Edge Intelligence‐Supported Programmable Hardware System for PV‐Storage‐Load Microgrids

This paper proposes an edge intelligence-supported programmable hardware system (EI-PHS) for photovoltaic (PV)-storage-load microgrids. The platform enables hardware-level validation of high-level microgrid optimization algorithms and artificial intelligence (AI) models, thereby bridging the gap between algorithm development and physical deployment. A unified hardware–software co-design framework is established by integrating programmable DC/DC converters, an edge AI module, and controller area network-based communication, supporting the real-time implementation of diverse scheduling, forecasting and energy-management algorithms. The proposed EI-PHS also provides highly programmable emulation of PV generation, battery charge/discharge dynamics and load behaviour, allowing flexible and reproducible construction of various microgrid operating scenarios using a simplified experimental setup. The effectiveness of the EI-PHS is validated through four hardware case studies. These include programmable emulation of PV, battery and load behaviours; conventional system-level power balancing; reinforcement learning-based battery arbitrage; and spatio-temporal graph neural network-based forecasting and power balancing. In the two AI-based cases, the hardware measurements closely match the simulation references, with mean absolute errors of 0.027 to 0.107 W and root mean square errors of 0.040 to 0.149 W. Only minor deviations are observed during rapid power transitions, confirming that the proposed EI-PHS can accurately and reproducibly implement both conventional and AI-based microgrid algorithms.