.00 158.85 25.00 140.00 224.09 25.00 77.80 68.95 51.12 68.95 25.00 25.00 68.95 140.00 -2.00 PSO 80.12 68.95 264.86 68.95 40.94 259.00 54.25 68.95 68.95 59.06 258.85 68.95 64.63 74.59 197.00 68.95 72.58 204.38 140.00 262.79 76.00 15.20 76.00 154.90 25.00 140.00 256.20 25.00 78.01 68.95 64.63 68.95 25.00 25.00 68.95 140.00 -5.00 7172 5.00 4.00 5.00 8283 -5.00 5.00 -5.00 Case 1 Case 2 Case 3 GA 28.71 68.95 295.29 72.27 79.36 257.28 54.25 68.95 78.63 54.25 257.15 68.95 27.72 90.09 197.00 125.86 123.65 140.00 140.00 257.62 44.60 26.16 76.00 140.00 25.00 140.00 239.84 25.99 68.49 68.95 27.72 68.95 77.42 25.00 68.95 140.00 -5.00 48684 48825 53705 48476 48557 53396 48960 49211 54082 40.27 40.61 39.68 TABLE V OPTIMAL SETTINGS OF CONTROL VARIABLE USING CPSO METHOD Selected Contingency Events Control variable Case 1 Case 2 Case 3 0.20 0.20 0.20 X c ( 2324) -0.60 -0.80 -0.80 Pg1 Pg4 Pg8 Pg10 Pg12 Pg24 Pg25 Pg26 Pg27 Pg31 Pg40 Pg42 Pg46 Pg49 Pg54 Pg59 Pg61 Pg65 Pg66 Pg69 Pg70 Pg72 Pg73 Pg80 Pg87 Pg89 Pg90 Pg91 Pg99 Pg100 Pg103 Pg107 Pg111 Pg112 Pg113 Pg116 2472 X c (92100) -0.60 0.20 -0.80 7172 5.00 5.00 4.00 8283 0.00 1.00 -5.00 X c ( 58) -0.40 0.10 -0.50 X c ( 2324) -0.40 -0.60 -0.60 X c (92100) 0.00 -0.50 -0.10 f 48476 46748 80.66 1.74 48557 46849 78.64 1.74 53396 51516 81.20 1.86 X c ( 58) f Fitness FT PLoss V V i ref 48373 46732 73.99 1.73 48421 46680 82.30 1.74 48555 46838 79.56 1.72 As can be seen in Table II, under a normal operating state, the average fitness that obtained by the CPSO, PSO and GA are 48670, 48680 and 48757, respectively. In Table III, the CPSO method has the best fitness of 48373, thus implying a total generation cost of $46732, a transmission Fitness FT PLoss V V i 5 ref 36.92 68.95 271.03 68.95 25.00 257.40 61.42 68.95 68.95 54.25 259.16 68.95 54.92 72.68 197.00 154.67 83.54 162.81 170.31 262.42 69.17 16.57 76.00 162.08 25.00 140.00 242.72 25.00 74.48 77.06 54.92 68.95 36.60 25.00 68.95 140.00 -1.00 25.00 75.02 267.96 68.95 34.83 249.24 54.25 68.95 83.19 54.25 249.50 68.95 41.07 194.97 179.75 132.08 68.95 191.21 140.00 253.47 49.82 58.32 76.00 140.00 25.00 140.00 226.65 25.00 25.00 68.95 41.07 68.95 36.56 25.90 68.95 140.00 -5.00 26.75 73.26 273.98 68.95 78.64 259.43 65.92 83.26 76.22 123.17 261.49 68.95 65.60 172.54 197.00 156.47 68.95 226.69 140.00 273.44 30.97 15.20 76.00 182.90 25.00 140.00 216.36 36.60 77.88 68.95 65.60 81.19 28.36 26.45 81.17 194.00 -5.00 As evaluated previously, the proposed CPSO method has the best performance in solving the SCED problem in normal operating state that has been proved. Furthermore, we also employed the method to solve the SCED problems with varied contingency constraints. The simulation results are summarized in Tables IV-V. As with Case 1, the best fitness that obtained by the CPSO method is 48476, thus implying a total generation cost of $46748, a transmission loss of 80.66MW, and a summation of bus voltage deviation of 1.74 p.u. As with Case 2, the best fitness that obtained by the CPSO method is 48557, thus implying a total generation cost of $46849, a transmission loss of 78.64MW, and a summation of bus voltage deviation of 1.74 p.u. As with Case 3, the best fitness that obtained by the CPSO method is 53396, thus implying a total generation cost of $51516, a transmission loss of 81.20MW, and a summation of bus voltage deviation of 1.86 p.u.. These results are quite reasonable, thus verifying the advantages of the proposed CPSO method. VI. limits of series FACTS devices is proposed. The objective of SCED is defined to not only to minimize total generation cost but also to enhance transmission security, reduce transmission loss, and improve the bus voltage profile. Many non-linear characteristics of the generator, such as ramp rate limits and valve-point loading effects are considered using the proposed method for practical generator operation. Therefore, the solution obtained by the proposed CPSO-based SCED method not only can minimize the objective function in pre-contingent state but can also enhance the security of the system even if the system suffers one transmission line outage. The experimental results show that the advantages of the CPSO-based method are greater than the PSO and GA. Due to the FACTS devices in a modern meshed network can be an alternative to reduce the flows in heavily loading lines, resulting in an increased loadability, low system loss, improved stability of the network, reduced cost of production, and fulfilled the security requirements by controlling the power flows in the network. Therefore, a steady-state SCED problem with considering the parameter settings of FACTS devices will be a more valuable investigation. CONCLUSION An efficient constriction PSO (CPSO) method with Gaussian mutation mechanism for solving the steady-state SCED problem with contingency constraints and operating Voltage profile (Pre-contingency) 1.1 Volt. (p.u.) 1 0.9 0.8 0.7 0.6 10 20 30 40 50 60 70 Bus Number 80 90 100 110 Fig.3. Bus voltage profile of system using CPSO method 1997. A. J. Elacqua and S. L. Corey, “Security Constrained Dispatch at the New York Power Pool,”IEEE Trans. on Power Apparatus and Systems, Vol. PAS-101 (8), pp. 2876-2883, Aug. 1982. [6] T.Ya l c i no za n dM.J .Sho r t ,“Ne ur a lNe t wo r ksAppr o a c h for Solving Economic Dispatch Problem with Transmission Capacity Co ns t r a i nt s ,” I EEE Tr a ns .o nPo we rSy s t e ms ,Vo l .13,No .2,p p. 307-313, May 1998. [7] G. C. Ejebe, C. Jing, J. G. Waight, V. tiital, G. Pirper, F. Jamshidian, P. Hirsch, and D. Sobajic, “Online Dynamic Security Assessment in an EMS,”IEEE Computer Applications in Power, pp. 43-47, Jan. 1998. [8] K. Morison, L. Wang and P. Kundur, “Power System Security Assessment,”IEEE Power & Energy Magazine, pp. 30-39, Sept./Oct. 2004. [9] Y. Xiao, Y. H. Song, and Y. Z. Sun, “Power Flow Control Approach to Power Systems with Embedded FACTS Devices,”IEEE Trans. on Power Systems, Vol. 17, No. 4, pp. 943-950, Nov. 2002. [10] J. A. Momoh, J. Z. Zhu, C. D. Boswell and S. Hoffman, “Power [5] VII. REFERENCES [1] [2] [3] [4] R. Lugtu, “Security Constrained Economic Dispatch,”IEEE Trans. on Power Apparatus and Systems, Vol. PAS-98 (1), pp. 270-274, Jan./Feb. 1979. A. Bakirtzis, V. Petridi sa n dS.Ka z a r l i s ,“ Ge ne t i cAl g o r i t hm So l ut i o n t o t he Ec o no mi c Di s pa t c h Pr o bl e m, ” I EE Pr o c . -Generation, Transmission and Distribution, Vol. 141, No. 4, pp.377-382, July 1994. B. Stott, O. Alsac, and A. J. Monticelli, “Security Analysis and Optimization,” Proceedings of the IEEE, Vol. 75, No. 12, pp. 1623-1987, Dec. 1987. M. Aganagic, B. Awobamise, G. Raina, and A. I. McCartney, “Economic Dispatch with Generation Contingency Constraints,” IEEE Trans. on Power Systems, Vol. 12, No. 3, pp. 1229-1236, Aug. 6 [11] [12] [13] [14] [15] [16] [17] [18] [19] System Security Enhancement by OPF with Phase Shifter,”IEEE Trans. on Power Systems, Vol. 16, No. 2, pp. 287-293, May. 2001. T. Orfanogianni and R. Bacher, “Steady-State Optimization in Power Systems with Series FACTS Devices,”IEEE Trans. on Power Systems, Vol. 18, No. 1, pp. 19-26, Feb. 2003. D.C.Wa l t e r sa n d G.B.She bl e ,“ Ge ne t i cAl go r i t hm So l ut i o no f Ec o no mi cDi s pa t c hwi t hVa l v ePo i ntLo a di ng , ”I EEETr a ns .o nPo we r Systems, Vol. 8, No. 3, pp. 1325-1332, Aug. 1993. Gerald B. Sheble, and Kristin Brittig,“ Re f i ne dGe ne t i cAl go r i t hm – Ec o no mi cDi s p a t c hEx a mpl e , ”I EEETr a ns .o nPo we rSy s t e ms ,Vo l . 10, No. 1, pp. 117-124, February 1995. Po-Hung Chen and Hong-Cha n Ch a ng ,“ La r g e -Scale Economic Di s pa t c hbyGe ne t i cAl g or i t hm, ”I EEEo nPo we rSy s t e ms ,Vol .10 , No. 4, pp. 1919-1926, Nov. 1995. T.Ya l c i o no z ,H.Al t un,a ndM.Uz a m,“ Ec o no mi cDi s pa t c hSo l ut i o n Us i ngA Ge ne t i cAl go r i t hm Ba s e do nAr i t hme t i cCr o s s o v e r , ”200 1 IEEE Proto Power Tech Conference, Proto, Portugal, Sep. 2001. Z. L.Ga i ng ,“Particle Swarm Optimization to Solving the Economic Di s pa t c h Co ns i de r i ng t heGe ne r a t o rCo ns t r a i n t s , ”I EEE Tr a ns .o n Power Systems, Vol. 18, No. 3, pp. 1187-1195, August 2003. R. K. Pancholi and K. S. Swarup, “ Pa r t i c l eSwa r m Opt i mi z a t i o nfor Security Constrained Economic Dispatch,” Proceedings of International Conference on Intelligent Sensing and Information Processing, pp. 7-12, 2004. D. B. Fogel, Evolutionary Computation: Toward a New Philosophy of Machine Intelligence, 2nd edition, IEEE Press, 2000. Y.Sh ia n d R.C.Ebe r ha r t ,“ Emp i r i c a lSt udy o fPa r t i c l eSwa r m Opt i mi z a t i o n , ”Pr o c e e di ng so ft he19 9 9Co ng r e s so nEv o l ut i o n a r y Computation, pp. 1945-1950, Piscataway, 1999. [20] R.C.Ebe r ha r ta ndY.Shi ,“Co mp a r i s o nbe t we e nGe ne t i cAl g o r i t hms and Pa r t i c l eSwa r m Opt i mi z a t i o n, ”Pr o c e e di ng so fI EEEI nt e r na t i o na l Conference on Evolutionary Computation, pp. 611-616, Anchorage, May 1998. [21] A. Stacey, M. Jancic, and I. Grundy, “ Pa r t i c l eSwa r m Opt i mi z a t i o n with Mutation, ”Pr o c e e di ng so fI EEE I nt e r na t i o n a l Conference on Evolutionary Computation, Vol. 2, pp. 1425-1430, Dec. 2003. [22] N. Higashi and H. Iba, “ Pa r t i c l eSwa r m Opt i mi z a t i o nwith Gaussian Mutation, ” Pr o c e e di ng so f the IEEE on Swarm Intelligence Symposium, pp. 72-79, April 2003. Zwe-Lee Gaing(M’ 02) received his M.S. and Ph.D. degree from National Sun Yat-Sen University, Kaohsiung, Taiwan in 1992 and 1996, respectively. Currently, he works as a professor in the Electrical Engineering Department at Kao Yuan University, Kaohsiung, Taiwan, R.O.C. His research interests are in the field of artificial intelligence with application to power system operation and control. Rung-Fang Chang received his B.S. degree in electrical engineering from the Chung Yuan Christian University, Chung-Li, Taiwan in 1990 and received the M.S. degree and the Ph.D. degree from the National Sun Yat-Sen University, Kaohsiung, Taiwan in 1992 and 2002 respectively. He is an associate professor of the Department of the Electrical Engineering, Kao Yuan University, Lu-Chu, Taiwan. His research interests are optimization of power system and load research. Appendix TABLE AI GENERATING UNIT CAPACITY AND COEFFICIENTS IN IEEE 118-BUS SYSTEM Bus No. 1 4 8 10 12 24 25 26 27 31 40 42 46 49 54 59 61 65 66 69 70 72 73 80 87 89 90 91 99 100 103 107 111 112 113 116 Pgi( 0 ) 25.00 70.00 220.00 60.00 65.00 225.00 54.25 68.95 68.95 54.25 212.00 68.95 20.00 68.95 140.00 80.00 68.95 140.00 140.00 270.00 30.00 30.00 50.00 140.00 25.00 140.00 165.00 50.00 40.00 70.00 16.00 68.95 25.00 25.00 68.95 140.00 Pgimin 25.00 68.95 100.00 68.95 25.00 100.00 54.25 68.95 68.95 54.25 100.00 68.95 15.20 68.95 68.95 68.95 68.95 140.00 140.00 100.00 15.20 15.20 15.20 140.00 25.00 140.00 100.00 25.00 25.00 68.95 15.20 68.95 25.00 25.00 68.95 140.00 Pgimax 100.00 197.00 400.00 197.00 100.00 400.00 155.00 197.00 197.00 155.00 400.00 197.00 76.00 197.00 197.00 197.00 197.00 350.00 350.00 400.00 76.00 76.00 76.00 350.00 100.00 350.00 400.00 100.00 100.00 197.00 76.00 197.00 100.00 100.00 197.00 350.00 Qgimin -55.00 -120.00 -300.00 -120.00 -55.00 -300.00 -50.00 -120.00 -120.00 -50.00 -300.00 -120.00 -50.00 -120.00 -120.00 -120.00 -120.00 -200.00 -200.00 -300.00 -50.00 -50.00 -50.00 -200.00 -55.00 -200.00 -300.00 -55.00 -55.00 -120.00 -50.00 -120.00 -55.00 -55.00 -120.00 -200.00 Qgimax 100.00 180.00 300.00 180.00 100.00 300.00 140.00 180.00 180.00 140.00 300.00 180.00 50.00 180.00 180.00 180.00 180.00 200.00 200.00 300.00 50.00 50.00 50.00 200.00 100.00 200.00 300.00 100.00 100.00 180.00 50.00 180.00 100.00 100.00 180.00 200.00 S gimax 110.0 200.0 420.0 200.0 110.0 420.0 160.0 200.0 200.0 160.0 420.0 200.0 80.0 200.0 200.0 200.0 200.0 360.0 360.0 420.0 80.0 80.0 80.0 360.0 110.0 360.0 420.0 110.0 110.0 200.0 80.0 200.0 110.0 110.0 200.0 360.0 7 URi 60 50 80 50 60 80 50 50 50 50 80 50 76 50 50 50 50 70 70 80 76 76 76 70 60 70 80 60 60 60 76 50 60 60 50 70 DRi 100 90 120 90 100 120 80 90 90 80 120 90 76 90 90 90 90 110 110 120 76 76 76 110 100 110 120 100 100 100 76 90 100 100 90 110 ai bi ci 190.00 220.00 240.00 220.00 190.00 240.00 200.00 220.00 220.00 200.00 240.00 220.00 200.00 220.00 220.00 220.00 220.00 220.00 220.00 240.00 200.00 200.00 200.00 220.00 190.00 220.00 240.00 190.00 190.00 220.00 200.00 220.00 190.00 190.00 220.00 220.00 12.10 10.50 7.10 10.52 12.10 7.10 11.20 10.50 10.53 11.21 7.10 10.50 11.10 10.50 10.50 10.50 10.50 8.50 8.50 7.10 11.20 11.20 11.20 8.50 12.10 8.50 7.10 12.10 12.10 10.50 11.20 10.50 12.10 12.10 10.50 8.50 0.0075 0.0080 0.0072 0.0081 0.0074 0.0071 0.0096 0.0080 0.0082 0.0095 0.0071 0.0080 0.0091 0.0080 0.0080 0.0080 0.0080 0.0090 0.0090 0.0072 0.0091 0.0092 0.0092 0.0090 0.0074 0.0090 0.0071 0.0075 0.0076 0.0080 0.0091 0.0080 0.0074 0.0075 0.0080 0.0090 di 51.1331 61.2111 80.3210 61.2111 51.1331 80.3210 32.7233 61.2111 61.2111 32.7233 80.3210 61.2111 20.1012 61.2111 61.2111 61.2111 61.2111 42.5122 42.5122 80.3210 20.1012 20.1012 20.1012 42.5122 51.1331 42.5122 80.3210 51.1331 51.1331 61.2111 20.1012 61.2111 51.1331 51.1331 61.2111 42.5122 ei 0.0600 0.0210 0.0200 0.0210 0.0600 0.0200 0.0410 0.0210 0.0210 0.0410 0.0200 0.0210 0.0110 0.0210 0.0210 0.0210 0.0210 0.0120 0.0120 0.0200 0.0110 0.0110 0.0110 0.0120 0.0600 0.0120 0.0200 0.0600 0.0600 0.0210 0.0110 0.0210 0.0600 0.0600 0.0210 0.0120