Photovoltaic Array Maximum Power Point Tracking System Based on Modified Fuzzy Neural Network

Abstract

The article considers the task of maximum power point tracking of a photovoltaic array. A photovoltaic array maximum power point tracking system based on a modified fuzzy neural network is developed. Compared to existing fuzzy neural networks a modified fuzzy neural network includes recurrent neural networks, one of which approximates multidimensional membership functions based on experimental data. In order to create an optimal architecture of a modified fuzzy neural network, we developed a modified multi-dimensional particle swarm algorithm. The algorithm generates an initial particle position based on the Nguyen–Widrow method and combines the stages of global optimization based on a multi-dimensional particle swarm optimization with the stages of gradient descent based on the Levenberg–Marquardt algorithm. The validity and advantages of the proposed photovoltaic array maximum power point tracking system based on a modified fuzzy neural network under random perturbations by numerical simulations in Octave are demonstrated. The simulation results show that the proposed tracking system achieves competitive performance and robustness, as compared to a classical control model with a PID controller based on perturbation and observation, or incremental conductance algorithm.