Research On Reactive Power Optimization Of Distribution Network Based On Improved Niche Genetic Algorithm

Reactive power optimization has been a hot topic for researchers and a number of methods have been put forward. The traditional numerical optimization methods have obtained good developments, but there are two problems to be resolved. First, more and more modern power systems are becoming complicated and variables are increasing. The solution space is multi-dimensional and complex. Those numerical methods depend on selection of a initial condition, therefore, the results of traditional methods may easily converge to the local optimal solution rather than the global optimal solution. Second, reactive power optimization problem usually contains both continuous variables and discrete variables in the optimization problem, such as, turns ratio connector of voltage transformer and switching combinations of compensation capacitors. Traditional methods generally require the objective function to be differentiable or linear in local, which may lead to great error. Thus, people are trying to apply artificial intelligence methods to the reactive power optimization problem.Compared with the traditional optimization methods, genetic algorithm has the following characteristics.1) Searching from many initial points along the many lines, in a rather large probability of sense the global optimal solution of optimization problems can be found.2) Genetic algorithms optimize the objective function directly and the objective function does not require being continuous and differentiable.3) It is easily used to deal with the optimization problem with discrete and continuous variables. Because conventional genetic algorithm has slow convergence and early premature, we introduce the niche technology in the conventional genetic algorithm. We design a finite number of individual sharing strategy within the limited niche sources. Since the existing Niche Genetic Algorithm is related to the number of niche radius, distance, size and other parameters of the default problem, it is difficult to predict these parameters in advance and lead to poor performance.This thesis mainly studies the method how to avoid pre-set parameters of any niche. Considering the genetic evolution of niche characteristics, we design the pairs to form individual niche under minimum Euclidean distance. Using the improved evolution operator, we design the competition with (2+2), and at the same time every niche evolution. Thus, a new niche algorithm is constructed. Through the selection mechanism, encoding, genetic operators and other control parameters, we form a niche based on improved adaptive genetic algorithm and solve such problems of reactive power optimization. By TongLiao distribution network reactive power optimization calculation, the proposed algorithm is shown to be superior to the existing approaches.