Multi-Objective Reactive Power Optimization On Improved Particle Swarm Algorithm

The main propose of power system reactive power optimization was to decrease the active losses of system, to enhance the voltage quality and system stability by reasonably adjusting reactive power equipment. To the regular reactive power optimization, the minimum system losses was commonly used target function, while each node voltage was close or even reaching the upper limit value in optimization results, so there was a conflict between target function of reactive power optimization and voltage safety of system. According to grid safety and economy, multi-objective model of reactive power optimization was put forward, which contained active losses of system and voltage levels.Reactive power optimization in power system was in the scale of non-linear optimization category, its characteristic included multi-objective, multi-control variables,multi-constraints, Continuous variable and integer variables, and uncertainty. Weighting method, which was a method to turn multi-objective optimization problem to single-objective problem, was adopted in the regular optimization of these kinds of problems, however, because of the different physical dimension, the value of weight factor was very hard to use in different target functions. The method based on fuzzy set can be adopted in transformation from multi-objective to single-objective by normalized calculation of membership function of every single target. This method could quickly increase the computational amount because of the complexities of computation parameters of the membership functions which belong to the target functions during searching process, so the calculation speed would slow down. Furthermore, due to the difference of dimension and contradiction of multi-objective problem, it was impossible to turn multi-objective problem to single-objective problem. In this thesis, particle swarm optimization and its application in power system were studied, and a novel method of multi-objective reactive power optimization is proposed.In this thesis, a Particle Swarm Optimization algorithm for multi-objective optimization problem is presented. In this algorithm, the local optimal extremum and global optimal extremum of the particles are renewed by the Pareto dominance relationship, and non-dominated sets are constructed through arena's principle. Farther, non-dominated solutions are preserved in storage pool during the searching process, and particles with the longest distance is selected as the global optimal extremum from the storage pool. The crowded degree algorithm is employed to clip non-dominated solutions in order to keep the distribution of solutions. Inertia weight value of every particle is fixed according its virtues or defects degree to keep the balance between the local and the global researching abilities. This method is applied to optimizing the reactive power of IEEE-14 and 30 bus systems, and a set of Pareto optimal solutions are figured out independently.