An Approach to Solar Radiation Prediction Using ARX and ARMAX Models

Abstract

In recent years, Brazilian meteorological networks have introduced numerous automatic stations to monitor global solar radiation at hourly intervals. Historically, large-scale climate data measurement has supported aviation and agricultural activities. The need for a good mathematical model to adequately describe a process is a great challenge, since the performance of control and simulation systems can significantly impact both system operation and/or automation and system planning. The design of control systems based on predictive models should allow for describing the dynamic behavior of the process or system under realistic conditions, as well as finding the simplest possible model to optimize the computational resources. The present work sought to predict solar radiation levels via ARX and ARMAX linear mathematical modeling. During the simulations, global horizontal radiation was defined as input, while the following parameters were outputs: extraterrestrial normal radiation, infrared horizontal radiation, extraterrestrial horizontal radiation, direct normal radiation, and diffuse horizontal radiation. It must be noted that a new simulation was performed for each variable. The use of linear modeling (ARX and ARMAX) to predict solar radiation data was efficient for extraterrestrial normal, infrared, and extraterrestrial horizontal radiation with the mean square error equal to 2.51, 1.40 and 7.15%, respectively.