Research On Reactive Power Optimization In Regional Power Grids By Genetic Algorithm Integrated With Quasi-Newton Method

As an important part of public utility, the industry of electric power is an essential basis for the whole national economy. The power system’s security and stability is of great significance for rapid development of economy and improving of people’s living standards. Reactive power optimization is one of the essential issues of regional power system dispatching, it is optimal control of transformers and other devices like parallel capacitors or FACTS to reduce active power losses, exaltation node voltage quality and the utilization of power grid assets simultaneously, which contributes to power system’s secure and economical operation.The ORP problem in regional power networks is a multi-objective nonlinear optimization problem with large numbers of variables, constraints and extreme values. Having summarized the current research on reactive power optimization, this paper aims at seeking a more efficient algorithm reactive power optimization in regional power grids. Firstly the BFGS quasi-newton method which is quite efficient with quadratic convergence is proposed here for the first time applied to reactive power optimization, and the simulation results demonstrate that it is extremely quick in local search while greatly influenced by initial values. Then the characteristics of improved genetic algorithm for reactive power optimization is studied. The performance analysis through experimental results indicate that the genetic algorithm has a strong capability of global search while unwarrantably convergent to the extreme. Hence proposing a hybrid algorithm easy for computer implementation for the solution of ORP problem in regional power networks. The global search of divisional genetic algorithm can obtain initial cycle solutions for the fast local search of BFGS. Compared with the improved GA and several other intelligent algorithms, the testing results on IEEE 118 bus system and two actual regional grids indicates that the algorithm obtains better node voltage profile and less power loss within less time and has better computational performance.