Analysis on Parameter Effect for Solar Radiation Prediction Modeling using NNARX

Abstract

The radiant energy from the sun is defined as solar radiation. It had been discovered as a renewable energy which can provide electricity supplies using a photovoltaic system. Before developing the system, a preliminary test must be carried out to perform the analysis of solar energy potential in that specific area. This preliminary test is known as a modeling technique. The technique will use the related parameters as an input to predict the solar radiation value. Since there are multiple parameters used for solar radiation prediction model development, there had been multiple attempts on using only certain parameters to produce predictions for solar radiation value. This paper will review and further analyzed several works presented by the previous studies on developing solar radiation prediction models using various parameters with their results. With the findings, the implementation of the Neural Network Autoregressive Model with Exogenous Input (NNARX) on solar radiation prediction carried out for the different input parameter configurations. Based on the results, it shows that the solar radiation prediction model development using more input parameters produced the best prediction performance with the R2 value of 0.9329.