Solar Photovoltaic Power Prediction Using Statistical Approach-Based Analysis of Variance

Abstract

With the increase in global demand for energy and the rise of environmental warnings supported by the United Nations and its sustainable development goals (SDGs) in 2015, transitioning from traditional energy systems to renewable ones, especially solar energy systems, has become necessary. However, this transition should be supported by prediction models that can help forecast these power outputs. This research aims to develop a data-driven model based on a statistical approach. Analysis of variance ANOVA and fit summary were the tools that were used in creating the model. Three input variables, namely Global Radiation, Ambient Relative Humidity, and Ambient Temperature, were utilized alongside one output variable, output power. The model utilized 360 readings during six hours from 10:00 am to 4:00 pm. Stat-ease software was used to develop the model. The quadratic statistical model shows significant results with five statistical terms. The Model’s F-value of 687.89 indicates that the model is highly significant, demonstrating only a small chance of 0.01% that such a large F-value could be caused by random variations. In addition, the P-values for the remaining model terms in the ANOVA table, all being less than 0.0500, confirm their significance. The developed model was validated by comparing the original experimental data with those obtained from the model. The validation showed an average percentage error of 7.35%.