The rapid integration of photovoltaic (PV) generation into modern power networks introduces significant operational challenges, including intermittent power production, uneven charge distribution, and reduced system reliability in multi-battery energy storage systems. Addressing these challenges requires intelligent, adaptive, and physically consistent control strategies capable of operating under uncertain environmental and load conditions. This study proposes a Physics-Informed Neural Network (PINN)-based charge allocation framework that explicitly embeds physical constraints—namely charge conservation and State-of-Charge (SoC) equalization—directly into the learning process, enabling real-time adaptive control under varying irradiance and load conditions. The proposed controller exploits real-time measurements of PV voltage, current, and irradiance to achieve optimal charge distribution while ensuring converter stability and balanced battery operation. The framework is implemented and validated in MATLAB/Simulink under Standard Test Conditions of 1000 W·m−2 irradiance and 25 °C ambient temperature. Simulation results demonstrate stable PV voltage regulation within the 230–250 V range, an average PV power output of approximately 95 kW, and effective duty-cycle control within the range of 0.35–0.45. The system maintains balanced three-phase grid voltages and currents with stable sinusoidal waveforms, indicating high power quality during steady-state operation. Compared with conventional Proportional–Integral–Derivative (PID) and Model Predictive Control (MPC) methods, the PINN-based approach achieves faster SoC equalization, reduced transient fluctuations, and more than 6% improvement in overall system efficiency. These results confirm the strong potential of physics-informed intelligent control as a scalable and reliable solution for smart PV–battery energy systems, with direct relevance to renewable microgrids and electric vehicle charging infrastructures.
Akinwola et al. (2026) studied this question.