Review of AI Based Maximum Power Point Tracking Techniques & Performance Evaluation of Artificial Neural Network based MPPT Controller for Photovoltaic Systems
Abstract
AI-based MPPT techniques suitable for Photovoltaic systems is investigated in this paper. Performance analysis of various training algorithms and network architectures are carried out to examine the suitable algorithm and network architecture for the function approximation of MPPT problem in PV. Artificial Neural network (ANN) based soft computing tools will be utilized for function approximation of the nonlinear dynamic nature of PV panels. Data set for training Neural Networks are generated by PV panel modelling and Perturb & Observed based MPPT algorithm simulated using MATLAB. Various Supervised network training approaches are investigated and compared to the feasibility of implementing MPPT. The performance of three main classes of network training functions namely Quasi-Newton, Gradient Descent and Conjugate Gradient algorithms are analyzed. Various neural network architectures such as Feedforward and Recurrent type neural networks are studied in this paper for the feasibility of implementing AAN based MPPT controller. The performance is evaluated for the developed ANN based MPPT using the MATLAB Neural Network Tool Box. The results indicate that the ANN based MPPT methods can ensure that Maximum Power Point (MPP) is efficiently tracked during fluctuating environmental conditions