Distributed Nonconvex Economic Dispatch Algorithm for Large-scale Power System

This paper proposes a distributed genetic algorithm to solve non-convex optimal power allocation problem in largescale power system. Such problem is usually solved by some heuristic search methods such as genetic algorithm (GA). It is noted that traditional genetic algorithms are usually centralized, and their genetic operators are directly operated on the entire combined solution by a single control center. When considered system scales up, its computational cost would be tremendously high. In order to overcome this problem, in this paper we aim to revise the GA algorithm to be implemented in a fully distributed way. The main idea of the proposed distributed GA is to treat the power output solution of a single distributed generator as an individual, and each local agent has a separate population and performs genetic operations independently. The fitness information used for the crossover comes from the average consensus of the local fitness. The power outputs of the entire generator set is considered as the combined individual, which are found to converge near the optimal solution in the formulation experiment At the end of the paper, a case study on 10 generator units shows that the distributed genetic algorithm is capable to solve nonconvex economic dispatch (ED) problem while perfectly fitting the distributed framework in most scenario.