Optimizing residential energy costs: a novel machine learning approach for solar PV and EV integration through heuristic price signal dispatch

The integration of solar Photovoltaic (PV) systems with Electric Vehicle (EV) technology is emerging as a sustainable and promising method to cope with increasing energy demands, mitigate environmental impact, and reduce carbon emissions within residential and transportation sectors. This study focuses on the residential application of a grid-connected "PV+EV"system in Sydney, Australia, underscoring the benefits of using bi-directional vehicle batteries in conjunction with rooftop PV systems. In addition to heuristic price signal dispatch algorithms in the System Advisor Model (SAM) software tool, which rely on manual dispatch and peak shaving analyses, a novel Q-Learning-Based-Model (QLBM) algorithm within the domain of machine learning methodology is employed to enhance the understanding of system dynamics. This novel approach is designed to predict optimal energy efficiency by prioritizing the most cost-effective energy source, thereby alleviating grid stress, minimizing energy costs. The results are then compared to other techniques employed in this paper, affirming the superiority of the proposed algorithm.