Neural network based global maximum power point tracking under partially shaded conditions

Partial shading changes characteristics of a solar panel and creates a number of local maximum power points (MPPs) that only one of them is the global MPP. On the other hand measurement of light intensity is quite hard and requires use of specialized sensors. In this paper, a new method for tracking the global MPP under partially shaded conditions using neural network is proposed in which instead of measuring light intensity, the network approximates it. For this purpose, at first by measuring the voltage, current and temperature of panels we estimate the radiation intensity and then a neural network is trained using radiation intensity and temperature of panels as inputs and MPP as output of the network. Finally, this method is simulated in MATLAB/Simulink environment and results show the effectiveness of the proposed method.