Artificial Neural Network Modeling for Efficient Photovoltaic System Design

Abstract

Efficiency and certainty of payback have not yet attained desired level for solar photovoltaic energy systems. Despite huge development in prediction of solar radiation data, a clear disconnect in extraction and effective engineering utilization of pertinent information from such data is acting as a major roadblock towards penetration of this emerging technology. It is crucial to identify and optimize the most significant statistics representing insolation availability by a solar PV installation for all necessary engineering and financial calculation. A MATLAB program has been used to build the annual frequency distribution of hourly insolation over any module plane at a given site location. Descriptive statistical analysis of such distributions is done through MINITAB. To make the analysis more meaningful, composite frequency distribution for a Building Integrated Photo Voltaic (BIPV) set up has been considered, which is formed by weighted summation of insolation distributions for different module planes used in the installation. The most influential statistics of the composite distribution have been optimized through Artificial Neural Network Computation. This novel approach is expected to be a very powerful tool for the BIPV system designers.