Adaptive neuro-Kpis approach in the evaluation of the performance and parameters of a PV module

Abstract

This paper presents an adaptive approach to performance analysis and estimation of unmeasurable and inaccessible parameters of a PV power plant based on the neural structures and mechanism proposed by Massachusetts Institute of Technology (MIT). This approach takes into account the single diode model, the empirical efficiency analysis model and an additional simple electronic circuit that allows to establish the voltage and current dynamics. The new scheme found here develops the observers using neural networks for the estimation of the unmeasurable and inaccessible parameters of the one-diode model. The MIT mechanism combined with the proposed ADALINE neural network has interesting features for the accurate determination of the k-pis parameters of the empirical model for PV plant performance analysis. The implementation of the proposed tool is done in the numerical environment Matlab/simulink and in real time via the Arduino target for the experimental part. The numerical and experimental results are presented in order to estimate the parameters and their performances. The results found here show that our approach stands for a robust tool to analyze defects. Another particularity of the new approach concerns the fault analysis based on the resistor behaviors in a three-dimensional frame. Hence, the proposed tools contribute to the fault analysis, the parameter estimation and the performance evaluation of the PV source whatever the operating conditions.