A Data-Driven Approach for Accurate Solar Energy Prediction Using Artificial Intelligence

Abstract

Specifically, to ensure the reliable inclusion of renewable energy sources in the smart power grids, there must be accurate prediction of solar energy production. Fluctuations in the amount of sunlight hitting the earth, due to weather and atmospheric processes, cause uncertainty to power production. The proposed study offers a method based on the data analysis to forecast the generation of solar power with the high precision with the help of the artificial intelligence (AI). A large body of data that included meteorological and photovoltaic (PV) parameters of the UNISOLAR dataset on Kaggle was to be trained and tested on a range of artificial intelligence models, such as the Random Forest (RF), Gradient Boosting Regressor (GBR), and Long Short-Term Memory (LSTM) networks. The enhancement in the prediction accuracy was carried out by means of feature selection, correlation analysis and time-series modeling. The findings show that LSTM model performed better with a RMSE of 17.3 W/m 2 and R 2 = 0.94 that was better than other methods such as ensemble and baseline. The proposed solution proves the use of the AI-based forecasting technique to optimize solar energy utilization and grid stability in dynamic environments.