A Distribution Network Restoration Decision Support Algorithm Based on Multi-Agent System

The goal of this paper is to provide a fast and effective restoration scheme after a fault happened in a distribution network. Many centralized strategies have been proposed to address this multi-objective problem, while in this paper, we present a proper function-decoupling approach by multi-agent system (MAS) to search the optimal restoration solution. The proposed three-layer agent framework decouples the procedure of generating switch operating sequence. We regard each objective function as a single agent, which are: 1) Agent 1 to maximize the amount of loads served; 2) Agent 2 to minimize switching operations; and 3) Agent 3 to minimize power loss. As the most prioritized goal is to restore loads as much as possible, we put Agent 1 in the primary layer. Applying the improved genetic algorithm, a solution set representing feasible reconfiguration topologies with maximum recovered loads can be generated. The secondary layer, consisting of Agent 2 and Agent 3, is to filter solution(s). Agent 2 is achieved by comparing the switch status before and after the reconfiguration, and Agent 3 is to calculate the power loss. Parallel computing of Agent 2 and Agent 3 is feasible to improve computational efficiency. Finally, the third layer is for coordinating each solution. With the Pareto Principle and analytic hierarchy process (AHP) applied, the optimal switching operation sequence can be chosen.