Optimal Scheduling of Grid Supply and Batteries Operation in Residential Building: Rules and Learning Approaches

Abstract

This article discusses and simulates a demand management algorithm in a building with a battery energy storage system (BESS) and on-grid supply scheduling using deep reinforcement learning algorithms (DRL) and rule-based controllers (Fuzzy logic). BESS is used for supplying the load profile and minimizing the electricity bills of the building. Deferrable loads of the building are controlled to reduce power consumption during peak time. Controllers used in this paper aim to optimize multi-objectives, including the cost of utility bills, state of health of battery systems (SOH), and reliability of the power source. First, this article works with optimizing BESS using the fuzzy logic controller and compares the results with DRL agent outputs. Secondly, commercial loads are modeled based on the deferrability index to be introduced into the optimization problem. Finally, the paper presents a model for controlling the battery and on-grid supply schedule, minimizing the annual electricity bill without draining the battery’s SOH and disturbing the residential comfort of household appliances.