Neural network estimation of a photovoltaic system based on the MPPT controller

Abstract

MPPT is necessary to achieve an optimal exploitation of the photovoltaic (PV) system. This paper deals with the problem of the optimization of the power, delivered by the photovoltaic panel (PVP). To achieve this aim, a neural network estimator (NNE), followed by a conversion coefficient and a calculation stage of the optimal duty cycle, has been developed. The NNE is used to calculate the open circuit voltage corresponding to each solar radiation and to a various value of temperature, based only on the standard open circuit voltage. A coefficient, determining for each solar radiation the voltage of the maximum power directly from the open circuit voltage, is estimated by a practical test. Finally, the optimal duty cycle is, next, determined by the input/output equation of the boost converter. The proposed MPPT is tested and compared with the most widely used MPPT methods by simulations using MATLAB/Simulink and real time hardware in the loop (HIL) implementation. The results obtained with the proposed MPPT show excellent dynamic performance under fast irradiation changes.