Research On Reactive Power Optimization Based On Memetic Algorithm

Under comprehensive attention nowadays, Reactive power optimization, an important measure to promote security and economy of power system, involves two research aspects including detailed modelization considering more practical requirements, and fast as well as accurate solution which is the main point and deeply discussed in this paper.It is difficult for traditional mathematical programming to manage the RPO, a non-linear, non-convex, and multimodal optimization which contains both continuous and discrete variables. With global convergence properties, Evolutionary Algorithm, the advanced representative of intelligent optimization methods which can deal with hybrid variables conveniently, has wide application in RPO. To improve the practicability of EA in RPO real-time system of power system, Local Search (LS) strategies are introduced to construct hybrid algorithms to enhance LS capabilities, and accelerate converging rate which constitutes the main content of this paper.Related theories and background information is summarized at first. After a detailed introduce of the development process of RPO and related research background, the characteristics of the traditional mathematical programming and intelligent optimization algorithms as well as mathematical models considering various factors and objectives are described and analyzed.In addition, the Evolutionary Algorithms and its Local Search strategies for RPO are analyzed. After a review of applications for EAs in RPO, the improved approaches is summarized, based on which the LS strategies in RPO are reviewed and classified into three parts: Knowledge-based LS, gradient-based LS, random-based LS, and the idea of combining these LS strategies to improve algorithm efficiency comes out.Finally, Memetic Algorithms is introduced to solve the RPO problem. The creation, biologic principle and development process of Memetic Algorithms are reviewed in the first place. Then, the process flow and the theory of integrating several LS strategies is discussed, according to which the Memes Pool for RPO and the basic framework of MA are proposed. Besides, to overcome the disadvantages of heavy computing work in Greedy Algorithm and of low efficiency in Random Algorithm, the Partial Greedy Algorithm is proposed. The simulation on IEEE 30 bus system shows the high efficiency of Memetic Algorithm in RPO and the advantages of combining various LS strategies. The feasibility of the Partial Greedy Algorithm is demonstrated as well.To sum up, memetic Algorithm for RPO, whose efficiency is validated in the simulation, provides a novel solving scheme for voltage and var control in the large-scale power system.