Novel Gaussian process method for photovoltaic maximum power point tracking

• A Novel and robust approach for maximum power point tracking in photovoltaic (PV) systems is proposed. • The approach adapts for changes in the system’s dynamic model and measurement model. • The use of the GP’s hyperparameter inference capability eliminates the need for manual tuning. • The approach is compared to common MPPT approaches. • The proposed algorithm is experimentally verified. Maximum power point tracking (MPPT) in photovoltaic (PV) systems remains challenging under model uncertainty, measurement noise, and changing operating conditions, where conventional methods often require careful tuning or suffer degraded performance. This study investigates whether a Gaussian Process (GP)-based framework can provide accurate and robust MPPT without relying on precise prior system tuning. A novel GP-inspired MPPT algorithm is proposed and experimentally validated using PV modules arranged in three configurations: parallel, series, and series–parallel. The method reformulates the estimation problem within a GP-based state estimation framework and uses Metropolis-Hastings Markov Chain Monte Carlo (MH-MCMC) to infer key hyperparameters, eliminating the need for manual tuning. Its performance is benchmarked against conventional Perturb and Observe (P&O), Adaptive P&O, and Kalman Filter (KF) methods using root mean square error (RMSE), integral time absolute error (ITAE), and convergence behavior. The results show that the proposed GP method matches KF performance under well-tuned conditions. Since the GP infers the tuning parameters, GP ITAE is lower than the KF by a factor of 10 in most experiments. The GP also outperforms traditional P&O and achieves lower ITAE than Adaptive P&O, with convergence in the microsecond-to-millisecond range depending on implementation hardware. These findings demonstrate that the proposed method is a reliable and adaptive MPPT solution for PV systems operating in uncertain and dynamic environments. The novelty of this study lies in integrating Gaussian Process based state estimation with MH-MCMC hyperparameter inference for MPPT, enabling accurate and adaptive tracking without manual tuning of the system or noise parameters.