Research On Reactive Power Optimization Based On Genetic Algorithms In Power Systems

Along with the development of modern industry, the quality of electric power becomes more and more important. The aim of voltage stabilization and reducing line loss can be make true by reactive power optimization(RPO). RPO is a kind of complicated nonlinear constrained optimization problem. A variety of methods have been devoted to solve the problem. The methods of RPO are presented in this paper,kinds of methods for solving the problem are summarized. The methods fall into two categories, one is the classical methods, the other is the modern methods. The classical methods mostly point to mathematical analysis programming methods. The modern methods include artificial intelligent(esp. genetic algorithm) ,Tabu search algorithm ,simulated annealing algorithm and so on. In this paper, a mathematical model involves all kinds of reactive power regulation facilities and operational constraints is established,it can reflects the actual requirements of power system. The mathematical model is matrix-based, the gradient formulas of object function and constraints also are established on the basis of the matrix. It is convenient to use the formulas in object-oriented program tools (esp. MATLAB). Mathematical analysis programming methods with mature and integrated theory can obtain the final result very quickly. Feasible direction methods is one type of the most efficient methods to solve constrained nonlinear optimization problem. In this paper, the application of three different feasible direction methods—Generalized Reduced Gradient(GRG) method, Feasible Point(FP) method and Sequential Quadratic Programming(SQP) method to RPO are discussed. In this paper, a program based on MATLAB optimization toolbox implements SQP method to solve RPO problem. Genetic algorithms(GAs) with self-adapting step size can search the best solution directly and more likely to acquire the global optimum result. The processes of genetic algorithm for RPO: modeling of networks, designing chromosome code and fit degree function for variable, and selecting, intercrossing, differentiating, reserving chromosome. Optimum results are satisfied, it has great effect in reducing line losses and eliminating voltage exceeding specified limits. In this paper, GENOCOPII/GENOCOPIII is cited to RPO.The GENOCOPII uses two-step selection method and annealing penalty function,searches the optimum result from one single point.After lots of numeric experiments of GENOCOPII ,some experiential principles are discussed in this paper. GENOCOPIII is a co-evolutionary genetic algorithm. It maintains two separate populations and uses the method of repairing infeasible individuals to search the feasible points. After examination of original GENOCOPIII, some genetic operations of the algorithm are improved, which yields more satisfactory results and speeds up the optimization computation. To overcome the poorly computational efficiency of GAs, an improvement of it by using local searching is applied to enhance its'convergence speed and searching efficiency.A heuristic GENOCOPIII based on the concept of feasible direction is discussed in this paper. The concept of PQ-decoupled method is cited in the heuristic GENOCOPIII in this paper too. In this paper , The proposed algorithms is applied to Ward&Hale6-bus system and IEEE 14-bus system, the numerical results of the simulation demonstrate the validity and effectiveness of these algorithms.