Residential Load Scheduling With Renewable Generation in the Smart Grid: A Reinforcement Learning Approach

The significance and need of demand response (DR) programs is realized by the utility as a means to reduce the additional production cast imposed by the accelerating energy demand. With the development in smart information and communication systems, the price-based DR programs can be effectively utilized for controlling the loads of smart residential buildings. Nowadays, the use of stochastic renewable energy sources like photovoltaic (PV) by a small domestic consumer is increasing. In this paper, a generalized model for the residential load scheduling or load commitment problem (LCP) in the presence of renewable sources for any type of tariff is presented. Reinforcement learning (RL) is an efficient tool that has been used to solve the decision making problem under uncertainty. An RL-based approach to solve the LCP is also proposed. The novelty of this paper lies in the introduction of a comprehensive model with implementable solution considering consumer comfort, stochastic renewable power, and tariff. Simulation experiments are conducted to test the efficacy and scalability of the proposed algorithm. The performance of the algorithm is investigated by considering a domestic consumer with schedulable and nonschedulable appliances along with a PV source. Guidelines are given for choosing the parameters of the load.