Study On Reactive Power System Optimization Based On Improved Evolutionary Programming Method

Reactive power system optimization, an important measure to ensure power system security, and to improve the quality and economic efficiency of power supply, refers to Optimizing control variables, including the generator terminal voltage, tap stall of load tap transformer or input capacity of reactive power compensation equipment, to make one or more performance measurement of the system achieve optimal, when the system parameters and load situation are settled, and all kinds of operation restrictions are satisfied. In mathematics, reactive power system optimization, a complex non-linear optimization problem which contains continuous and discrete variables, non-linear objective functions, non-linear equalities and inequalities restrictions, is non-convex and multi-extremum and very difficult to get solution fast. Many research results showed that there are still a lot of issues to be resolved. This paper studied on resolving methods of reactive power system optimization, including three parts, namely, analyzing of the voltage and reactive power theory, improving of the evolutionary programming methods, and proposing of a new reactive power system optimization algorithm.First, theories related reactive power optimization is summarized. After a detailed introduce of the reactive power and voltage control and research background, as well as basic mathematical models of reactive power optimization, this paper summarized many algorithms of resolving reactive power optimization problems about their advantages and disadvantages and applications. Classic Mathematical Programming Optimization algorithms which based on differential coefficient have tight theory and fast convergence, but it is difficult to adapt practical problems for complex mathematical models and differential calculations. Modern artificial intelligence algorithms which based on random search have an adaptive search capability without the request of continuously differentiable, they can converge to the global optimal solution in a greater probability, especially the evolutionary programming (EP) algorithm which belongs to the important branch of evolutionary algorithms. Compared with other intelligent algorithms, this EP algorithm has a stronger global searching capability and many other advantages, such as, making encoding unnecessary, implementing parallel computing easily, and having fewer parameters to be determined. Then, the problems existed in traditional evolutionary programming operator is analyzed. A new mutation operator is constructed using mixed random distribution function and ICSSP. Metropolis Criteria instead of random "q competition" selection method is introduced to construct a new selection operator. It's verified that the improved evolutionary programming method overcomes disadvantages of the slow searching speed and premature phenomenon.Finally, it proposed an improved method of evolutionary programming (IMEP) suitable for reactive power optimization. According to the practical operation situation of power system, this new method makes a practical mathematical model of reactive power optimization, and revised the model in setting evolution parameters, conforming fitness calculation functions and punishment coefficient depended on practical requirements of reactive power optimization. The simulation on IEEE-30 and IEEE-57 system demonstrate the advantages of high speed, high efficiency and stability and solutions with good quality, etc. It can conduct reliable reactive power optimization calculation of actual systems, and has a certain level of practical value.