Artificial intelligence prediction of maximum power point tracking voltage and current based on battery for sensor reduction and complexity minimization for photovoltaic charge controller

This research works on an Artificial Intelligence (AI)–based approach for Maximum Power Point Tracking (MPPT) in photovoltaic (PV) systems, focusing on the prediction of PV voltage (Vpv) and current (Ipv) from battery-side parameters rather than direct PV-side sensing. By applying Machine Learning (ML) and Deep Learning (DL) algorithms, the proposed framework eliminates the requirements for dedicated PV voltage–current sensors inside MPPT charge controllers, thereby reducing hardware cost, calibration requirements, and system complexity. An MPPT charge controller was employed to provide V_MPPT and I_MPPT values as ground truth for validation. Two experimental scenarios were designed: (i) using battery parameters alongside ambient temperature and humidity, and (ii) relying solely on battery parameters. A comprehensive evaluation of 10 ML and 7 DL algorithms was conducted, with the best-performing models selected via K-fold cross-validation. Results demonstrate that the Extra Trees Regressor achieved a root mean square error as low as 0.02 (normalized scale of 1), indicating strong accuracy in predicting PV operating points. The proposed approach highlight the practical system of PV sensor reduction, AI-driven MPPT strategy, offering a cost-effective and scalable alternative to traditional MPPT methods for both small-scale and potentially larger PV systems.