Distrituion Network Service Restoration Based On Multi-Agent Genetic Algorithm

With the development of the distribution network, distribution network becomes more complex and the probability of failure increases. The distribution network service restoration strategies are the recovery the de-energized loads power supply, according to the topology and the electrical relationships of the distribution network. It can reduce the scope and the time of the power failure, improve the power quality, and reduce economic losses. Therefore, the study of fast and accurate distribution network service restoration strategy is meaningful.Firstly, describe the basic principles and steps of the distribution network service restoration and multi-agent genetic algorithm, as the theoretical foundation of distribution network service restoration; Secondly, proposes a fast topology analysis and power flow calculation method to the distribution network service restoration. The distribution network structure is stored by SQL relational database. The data maintenance convenient and the topology reconstruction flexibility high are presented. Both of the node list and branch list store the data loaded from a relational database, which forms the object-oriented real-time database and improves real-time capability. The topology based on the breadth first search method is analyzed, and the power flow combined with the forward and backward substitution method is calculated. Based on C++programme on Qt, analyze the topology of the IEEE16nodes system and IEEE33nodes system and calculate their power flow, the results of topology analysis and power flow on GUI are provided. Results show that the algorithm can quickly detect the distribution network connectivity, judge whether there are loops or isolations, and on the premise of the convenient data maintenance. Finally, according to the characteristics of distribution network service, propose a based on multi-agent genetic algorithm distribution network service restoration. The simplied topology analysis makes the code feasible. The agent of muliple energy ranks based on the fast elitist non-dominated sorting algorithm and crowding distance, reduce energy comparison and well weigh the multiple objectives. Construct the Pareto optimal solution set with the efficient contest method. Based on C++programme on Qt, simulate the distribution network service restoration of the IEEE16nodes system and IEEE33nodes system. Results show that the algorithm has a higher convergence, well keeps the population diversity, and avoides falling into local optimal solution.