Toward sustainable PV–EV integrated smart homes: A critical review and data-driven reinforcement learning–enabled cyber–physical framework

The accelerating adoption of rooftop photovoltaic (PV) systems and residential electric vehicles (EVs) is reshaping households into active energy prosumers, intensifying the need for intelligent and sustainable Smart Home Energy Management Systems (SHEMS). Despite extensive research on optimization- and machine learning–based control, existing studies remain fragmented, often focusing on algorithmic performance while overlooking cyber–physical integration, scalability, and system-level sustainability. This paper presents a critical review of PV–EV integrated SHEMS from a control-oriented and sustainability-driven perspective. The literature is systematically analyzed across architectural paradigms, uncertainty modeling, energy storage coordination, EV charging strategies, and reinforcement learning (RL) applications. The review identifies a persistent gap between advanced learning algorithms and deployable, grid-interactive residential implementations aligned with decarbonization objectives. To bridge this gap, a reinforcement learning–enabled cyber–physical framework is proposed that integrates data-driven forecasting, structured state–action modeling, degradation-aware reward design, and hierarchical coordination of PV, ESS, and EV resources. The proposed framework emphasizes transparency, computational feasibility, and compatibility with real-time operation. By linking algorithmic intelligence to renewable integration, peak demand mitigation, and emission-reduction goals, this work outlines a scalable pathway toward resilient, sustainability-oriented PV–EV-integrated smart homes.