SCHEDULING OF DIRECT LOAD CONTROL USING GENETIC PROGRAMMING

摘要

A modified Genetic Programming (GP) called block deepening GP (BDGP) is proposed in this paper to optimize the scheduling of direct load control (DLC). A search scheme similar to the breadth first search approach in artificial intelligence is adopted in the BDGP to search for the optimal scheduling arrangements. BDGP is composed of a master GP and a slave GP. As the master GP evaluates the status combination of all load groups at every time step, it calls upon the slave GP to simultaneously look ahead an additional D steps to evaluate the best possible load difference that could result. The best status combinations in the following D steps associated with the status combination under evaluation are determined globally in the following D-step block. Since the proposed BDGP optimizes the scheduling for DLC aiming to minimize the load difference in the next time step, and the following D time steps, the scheduling results obtained by BDGP are closer to the globally optimal solution.