A Novel Maximum Power Point Tracking Approach Based on Fuzzy Logic Control and Optimizable Gaussian Process Regression for Solar Systems Under Complex Environment Conditions

Abstract

Photovoltaic (PV) systems have multiple peaks in the power-voltage (P-V) curve due to partial shading conditions (PSC). Multiple peaks make determining and tracking the global maximum power point (GMPP) more complex. More powerful algorithms and controller structures are needed to track GMPP in PV systems operating in complex environmental conditions.Therefore, this paper introduces a machine learning based fuzzy logic controller (MLBFLC) method to determine and track GMPP. MLBFLC is proposed to determine the optimal duty cycle of the DC-DC converter used in PV systems operating under PSCs. In order to test this method, real-time temperature and irradiance data for one month (February, May, August and November) from different seasonal conditions were used. The reference voltage values at the maximum power point (MPP) were obtained from the hyperparameter optimized Gaussian Process Regression (GPR) method. The Fuzzy Logic Controller (FLC) method was used to determine the optimum duty cycle of the converter. The proposed method was compared with the metaheuristic optimization algorithms such as particle swarm optimization (PSO) and the flying squirrel search optimization (FSSO) algorithm, for four different scenarios, using real-time temperature and irradiance data. Consequently, it is observed that the proposed MLBFLC method successfully tracks the GMPP with higher speed and higher accuracy for all scenarios. Under different PSCs determined in the scenarios, an efficiency value of 99.916% was achieved with the MLBFLC method and it was observed that it successfully followed the MPP with a tracking time of 0.123 s.