Modelling Nonseasonal Daily Clearness Index for Solar Energy Estimation in Ilorin, Nigeria Using Support Vector Regression

Abstract

Solar radiation is the primary energy source for the planet and is crucial for energy generation in technologies such as photovoltaic systems and solar thermal food dryers. However, accurately quantifying solar radiation poses challenges due to its variability and the lack of appropriate instrumentation, among other factors. To address this, support vector regression (SVR), a machine learning (ML) algorithm, was employed using various kernel functions such as linear, radial basis function (RBF), and sigmoid, with hyperparameter tuning. This approach aimed to estimate the daily clearness index (K_T), which is a key metric for estimating global solar radiation at Ilorin (8° 32′ N, 4° 34′ E), Nigeria. The SVR models were developed and assessed by considering statistical measures such as the correlation coefficient and mean absolute error. The input parameters used in the model included sunshine hours, maximum temperature, minimum temperature, and the ratio of both temperatures. The correlations between K_T and its estimators, and between its actual and calculated values were all below 70%. SVR-RBF outperformed the others, including the traditional regression model, under the statistical assessment measures, both with the training dataset and the testing dataset. Although the regression model obtained under the same conditions surpassed the other kernel functions in some areas and is highly competitive, SVR-RBF is recommended for the estimation of the daily clearness index in this vicinity.