Optimizing Home Energy Flows and Battery Management with Supervised and Unsupervised Learning in Renewable Systems

This study examines reinforcement learning (RL) and fuzzy logic control (FLC) for optimizing battery energy storage in residential systems with photovoltaic (PV) power, grid interconnection, and dynamic or fixed electricity pricing. Effective management strategies are crucial for reducing costs, extending battery lifespan, and ensuring reliability under fluctuating demand and tariffs. A 24 h simulation with minute-level resolution modeled diverse conditions, including random household demand and ten initial state of charge (SOC) levels from 0% to 100%. RL employed proximal policy optimization (PPO) for adaptive energy scheduling, while FLC used rule-based logic for charge–discharge cycles. Results showed that FLC rapidly restored SOC at low levels, ensuring immediate availability but causing cost fluctuations and increased cycling, particularly under stable pricing or low demand. RL dynamically adjusted charging and discharging, reducing costs and smoothing energy flows while limiting battery cycling. Feature importance analysis using multiple linear regression (MLR) and random forest regression (RFR) confirmed SOC and time as key performance determinants. The findings highlight a trade-off between FLC’s rapid response and RL’s sustained cost efficiency, providing insights for optimizing residential energy management to enhance economic and operational performance.