Optimizing Solar Radiation Prediction Based on The Internet of Things Platform in Photovoltaic Power Plant

Abstract

The solar radiation value parameter is one of the most important parameters in determining the output power value of photovoltaic panels. Accurate prediction of this parameter is crucial for dispatching and load management planning. Managers and designers encounter economic and managerial challenges due to the uncertainty and difficulty in predicting solar radiation levels. This research introduces a highly accurate prediction method utilizing tree-based methods, enhanced by meta-heuristic algorithms to boost performance. The proposed method emphasizes preventing overfitting and ensuring high reliability for use in Internet of Things systems. Meta-heuristic algorithms are utilized for optimizing tree-based methods, as well as for feature and instance selection. Employing meta-heuristic methods as the main innovation in this research not only optimizes machine learning model settings but also mitigates the impact of noise, outliers, and ineffective inputs, thereby enhancing the final output quality. Utilizing an innovative fitness function in model optimization enhances prediction accuracy and adaptability to real photovoltaic power plant environments. The final outcome is a strong model that has a score of 0.95 with the R-square criterion and is optimal model.

Keywords

• internet of things
• decision tree
• machine learning
• bat algorithm
• photovoltaic power plants.