Virtual Population Genetic Algorithm-based Reactive Power Optimization

With the fast development of national economy, demand for electricity has been increasing rapidly as well. Consequently, it is very important to ensure security and economy of power system operation under such a demanding situation. Reactive power optimization (RPO) is a very important part of power system security and economic operation research. RPO is essential for system voltage stability enhancement, power loss minimization and therefore essential for power system security and reliability as well.Genetic Algorithms (GAs) are search algorithms based on of natural evolution processing including selection, mutation and crossover operations on the genes of individuals or potential solutions. GAs are stochastic optimization algorithms and can be carried out through parallel computing. GAs do not require the optimization problem to be convex and differentiable as required by other traditional optimization techniques to find the global optimal solution. GAs have been widely applied to solving many power engineering problems over the past decades including power system planning, economic dispatch, unit commitment and fault analysis etc.Detailed analysis of RPO problem is given in this thesis based on research content, methods, RPO modeling and RPO objective function used. AS important background knowledge, the fundamental and advanced techniques in GAs are given in the background chapters of the thesis. An improved GA is used for RPO problem in the thesis. The advanced GA uses two virtual population methodologies to process the population of standard binary and real code GA for RPO problem. These techniques are then applied to realist power system problems. I also compared the simulation results with standard binary and real code GAs. The main research topics of the thesis have been followed:Firstly, according to the investigation of research results of RPO in the literature, a full scheme of improved genetic algorithm (IGA) was proposed. In this thesis, the mathematic model of RPO was improved, including the establishment of objective function and the selection of constrains.Secondly, based on the characteristics of RPO problem, I modified the coding strategy and genetic operators including the application of elitism, hybrid crossover method and high probability of mutation. Special attention is paid to enhance the searching ability, speed up the computation speed and improving the search accuracy.Thirdly, when applied to solving large-scale network problems GAs have some common shortcomings, such as consuming large computation time, easy to be trapped by local minimal and difficulties in searching for the global optimal solution. With regard to these shortcomings, research has been conducted on searching route to ensure global optimality. From the application on practical power system, the result shows that the IGA has the advantages over standard genetic algorithm (SGA) on computational cost and the quality of optimal solutions.Finally, analysis and discussion is given on the results generated by all algorithms synthetically. The pros and cons of each algorithm are compared and discussed. Future work on further development of GAs is also proposed in the thesis.On solving the RPO problem, the application of GAs performs the its parallel, stochastic advantages fully. Further more, there is no requirement for fitness and objective function using GAs, which are also convenient to be used in solving the optimization problem. This thesis shows that the proposed approaches has clear advantages over other methodologies.