Employing Demand Response in Energy Management of Microgrid using Q learning

This paper presents the energy management of a microgrid in conjunction with the conventional grid. The microgrid incorporates wind, solar and diesel generators. An incentive based demand response program is employed which comprises of penalties and incentives in response to customer action. The objective is to optimize utility and customer benefits along with minimization of transferred power and fuel costs. A Q-learning based approach has been employed to accomplish the objectives. A case study considering three customers has been carried out with an assumption that they employ wind, solar and conventional generators. To establish the superiority of Q-learning, the obtained results have been compared with conventional particle swarm optimization technique. The results indicate considerable decrease in fuel cost and transferred power cost alongside high utility and customer benefits. It is also evident from results that with Q-learning offers 69 percent more customer benefits as compared to PSO.