Optimal Load Scheduling for Smart Home Energy Management using Deep Reinforcement Learning

This article presents an approach to optimal energy management in Smart Home Energy Management (SHEM) systems, utilizing the technique of optimal load shifting under the Time-of-Use (TOU) electricity tariff. The study employs Deep Reinforcement Learning (DRL) algorithms based on the Markov Decision Process (MDP), including Deep QNetwork (DQN), Advantage Actor-Critic (A2C), and Proximal Policy Optimization (PPO). These techniques aim to identify optimal electricity load changes to minimize energy costs following the TOU tariff of the Provincial Electricity Authority (PEA). Simulation results show that using DQN, electricity costs decreased by 36.167% and the Peak Average Ratio (PAR) by 2.067 compared to the original load profile. However, the outcomes varied slightly among the different techniques, demonstrating that appropriate energy management in smart homes can effectively reduce electricity costs under the TOU tariff.