Académique Documents
Professionnel Documents
Culture Documents
CHAPTER-1 INTRODUCTION
1.1 GENERAL
The main objective of electric power utilities is to provide high quality reliable supply to the consumers at the lowest possible cost while operating to meet the limits and constraint imposed on the generating units. This formulates the well-known Economic Load Dispatch (ELD) problem for finding the optimal combination of the output power of all online generating units that minimizes the total fuel cost, while satisfying all constraints. The economic load dispatch (ELD) is an important function in modern power system like unit commitment, Load Forecasting, Security Analysis, Scheduling of fuel purchase etc. A bibliographical survey on ELD methods reveals that various numerical optimization techniques have been employed to approach the ELD problem. The Optimal Power Flow (OPF) is an important criterion in todays power system operation and control due to scarcity of energy resources, increasing power generation cost and ever growing demand for electric energy. As the size of the power system increases, load may be varying. The generators should share the total demand plus losses among themselves. The sharing should be based on the fuel cost of the total generation with respect to some security constraints. Generally, most of the approaches apply sensitivity analysis and gradient-based optimization algorithms by linearizing the objective function and system constraints around an operating point. Unfortunately, the problems of OPF are highly nonlinear and a multi model optimization problems, i.e. there exist more than one local optimum. Therefore, conventional optimization methods that make use of derivatives and gradients are, in general, not able to locate or identify the global optimum. Heuristic algorithms such as genetic algorithms (GA) and evolutionary programming have been recently proposed for solving the OPF problem. The results reported were promising and encouraging for further research in this direction. Unfortunately, recent research has identified some deficiencies in GA performance. This degradation in efficiency is apparent in applications with highly epistatic objective function, i.e. where the parameters being optimized are highly correlated. In
2
addition, the premature convergence of GA degrades its performance and reduces its search capability. Recently, a new evolutionary computation technique, called Particle Swarm Optimization (PSO), has been proposed and introduced. This technique combines social psychology principles in socio-cognition human agents and evolutionary computations. PSO has been motivated by the behavior of organisms such as fish schooling and bird flocking. Generally, PSO is characterized as simple in concept, easy to implement, and computationally efficient. Unlike the other heuristic techniques, PSO has a flexible and well-balanced mechanism to enhance and adapt to the global and local exploration abilities. ELD is solved traditionally using mathematical programming based on optimization techniques such as lambda iteration, gradient method and so on. Economic load dispatch with piecewise linear cost functions is a highly heuristic, approximate and extremely fast form of economic dispatch. Complex constrained ELD is addressed by intelligent methods. Among these methods, some of them are genetic algorithm (GA) and, evolutionary programming (EP), dynamic programming (DP), tabu search, hybrid EP, neural network (NN), adaptive Hopfield neural network (AHNN), particle swarm optimization (PSO) etc. For calculation simplicity, existing methods use second order fuel cost functions which involve approximation and constraints are handled separately, although sometimes valve-point effects are considered. Lambda iteration, gradient method can solve simple ELD calculations and they are not sufficient for real applications in deregulated market. However, they are fast. There are several Intelligent methods among them genetic algorithm applied to solve the real time problem of solving the economic load dispatch problem, whereas some of the works are done by Evolutionary algorithm. Few other methods like tabu search are applied to solve the problem. Artificial neural network are also used to solve the optimization problem. However many people applied the swarm behavior to the problem of optimum dispatch as well as unit commitment problem are general purpose. However, they have randomness. For a practical problem, like ELD, the intelligent methods should be modified accordingly so that they are suitable to solve economic dispatch with more accurate multiple fuel cost functions and constraints, and they can reduce randomness.
Intelligent methods are iterative techniques that can search not only local optimal solutions but also a global optimal solution depending on problem domain and execution time limit. They are general-purpose searching techniques based on principles inspired from the genetic and evolution mechanisms observed in natural systems and populations of living beings. These methods have the advantage of searching the solution space more thoroughly. The main difficulty is their sensitivity to the choice of parameters. Among intelligent methods, PSO is simple and promising. It requires less computation time and memory. It has also standard values for its parameters. In this, the Particle Swarm Optimization (PSO) is proposed as a methodology for economic load dispatch.
In recent years, many heuristic algorithms, such as genetic algorithms [17-20] and evolutionary programming [21-26], simulated annealing [27], particle swarm optimization [28-35], chaos optimization algorithm [36, 37], tabu search [38, 40] have been proposed for solving the OPF problem, without any restrictions on the shape of the cost curves. A genetic algorithm approach is applied for ac-dc optimal power flow problem in [17]. In [18], improved genetic algorithm for optimal power flow solutions under both normal and contingent operation states is proposed. An initialization procedure in solving optimal power flow by genetic algorithm is proposed in [19]. In [20], the simple GA with an added set of advanced and problem specific genetic operators in order to increase its convergence speed and improve the quality of solutions is applied to OPF. Multi-objective optimal power flow is solved using strength pareto evolutionary algorithm in [22]. Improved evolutionary programming is applied for OPF in [23]. Improved evolutionary programming with various crossover techniques is used in [24] to solve OPF problem. Meanwhile, an improved evolutionary programming [25] was successfully used to solve combinatorial optimization problems. In [26], a multi-objective hybrid evolutionary strategy is presented for the solution of the comprehensive model of OPF. Optimal power flow subject to security constraints id solved in [29] with a particle swarm optimizer. In [30], improved particle swarm optimization algorithm for OPF problems is proposed. In [31], modified particle swarm optimization algorithm is presented in which particles not only studies from itself and the best one but also from other individuals. An efficient mixed-integer particle swarm optimization with mutation scheme for solving the constrained optimal power flow with a mixture of continuous and discrete control variables and discontinuous fuel cost functions is presented in [32]. In [33], an improved particle swarm optimization algorithm for OPF problems, which incorporates non-stationary multistage assignment penalty function is proposed. Optimal power flow constrained by transient stability is solved with improved particle swarm optimization in [34]. Authors in [35] used particle swarm optimization to solve OPF with generator capability curve constraint. A hybrid algorithm using the chaos optimization and the linear interior point algorithm is developed in [36] for optimal power flow. In [37], a hybrid algorithm of chaos optimization and slp for optimal power flow problems with multimodel
characteristic is considered. The results reported by the above methods were promising and encouraging for further research in this direction. Moreover, to enhance the search efficiency, many hybrid algorithms have been introduced for solving the power system optimization problems. For instance, a hybrid tabu search and simulated annealing [39] was applied to solve the OPF with flexible alternating current transmission systems (FACTS) device problem; a hybrid evolutionary programming and tabu search or improved tabu search [40] was used to solve the economic dispatch problem with non-smooth cost functions. An ordinal optimization theory based algorithm to solve OPF problem is proposed in [41]. Optimal power flow for a system of micro grids with controllable loads is solved with particle swarm optimization in [42]. In recent past, power full evolutionary algorithm such as differential evolution techniques are employed for power system optimization problems. Storn and Price [43] developed the DE, and it is a numerical optimization approach that is easy to implement, significantly faster and robust. DE can be used to minimize nonlinear and non-differentiable continuous space functions with real valued parameters. The most important characteristics of differential evolution is that it uses the differences of randomly sampled pairs of object vectors to guide the mutation operation instead of using the probability distribution function as other evolutionary algorithms. Differential evolution combines simple arithmetic operators with the classical operators of crossover mutation and selection to evolve from a randomly generated starting population to a final solution. The fittest of an offspring competes on-to-one with that of corresponding parent, which is different from other evolutionary algorithms. This on-to-one competition gives rise to faster convergence rate. The differential evolution has been successfully applied to various power system optimization problems such as generation expansion planning [44], hydrothermal scheduling [45]. Figueroa and Cederio [46] applied DE for power system state estimation. Coelho and Mariani [47] used this algorithm for economic dispatch with valve-point effect. M.Basu [48] applied DE for OPF incorporating FACTS devices. In this project report, evolutionary programming and particle swarm optimization algorithms are developed to effectively solve the optimal power flow
6
problem incorporating a set of constraints. Simulations for evolutionary OPF are carried out on various IEEE test systems with different objective functions.
2.1 INTRODUCTION
The great importance of power flow or load-flow studies is in the planning the future expansion of power systems as well as in determining the best operation of existing systems. The principal information obtained from the power flow study is the magnitude and phase angle of the voltage at each bus and the real and reactive power flowing in each line. Electrical transmission systems operate in their steady state mode under normal conditions. Three major problems encountered in steady state mode of operations are listed below in their hierarchical order of importance: 1. Load flow problem 2. Optimal load dispatch problem 3. Systems control problem The computational procedure required to determine the steady-state operating characteristics of a power system network is termed load flow or power flow. The aim of the power flow calculations is to determine the steady-state operating characteristics of a power generation/transmission system for a given set of bus bar loads. Active power generations are specified according to economic dispatching. The magnitude of generation voltage is maintained at the specified level by an automatic voltage regulator acting on the machine excitation. Loads are specified by their constant active and reactive power requirements. The loads are assumed to be
9
unaffected by the small variations of voltage and frequency expected during normal steady-state operation. The direct analysis of the network is not possible, as the loads are given in terms of complex powers rather than impedances. The generators behave more like power sources than voltage sources. The main information obtained from the load flow study consists of 1. Magnitudes and phase angles of load bus voltages. 2. Reactive powers and voltage phase angles at generator buses. 3. Real and reactive power flow on transmission lines. 4. Power at the reference bus. This information is essential for the continuous monitoring of the current state of the system. The information is also important for analyzing the effectiveness of the alternative plans for the future, such as adding new generator sites, meeting increased load demand and locating new transmission sites. The single line diagram of a power system having four buses is shown in the figure. In the power system, the variables defined on each bus are 1. Complex powers, and , . ,
10
There results a net injection of power into the transmission system. The transmission system may be a primary transmission system or sub transmission system. The primary transmission system transmits bulk power from the generators to the bulk power stations. The sub-transmission system transmits power from the substations or some old generators to the distribution systems. The transmission system has to be designed in such a manner that the power system operation is reliable and economic and no difficulties are encountered in its operation. The difficulties are involved, however are: 1. One or more transmission lines becoming over loaded 2. Generators becoming over loaded 3. The stability margins for a transmission link being too small There may be emergencies such as 1. The loss of one or more transmission links 2. Shut down of some generators which give rise to overloading of other generators and transmission lines. In system operation and planning, the voltages and powers are kept within certain limits. The power system networks of today are highly complicated consisting of hundreds of buses and transmission links. Thus, the load flow study involves extensive calculations. With the advent of fast digital computers with huge memory, all kinds of power system studies including the load flow study can now be used which can be: 1. Accurate or approximate 2. Unadjusted or adjusted 3. Offline or online 4. Single case or multiple cases. 2.2 LOAD FLOW PROBLEM The complex power injected by the source into the bus of a power system is (2.1) Substituting the value of ( Where Is the voltage of the bus ) in the above power equation. (2.2)
Equating the real and imaginary parts ( ( ( ( Where Is the real power Is the reactive power Let Where | | Is the magnitude of voltage Is the angle of the voltage Is the load angle Substituting for , and | | = | | | | Or | || Or | || | Or | | | | | | ) (2.8) (2.7) | (2.6) | | (2.5) | | | | , =| | , | | )) )) (2.3) (2.4)
Separating the real and imaginary parts of the above equation to get real and reactive powers, | | | | Where (2.11) (2.12) | | =| | | | )| | (i=1,2,3,.........NB) (2.13) (2.14) | || | || | | (2.9) (2.10)
| | | |
[ [
] ]
(i=1,2,3,.........NB) (i=1,2,3,.........NB)
(2.15) (2.16)
Equations 2.15 and 2.16 are called power flow equations. These are NB real and NB reactive power flow equations giving a total of 2 NB power flow equations. At each bus these are four variables, namely| |, , and , giving a total of 4 NB
variables(for NB buses). If at every bus two variables are specified (thus specifying a total of 2 NB variables), the remaining two variables at every bus (a total of 2 NB remaining variables) can be found by solving 2 NB power flow equations. In a physical system, the variables are specified depending upon what kind of devices are connected to that bus. In general fur types of buses are defined. 2.2.1 Classification of Buses Depending upon which quantities have been specified at each bus, buses are classified into four categories which are given below 1. Slack bus/Swing bus/Reference bus 2. PQ bus/Load Bus 3. PV Bus /Generator Bus 4. Voltage-controlled buses Slack bus/Swing bus/Reference bus In a load flow study, real and reactive powers cannot be fixed a priori at all the buses as the net complex power flow into the network is not known in advance. So, the system power losses are unknown till the load flow study is complete. It is therefore necessary to have one bus at which complex power is unspecified so that it supplies the difference in the total system load plus losses and the sum of the complex powers specified at the remaining buses. Such a bus is known as slack bus and must be a generator bus. If slack bus is not specified, the bus connected to the largest generating station is normally selected as the slack bus. Usually this bus is numbered 1 for the load flow studies. This bus sets the angular reference for all the other buses. Since it is the angle difference between two voltage sources that dictates the real and reactive power flow between them, the particular angle of the slack bus is not important. However it sets the reference against which angles of all the other bus voltages are measured. For this reason the angle of
13
this bus is usually chosen as 0. Furthermore it is assumed that the magnitude of the voltage of this bus is known. Voltage magnitude and voltage phase angle are specified. Normally, voltage magnitude is set to 1pu and voltage angle is set to zero. The real and reactive powers are not specified. The known parameters are voltage magnitude | | and voltage angle . The unknown parameters are real power PQ bus/Load Bus In these buses no generators are connected and hence the generated real power and reactive power defined by real power are taken as zero. The load drawn by these buses are and reactive power in which the negative sign and reactive power .
accommodates for the power flowing out of the bus. This is why these buses are sometimes referred to as P-Q bus. The objective of the load flow is to find the bus voltage magnitude |Vi| and its angle i. In a power system 80% of the buses are P-Q buses. A pure load bus is a PQ bus. A load bus has no generating facility (i.e . At this type of bus, the net real power are known as = Where , and = (2.17) , are and the reactive power
are the real and reactive power generations at the bus respectively. and
are the real and reactive power demands at the respectively. known from the load forecasting and on bus are real power and
magnitude and voltage angle. The PQ buses are the most common comprising almost 85% of all the buses in a given power system. PV Bus /Generator Bus These are the buses where generators are connected. Therefore the power generation in such buses is controlled through a prime mover while the terminal voltage is controlled through the generator excitation. Keeping the input power constant through turbine-governor control and keeping the bus voltage constant using automatic voltage regulator, we can specify constant PGI and | Vi | for these buses. This is why such buses are also referred to as P-V buses. It is to be noted that the reactive power supplied by the generator QGI depends on the system configuration and
14
cannot be specified in advance. Furthermore we have to find the unknown angle i of the bus voltage. In a power system 10% of the buses are P-V buses. A generator is always connected to a PV bus. Hence the net power as is also known from load forecasting. The knowns are real power and voltage angle magnitude | |. The unknowns are reactive power , known
comprise about 15% of all the buses in o power system. Voltage-controlled buses Generally the PV buses and the voltage-controlled buses are grouped together but these buses have physical difference. The voltage-controlled bus has also voltage control capabilities, and uses a tap-adjustable transformer and/or a static VAR compensator a instead of a generator. Hence, =0 at these buses. Thus =and = at these buses. The
The limit arises due to the fact that the power system equipment is designed to operate values. Certain of the voltage angles (state variables) must satisfy | | | | (2.19) at fixed voltages with in allowable variations of % of rated
This constraint limits the maximum permissible power angle of transmission line connecting buses i and k and is imposed by considerations of stability. Owing to physical limitations of P and/or Q, generation sources are constrained. (i=1,2,.NB) (2.20)
(i=1,2,.NB)
(2.21)
15
The load flow equations are non-linear algebraic equations and have to be solved through iterative numerical techniques, etc. At the cost of solution accuracy, it is possible to linearize load flow equations by making suitable assumptions and approximations so that fast and explicit solutions become possible.
3. Effect on convergence due to choice of slack bus. Because of the above drawbacks, use of Gauss-Seidel method is limited only to systems with smaller number of buses.
17
4. The number of iterations required in this method is almost independent of the system size. 5. N-R method is suitable for large power systems. Disadvantages 1. Difficult solution technique. 2. More calculations in each iteration resulting in large computer time per iteration and the large requirement of computer memory but the last drawback can be overcome through a compact storage scheme.
R (maximum number of iterations) , (tolerance in convergence). 2. Form the 3. Assume initially | | (i=NV+1, NV+2NB) and (i=2, 3NB)
4. Set the iteration count, r=0. 5. Compute , using the equations. [ (i=2,3,..NB) Compute NV+2, ..NB) 6. If maximum { then GOTO step15. 7. Compute Jacobian matrix elements using the equations When i=k } [ ] (i=NV+1, ] (i=2,3,.NB)
18
, = =
] ] ] ] ]
]=[
10. Set bus count i=2. 11. If PQ bus then check the limits of = = if if and check the limits of and set. and set.
If the limits are violated then PV bus is temporarily converted to PQ bus. So, compute and With updated values of = [ , and . Then calculate the change in voltage i.e ]
13. Increment the bus count, i=i+1 If i NB, then GOTI step11. 14. Advance the count, r=r+1 If r R then GOTO step5 and repeat. 15. Compute active and reactive power on slack bus i.e = = [{ [{ [ [ } } ] ] ] ]
2.4 CONCLUSION
In this chapter, information on load flow studies, their need to power system operation and different methods of load flow studies were presented.
20
21
3.1 INTRODUCTION
In past three decades, various optimization techniques have been proposed to solve the optimal power flow (OPF) problem. They range from improved mathematical techniques to more efficient problem formulation. According to difficult models in use, the OPF methods can be classified as non-compact methods where network sparsity is retained, or compact ones in which the state variables are expressed in terms of control variables using various sensitivities. Based on the applied mathematical optimization, the OPF methods can be categorized as 1. Nonlinear Programming (NLP), 2. Successive Linear Programming (SLP), and 3. Non-conventional techniques. The gradient methods, using only first order information, were initially used for the solution of OPF problem. These methods were characterized by slow and unreliable convergence soon after, the quadratic programming (QP) approaches were proposed, which use the second-order derivatives to improve the convergence of the gradient methods. Their distinct feature is that they use the Quasi-Newton process to iteratively approximate the Hessian matrix and, thus avoid the difficulty in explicitly calculating the second derivatives of the load flow equations. However, the reduced Hessian so created is dense, which may make these methods too slow as the number of control variables becomes very large As the demand increases, faster and more stable techniques grow more accurate representation of the second-order information became essential. Lagrangian techniques with the exact Hessian matrix regained engineers interest. Although these
22
methods were proposed as earlier as 1960s, few were either reliable or fast until SUN introduced a Newton approach combined with Lagrangian techniques and penalty functions. The major difficulty in this algorithm development was turned out to be the efficient identification of binding inequality constraints. In some cases the convergence is affected by the chosen initial point. Sometimes the problem becomes divergence due to chosen initial point. The initial point should be the feasible point. To overcome the above difficulties Karmarkar introduced a method called Interior point Method. He introduced a Fiacco &McCormics logarithmic barrier method for optimization with inequalities, Lagranges method for optimization with inequalities and Newton method for solving the nonlinear equations of Karush-Kuhn Tucker (KKT) optimality conditions. With their nice polynomial complexity plus computational efficiency, interior point methods have proved much faster than the traditional methods. The name interior point comes from LP notation. Namely, IPM move through the interior of the feasible region towards the optimal solution. In past fifteen years, researches on Interior Point (IP) methods experience an awesome expansion. Both Interior Point theory and computational implementation have evolved extremely fast. Interior point method variants are being extended to solve all kinds of problems from linear to nonlinear and from convex to non convex (the later with no guarantee regarding convergence). In the same way they are also being applied to solve all sorts of practical problems. Optimization of power system operation is one of the areas where Interior Point (IP) methods are being applied extensively due to size and special features of these problems.
In artificial intelligence, an evolutionary algorithm (EA) is a subset of evolutionary computation, a generic population-based metaheuristic optimization algorithm. An EA uses some mechanisms inspired by biological evolution: reproduction, mutation, recombination, and selection. Candidate solutions to the optimization problem play the role of individuals in a population, and the fitness function determines the environment within which the solutions "live". Evolution of the population then takes place after the repeated application of the above operators. Evolutionary algorithms often perform well approximating solutions to all types of problems because they ideally do not make any assumption about the underlying fitness landscape; this generality is shown by successes in fields as diverse as engineering, art, biology, economics, marketing, genetics, operations research, robotics, social sciences, physics, politics and chemistry. Apart from their use as mathematical optimizers, evolutionary computation and algorithms have also been used as an experimental framework within which to validate theories about biological evolution and natural selection, particularly through work in the field of artificial life. Techniques from evolutionary algorithms applied to the modeling of biological evolution are generally limited to explorations of microevolutionary processes, however some computer simulations, such as Tierra and Avida, attempt to model macroevolutionary dynamics. In most of real applications of EAs, computational complexity is a prohibiting factor. In fact, this computational complexity is due to fitness function evaluation. Fitness approximation is one of the solutions to overcome this difficulty. Another possible limitation of many evolutionary algorithms is their lack of a clear genotype-phenotype distinction. In nature, the fertilized egg cell undergoes a complex process known as embryogenesis to become a mature phenotype. This indirect encoding is believed to make the genetic search more robust (i.e. reduce the probability of fatal mutations), and also may improve the evolvability of the organism. Such indirect (aka generative or developmental) encodings also enable evolution to exploit the regularity in the environment. Recent work in the field of artificial embryogeny, or artificial developmental systems, seeks to address these concerns. Its main variation operator is mutation; members of the population are viewed as part of a specific species rather than members of the same species therefore each parent generates an offspring, using a ( + ) survivor selection.
24
Evolutionary Programming has developed into an extraordinary subject to study in the field of Artificial Intelligence. Evolutionary Programming is the study of programs that use simulations of biological functions in order to evolve to a specified environment. Scientists have used Evolutionary Programming techniques in order to find solutions to very complex problems. These evolving programs find good solutions to these complex problems, but not perfect solutions. The perfect solutions would take even the most advanced super computer decades to figure out. Evolutionary Programming is a method for finding solutions in smaller intervals of time, because the solutions themselves evolve over time. An initially random population of individuals (trial solutions) is created. Mutations are then applied to each individual to create new individuals. Mutations vary in the severity of their effect on the behavior of the individual. The new individuals are then compared in a "tournament" to select which should survive to form the new population. EP is similar to a genetic algorithm, but models only the behavioral linkage between parents and their offspring, rather than seeking to emulate specific genetic operators from nature such as the encoding of behavior in a genome and recombination by genetic crossover. EP is also similar to an evolution strategy (ES) although the two approaches developed independently. In EP, selection is by comparison with a randomly chosen set of other individuals whereas ES typically uses deterministic selection in which the worst individuals are purged from the population. Evolutionary algorithms, such as evolutionary programming (EP), evolution strategies (ES), genetic algorithms (GA), and genetic programming (GP), have attracted considerable interest as optimization heuristics during the past 10-15 years. Because of the exponential increase in computer power during the last decade, they are able to deal with real-world problems and new application domains are still arising. Although evolutionary algorithms are easy to implement, the underlying process is complicated and stochastic, depending on the fitness function and the free parameters controlling variation and selection. The analysis of these stochastic processes seems to be much more difficult than the analysis of randomized algorithms for special purpose. Evolutionary algorithms (EAs) are stochastic optimization methods that are based on principles derived from natural evolution. From a more general perspective, EAs are one instance of bio-inspired search heuristics. Other examples include Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO), where the search behaviors of ant colonies or insect swarms inspired a randomized search
25
technique. Since the underlying ideas of bio-inspired search are easy to grasp and easy to apply, EAs and different bio-inspired search heuristics are widely used in many practical disciplines, mainly in computer science and engineering. It is a central goal of theoretical investigations of search heuristics to assist practitioners with the tasks of selecting and designing good strategy variants and operators. Due to the rapid pace at which new strategy variants and operators are being proposed, theoretical foundations of EAs and other bio-inspired search heuristics still lag behind practice. However, EA theory has gained much momentum over the last few years and has made numerous valuable contributions to the field of evolutionary computation. Much of this momentum is due to the Dagstuhl seminars on ``Theory of Evolutionary Algorithms'', which have been held biannually since 2000. The theory of EAs today consists of a wide range of different approaches. Runtime analysis, schema theory, analyses of the dynamics of EAs, and systematic empirical analysis consider different aspects of EA behavior. Moreover, they employ different methods and tools for attaining their goals, such as Markov chains, infinite population models, or ideas based on statistical mechanics or population dynamics. In the most recent seminar, more recent types of bio-inspired search heuristics were discussed. Results regarding the runtime have been generalized from EAs to ACO and PSO. Although the latter heuristics follow a different design principle than EAs, the theoretical analyses reveal surprising similarities in terms of the underlying stochastic process. Theoretical studies of EAs in continuous domain have recently evoked interest of people working in the field of classical numerical optimization. Although stochastic and deterministic optimization algorithms address optimization of different types of problems---mainly convex and smooth for deterministic algorithms and noisy, multimodal, irregular for stochastic algorithms---the focuses of both fields became closer and closer: on the one hand many hybridizations of stochastic search and gradient-based algorithms have been proposed, on the other hand, derivative-free optimization is now a well established part of the research in the classical optimization community.
offspring (new candidates) by copying their genes, with two possible changes: crossover recombines the parental genes and mutation alters the genotype of an individual in a random way. These new candidates compete with old candidates for their place in the next generation (survival of the fittest). This process can be repeated until a candidate with sufficient quality (a solution) is found or a previously defined computational limit is reached.
3.4.1. Reproduction
Reproduction (or procreation) is the biological process by which new "offspring" individual organisms are produced from their "parents". Reproduction is a fundamental feature of all known life; each individual organism exists as the result of reproduction. The known methods of reproduction are broadly grouped into two main types: sexual and asexual. In asexual reproduction, an individual can reproduce without involvement with another individual of that species. The division of a bacterial cell into two daughter cells is an example of asexual reproduction. Asexual reproduction is not,
27
however, limited to single-celled organisms. Most plants have the ability to reproduce asexually and the ant species Mycocepurus smithii is thought to reproduce entirely by asexual means. Sexual reproduction typically requires the involvement of two individuals or gametes, one each from opposite type of sex.
3.4.2. Initialization
Initial population is one of the deciding factors for reaching the optimum, it should be carefully generated. The initial population is composed by K parent individuals. The elements of the parent are the randomly created permutation of the input variables of the generated units. Each element in a population is uniformly distributed within its feasible range.
3.4.3. Mutation
The most commonly used evolutionary operator is mutation. Mutation is the random occasional alteration of the information contained in the individual. It is performed on each element by adding a normally distributed random number with mean few and standard deviation. In molecular biology and genetics, mutations are changes in a genomic sequence: the DNA sequence of a cell's genome or the DNA or RNA sequence of a virus. They can be defined as sudden and spontaneous changes in the cell. Mutations are caused by radiation, viruses, transposons and mutagenic chemicals, as well as errors that occur during meiosis or DNA replication. They can also be induced by the organism itself, by cellular processes such as hypermutation. Mutation can result in several different types of change in DNA sequences; these can either have no effect, alter the product of a gene, or prevent the gene from functioning properly or completely. Studies in the fly Drosophila melanogaster suggest that if a mutation changes a protein produced by a gene, this will probably be harmful, with about 70 percent of these mutations having damaging effects, and the remainder being either neutral or weakly beneficial. Due to the damaging effects that mutations can have on genes, organisms have mechanisms such as DNA repair to remove mutations. Therefore, the optimal mutation rate for a species is a trade-off between costs of a high mutation rate, such as deleterious mutations, and the metabolic costs of maintaining systems to reduce the mutation rate, such as DNA repair enzymes. Viruses that use RNA as their genetic material have rapid mutation rates, which can
28
be an advantage since these viruses will evolve constantly and rapidly, and thus evade the defensive responses of e.g. the human immune system.
natural selection the environment acts as a sieve through which only certain variations can pass.Natural selection is one of the cornerstones of modern biology. The term was introduced by Darwin in his influential 1859 book On the Origin of Species, in which natural selection was described as analogous to artificial selection, a process by which animals and plants with traits considered desirable by human breeders are systematically favored for reproduction. The concept of natural selection was originally developed in the absence of a valid theory of heredity; at the time of Darwin's writing, nothing was known of modern genetics. The union of traditional Darwinian evolution with subsequent discoveries in classical and molecular genetics is termed the modern evolutionary synthesis. Natural selection remains the primary explanation for adaptive evolution.
recombination are also applied in protein engineering to develop new proteins of biological interest.
v. The selected solutions are considered as parents for the next iteration vi. After the required number of iterations, the best optimal solution is obtained.
Calculate fitness
Mutation
Calculate fitness
Selection
NO Convergence test YES End Fig 3.1. Flow Chart for Implementation of EP
31
optimization (ACO) and particle swarm optimization (PSO). ACO was inspired by the behaviors of ants and has many successful applications in discrete optimization problems. The particle swarm concept originated as a simulation of simplified social system. The original intent was to graphically simulate the choreography of bird of a bird block or fish school. However, it was found that particle swarm model could be used as an optimizer.
The behavior of each agent inside the swarm can be modeled with simple vectors. The research results are one of the basic backgrounds of PSO. Boyd and Richardson examined the decision process of humans and developed the concept of individual learning and cultural transmission. According to their examination, people utilize two important kinds of information in decision process.
33
The first one is their own experience; that is, they have tried the choices and know which state has been better so far, and they know how good it was. The second one is other peoples experiences, i.e., they have knowledge of how the other agents around them have performed. Namely, they know which choices their neighbors have found most positive so far and how positive the best pattern of choices was. Each agent decides its decision using its own experiences and the experiences of others. The research results are also one of the basic background elements of PSO. According to the above background of PSO, Kennedy and Eberhart developed PSO through simulation of bird flocking in a two-dimensional space. The position of each agent is represented by its x, y axis position and also its velocity is expressed by (the velocity of x axis) and (the velocity of y axis). Modification of the agent
position is realized by the position and velocity information. Bird flocking optimizes a certain objective function. Each agent knows its best value so far (pbest) and its x, y position. This information is an analogy of the personal experiences of each agent. Moreover, each agent knows the best value so far in the group (gbest) among pbests. This information is an analogy of the knowledge of how the other agents around them have performed. Each agent tries to modify its position using the following information: The current positions (x, y), The current velocities ( , )
The distance between the current position and pbest The distance between the current position and gbest
This modification can be represented by the concept of velocity (modified value for the current positions). Velocity of each agent can be modified by the following equation: ( ) (3.1)
where vik is velocity of agent i at iteration k, w is weighting function, c1 and c2 are weighting factors, rand1 and rand2 are random numbers between 0 and 1,
sik is current position of agent i at iteration k, pbesti is the pbest of agent i, and gbest
is gbest of the group. Namely, velocity of an agent can be changed using three vectors such like boid. The velocity is usually limited to a certain maximum value. PSO using eqn. (3.1) is called the Gbest model.
34
The following weighting function is usually utilized in eqn. (3.1): (3.2) Where wmax is the initial weight, wmin is the final weight, itermax is maximum iteration number and iter is current iteration number. The meanings of the right-hand side (RHS) of eqn. (3.1) can be explained as follows. The RHS of eqn. (3.1) consists of three terms (vectors). The first term is the previous velocity of the agent. The second and third terms are utilized to change the velocity of the agent. Without the second and third terms, the agent will keep on flying in the same direction until it hits the boundary. Namely, it tries to explore new areas and, therefore, the first term corresponds with diversification in the search procedure. On the other hand, without the first term, the velocity of the flying agent is only determined by using its current position and its best positions in history. Namely, the agents will try to converge to their pbests and/or gbest and, therefore, the terms correspond with intensification in the search procedure. As shown below, for example, wmax and wmin are set to 0.9 and 0.4. Therefore, at the beginning of the search procedure, diversification is heavily weighted, while intensification is heavily weighted at the end of the search procedure such like simulated annealing (SA). Namely, a certain velocity, which gradually gets close to pbests and gbest, can be calculated. PSO using eqns.(3.1) & (3.2) is called inertia weights approach (IWA).
Fig 3.2: concept of modifications of a searching point by PSO =current searching point = modified searching point = current velocity = modified velocity = velocity based on pbest = velocity based on gbest
35
The current position (searching point in the solution space) can be modified by the following equation (3.3): (3.3) Fig3.2 shows a concept of modification of a searching point by PSO, and it shows a searching concept with agents in a solution space. Each agent changes its current position using the integration of vectors as shown in Fig3.2.
36
Start
Initialize particles with random position and velocity vectors For each particle position (p) evaluate the fitness
NO
YES end Fig 3.3 Flow Chart for Implementation of PSO The features of the searching procedure of PSO can be summarized as follows: As shown in eqns. (3.1), (3.2), and (3.3), PSO can essentially handle continuous optimization problems.
37
PSO utilizes several searching points, and the searching points gradually get close to the optimal point using their pbests and the gbest. The first term of the RHS of eqn. (3.1) corresponds with diversification in the search procedure. The second and third terms correspond with intensification in the search procedure. Namely, the method has a well-balanced mechanism to utilize diversification and intensification in the search procedure efficiently.
The above concept is explained using only the x, y axis (two-dimensional space). However, the method can be easily applied to n-dimensional problems. Namely, PSO can handle continuous optimization problems with continuous state variables in an n-dimensional solution space.
Shi and Eberhart tried to examine the parameter selection of the above parameters. According to their examination, the following parameters are appropriate and the values do not depend on problems:
The values are also proved to be appropriate for power system problems. The basic PSO has been applied to a learning problem of neural networks and Schaffer f6, a famous benchmark function for GA, and the efficiency of the method has been observed.
3.11 CONCLUSION
In this chapter, the methodology, algorithms and flow charts of EP/PSO methods for solving optimal power flow problems has been presented.
38
39
4.1 INTRODUCTION
Evolutionary Programming (EP) based approach is proved to be quite encouraging in solving OPF problem. The EP technique is a stochastic optimization method in the area of evolutionary computation, which uses the mechanics of evolution to produce optimal solutions to a given problem. It works by evolving a population of candidate solutions towards the global minimum through the use of a mutation operator and selection scheme. The EP technique is particularly well suited to non-monotonic solution surfaces where many local minima may exist. Another evolutionary computation technique, called Particle Swarm Optimization (PSO), has been proposed and introduced. This technique combines social psychological principles in socio-cognition human agents and evolutionary computations. PSO has been motivated by the behavior of organisms such as fish schooling and bird flocking. Generally, PSO is characterized as simple in concept, easy to implement and computationally efficient. Unlike other heuristic techniques, PSO has a flexible and well-balanced mechanism to enhance and adapt to the global and local exploration abilities.
bus voltages V L , generator reactive power outputs QG , and the transmission line loadings S l , Hence, x can be expressed as given [ ] (4.4)
Where NL, NG and nl are number of load buses, number of generators and number of transmission line respectively. u is the vector of independent variables consisting of generator voltages VG, generator real power outputs PG except at the slack bus PG , transformer tap settings
1
Where NT and NC are the number of the regulating transformers and shunt compensators, respectively. F is the objective function to be minimized. g is the equality constraints that represents typical load flow equations and h is the system operating constraints Objectives The objectives considered for minimization are as follows. Objective Function 1: Fuel cost of generating units ( ) Objective Function 2: Active power loss ( ) Objective Function 3: Weighted multi objective function ( ) where ( ) (4.6) (4.7) (4.8) Here Constraints The OPF problem has two categories of constraints: Equality Constraints: These are the sets of nonlinear power flow equations that govern the power system, i.e, | | | || | (4.9)
41
Where
| | | || |
(4.10)
PGi and QGi are the real and reactive power outputs injected at bus- i respectively, the
load demand at the same bus is represented by PDi and QDi , and elements of the bus admittance matrix are represented by Yij and ij . Inequality Constraints: These are the set of constraints that represent the system operational and security limits like the bounds on the following: 1) Generators real and reactive power outputs (4.11) (4.12) 2) Voltage magnitudes at each bus in the network (4.13) 3) Transformer tap settings (4.14) 4) Reactive power injections due to capacitor banks (4.15) 5) Transmission lines loading (4.16) 6) voltage stability index (4.17) Handling of Constraints: There are different ways to handle constraints in evolutionary computation optimization algorithms. In this thesis, the constraints are incorporated into fitness function by means of penalty function method, which is a penalty factor multiplied with the square of the violated value of variable is added to the objective function and any infeasible solution obtained is rejected. To handle the inequality constraints of state variables including load bus voltage magnitudes and output variables with real power generation output at slack bus, reactive power generation output, and line loading, the extended objective function can be defined as: where
42
| |
| | (4.18)
K p , K q , K v , K s are penalty constants for the real power generation at slack bus, the
reactive power generation of all generator buses or PV buses and slack bus, the voltage magnitude of all load buses or PQ buses, and line or transformer loading, respectively. h( PG1 ) , h(QGi ) , h( Vi ) , h( S i ) are the penalty function of the real power generation at slack bus, the reactive power generation of all PV buses and slack bus, the voltage magnitudes of all PQ buses, and line or transformer loading, respectively. NL is the number of PQ buses. The penalty function can be defined as:
(4.19) where h(x) is the penalty function of variable x , and lower limit of variable x , respectively. and are the upper limit
43
Step 10: For each new individual in the archive, run power flow to determine load bus voltages, angles, load bus voltage stability indices, generator reactive power outputs and calculate line power flows. Step 11: Evaluate the penalty functions Step 12: Evaluate the objective function values and the corresponding fitness values for each new individual. Step 13: Apply the selection operator of EP/PSO and update the individuals. Step 14: Update the generation local best xlocal and global best xglobal and store them. Step 15: If one of stopping criterion have not been met, repeat steps 4-15. Else go to step 16 Step 16: Print the results There are two stopping criterion for the optimization algorithm. The algorithm can be stopped if the maximum number of generations is reached (Gen = Genmax) or there is no solution improvement over a specified number of generations. The first criterion is used in this thesis.
44
910 900
Objective function-1
890 880 870 860 850 840 830 1 14 27 40 53 66 79 92 105 118 131 144 157 170 183 196 209 222 235 248
No. of iterations
Objective function-2
0.0175 0.0174 0.0173 0.0172 0.0171 0.017 1 15 29 43 57 71 85 99 113 127 141 155 169 183 197 211 225 239
No. of iterations
990 970 950 930 910 890 870 850 830 810 1 15 29 43 57 71 85 99 113 127 141 155 169 183 197 211 225 239
No. of iterations
46
0.0515
0.051 0.0505 0.05 0.0495 0.049 0.0485 1 15 29 43 57 71 85 99 113 127 141 155 169 183 197 211 225 239
No. of iterations
Objective function-1
1100 1050 1000 950 900 850 800 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145
No. of iterations
Objective function-2
0.04 0.035 0.03 0.025 0.02 0.015 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145
No.of iterations
47
1000
980 960 940 920 900 880 860 840 820 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145
No.of iterations
0.055
0.05
0.045
No. of iterations
48
Table 4.2 Optimal settings of control variables Control variables Objective function-1 EP PSO 1.1242 1.1279 0.7000 0.7000 0.2783 0.2749 0.2605 0.2613 0.2778 0.2746 0.0100 0.1108 0.1760 0.1283 0.0995 1.0700 1.0567 1.0275 1.0137 1.0480 0.9543 1.0879 1.0157 0.0521 0.0975 0.0191 0 0.0306 Cost ($/hr) Loss (p.u.M.W) CPU Time (sec) -0.0929 0.0534 0.0373 0.0924 0.1459 1.0700 1.0621 1.0389 1.0598 1.0688 1.0322 0.9000 0.9922 0.0260 0.0500 0.0351 0.0500 0.0500 Objective function-2 EP PSO 0.2485 0.6138 0.7000 0.3889 0.8000 0.8000 0.5424 0.5019 0.3163 0.3049 0.0741 0.0506 0.1786 0.0044 0.0505 1.0700 1.0644 1.0541 1.0491 1.0326 1.0260 0.9665 0.9669 0.0016 0.0105 0.0389 0.0824 0.0787 1105.4 0.0583 60.9060 0.1796 -0.1906 -0.0342 0.1807 0.0628 1.0700 1.0541 1.0425 1.0309 1.0387 1.0064 0.9360 1.0473 0.0183 0.0098 0.0348 0.0178 0.0399 1072.1 0.0194 39.7190 Objective function-3 EP PSO 1.1191 1.1278 0.7000 0.7000 0.2728 0.2748 0.2746 0.2627 0.2734 0.2739 -0.0789 0.1838 0.1688 0.0322 0.1177 1.0700 1.0609 1.0321 1.0564 1.0571 1.0331 0.9017 0.9617 0.0369 0.1000 0.0658 0.0364 0.0551 -0.0936 0.0574 0.0370 0.1177 0.1782 1.0715 1.0621 1.0387 1.0520 1.0784 1.0118 0.9201 1.0004 0.0500 0.0500 0.0218 0.0208 0
49
L-indices
Load buses Fig 4.9 L-indices of 14 bus by EP and PSO for objective function-1
120 100
Line loadings
80 60 40 20 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 PSO EP
Number of Lines Fig 4.10 Line loadings of 14 bus by EP and PSO for objective function-1
1.08
Load voltages
PSO EP
Load buses Fig 4.11 Load voltages of 14 bus by EP and PSO for objective function-1
50
L-indices
PSO
EP
Load buses Fig 4.12 L-indices of 14 bus by EP and PSO for objective function-2
Line loadings
PSO EP
Number of Lines Fig 4.13 Line loadings of 14 bus by EP and PSO for objective function-2
Loadvoltages
PSO EP
Load buses Fig 4.14 Load voltages of 14 bus by EP and PSO for objective function-2
51
L-indices
PSO EP
Loadbuses Fig 4.15 L-indices of 14 bus by EP and PSO for objective function-3
120 100
Line loadings
80 60 40 20 0 1 2 3 4 5 6 7 8 9 1011121314151617181920
PSO EP
Number of Lines Fig 4.16 Line loadings of 14 bus by EP and PSO for objective function-3
Chart Title 1.06
Load voltages
1.05
PSO EP
Loadbuses
10
11
12
13
14
Fig 4.17 Load voltages of 14 bus by EP and PSO for objective function-3
52
53
835 830
Objective function-1
825 820 815 810 805 800 795 1 14 27 40 53 66 79 92 105 118 131 144 157 170 183 196 209 222 235 248
No.of iterations
825
820
815
810
805 1 14 27 40 53 66 79 92 105 118 131 144 157 170 183 196 209 222 235 248
No. of iterations
0.11 0.109 0.108 0.107 0.106 0.105 0.104 0.103 0.102 1 15 29 43 57 71 85 99 113 127 141 155 169 183 197 211 225 239
No. of iterations
54
910 890 870 850 830 810 790 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145
Objective function-1
No. of iterations
Objective function-2
0.075 0.07 0.065 0.06 0.055 0.05 0.045 1 10 19 28 37 46 55 64 73 82 91 100 109 118 127 136 145
No. of iterations
880 870 860 850 840 830 820 810 800 790 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 121 129 137 145
No. of iterations
0.12 0.115 0.11 0.105 0.1 0.095 0.09 0.085 0.08 1 10 19 28 37 46 55 64 73 82 91 100 109 118 127 136 145
No. of iterations
Control variables
Table 4.3 Optimal settings of control variables Objective function-1 Objective function-2 Objective function-3 EP PSO EP PSO EP PSO 1.7814 1.7616 0.5304 0.9421 1.7778 1.7618 0.4990 0.4885 0.7940 0.6868 0.4985 0.4884 0.1961 0.2207 0.3500 0.2480 0.2008 0.2212 0.1200 0.1217 0.3000 0.1768 0.1167 0.1220 0.2242 0.2150 0.5000 0.4870 0.2277 0.2151 0.1206 0.1200 0.4000 0.3405 0.1200 0.1200 -0.1967 0.1451 0.3189 0.1496 0.4744 0.1218 1.0500 1.0364 1.0284 1.0649 1.0148 1.0634 -0.1622 0.1268 -0.0121 0.2631 0.0988 0.2056 1.0500 1.0379 1.0189 1.1000 1.0109 1.0768 1.0049 0.9591 1.0064 0.9592 0.0282 0.0443 0.0365 0.0414 0.0389 0.0415 0.0339 0.0500 0.0326 801.9525 0.0936 92.1720 - 0.1861 0.1183 0.3264 0.0022 0.3885 0.0912 1.0500 1.0500 1.0444 1.0330 1.0349 1.0571 1.0115 0.9544 1.0041 0.9842 0.0154 0.0237 0.0605 0.0560 0.0461 0.0913 0.0410 0.0243 0.0283 968.6317 0.0494 134.0320 -0.0569 -0.0017 0.1036 0.1485 0.0538 0.2138 1.0500 1.0418 1.0282 1.0538 1.0209 1.0432 1.0000 0.9964 1.0473 0.9616 0.0166 0.0239 0.0259 0.0321 0.0168 0.0378 0.0205 0.0269 0.0257 899.1088 0.0472 129.6410 -0.1432 0.1859 0.2465 0.0308 0.4489 0.2000 1.0500 1.0345 1.0175 1.0375 1.0080 1.0577 0.9769 1.0058 1.0240 0.9812 0.0342 0.0238 0.0709 0.0415 0.0491 0.0544 0.0419 0.0200 0.0761 806.6193 0.1075 120.2630 -0.2000 0.0698 -0.0956 0.2648 0.0889 0.4472 1.0500 1.0385 1.0196 1.1000 1.0121 1.1000 0.9956 0.9819 1.1000 0.9755 0 0.0231 0.0495 0.0447 0.0336 0.0500 0.0234 0.0500 0.0500 802.2686 0.0944 90.9330
T1 T2 T3 T4 Qsh1 Qsh2 Qsh3 Qsh4 Qsh5 Qsh6 Qsh7 Qsh8 Qsh9 Cost Loss CPU Time
1.0331 0.9327 0.9891 0.9611 0.0054 0.0057 0.0127 0.0641 0.1000 0.0581 0.0507 0.0495 0.0275 806.3633 0.1703 145
56
PSO EP
PSO EP
Fig 4.26 Line loadings of 30 bus by EP and PSO for objective function-1 1.06
1.04
1.02 PSO 1 0.98 0.96 7 8 9 101112131415161718192021222324252627282930 Fig 4.27 Load voltages of 30 bus by EP and PSO for objective function-1 EP
57
PSO EP
PSO EP
Fig 4.29 Line loadings of 30 bus by EP and PSO for objective function-2
1.06 1.05 1.04 1.03 1.02 1.01 1 0.99 0.98 0.97 0.96 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
PSO EP
Fig 4.30 Load voltages of 30 bus by EP and PSO for objective function-2
58
PSO EP
PSO EP
Fig 4.32 Line loadings of 30 bus by EP and PSO for objective function-3 1.06 1.04 1.02 1 0.98 0.96 0.94 7 8 9 101112131415161718192021222324252627282930 Fig 4.33 Load voltages of 30 bus by EP and PSO for objective function-3 PSO EP
59
Here an IEEE-30 bus system is considered. Evolutionary programming (EP) method has been applied on the IEEE-30 bus system. Here we have considered three objective functions. Objective function1 is the cost objective function. Objective function2 is the loss objective function. Objective function3 is the multi objective function i.e. both cost and losses are taken as objectives. The fitness function is taken as the reciprocal of the objective function.
4.6 CONCLUSION
In this chapter, the application of EP/PSO methods for solving optimal power flow problems has been presented. The generation cost and real power losses were reduced through adjustment of generator outputs, generator voltages, tap changing transformers, and shunt compensation. The simulation results on IEEE 14- and 30-bus systems have been presented for illustration purpose. The algorithms EP and PSO were accurately and reliably converged to the global optimum solution in each case. Moreover, the PSO-algorithm is capable of producing better results compared with other algorithms.
60
CHAPTER 5 CONCLUSION
61
CHAPTER 5 CONCLUSION
5.1 CONCLUSIONS
This chapter summarizes the work presented in the earlier chapters. The project work carried out broadly touches upon the aspects of load flow studies, and power system optimal power flow with a view of reducing cost and loss of the systems. An EP and PSO based optimization algorithms have been proposed for solving optimal power flow problems with different objective functions. These algorithms take into consideration all the equality and inequality constraints. The improvement in system performance is based on reduction in cost of power generation and active power loss. The proposed algorithms have been compared with the other methods reported in the literature. Simulation studies have been carried out for the optimal solutions of the IEEE 14-bus and IEEE 30-bus systems. It was observed that the results obtained by the proposed algorithms can be implemented in real life power systems for operation and analysis. Based on the overall observations from the results obtained on various IEEE test systems, it can be concluded that the proposed methods for optimal solutions are suitable for implementing in modern power system operation.
References
63
References
[1] E.Ewald,D.W.Angland, Regional integration of electric power systems,IEEE Spectrom, April 1964 ,pp.96-101. [2] D.Watts, Security & vulnerability in electric power system,NAPS 2003, 35th North American Power Symposium, University of Missouri-Rolla in Rolla, Missouri, October 20-21, 2003. pp. 559-566. [3] Bullock, G.C., Cascading Voltage Collapse in Tennese, August 22, 1987, Proceedings of 17th Annual Western Protective Relay Conference", Spokane, Washington, October 1990. [4] IEEE Special Publication 90TH0358-2PWR, Voltage Stability of Power Systems: Concepts, Analytical Tools and Industry Experience", The IEEE Working Group on Voltage Stability, 1990. [5] North American Electric Reliability Council, Survey of the voltage collapse phenomenon, August 1991. [6] T. K. P. Medicherla, R. Billinton, and M. S. Sachadev, Generation rescheduling and load shedding to alleviate line over loads: analysis, IEEE Trans. on PAS, Vol.98, No.12, 1979, pp .1876-1884. [7] T K P Medicherla, R Billinton, and M S Sachadev, Generation rescheduling and load shedding to alleviate line over loads: system studies, IEEE Transactions on PAS, Vol.100, No.1, 1981, pp .36-42. [8] IEEE Power Engineering Society/CIGRE: FACTS applications, IEEE service center, Piscataway, N.J., 1996. Special Issue.96TP116-0. [9] A J Wood and B F Wollenberg, Power generation operation and control, John Wiley and Sons Inc.Singapore, 1984. [10] [11] Bose, A., Power System Security, 1998 (WA 99164-2752). Brian Stott, Ongun Alsac, and Alcir J. Monticelli, Security analysis and optimization, Proc. of IEEE, Vol.75, No.12, December 1987, pp.1623-1644. [12] Hsiao-Dong Chiang and Rene Jean Jumeau, Toward a practical performance index for detecting voltage collapse in electric power systems, IEEE Transactions on power systems, Vol.10, No.21, 1992, pp.584-592. [13] Claudio A. Canizares, Antonio C. Z de Souza, and Victor H. Quintana, Comparison of performance indices for detection of proximity to voltage collapse, Vol.11, No.3, August 1996, pp.1441-1450.
64
[14]
Miroslav M Begovic, and Arun G Phadke, Dynamic simulation of voltage collapse, IEEE Trans. on PS, Vol.5, No.4, November 1990, pp.1529-1534. Van Cutsem, T., A method to compute reactive power margins with respect to voltage collapse, IEEE Trans. on power systems, Vol.6, No.1, February 1991, pp.145-156. Parker, C.J., I. F Morrison, and D. Sutanto, Application of an optimization method for determining the reactive margin from voltage collapse in reactive power planning, IEEE Trans. on Power Systems, Vol.11, No. 3, August 1996, pp. 1473-1481. Venikov, V.A., V.A Stroev, V.I. Idelchick, and V.I. Tarasov, Estimation of electric power system steady state stability in load flow calculations", IEEE Trans. on PAS, Vol.PAS-94, No.3, May/June 1975, pp.1034-1040. Thomas, R.J. and A. Tiranuchit, Voltage Instabilities in Electric Power Network, Proceedings of the 18th Southeastern Symposium on System Theory, Knoxville, Tennessee, April 1986. A. Tiranuchit, and R.J. Thomas, A posturing strategy against voltage instabilities in electric power systems, IEEE Transactions on Power Systems, Vol. 3, No.1, February 1988, pp.87-93. P A Lof,T Smed,G Anderson,and D J Hill, Fast calculation of a voltage stability index, IEEE Trans. on PS, Vol.7, No.1, February 1992,pp.54-64.
[15]
[16]
[17]
[18]
[19]
[20]
[21]
Lof, P.A., David J. Hill, Stefan Arnborg, and Goran Andersson, "On the Analysis of Long-term Voltage Stability", Electric Power & Energy Systems, Vol.15, No.4, 1993, pp.229-237. Illiceto, F., E. Cineiri, and L. Casely-Hayford, Long lightly loaded HV transmission lines to expand electrification of developing countries, application in Ghana", CIGRE, International Conference on Large High Voltage Electric Systems, 29th August-6th September, 1984. Carson W. Taylor, Survey of effective and practical solutions for long-term voltage stability, Electric power and energy systems, Vol. 15, No.4, 1993, pp.217-220.
[22]
[23]
[24]
V. Ajjarpu, C. Christy, "The Continuation Power Flow: A Tool for Steady State Voltage Stability Analysis", IEEE Trans. On Power System, Vol.7, No.1, February 1992.
65
[25]
R. Seydel, From equilibrium to chaos; practical bifurcation and stability analysis, Elsevier, NY, 1988. K. Iba et al., Calculation of critical loading condition with nose curve using homotopy continuation method, IEEE Trans. Power Systems, Vol. 6, May 1991, pp. 584-593. C.A. Canizares, and F.L. Alvarado, Point of collapse and continuation methods for large AC-DC systems, IEEE Transactions on power systems, Vol.7, No.1, February 1993, pp.1-8. H.D. Chiang, A. Flueck, K.S. Shah, N. Balu, CPFLOW: A practical tool for tracing power system steady-state stationary behavior due to load and generation variations, IEEE Trans. Power Systems, vol. 10, no. 2, May 1995, pp. 623-634. S. Greene, I. Dobson, F.L. Alvarado, Sensitivity of the loading margin to voltage collapse with respect to arbitrary parameters, IEEE Trans. on power systems, Vol. 12, No. 1, February 1997, pp. 262-272. S. Greene, I. Dobson, and F.L. Alvarado, Contingency analysis for voltage collapse via sensitivities from a single nose curve, IEEE PES Summer Meeting, Berlin, Germany, July 1997. C. Rajagopalan, S. Hao, D. Shirmohammadi, M.K. Celik, Voltage collapse operating margin analysis using sensitivity techniques, Proceedings Athens Power Tech 1993, Athens, Greece, Sep. 5-8, 1993, pp.332-336. F.L. Alvarado, T.H. Jung, Direct detection of voltage collapse conditions, Proceedings: Bulk Power System Voltage Phenomena Voltage Stability and Security, EPRI-6183, Jan. 1989, pp. 523-538. I. Dobson, L. Lu, Computing an optimum direction in control space to avoid saddle node bifurcation and voltage collapse in electric power systems, IEEE Trans. Automatic Control, Vol. 37, no. 10, Oct. 1992, pp. 1616-1620. T. Van Cutsem, Extensions and applications of a mid-term voltage stability analysis method,1994. T. Van Cutsem, R. Mailhot, Validation of a fast stability analysis method on the Hydro-Quebec system, IEEE Transactions on power systems, Vol. 12, No. 1, Feb. 1997, pp. 282292.
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
66
[36]
C. L. DeMarco, G. C. Verghese, Bringing phasor dynamics into the power system load flow, Proceedings 1993 North American Power Symposium, Washington D.C., Oct. 11-12, 1993, pp. 463-471. P.A. Lof,G. Andersson, D. Hill, Voltage dependent reactive power limits for voltage stability studies, IEEE Trans.on power systems, Vol. 10, No. 1, Feb. 1995, pp. 220-228. P Kessel and H Glavitsch, Estimating the voltage stability of a power system, IEEE Trans. on PD, Vol.1, No.3, pp.346-354, 1986. A.M.Chebbo, M.R.Irving, and M.J.H.Stirling, Voltage collapse proximity indicator:behaviour and implications, Proc.IEE, Gen.Tr.Distrib, Vol.139, No.3, 1992, pp.584-592. Gao, B, G. K. Morison, and P. Kundur, Voltage stability evaluation using modal analysis, IEEE Trans actions on Power Systems, Vol.7, No.4, November 1992. Begovic, M, and A.G. Phadke, Control of Voltage Stability Using Sensitivity Analysis, IEEE Trans. on power systems, Vol.7, No.11, 1992, pp.114-123. O.Crisan and M.Liu, Voltage collapse prediction using an improved sensitivity approach, Electric Power System Research, Vol.28, No.3, 1994, pp, 181-190. F Gubina, and B Strmcnik, Voltage collapse proximity index determination using voltage phasors approach IEEE Trans actions on Power Systems, Vol.10, no.2, 1995, pp.788-793. C.J.Huang,C.E.Lin,and C.L.Huang, Fuzzy approach for generator
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
maintenance scheduling, Electric power system research, Vol.24, 1992, pp.3138. [45] J.J.Saraiva, V.Miranda, and L.M.V.G.Pinto, Impact on some planning decisions from a fuzzy modeling of power systems,IEEE Trans.on power systems, Vol.9, No.2, 1994, pp.819-825. [46] A.K. David and R. Zhao, An expert system with fuzzy sets for optimal planning. IEEE Trans. on PS, 6(2), 1995, pp.213-219. [47] S K Tso, T X Zhu, Q Y Zeng, and K L Lo, Fuzzy set approach to dynamic voltage security assessment, IEE Proc.Gen, Trans.Distrib., Vol.142, No.2,March 1995.
67
[48]
J.L.Souflis,AV Machias,and B.C.Papadias, An application of fuzzy concepts to transient stability evaluation, IEEE Trans.on power systems, Vol.4, No.3, August1989. W. Ongsakul and T. Tantimaporn, Optimal power flow by improved
[49]
evolutionary programming, Electric Power Components and Systems, 2006, 34:pp.79-95,. [50] Abido MA. Optimal power flow using particle swarm optimization, Electric Power Energy Syst 2002; 24(7): pp.563-71. [51] Peerapol Jirapong and Weerakorn Ongsakul Optimal placement of multi type FACTS devices for total transfer capability enhancement using hybrid evolutionary algorithm, Electric power componenets and systems, 01 September 2007, 35:pp.981-1005. [52] Ramasubramanian Jayashree and Mohammed Abdullah Khan A unified optimization approach for the enhancement of available transfer capability and congestion management using unified power flow controller, Serbian journal of electrical engineering, Vol.5, No.2, November 2008, pp.305-324. [53] D.Devaraj and B.Yegnanarayana, Genetic Algorithm-Based Optimal Power Flow for Security Enhancement, IEE Proceedings on Generation, Transmission and Distribution 2005, 152(6), pp 899-905. [54] X.P.Zhang, S.G.Petoussis and K.R.Godfrey Nonlinear interior point optimal power flow method based on a current mismatch formulation, IEE Proc.Gener. Transm. Distrb. Vol.152, No.6, January 2005, pp.795-805.
68
69
Fig. A.1 Single line diagram of the standard IEEE thirty bus system Line data The details about the line parameters of the IEEE standard thirty bus system are shown in the Table A.1
70
Table A.1 Line data of a standard IEEE 30 bus system Line From 1 1 2 8 2 2 11 5 13 13 9 9 12 12 12 12 14 16 15 18 19 10 10 10 10 21 15 22 23 24 25 25 27 27 29 3 13 13 13 11 28 To 2 8 11 11 5 13 13 7 7 3 4 10 6 14 15 16 15 17 18 19 20 20 17 21 22 22 23 24 24 25 26 27 29 30 30 28 28 9 10 12 27 Rij(p.u.) 0.0192 0.0452 0.0570 0.0132 0.0472 0.0581 0.0119 0.0460 0.0267 0.0120 0.0000 0.0000 0.0000 0.1231 0.0662 0.0945 0.2210 0.0524 0.1073 0.0639 0.0340 0.0936 0.0324 0.0348 0.0727 0.0116 0.1000 0.1150 0.1320 0.1885 0.2544 0.1093 0.2198 0.3202 0.2399 0.0636 0.0169 0.0000 0.0000 0.0000 0.0000 Line Parameters Xij(p.u.) 0.0575 0.1652 0.1737 0.0379 0.1983 0.1763 0.0414 0.1160 0.0820 0.0420 0.2080 0.1100 0.1400 0.2559 0.1304 0.1987 0.1997 0.1923 0.2185 0.1292 0.0680 0.2090 0.0845 0.0749 0.1499 0.0236 0.2020 0.1790 0.2700 0.3292 0.3800 0.2087 0.4153 0.6027 0.4533 0.2000 0.0599 0.2080 0.5560 0.2560 0.3960 Bsh/2(p.u.) 0.0264 0.0204 0.0184 0.0042 0.0209 0.0187 0.0045 0.0102 0.0085 0.0045 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0214 0.0065 0.0000 0.0000 0.0000 0.0000 Tap Ratios n(p.u.) 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.978 0.969 0.932 0.968 Line MVA Limits Smax(p.u.) 1.30 1.30 0.65 1.30 1.30 0.65 0.90 0.70 1.30 0.32 0.65 0.65 0.65 0.32 0.32 0.32 0.16 0.16 0.16 0.16 0.32 0.32 0.32 0.32 0.32 0.32 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.16 0.32 0.32 0.65 0.32 0.65 0.65
71
Bus data The bus details of the standard IEEE 30 bus system is given in the TableA.2. Bus code is used to identify the type of the bus in the system. Generators are connected to the buses 1, 2, 5, 8, 11, &13. Table A.2 Bus data of a standard IEEE 30 bus system
B. B V Del No. Co. 1 1 1.06 0.00 2 2 1.04 0.00 3 0 1.00 0.00 4 0 1.06 0.00 5 2 1.01 0.00 6 0 1.00 0.00 7 0 1.00 0.00 8 2 1.01 0.00 9 0 1.00 0.00 10 0 1.00 0.00 11 2 1.08 0.00 12 0 1.00 0.00 13 2 1.07 0.00 14 0 1.00 0.00 15 0 1.00 0.00 16 0 1.00 0.00 17 0 1.00 0.00 18 0 1.00 0.00 19 0 1.00 0.00 20 0 1.00 0.00 21 0 1.00 0.00 22 0 1.00 0.00 23 0 1.00 0.00 24 0 1.00 0.00 25 0 1.00 0.00 26 0 1.00 0.00 27 0 1.00 0.00 28 0 1.00 0.00 29 0 1.00 0.00 30 0 1.00 0.00 Pl 0.0 21.7 2.4 7.6 94.2 0.0 22.8 30.0 0.0 5.8 0.0 11.2 0.0 6.2 8.2 3.5 9.0 3.2 9.5 2.2 17.5 0.0 3.2 8.7 0.0 3.5 0.0 0.0 2.4 10.6 Ql Pg Qg 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Qmin Qmax Qsh
0.00 0.0 12.7 40.0 1.2 0.0 1.6 0.0 19.0 0.0 0.0 0.0 10.9 0.0 30.0 0.0 0.0 0.0 2.0 0.0 0.0 0.0 7.5 0.0 0.0 0.0 1.6 0.0 2.5 0.0 1.8 0.0 5.8 0.0 0.9 0.0 3.4 0.0 0.7 0.0 11.2 0.0 0.0 0.0 1.6 0.0 6.7 0.0 0.0 0.0 2.3 0.0 0.0 0.0 0.0 0.0 0.9 0.0 1.9 0.0
0.00 0.00 0.00 -40.00 50.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -40.00 40.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -10.00 40.00 0.00 0.00 0.00 0.00 0.00 0.00 19.00 -6.00 24.00 0.00 0.00 0.00 0.00 -6.00 24.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 4.30 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
72
Transformer Data Table A.3 Transformer Data of a standard IEEE 30 bus system Transformer From 13 13 11 28 To 9 10 12 27 Rij(p.u.) 0 0 0 0 Parameters Xij(p.u.) 0.2080 0.5560 0.2560 0.3960 Bsh/2(p.u.) 0 0 0 0 Tap Ratios n(p.u.) 0.978 0.969 0.932 0.968 Line MVA Limits Smax(p.u.) 0.65 0.32 0.65 0.65
Load Data Table A.4 Load Data of a standard IEEE 30 bus system Bus No. 2 3 5 7 8 10 11 12 14 15 16 17 18 19 20 21 23 24 26 29 30 P (MW) 21.7 30.0 94.2 22.8 2.4 5.8 7.6 11.2 6.2 8.2 3.5 9.0 3.2 9.5 2.2 17.5 3.2 8.7 3.5 2.4 10.6 Q (MVAR) 12.7 30.0 19.0 10.9 1.2 2.0 1.6 7.5 1.6 2.5 1.8 5.8 0.9 3.4 0.7 11.2 1.6 6.7 2.3 0.9 1.9
73
Generation Data a, b, c Pmin, Pmax Qmin, Qmax - Cost coefficients related to quadratic fuel cost function. - Active power Generation limits - Reactive power Generation limits
Table A.5 Generator Data of a standard IEEE 30 bus system Bus No. 1 2 3 4 5 6 PG (MW) 260.2 80.0 20.0 20.0 50.0 20.0 QGmin (MVAR) -20.0 -20.0 -15.0 -10.0 -15.0 -15.0 QGmax (MVAR) 150.00 60.00 48.73 45.00 62.45 44.72 Vsp (p.u) 1.050 1.045 1.010 1.050 1.010 1.050 a $/p.u2h 37.5 175 83.4 250 625 250 b $/p.u.h 200 175 325 300 100 300 c $/h 0 0 0 0 0 0
: 14 : 5 : 17 : 3 : 1 : 0.0001
G
G
1 5 4
G
8 7
G
6 12 11 10 9
13
14
Transmission Line data: Bsh/2 : Half Line Charging Admittance Rij Xij Smax : Resistance between buses i and j
: Reactance
Note:- Old Bus Numbers : 6,8 New Bus Numbers :4,5 Table A.6 Transmission line Data of a standard IEEE 14 bus system Line Line Parameters Tap Ratios Line MVA Limits Smax(p.u.) 1.30 1.30 1.30 0.65 0.65 0.32 0.70 0.32 0.32 0.32 0.32 0.65 0.32 0.32 0.32 0.32 0.32 0.65 0.32 0.65
From 1 1 2 2 2 3 6 4 4 4 7 7 9 9 10 12 13 6 6 8
To 2 8 3 6 8 6 8 11 12 13 5 9 10 14 11 13 14 7 9 4
Rij(p.u.) Xij(p.u.) Bsh/2(p.u.) n(p.u.) 0.01938 0.05917 0.0264 1.000 0.05403 0.04699 0.05811 0.05695 0.06701 0.01335 0.09498 0.12291 0.06615 0.00000 0.00000 0.03181 0.12711 0.08205 0.22092 0.17093 0.00000 0.00000 0.00000 0.22304 0.19797 0.17632 0.17388 0.17103 0.04211 0.19890 0.25581 0.13027 0.17615 0.11001 0.08450 0.27038 0.19207 0.19988 0.34802 0.20912 0.55618 0.25202 0.0246 0.0219 0.0170 0.0173 0.0064 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.978 0.969 0.932
75
Load Data
Table A.7 Load Data of a standard IEEE 14 bus system Bus No. PDo (MW) PDmin (MW) 1 0 0 2 21.7 18.4450 3 94.2 80.0700 4 11.2 9.5200 5 0 0 6 47.8 40.6300 7 0 0 8 7.6 6.4600 9 29.5 25.0750 10 9.0 7.6500 11 3.5 2.9750 12 6.2 5.2700 13 13.5 11.4750 14 14.9 12.6650 Generation Data for 14-Bus System PDmax (MW) 0 24.9550 108.3300 12.8800 0 54.9700 0 8.7400 33.9250 10.3500 4.0250 7.1300 15.5250 17.1350 Q (MVAR) 0 12.7 19.0 7.5 0 -3.9 0 1.6 16.6 5.8 1.8 1.6 5.8 5.0
Table A.8 Generation Data of a standard IEEE 14 bus system Bus No. 1 2 3 4 5 PG (MW) 232.4 40.0 50.0 20.0 20.0 PGmin (MW) 0 0 0 0 0 PGmax (MW) 340 70 80 90 70 QGmin
(MVAR)
76
APPENDIX-B PUBLICATION
77