Electric Power Systems Research 79 (2009) 694702

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Applications of computational intelligence techniques for solving the revived optimal power ow problem
M.R. AlRashidi a, , M.E. El-Hawary b
a b

Electrical Engineering Department, College of Technological Studies, Shuwaikh, Kuwait Department of Electrical and Computer Engineering, Dalhousie University, Halifax, NS B3J 2X4, Canada

a r t i c l e

i n f o

a b s t r a c t
Computational intelligence tools are attracting added attention in different research areas and research in power systems is not different. This paper provides an overview of major computational issues with regard to the optimal power ow (OPF). Then, it offers a brief summary of major computational intelligence tools. A detailed coverage of most OPF related research work that make use of modern computational intelligence techniques is presented next. 2008 Elsevier B.V. All rights reserved.

Article history: Received 11 November 2007 Received in revised form 30 August 2008 Accepted 11 October 2008 Available online 28 November 2008 Keywords: Optimal power ow Computational intelligence Metaheuristic Power system optimization

1. Introduction Power engineers require special tools to optimally analyze, monitor, and control different aspects of power systems operation and planning. Most of these tools are properly formulated as some sort of optimization problems. The optimal power ow (OPF) is the backbone tool that has been extensively researched since its rst introduction in the early 1960s [1,2]. It appears that the term optimal power ow was rst introduced by Dommel and Tinney in 1968 [3]. Originally, the OPF was formulated as a natural extension of the traditional economic dispatch. Differences between the two optimization functions exist even though both of them may share the same objective function. In economic dispatch, the entire power network is reduced to a single equality constraint. By contrast, all major elements of the modeled system are explicitly presented in the OPF problem. The generic term OPF is no longer associated exclusively with the extended economic dispatch calculation. Rather, it presents a wide range of optimization problems commonly formulated in power systems related studies. OPF studies are evolving over time from its basic form to cope with the continuous changes that are taking place in power systems. Deregulation of the electric power industry, advances being made in the area of power electronics, and the environmental regulations that are

being imposed on power plants are some of the main factors that have played major role in constantly reformulating the OPF. The historical development of the OPF is closely correlated with the advances made in the area of numerical optimization techniques [4]. Researchers have attempted to apply most optimization techniques to solve the OPF. The purpose of OPF is to nd the optimal settings of a given power system network that optimize a certain objective function while satisfying its power ow equations, system security, and equipment operating limits. Different control variables are manipulated to achieve an optimal network setting based on the problem formulation. The main control variables typically used in optimizing the OPF are as follows: Generators real power outputs and voltages. Transformer tap changing settings. Phase shifters settings and placement for expansion planning. Switched capacitors and reactors. FACTS devices settings and placement for expansion planning.

Corresponding author. Tel.: +965 4847100x4312 fax: +965 481 6568. E-mail address: malrash2002@yahoo.com (M.R. AlRashidi). 0378-7796/$ see front matter 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.epsr.2008.10.004

A main obstacle of the OPF problem is the nature of the control variables since some of them are continuous (e.g. real power outputs and voltages) and others are discrete (e.g. transformer tap setting, phase shifters, and reactive injections). The presence of discrete variables makes the optimization problem a non-convex one, which in turn complicates the solution methodology. Objective functions of the OPF studies can be arranged into two main categories: traditional and newly emerged objectives. The latter group

M.R. AlRashidi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 694702

695

Fig. 1. Most commonly used objective functions in the OPF studies.

has emerged mainly as a result of restructuring the electric power industry. Fig. 1 summaries most used objectives in the OPF studies. It is noted from this study that the most commonly used objective is the minimization of the overall fuel cost function (convex and non-convex). Researchers proposed different mathematical formulations of the OPF problem, which can be broadly classied as follows: 1. Linear programming problem in which objectives and constraints are given in linear forms with continuous control variables. 2. Non-linear programming problem where either objectives or constraints or both combined are non-linear with continuous control variables. 3. Mixed integer linear and non-linear programming problems when control variables are both discrete and continuous. Many conventional optimization techniques were developed to solve the OPF problem, the most popular being linear programming, sequential quadratic programming, generalized reduced gradient method, and the Newton method. References [57] offer a complete list of the most commonly used conventional optimization algorithms with regard to the OPF. Despite the fact that some of these techniques have excellent convergence characteristics and various among them are widely used in the industry, some of their drawbacks are: 1. Convergence to the global or local solution is highly dependant on the selected initial guess, i.e. they might converge to local solutions instead of global ones if the initial guess happens to be in the vicinity of a local solution. 2. Each technique is tailored to suit a specic OPF optimization problem based on the mathematical nature of the objectives and/or constraints. 3. They are developed with some theoretical assumptions, such as convexity, differentiability, and continuity, among other things, which may not be suitable for the actual OPF conditions. The rapid developments of recent computational intelligence tools have attracted many researchers to employ them in solving the ever changing OPF. This paper provides an extensive coverage of the major research work conducted in applying modern computational tools to the OPF problem. Section II briey discusses the main

computational intelligence tools. A general mathematical formulation of the OPF is offered in Section III. Section IV presents a detailed coverage of most applications of computational intelligence tools in solving the OPF. The paper is then concluded in Section V. 2. Modern optimization techniques A new category of computational intelligence tools has emerged to cope with some of the traditional optimization algorithms shortcomings. The main modern techniques include evolutionary programming (EP), genetic algorithm (GA), evolutionary strategies (ES), articial neural network (NN), simulated annealing (SA), ant colony optimization (ACO), fuzzy set theory (FST), and particle swarm optimization (PSO). Most of these relatively new developed tools mimic a certain natural phenomenon in its search for an optimal solution like species evolution (EP, GA, and ES), human neural systems (NN), thermal dynamics of a metal cooling process (SA), data processing and interpretation in human brain (FST), or social behavior (ACO and PSO). They have been successfully applied to a wide range of optimization problems in which global solutions are more preferred than local ones or when the problem has nondifferentiable regions. Also, they are known for their capabilities of fast search of large solution spaces and their ability to account for uncertainty in some parts of the power system networks. The following is a brief description of these optimization techniques: 1. Evolutionary programming: EP and ES are probably the oldest evolutionary algorithms used to solve optimization problems. EP along with GA and ES adapt the concept of natural evolution or survival of the ttest founded in Darwinian Theory of evolution to enhance the population or solutions quality. It was rst introduced by Fogel in 1964 as ndings of his doctorate work [8]. The EP paradigm emphasizes the relationship between ancestors (parents) and their descendants (offspring) and it relies exclusively on a mutation operator to produce offspring, i.e. there is no recombination operator. 2. Evolutionary strategies: The German researchers Rechenberg et al. developed ES in 1965 to successfully solve difcult engineering optimization problems when they encountered difculties in applying gradient based methods [9]. ES make use of both mutation and recombination operators to produce new population. Typically, different sizes of parent and offspring populations are used in ES implementation.

696

M.R. AlRashidi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 694702

3. Genetic algorithm: GA (originally it was termed reproductive plans) is by far the most commonly used and well developed evolutionary technique with abundant number of applications and developments found in technical articles and books. It was introduced by Holland in 1975 in an attempt to develop a technique that enables computer programs to mimic the evolution process [10]. In addition to the selection, crossover, and mutation operators, some GA-based approaches implement memory elements to preserve the elite solutions [11]. In most GA approaches, the binary representation is used for parameters encoding to unify the recombination and mutation operators. However, more recent advances in GA approaches elevated its capabilities to represent solutions using oating point arithmetic. Genetic programming is an extension of GA in which individuals are represented as trees. 4. Articial neural network: Hopeld and Tank introduced a special type of NNs to solve the famous traveling salesman combinatorial problem in 1985 [12]. In this NN architecture, some of the outputs are fed back to the neurons in the input layer. Parallel data processing and fast network convergence are key features of this approach. 5. Simulated annealing: This optimization technique was proposed independently by Kirkpatrick et al. in 1983 [13] and by Cerny in 1985 [14]. SA emulates the physical gradual cooling process (called annealing) to produce high quality crystals, i.e. better strength properties, in metals. In both papers, SA was introduced to solve combinatorial problems by adapting the crystallization process model developed by Metropolis et al. [15]. 6. Fuzzy Set Theory: Since its rst introduction by Zadeh in 1965 [16], FST has been applied in enormous number of engineering applications. It is a mathematical tool intended to deal with uncertainty and vagueness that exist in most systems modeling. FST provides better modeling scheme for complex systems, when ambiguity is accounted for, by constructing fuzzy rather than crisp relationships among different system elements. Within the scope of OPF optimization, fuzziness can be introduced in either the objectives and/or the constraints to model the uncertainty associated with some of the OPF data such as fuel cost and emission coefcients, forecasted demand, and transmission line limits. The reader is referred to reference [17] for more general survey of the applications of FST in power systems. 7. Ant colony optimization: Dorigo invented ACO as a new metaheuristic optimization tool in the early 1990s as he was inspired by the foraging behavior of some ant species. Solution candidates, called ants in ACO, communicate with other members of the ant colony by depositing pheromone to mark a path. High concentrations of pheromones indicate more favorable paths that other members should follow in order to reach the optimal solution. The original version of ACO published in references [18,19] was called the ant system and it inspired many researchers to modify and adapt the original ACO to suit various optimization problems. 8. Particle swarm optimization: Kennedy and Eberhart rst introduced PSO in 1995 as a new metaheuristic method [20,21]. They studied a stochastic non-linear model that was developed by Reynolds and Heppner and Grenander to simulate species movement traveling in groups [22,23]. The original motivation of the conducted research was to create a computer model that simulates the social behavior of bird ocks and sh schools. As Kennedy and Eberhart progressed in their research, they discovered that with some modications, the social behavior model can also serve as a powerful optimizer. They realized that such species try to approach their target in an optimal manner which resembles nding the optimal solution to any mathematical optimization problem. References [24,25] present more

general reviews of PSO applications in various areas of power systems. Other computational intelligence tools are Tabu search, pattern search, differential evolution, and cultural algorithms. However, these techniques are currently not as popular especially with regard to the OPF. Thus, only the main tools are briey highlighted in this section. It is obvious that most of modern computational intelligence tools are quite versatile in terms of their applications. These techniques have been gaining more popularity mainly because of their robustness, simplicity, and their ability to deal with more exact models instead of making intolerable approximations. They are similar in some perspectives such as using probabilistic transition rules to move from one point to the next point during their search for optimal solution and the use of objective function to guide the search process. The major drawbacks of these techniques are the lack of solid mathematical foundation and failure to assure global optimal solution. The reader is referred to online references [2631] for additional information regarding computational intelligence tools. As far as the computation speed is concerned, several factors may inuence the execution time of these modern solution methods such as population size and constraint handling mechanism. Some solution techniques require larger population sizes which in turn increase the execution time. Constraints are typically handled in different methods and the overall performance of any optimization algorithm is usually affected by the type of constraint treatment. With special regard to the OPF, power ow equations are the main equality constraints in the OPF problem. Different methods are typically used to solve the power ow equations with the main ones being Newton-Raphson, Guass Seidel, fast decouple, and radial power ow. These methods solve the same problem in completely different manners and execution time varies from one method to the other. 3. Problem formulation The desired goal of the OPF is to optimize a certain objective subject to different sets of equality and inequality constraints. The problem can be formulated as follows: Min F(x, u) (1)

Subject to g(x, u) = 0 h min h(x, u) h max (2) (3)

where vector x denotes the dependant or state variables of a power system network that contains the slack bus real power output (PG1 ), voltage magnitudes and phase angles of the load buses (VLk , Lk ), and generator reactive power outputs (QG ). Vector u represents both integer and continuous control variables that consist of real power generation levels (PGN ) and voltage magnitudes (|VGN |), transformer tap setting (Tk ), and reactive power injections (QCk ) due to voltamperes reactive (VAR) compensations; i.e.

OPF Control Variables

u = ucontinuous , udiscrete where ucontinuous = PG2 . . . PGN , VG2 . . . VGN and udiscrete = T1 . . . TN , QC1 . . . QCN

(4)

(5)

M.R. AlRashidi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 694702

697

3.1. Constraints The OPF problem has two categories of constraints. 3.1.1. Equality constraints These are the sets of non-linear power ow equations that govern the power system, i.e. PGi PDi PTi (V, ) = 0 QGi QDi QTi (V, ) = 0 (6) (7)

where PGi and QGi are the generated real and reactive power at bus i, respectively, the load demand at the same bus is represented by PDi and QDi , and subscript T denotes the total sending and receiving power at each bus. 3.1.2. Inequality constraints These are the set of continuous and discrete constraints that represent the system operational and security limits like the bounds on: 1. The generators real and reactive power outputs;
min max PG PGi PG ,
i i

i = 1, . . . , GN i = 1, . . . , GN

(8) (9)

min max QG QGi QG ,

i i

2. Voltage magnitudes at each bus in the network; Vimin Vi Vimax , i = 1, . . . , N (10)

3. The discrete transformer tap settings; Timin Ti Timax , i = 1, . . . , TN (11)

4. The discrete reactive power injections due to capacitor banks;

min max QC QCi QC ,
i i

i = 1, . . . , CN

(12)

Note that PGi , QGi , and Vi are continuous variables while Ti and QCi are discrete ones. 5. The transmission lines loading;
max SLi SL ,
i

i = 1, . . . , LN

(13)

Additional inequality constraints may include prohibited zones of the generating units, measures of transient stability, electromagnetic eld levels, etc. 4. Modern solution methods to the OPF This section offers a general recent review of most of the research work conducted with regard to the OPF. Due to the abundant number of reported OPF articles, discussion is exclusively limited to the OPF related research works that employ modern computational intelligence tools. Reviewed papers are extracted from major journals associated with power system studies and categorized based on the main computational intelligence tool used as follows. 4.1. Genetic algorithms-based approach A real-coded GA was developed by Gaing and Huang to solve the OPF problem with the objective function being the quadratic fuel cost function [32]. In their formulation, the authors accounted for both discrete and continuous optimization variables and introduced prohibited zones of the generation units as additional constraints. Lai et al. adopted an enhanced binary GA to optimize the same objective function for a given power system under normal

and contingent operation conditions [33]. The authors of reference [34] proposed a rened GA in an attempt to solve identical OPF problem. Comparison among three different solution techniques, namely GA, rended GA and decoupled OPF, were made and results were in favor of the rened GA. In reference [35], a similar OPF formulation was considered with the inclusion of the non-convex fuel cost function to better model the rippling effects in the I/O curve of the generating units due to the valve admission. The prohibited zones constraints were relieved in this formulation. Zhang et al. introduced a transient stability index as an additional inequality constraint, to account for the impacts of different contingencies on the transmission network stability, to their OPF formulation [36]. They employed a binary version of the GA to test their proposed approach in the deregulated market of the United Kingdom. An enhanced binary version of the GA with advanced and problem-specic operators was developed by Bakirtzis et al. to solve the OPF problem [37]. This enhancement improved the convergence rate and the quality of solutions. The proposed approach was tested using the quadratic fuel cost function as an objective of the OPF. Using the same objective function, a comparison of the performance of two metaheuristic methods (namely PSO and enhanced GA) and two available commercial grade optimization software packages that employ non-linear mathematical programming methods (LINDO and GAMS) was made in references [38,39]. In this problem formulation, both the discrete and continuous nature of the optimization variables were accounted for. In the case of the metaheuristic methods, the authors transformed the constrained OPF problem into an unconstrained problem by using the penalty factors approach. The authors have concluded that metaheuristic methods performed well in small test systems and failed to provide feasible solutions for medium-size test systems. Todorovski and Rajicic have proposed a special initialization procedure for the control variables of the OPF problem to generate an initial set of solution vectors with no or few constraint violations [40]. A real-coded GA based approach was used to solve the optimization problem and results indicated signicant reduction in the computation time. In their proposed solution method, they used a distinct and unconventional set of control variables to optimize the quadratic fuel cost function. They chose the complex voltages at the generator-buses (i.e. both magnitudes and phase angles), reactive power of synchronous condensers, transformer tap settings, and shunt device settings to be the control variables. The unusual usage of the phase angles at the voltage controlled buses as control variables is based on earlier research conducted by the same authors in which they have developed a new innovative way to solve the power ow equations using GA [41]. Their proposed approach appears to be faster than the fast decoupled method in solving the power ow equations and it eliminates the concept of predetermined slack bus. Devaraj and Yegnanarayane have used a real-coded GA technique to solve the OPF problem to improve system security [42]. They used a severity index, a measure of the severity of the contingency to line overloading condition, as an objective function in their formulation. The control variables selected were the phase angle of the phase-shifting transformer (discrete) and the real power outputs and voltage magnitudes of the generation units (continuous). Proper placement of the phase-shifting transformers was achieved based on sensitivity analysis. Das and Patvardhan treated the OPF as a multi-objective optimization problem with four different objective functions, i.e. fuel cost and emissions of the generating units, transmission real power losses, and the security margin index [43]. They combined GA with SA to form a hybrid solution tool with enhanced capability of handling both discrete and continuous control variables. The shape of the Pareto front was detected and they determined the best comprising solution of the OPF multi-objective optimization problem.

698

M.R. AlRashidi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 694702

Similarly, Abido proposed a strength Pareto evolutionary algorithm to solve the multi-objective OPF problem [44]. In his formulation, Abido considered two objective functions, namely the fuel cost and the voltage stability indicator. An entire set of Pareto optimal solutions were calculated and the Pareto shape was captured. Lai and Ma proposed a binary-coded GA approach to solve the OPF problem in FACTS [45]. They used a unied power ow controller (UPFC) to regulate the power ow such that the real power losses were minimized. Several contingencies were considered to successfully test the proposed approach. Padhy et al. used a different FACTS device, namely a thyristor-controlled series compensation (TCSC), to minimize an objective function that sums the complex power lost in transmission lines (i.e. real and reactive line losses) and the generated reactive power [46]. A hybrid tool of which GA was used to locate the optimal placements and settings of TCSC devices and a quasi-Newton algorithm was then employed to solve the OPF problem. Reference [47] presents a GA approach that makes use of both UPFC and TCSC devices to maximize the transmission lines capacity. Two similar applications of employing GA in solving the OPF problem are presented in references [48,49]. In both cases, a binary-coded GA coupled with a power ow algorithm is used to minimize the quadratic fuel cost function with the inclusion of UPFC settings as additional control variables. Chung and Li incorporated settings of the TCSC and thyristor-controlled phase shifter (TCPS) into the OPF problem that minimizes the quadratic fuel cost function [50]. A GA was used to determine the optimal settings of the used FACTS devices. With the deregulation of the electric power industry, new OPF formulations have emerged to represent the new environment in which GA versatile capabilities were further employed. Gountis and Bakirtzis have formulated a two level optimization problem to optimize the bidding strategies of electricity producers [51]. They developed a model that takes into account the uncertainty of nodal loading conditions and the market risk factors. The problem was solved by forming a hybrid tool of Monte Carlo simulation and GA. Dong and Hill used a risk analysis to optimally schedule the reactive power, a form of ancillary services, in the electricity market [52]. They combined linear programming and GA to minimize an objective that represents the future risk of possible scenarios. Reference [53] presents an application of GA in calculating the available transfer capability of a power network in a deregulated market. 4.2. Evolutionary programming In reference [54], EP was employed in an attempt to solve the OPF problem under different contingency cases. In this formulation, the objective was to minimize the quadratic fuel cost function by properly adjusting both discrete and continuous control variables. Similar OPF formulation was used in reference [55] to validate EP potential in deregulated electric market. In terms of solution accuracy, results obtained using EP approach were comparable to those of steepest descent method and better than GA outcomes. Improved EP was proposed in reference [56] to optimize the non-convex generator fuel cost curve. To validate the proposed approach, different metaheuristic methods were developed to solve the same problem. Comparison ndings were in favor of the improved EP in terms of solution accuracy and execution time. Lai and Ma introduced EP as a viable tool to minimize the real power losses in power networks by regulating the power ow via optimal UPFC settings [57]. The proposed approach was tested under various contingencies scenarios. Venkatesh et al. made use of EP to solve the OPF problem while accounting for UPFC in their formulation [58]. They addressed the problem considering two fuzzy objective functions, i.e. minimum of the real power losses and the best voltage prole, while accounting for the nature of control variables used. Ongsakul and Jirapong

used EP to maximize the total transfer capability between generation and load center areas [59]. The proposed approach optimally adjusts the real power outputs and voltage magnitudes at generation buses such that the total load in the sink area is maximized. They further enhanced their approach to determine the optimal settings and locations of four types of FACTS devices in addition to the other previously considered control variables to optimize the same objective [60]. Reference [61] presents a multi-objective formulation of security constrained OPF problem with the objective functions being the quadratic fuel cost and active power losses. In this application, a hybrid tool of EP and SQP was developed in which the former method is used to nd good initial solution for the latter method. This hybridization reduced the computation time and outperformed the performance of each individual method. The developed algorithm was tested under different contingency cases on a test system composed of two interconnected areas. Shi et al. proposed a hybrid method that combines EP and a classical gradient search method to minimize the quadratic fuel cost function [62]. Lo et al. developed a parallel EP approach to minimize the quadratic fuel cost function [63]. A master-slave set up of 31 computers was used to test their proposed approach on two test systems. They compared their results to those of sequential EP and concluded that both approaches were comparable in terms of the solution quality. However, parallel EP reduced the execution time by a factor of 1012 times. The same authors extended their work by incorporating the steepest descent as a local search mechanism for a portion of the population within EP to improve the speed of convergence [64]. In addition, three congurations of parallel EP, namely master-slave, dual-ring, and 2D-mesh, were implemented to solve the OPF problem. Comparison results reveal that 2D-mesh conguration of parallel EP provided a better solution at a faster convergence rate. Padhy solved the OPF problem using EP in order to calculate the wheeling rates of active power at various parts of the transmission network [65]. He used the quadratic fuel cost function to test his approach. Reference [66] employed EP to determine short run marginal cost for bilateral transactions. The cost is computed based on the OPF solution with the objective being the non-smooth fuel cost function. Sood et al. presented a hybrid method to calculate the wheeling rates of both real and reactive power based on the solution of the OPF [67]. EP was combined with the steepest decent method to minimize different types of fuel cost functions such as continuous quadratic, piecewise quadratic, piecewise linear with prohibited operating zones, piecewise quadratic with prohibited operating zones, and quadratic with a superimposed sine term to account for valve loading effects. A similar hybrid method was presented in reference [68] to optimize the voltage prole with three types of fuel cost functions. In reference [69], a hybrid method of EP and the Newton-Raphson method was formed to optimally select the best wheeling option when a privately owned generator is introduced in an existing network. The incremental costs of three options were computed based on the OPF calculations to make the proper judgment about the best wheeling option. 4.3. Articial neural network Hartati and El-Hawary presented an application of using a modied Hopeld NN to solve the OPF problem with the objective being the incremental fuel cost [70]. Two types of neuron transfer functions were used to investigate the robustness and accuracy of the proposed approach with regard to the OPF. Nguyen developed a NN that performs Newton-Raphson based OPF calculations to minimize the real power losses [71]. A key feature of his developed network is the parallel computation, i.e. reduced computation time,

M.R. AlRashidi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 694702

699

which makes use of the sparse nature of the OPF matrices. Dondo and El-Hawary proposed a methodology for real-time electricity spot pricing using NNs [72]. They were able to express the electricity rate in terms of the solution of the OPF problem. Results of the NN based solution were comparable to those of MINOS, an optimization software package, in terms of solution accuracy and better in terms of execution time. Luo et al. used a faster training algorithm, i.e. Quickprop, to train a NN to maximize the transfer capability of electric power among different areas [73]. The problem was formulated as an OPF where different contingencies and loading conditions were considered during the training process. The proposed network was heavily tested and its performance was compared to the exact solution of the OPF. The same authors extended their work in reference [74] where they compared their NN performance when considering two different training algorithms in addition to adding more contingencies and loading proles. In a different application, Luo et al. combined a NN with a Monte Carlo simulation to present a fast computing method to evaluate power system reliability [75]. The problem was formulated as an OPF problem with the objective being minimizing the sum of the load shedding at each bus. In reference [76], the authors proposed an online ANN that maximizes the voltage stability margin and simultaneously minimizes the generation cost and transmission network losses. 4.4. Fuzzy set theory Abdul-Rahman et al. developed a fuzzy linear model to include electromagnetic eld effects as additional constraints imposed on the OPF problem [77]. The objective was to reach a compromise solution that minimized the fuzzy incremental fuel cost while satisfying the traditional OPF crisp constraints by redirecting the generated power such that the electromagnetic eld fuzzy levels at some critical buses are kept within acceptable limits. In a different formulation, a fuzzy representation was used to model the uncertainty of loading conditions in reference [78]. Different membership functions were used to model different elements of the OPF problem that account for contingencies. Various objective functions, namely fuel cost, static security, and emissions, were aggregated using a fuzzy operator to reach a compromise solution of the conicting objectives. Ramesh and Li modeled the OPF problem as a multi-objective optimization problem with the conicting objectives being the pre-contingency operating cost and post-contingency correction time [79]. The two functions were formulated as fuzzy objectives subject to hard constraints and SQP was used to solve the problem. Similar problem formulation and solution technique were used in reference [80] to solve the multiobjective OPF problem. However, larger test systems were used in this paper to test the proposed approach. A non-linear predictorcorrector primal-dual interior point method was used in reference [81] to optimize a multi-objective OPF problem. The authors considered optimizing fuzzy functions that model the total generation cost and the deviation of actual loading conditions from an ideal loading condition. Wu used fuzzy rules to improve the efciency of a predictor-corrector interior point algorithm for the OPF [82]. The proposed approach was tested on a large power system and an average of 20% improvement in computation time was reported. A fuzzy dynamic programming approach was employed to solve the OPF problem that was modeled with fuzzy objective function and fuzzy loads in reference [83]. Padhy made use of the gradient descent method to solve the traditional OPF in a deregulated power environment with the objective being the quadratic fuel cost [84]. Then, he introduced the so called fuzzy opinion matrix approach to critically select the most suitable feasible transaction such that the network congestion is reduced.

4.5. Particle swarm optimization Abido is credited with introducing PSO to solve the OPF problem [85]. He used different objectives to test the proposed approach such as minimizing the fuel cost, improving the voltage proles, and enhancing the voltage stability. In a different approach to the problem, Zhao et al. solved the highly constrained OPF optimization problem by minimizing a non-stationary multi-agent assignment penalty function [86]. In this formulation, PSO was used to solve the highly constrained OPF optimization problem in which the penalty values were dynamically modied in accordance with system constraints. In reference [87], the passive congregation concept was incorporated in PSO to solve the OPF problem. This hybrid technique improved the convergence characteristics over the traditional PSO in solving the same OPF problem. Wang et al. developed a modied PSO to solve the OPF problem with the objective being the minimization of the quadratic fuel cost function [88]. The proposed algorithm mainly relied on the idea of randomly exchanging information among the entire swarm rather than only the best member in the swarm. The environmental-economic transaction planning problem in the electricity market was formulated as a multi-objective OPF in reference [89]. A multi-objective PSO algorithm was developed to solve the problem via a non-stationary multi-stage assignment penalty function. Different versions of PSO were developed in reference [90] in an attempt to construct a comparison of their performance with regard to the OPF. The objective functions selected in this study were the real power losses and voltage prole improvement. Gaing introduced an enhanced PSO to solve a multi-objective OPF problem with the objective functions being the fuel cost, real power losses, and voltage deviation [91]. Mo et al. made use of PSO to solve the OPF while accounting for transient stability in their problem formulation [92]. Reference [93] presented a modied PSO algorithm by incorporating a reconstruction operator that permits only feasible solution to exist in the population. The proposed approach was tested by solving the OPF while accounting for different contingencies. AlRashidi and El-Hawary developed a hybrid PSO algorithm that preserves only feasible solutions to solve the discrete OPF [94]. They considered different objectives in their formulation such as emission, real power losses, and the non-convex fuel cost that incorporates the valve point loading effects. Reference [95] used PSO to nd optimal locations and settings of different FACTS devices such that the total installation cost of FACTS devices is minimized. The authors made use of penalty factor approach to incorporate problem constraints. 4.6. Other techniques A bacteria foraging algorithm was developed in reference [96] to solve the multi-objective OPF problem that optimizes the real power losses and voltage stability limit simultaneously. The control variables considered were the UPFC optimal placement and setting in addition to the existing transformer settings. A dynamic version of the bacteria foraging algorithm was developed by Tang et al. to minimize the quadratic fuel cost function [97]. The OPF was solved considering a dynamic loading environment using the traditional control variables. A Tabu search algorithm, was employed to solve the OPF with non-convex fuel cost functions in reference [98]. Similar solution approach was developed in reference [99] to solve the OPF problem while considering various objectives such as voltage proles enhancement and the convex and non-convex fuel cost functions. SA technique was proposed to minimize the convex fuel cost function in solving the OPF in reference [100,101]. The optimal settings and placements of FACTS devices were included in the OPF formulation considered by Bhasaputra and Ongsakul [102104]. A hybridization of the Tabu search, SA, and quadratic programming

700

M.R. AlRashidi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 694702 [7] M. Huneault, F.D. Galiana, A survey of the optimal power ow literature, IEEE Transactions on Power Systems 6 (2) (1991) 762770. [8] L.J. Fogel, On the organization of intellect, Ph.D. thesis, University of California, 1964. [9] J. Kennedy, R.C. Eberhart, Swarm Intelligence, Morgan Kaufmann, San Francisco, 2001. [10] J.H. Holland, Adaptation in Natural and Articial Systems, University of Michigan Press, 1975. [11] W.A. Chang, R.S. Ramakrishna, Elitism-based compact genetic algorithms, IEEE Transactions on Evolutionary Computation 7 (4) (2003) 367385. [12] J.J. Hopeld, D.W. Tank, Neural computation of decisions in optimization problems, Biological Cybernetics 52 (1985) 141152. [13] S. Kirkpatrick, C.D. Gelatt Jr., M.P. Vecchi, Optimization by simulated annealing, Science 220 (4598) (1983) 671680. [14] V. Cerny, Thermodynamical approach to the traveling salesman problem: an efcient simulation algorithm, Journal of Optimization Theory and Application 45 (1) (1985) 4151. [15] N. Metropolis, A.W. Rosenbluth, M.N. Rosenbluth, A.H. Teller, Equation of state calculation by fast computing machines, Journal of Chemical Physics 21 (1953) 10871092. [16] L.A. Zadeh, Fuzzy sets, Journal of Information and Control 8 (1965) 338 353. [17] R.C. Bansal, Bibliography on the fuzzy set theory applications in power systems (19942001), IEEE Transactions on Power Systems 18 (4) (2003) 12911299. [18] M. Dorigo, V. Maniezzo, A. Colorni, Positive feedback as a search strategy, Dipartimento di Elettronica, Politecnico di Milano, Italy, Tech. Rep. 91-016, 1991. [19] M. Dorigo, Optimization, learning and natural algorithms (in italian), Ph.D. dissertation, Dipartimento di Elettronica, Politecnico di Milano, Italy, 1992. [20] J. Kennedy, R. Eberhart, Particle swarm optimization, in: IEEE International Conference on Neural Networks, vol. 4, Perth, Australia, 1995, pp. 19421948. [21] R. Eberhart, J. Kennedy, A new optimizer using particle swarm theory, in: Proceedings of the Sixth International Symposium on Micro Machine and Human Science, Nagoya, Japan, 1995, pp. 3943. [22] C.W. Reynolds, Flocks, herds, and schools: a distributed behavioral model, Computer Graphics 21 (4) (1987) 2534. [23] F. Heppner, U. Grenander, A stochastic nonlinear model for coordinated bird ocks, The Ubiquity of Chaos (1990) 233238. [24] M.R. AlRashidi, M.E. El-Hawary, A survey of particle swarm optimization applications in electric power systems, IEEE Transactions on Evolutionary Computation, (2006), in press. [25] Y. Valle, G.K. Venayagamoorthy, S. Mohagheghi, J. Hernandez, R.G. Harley, Particle swarm optimization: basic concepts, variants and applications in power systems, IEEE Transactions on Evolutionary Computation 12 (2) (2008) 171195. [26] Computational Intelligence Library [Online], Available: http://cilib. sourceforge.net/, 2008. [27] Decision Tree for Optimization Software [Online], Available: http://plato. asu.edu/sub/global.html, 2008. [28] Genetic Programming [Online], Available: http://www.geneticprogramming.org/, 2008. [29] Articial Neural Network [Online], Available: http://www. learnarticialneuralnetworks.com/, 2008. [30] Particle Swarm Optimization [Online], Available: http://www. swarmintelligence.org/, 2008. [31] Differential Evolution [Online], Available: http://www.icsi.berkeley. edu/storn/code.html, 2008. [32] Z. Gaing, H.-S. Huang, Real-coded mixed-integer genetic algorithm for constrained optimal power ow, in: IEEE TENCON Region 10 Conference, 2004, pp. 323326. [33] L.L. Lai, J.T. Ma, R. Yokoyama, M. Zhao, Improved genetic algorithms for optimal power ow under both normal and contingent operation states, International Journal of Electrical Power & Energy Systems 19 (June (5)) (1997) 287292. [34] S.R. Paranjothi, K. Anburaja, Optimal power ow using rened genetic algorithm, Electric Power Components & Systems 30 (October (10)) (2002) 10551063. [35] Z. Gaing, R.-F. Chang, Security-constrained optimal power ow by mixedinteger genetic algorithm with arithmetic operators, IEEE Power Engineering Society General Meeting (2006) 18. [36] X. Zhang, R.W. Dunn, F. Li, Stability constrained optimal power ow for the balancing market using genetic algorithms, IEEE Power Engineering Society General Meeting (2003) 932937. [37] A.G. Bakirtzis, P.N. Biskas, C.E. Zoumas, V. Petridis, Optimal power ow by enhanced genetic algorithm, IEEE Transactions on Power Systems 17 (2) (2002) 229236. [38] P.N. Biskas, N.P. Ziogos, A. Tellidou, C.E. Zoumas, A.G. Bakirtzis, V. Petridis, A. Tsakoumis, Comparison of two metaheuristics with mathematical programming methods for the solution of OPF, in: Proceeding of the 13th International Conference on Intelligent Systems Application to Power Systems, 2005, pp. 510515. [39] P.N. Biskas, N.P. Ziogos, A. Tellidou, C.E. Zoumas, A.G. Bakirtzis, V. Petridis, Comparison of two metaheuristics with mathematical programming methods for the solution of OPF, in: IEE Proceedings-Generation, Transmission and Distribution, vol. 153, no. 1, 2006, pp. 1624.

was formed to minimize the quadratic fuel cost function. Lin et al. developed an ordinal optimization theory-based algorithm to solve the discrete OPF [105]. They investigated the validity of the proposed algorithm with the objectives being fuel cost function and real power losses. The authors in reference [106] augmented the traditional quadratic fuel cost function by adding a linear term to account for the gas supply cost. This OPF model is based on an integrated gas and electric network that represents part of the Belgian energy transport system. A hybrid tool that combines evolutionary strategy algorithm with Newtons and interior point method was proposed to solve the problem. Differential evolution approach was developed in reference [107] to minimize the transmission network losses via proper adjustments of the OPF control variables. The same technique was used in reference [108] to minimize the fuel cost function while accounting for FACTS devices. Hugang et al. proposed an adaptive immune algorithm for optimal reactive power dispatch [109]. They formulated the OPF problem as a multiobjective one with the objectives being minimizing the system active power losses while simultaneously enhancing the voltage proles. 5. Conclusions This paper addresses major research work applying computational intelligence tools to solve the OPF problem. It highlights the main difculties facing the continuously changing OPF that led to the use of computational intelligence tools as supplements to traditional tools, which are commonly used by power engineers. This utilization alleviated some difculties previously encountered in analyzing and solving power system problems such as the need to use oversimplied modeling schemes to make theses problems suitable for traditional computational methods and to account for modeling uncertainty in different elements of power systems. A brief discussion of major computational intelligence tools is presented in addition to the OPF general problem formulation. The reviewed articles are organized based on the main computational intelligence tool used which helps the potential OPF researchers in easily identifying future research directions. Most of the computational intelligence tools are population based methods which, as a result, make them suitable for parallel processing. This key feature can signicantly reduce the OPF computation time, hence; increase the possibility of making the OPF an online tool for power systems analysts. Acknowledgement The rst author would like to thank the Public Authority of Applied Education and Training for their moral and nancial support. References
[1] R.B. Squires, Economic dispatch of generation directly from power system voltages and admittances, AIEE Transactions on Power Apparatus and Systems PAS-79 (3) (1961) 12351245. [2] J. Carpentier, Contribution e ltude do Dispatching Economique, Bulletin Society Francaise Electriciens (1962) 431447. [3] H.W. Dommel, W.F. Tinney, Optimal power ow solutions, IEEE Transactions on Power Apparatus and Systems PAS-87 (10) (1968) 18661876. [4] J.A. Momoh, Electric Power System Applications of Optimization, Marcel Dekker, New York, 2001. [5] J.A. Momoh, R. Adapa, M.E. El-Hawary, A review of selected optimal power ow literature to 1993. I. Nonlinear and quadratic programming approaches, IEEE Transactions on Power Systems 14 (1) (1999) 96104. [6] J.A. Momoh, M.E. El-Hawary, R. Adapa, A review of selected optimal power ow literature to 1993. II. Newton, linear programming and interior point methods, IEEE Transactions on Power Systems 14 (1) (1999) 105111.

M.R. AlRashidi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 694702 [40] M. Todorovski, D. Rajicic, An initialization procedure in solving optimal power ow by genetic algorithm, IEEE Transactions on Power Systems 21 (2) (2006) 480487. [41] M. Todorovski, D. Rajicic, A power ow method suitable for solving OPF problems using genetic algorithms, The IEEE Region 8 EUROCON 2 (2003) 215219. [42] D. Devaraj, B. Yegnanarayana, Genetic-algorithm-based optimal power ow for security enhancement, in: IEE Proceedings-Generation, Transmission and Distribution, vol. 152, no. 6, 2005, pp. 899905. [43] D.B. Das, C. Patvardhan, Useful multi-objective hybrid evolutionary approach to optimal power ow, in: IEE Proceedings-Generation, Transmission and Distribution, vol. 150, no. 3, 2003, pp. 275282. [44] M.A. Abido, Multiobjective optimal power ow using strength Pareto evolutionary algorithm, in: The 39th International Universities Power Engineering Conference, 2004, pp. 457461. [45] L.L. Lai, J.T. Ma, Power ow control with UPFC using genetic algorithms, in: International Conference on Intelligent Systems Applications to Power Systems, 1996, pp. 373377. [46] N.P. Padhy, M.A. Abdel-Moamen, B.J. Kumar, Optimal location and initial parameter settings of multiple TCSCs for reactive power planning using genetic algorithms, IEEE Power Engineering Society General Meeting (2004) 11101114. [47] K.R.S. Reddy, N.P. Padhy, R.N. Patel, Congestion management in deregulated power system using FACTS devices, in: IEEE Power India Conference, 2006, pp. 815. [48] H.C. Leung, T.S. Chung, Optimal power ow with a versatile FACTS controller by genetic algorithm approach, IEEE Power Engineering Society Winter Meeting (2000) 28062811. [49] H.C. Leung, T.S. Chung, Optimal power ow with a versatile FACTS controller by genetic algorithm approach, in: International Conference on Advances in Power System Control, Operation and Management, 2000, pp. 178183. [50] T.S. Chung, Y.Z. Li, A hybrid GA approach for OPF with consideration of FACTS devices, IEEE Power Engineering Review 21 (2) (2001) 4750. [51] V.P. Gountis, A.G. Bakirtzis, Bidding strategies for electricity producers in a competitive electricity marketplace, IEEE Transactions on Power Systems 19 (1) (2004) 356365. [52] Z.-Y. Dong, D.J. Hill, Power system reactive scheduling within electricity markets, in: International Conference on Advances in Power System Control, Operation and Management, 2000, pp. 7075. [53] B. Mozafari, A.M. Ranjbar, A.R. Shirani, A. Barkeseh, A comprehensive method for available transfer capability calculation in a deregulated power, in: IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies, 2004, pp. 680685. [54] P. Somasundaram, K. Kuppusamy, R.P. Kumudini Devi, Evolutionary programming based security constrained optimal power ow, Electric Power Systems Research 72 (December (2)) (2004) 137145. [55] Y.R. Sood, Evolutionary programming based optimal power ow and its validation for deregulated power system analysis, International Journal of Electrical Power & Energy Systems 29 (January (1)) (2007) 6575. [56] W. Ongsakul, T. Tantimaporn, Optimal power ow by improved evolutionary programming, Electric Power Components & Systems 34 (January (1)) (2006) 7995. [57] L.L. Lai, J.T. Ma, Power ow control in FACTS using evolutionary programming, in: IEEE International Conference on Evolutionary Computation, 1995, pp. 109113. [58] B. Venkatesh, M.K. George, H.B. Gooi, Fuzzy OPF incorporating UPFC, in: IEE Proceedings-Generation, transmission and Distribution, vol. 151, no. 5, 2004, pp. 625629. [59] W. Ongsakul, P. Jirapong, Calculation of total transfer capability by evolutionary programming, in: IEEE Region 10 Conference, 2004, pp. 492495. [60] W. Ongsakul, P. Jirapong, Optimal allocation of FACTS devices to enhance total transfer capability using evolutionary programming, in: IEEE International Symposium on Circuits and Systems, 2005, pp. 41754178. [61] K.-H. Kim, J.-K. Lee, S.-B. Rhee, S.-K. You, Security constrained OPF by hybrid algorithms in interconnected power systems, in: IEEE Power Engineering Society Summer Meeting, 2001, pp. 15911596. [62] L. Shi, G. Xu, Z. Hua, Part II.-application of heuristic evolutionary programming in solution of the optimal power ow, in: International Conference on Power System Technology, 1998, pp. 767770. [63] C.H. Lo, C.Y. Chung, D.H.M. Nguyen, K.P. Wong, Parallel evolutionary programming for optimal power ow, in: IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies, 2004, pp. 190195. [64] C.H. Lo, C.Y. Chung, D.H.M. Nguyen, K.P. Wong, A parallel evolutionary programming based optimal power ow algorithm and its implementation, in: International Conference on Machine Learning and Cybernetics, 2004, pp. 25432548. [65] N.P. Padhy, Wheeling using evolutionary programming based optimal power ow algorithm, in: International Conference on Advances in Power System Control, Operation and Management, 2000, pp. 144148. [66] R. Gnanadass, T.G. Palanivelu, N.P. Padhy, Determination of SRMC with nonsmooth fuel cost functions using evolutionary programming, Electric Power Components & Systems 33 (January (1)) (2005) 85102. [67] Y.R. Sood, N.P. Padhy, H.O. Gupta, A new hybrid model for wheeling cost analysis under deregulated environment, in: IEEE PES Transmission and Distribution Conference and Exposition, 2003, pp. 97102.

701

[68] J. Yuryevich, K.P. Wong, Evolutionary programming based optimal power ow algorithm, IEEE Transactions on Power Systems 14 (4) (1999) 12451250. [69] Y.R. Sood, S. Verma, N.P. Padhy, H.O. Gupta, Evolutionary programming based algorithm for selection of wheeling options, IEEE Power Engineering Society Winter Meeting (2001) 545552. [70] R.S. Hartati, M.E. El-Hawary, Optimal active power ow solutions using a modied Hopeld neural network, in: Canadian Conference on Electrical and Computer Engineering, 2001, pp. 189194. [71] T.T. Nguyen, Neural network optimal-power-ow, in: Fourth International Conference on Advances in Power System Control, Operation and Management, 1997, pp. 266271. [72] M.C. Dondo, M.E. El-Hawary, An approach to implement electricity metering in real-time using articial neural networks, IEEE Transactions on Power Delivery 18 (2) (2003) 383 386. [73] X. Luo, A.D. Patton, C. Singh, Quickprop algorithm for transfer capability calculations, IEEE Power Engineering Society Winter Meeting (1999) 890894. [74] X. Luo, A.D. Patton, C. Singh, Real power transfer capability calculations using multi-layer feed-forward neural networks, IEEE Transactions on Power Systems 15 (2) (2000) 903908. [75] X. Luo, C. Singh, A.D. Patton, Power system reliability evaluation using self organizing map, IEEE Power Engineering Society Winter Meeting (2000) 11031108. [76] B. Venkatesh, Online ANN memory model-based method for unied OPF and voltage stability margin maximization, Electric Power Components & Systems 31 (May (5)) (2003) 453. [77] K.H. Abdul-Rahman, S.M. Shahidehpour, N.I. Deeb, Effect of EMF on minimum cost power transmission, IEEE Transactions on Power Systems 10 (1) (1995) 347355. [78] K.H. Abdul-Rahman, S.M. Shahidehpour, Static security in power system operation with fuzzy real load conditions, IEEE Transactions on Power Systems 10 (1) (1995) 7787. [79] V.C. Ramesh, X. Li, A fuzzy multiobjective approach to contingency constrained OPF, IEEE Transactions on Power Systems 12 (3) (1997) 13481354. [80] T.K. Hahn, M.K. Kim, D. Hur, J.K. Park, Y.T. Yoon, Evaluation of available transfer capability using fuzzy multi-objective contingency-constrained optimal power ow, Electric Power Systems Research 78 (May (5)) (2008) 873882. [81] X. Liu, J. Li, H. Li, H. Peng, Fuzzy modeling and interior point algorithm of multi-objective OPF with voltage security margin, in: IEEE/PES Transmission and Distribution Conference and Exhibition: Asia and Pacic, 2005, pp. 16. [82] Y.-C. Wu, Fuzzy second correction on complementarity condition for optimal power ows, IEEE Transactions on Power Systems 16 (3) (2001) 360366. [83] N.P. Padhy, M.A.R. Abdel-Moamen, P.K. Trivedi, B. Das, A hybrid model for optimal power ow incorporating FACTS devices, in: IEEE Power Engineering Society Winter Meeting, 2001, pp. 510515. [84] N.P. Padhy, Congestion management under deregulated fuzzy environment, in: IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies, 2004, pp. 133 139. [85] M.A. Abido, Optimal power ow using particle swarm optimization, International Journal of Electrical Power & Energy Systems 24 (7) (2002) 563571. [86] B. Zhao, C.X. Guo, Y.J. Cao, Improved particle swam optimization algorithm for OPF problems, in: IEEE/PES Power Systems Conference and Exposition, 2004, pp. 233238. [87] S. He, J.Y. Wen, E. Prempain, Q.H. Wu, J. Fitch, S. Mann, An improved particle swarm optimization for optimal power ow, in: International Conference on Power System Technology, vol. 2, The Pan Pacic, Singapore, 2004, pp. 16331637. [88] C. Wang, H. Yuan, Z. Huang, J. Zhang, C. Sun, A modied particle swarm optimization algorithm and its application in optimal power ow problem, in: Proceedings of 2005 International Conference on Machine Learning and Cybernetics, vol. 5, Guangzhou, China, 2005, pp. 28852889. [89] R. Ma, P. Wang, H. Yang, G. Hu, Environmental/economic transaction planning using multiobjective particle swarm optimization and non-stationary multi-stage assignment penalty function, in: 2005 IEEE/PES Transmission and Distribution Conference and Exhibition: Asia and Pacic, Dalian, China, 2005, pp. 16. [90] J.G. Vlachogiannis, K.Y. Lee, A comparative study on particle swarm optimization for optimal steady-state performance of power systems, IEEE Transactions on Power Systems 21 (4) (2006) 17181728. [91] Z.L. Gaing, Constrained optimal power ow by mixed-integer particle swarm optimization, in: IEEE Power Engineering Society General Meeting, 2005, pp. 243250, San Francisco, USA. [92] N. Mo, Z.Y. Zou, K.W. Chan, T.Y.G. Pong, Transient stability constrained optimal power ow using particle swarm optimisation, in: IET Proceedings of Generation, Transmission & Distribution, vol. 1, no. 3, 2007, pp. 476 483. [93] P.E. Onate, J.M. Ramirez, Optimal power ow solution with security constraints by a modied PSO, in: IEEE Power Engineering Society General Meeting, 2007, pp. 16. [94] M.R. AlRashidi, M.E. El-Hawary, Hybrid Particle Swarm Optimization Approach for Solving the Discrete OPF Problem Considering the Valve Loading Effects, IEEE Transactions on Power Systems 22 (4) (2007) 20302038. [95] M. Saravanan, S.M.R. Slochanal, P. Venkatesh, J.P.S. Abraham, Application of particle swarm optimization technique for optimal location of FACTS devices

702

M.R. AlRashidi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 694702 considering cost of installation and system loadability, Electric Power Systems Research 77 (March (34)) (2007) 276283. M. Tripathy, S. Mishra, Bacteria foraging-based solution to optimize both real power loss and voltage stability limit, IEEE Transactions on Power Systems 22 (1) (2007) 240248. W.J. Tang, M.S. Li, S. He, Q.H. Wu, J.R. Saunders, Optimal power ow with dynamic loads using bacterial foraging algorithm, in: International Conference on Power System Technology, 2006, pp. 15. T. Kulworawanichpong, S. Sujitjorn, Optimal power ow using tabu search, IEEE Power Engineering Review 22 (6) (2002) 3755. M.A. Abido, Optimal power ow using tabu search algorithm, Electric Power Components & Systems 30 (2002) 469483. C.A. Roa-Sepulveda, B.J. Pavez-Lazo, A solution to the optimal power ow using simulated annealing, in: IEEE Porto Power Tech Conference, 2001, pp. 510. C.A. Roa-Sepulveda, B.J. Pavez-Lazo, A solution to the optimal power ow using simulated annealing, International Journal of Electrical Power & Energy Systems 25 (January (1)) (2003) 4757. P. Bhasaputra, W. Ongsakul, Optimal placement of multi-type FACTS devices by hybrid TS/SA approach, in: Proceedings of the 2003 International Symposium on Circuits and Systems, 2003, pp. 375378. [103] P. Bhasaputra, W. Ongsakul, Optimal power ow with multi-type of FACTS devices by hybrid TS/SA approach, in: IEEE International Conference on Industrial Technology, 2002, pp. 285290. [104] W. Ongsakul, P. Bhasaputra, Optimal power ow with FACTS devices by hybrid TS/SA approach, International Journal of Electrical Power & Energy Systems 24 (December (10)) (2002) 851857. [105] S.-Y. Lin, Y.-C. Ho, C.-H. Lin, An ordinal optimization theory-based algorithm for solving the optimal power ow problem with discrete control variables, IEEE Transactions on Power Systems 19 (1) (2004) 276286. [106] C. Unsihuay, J.W.M. Lima, A.C.Z. de Souza, Modeling the integrated natural gas and electricity optimal power ow, in: IEEE Power Engineering Society General Meeting, 2007, pp. 17. [107] M. Varadarajan, K.S. Swarup, Network loss minimization with voltage security using differential evolution, Electric Power Systems Research 78 (May (5)) (2008) 815823. [108] M. Basu, Optimal power ow with FACTS devices using differential evolution, International Journal of Electrical Power & Energy Systems 30 (February (2)) (2008) 150156. [109] X. Hugang, C. Haozhong, L. Haiyu, Optimal reactive power ow incorporating static voltage stability based on multi-objective adaptive immune algorithm, Energy Conversion and Management 49 (5) (2007) 11751181.

[96]

[97]

[98] [99] [100]

[101]

[102]