Reactive Power Optimization Based On Multi-Agent Crisscross Optimization Algorithm

With the development of economy and the improvement of industry, electricity consumption, power grid and installed capacity have been expanding, which put forward higher requirements to the security, reliability and economy of power system. Therefore, a reasonable and and effective reactive power optimization of power system plays a very important role in human’s life and industrial production.Reactive power optimization of power system is a complex optimization problem with multi-variables, multi-constraints and nonlinear problems, whose essence is to adjust the distribution of reactive power reasonably, reduce the power system active power loss, improve the level of the node voltage and electric energy quality by adjusting the terminal voltage of generators and transformer tap position as well as the capacitor bank switching number considering all kinds of constraints. In view of actual situation, three kinds of objective functions including single objective function of optimal network loss, based on fuzzy membership degree of multi-objective function and based on the theoretical Pareto multi-objective function are proposed to verify the comprehensive performance of our algorithm.Because the conventional optimization methods have some limitations, a novel approach called crisscross optimization algorithm is introduced in this paper. The method consists of two unique search operators, namely horizontal crisscross enhancing the global search ability and vertical crisscross preventing form convergence stagnancy. The MACSO algorithm by introducing multi-agent neighborhood cooperative competition and self learning operation is proposed to improve the convergence speed and precision of CSO. Finally, in order to adapt to the multi-objective optimization of Pareto, a MOMACSO algorithm by combining fast non-dominated sorting with elite selection strategy is put forward to further expand the scope of application of MACSO algorithm.In this paper, the experiments are divided into two parts. The former carries out simulation experiments through the standard nodes of different scale and flexible objective function reactive power optimization. The result shows that the MACSO algorithm performs powerful global searching ability and faster convergence speed. It should be noted that we found the proposed algorithm performs a better effect on large node and multiple target optimization problems. Other trails about reactive power optimization of multi-class objects in regional power grid are simulated in the second part. The results further substantiate the universal applicability of MACSO for solving reactive power optimization problem of super large scale real power grid. Moreover, it demonstrates that Pareto multi objective optimization method is reasonable and superior and the MOMACSO algorithm practical and effective.