Output Power Estimation of High Concentrator Photovoltaic using Radial Basis Function Neural Network

Abstract

High Concentrator PhotoVoltaic (HCPV) is a recent PV technology generating electricity from solar radiation. Unlike conventional PV systems, it uses lenses and curved mirrors to focus solar rays onto small, but highly efficient Multi-junction (MJ) solar cells. Solar tracker and cooling systems are part of a standard CPV facility. Due to the complex design of an HCPV system, the output power estimation becomes a very hard task. In contrast, Machine Learning (ML) methods, and more specifically Artificial Neural Networks (ANNs), provide very suitable solutions for modelling complicated systems. The aim of this work is to develop a Radial Basis Function Neural Network (RBFNN) model to predict the output power of an HCPV facility. RBFNNs have a simple topological structure and their ability to reveal how learning proceeds in an explicit manner. Our results showed that the RBFNN model provides more accurate estimation of output power compared to the ASTM-E2527 based on the same dataset.