ANFIS-based predictive control for photovoltaic systems under partial shading conditions

This paper proposes a novel hybrid Maximum Power Point Tracking (MPPT) strategy for photovoltaic (PV) systems operating under variable irradiance and partial shading conditions by integrating an offline-trained Adaptive Neuro-Fuzzy Inference System (ANFIS) with Model Predictive Control (MPC). The key innovation lies in the control architecture, where the ANFIS first learns the nonlinear dynamics of the PV generator during an offline training phase using irradiance and temperature data, enabling it to generate optimal voltage and current references. These references are then fed to the MPC, which computes the appropriate switching signals for the DC-DC converter. This approach effectively combines the learning and generalization capabilities of ANFIS with the predictive accuracy and constraint management of MPC, ensuring robust and efficient power extraction under changing environmental conditions. The methodology is validated through a comprehensive simulation framework in MATLAB/Simulink. The model incorporates a detailed PV array, a DC-DC converter, and the proposed controller. Performance is evaluated under Standard Test Conditions (STC), dynamic irradiance ramps, and the EN50530 European efficiency test profile, with benchmarking against conventional P&O, IncCond, and standalone MPC techniques. Results demonstrate that the proposed ANFIS-MPC approach achieves superior tracking accuracy, with an average efficiency of 99.58% under standard conditions and up to 99.99% under partial shading. The controller minimizes steady-state oscillations and significantly reduces convergence time, as reflected in the excellent integral performance indices (ITAE = 0.00395, IAE = 0.834). These high efficiency values are attributed to the offline-trained ANFIS’s ability to provide accurate Global Maximum Power Point (GMPP) references to the MPC, even under partial shading conditions. The findings confirm that the hybrid ANFIS-MPC controller offers a robust, accurate, and viable solution for real-world PV applications subjected to highly variable environmental conditions.