Solar Radiation Prediction Enhancement Via Ensemble Learning Method

Efficient forecasts of solar radiation are very important for optimizing solar energy systems, enhancing energy management, and ensuring the reliability of all renewable energy solutions. This study scrutinizes the performance evaluation of machine-learning algorithms as Decision Trees (DT), Support Vector Machines (SVM) and K - Nearest Neighbors (KNN), Linear Regression (LR), and Logistic Regression (LRg) to forecast solar radiation in relation to urban energy planning. Logistic Regression, usually used in binary classification problems, is modified in this study to demonstrate its regressive capacity. Its simplicity and modeling of linear relationships provide in-depth hungry comparative insights in conjunction with an ensemble. The analysis is made applying the year 2020 data at a high resolution sourced from the Solar Radiation Research Laboratory (SRRL) at the National Renewable Energy Laboratory (NREL). And the dataset considered metropolitan cities, including Delhi, Bangalore, and Hyderabad in the year 2020, where Global Horizontal Irradiance (GHI) is the core predictor variable in this study. Metrics of performance such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and Mean Squared Error (MSE), are similarly utilized to appraise the accurate model of the highest predictive power. Further, the current study incorporates weather parameters, such as temperature, humidity, and wind speed, to augment the forecasting operation and make a thoroughly vast analysis. The findings present an exhaustive comparison of the algorithms along with a highlighter showing their strengths and limitations within an urban energy planning context. The research has brought to light the potential for implementing machine learning algorithms in forecasting solar energy while offering key insights to enhance the efficiency, scalability, and reliability of solar energy systems in highly populated areas.