Maximum Power Point Tracking During Partial Shading Effect in PV System Using Machine Learning Regression Controller

Abstract

Maximum Power Point Tracking (MPPT) algorithm performs for maximizing the efficiency of solar Photo Voltaic (PV) system. The solar photovoltaic system efficiency reduces due to partial shading and ambient atmospheric condition, which varies with geographic locations. Traditional MPPT systems solve the above problem through different soft computing algorithms such as Perturb and observe (P&O), Flower pollination algorithm (FPA) and Particle swarm optimization (PSO). In P&O, FPA and PSO algorithms, duty cycle of boost converter varies to attain MPPT. The soft computing algorithms in MPPT perform less during the partial shading effect or rapid insolation, fluctuation condition of solar energy. The performance of MPPT with traditional algorithms is reduced due to slow convergence speed and oscillations in tracking by computing algorithms. In this paper, Regression controller based MPPT achieve maximum peak voltage during partial shading effect is developed. The regression controller predicts the duty cycle for boost converter based on stored dataset of PV system output voltage and load, during partial shading effect or rapid isolation for that particular geographic location. The regression based duty cycle prediction controller is programmed in MATLAB R2018a Simulink. Furthermore, Regression controller is implemented in PV system test bed. The simulation and hardware results of Regression controller based MPPT perform more of about 20%, 16.96% and 15% in efficiency respectively than PSO, FPA and P&O algorithms during partial shading condition in PV.