Proximal Policy Optimization-Driven Real-Time Home Energy Management System with Storage and Renewables

Intelligent energy management methods for modern smart homes are highly efficient in reducing residential energy expenditures through strategic scheduling of diverse energy utilization tasks. However, defining optimal usage schedules while ensuring user comfort remains an unsolved problem due to various variables, such as different appliance usage patterns, renewable generation variations, atmospheric conditions, and prevailing electricity tariffs. To tackle this challenge, this paper proposes a real-time home energy management system (HEMS) utilizing a proximal policy optimization (PPO)-based deep reinforcement learning (DRL) algorithm, developed for smart homes integrated with energy storage system (ESS), photovoltaic, and IoT-integrated appliances. The primary objective of this algorithm is to reduce total energy consumption costs while ensuring user preferences. A deep learning load regulations algorithm is developed to determine the optimal operation of various types of households, computing optimal usage schedules. The proposed DRL agent is primarily trained by historical data and subsequently deployed for real-time scheduling tasks mainly to react to the real-time information received from IoT sensors. Extensive simulation tests are carried out using authentic real-life data, proving the efficacy and resilience of the proposed algorithm.