Photovoltaic Solar Radiation Prediction Using Advanced Machine and Deep Learning Models

The increasing demand for energy has led to the worldwide adoption of solar photovoltaic (PV) systems. Accurate predictions of solar irradiance are necessary for the integration of PV systems into the electrical grid. Through utilizing of advanced machine learning (ML) and deep learning (DL) techniques, this study aims to improve solar irradiance prediction. XGBoost, Multilayer Perceptron (MLP), Random Forest (RF), Support Vector Regression (SVR), Gradient Boosting, K-Nearest Neighbors (KNN), LightGBM, CatBoost, Long Short-Term Memory (LSTM), and Bidirectional LSTM (BiLSTM) are among the state-of-the-art models whose performance is assessed by a thorough framework. The HI-SEAS weather station provides a strong dataset of 32,686 entries with 11 meteorological parameters, including temperature, pressure, humidity, wind direction, and wind speed, which is used to train and test these models. The dataset is preprocessed to ensure high-quality input for the models, and feature engineering techniques are applied to enhance predictive accuracy. Standard metrics, such as Mean Squared Error (MSE), Mean Absolute Error (MAE), and statistical R2, are used to thoroughly assess each model's performance. Our results show that Gradient Boosting and LightGBM are two of the ML models that perform well, demonstrating their capacity to process high-dimensional data and produce precise predictions. LSTM, and BiLSTM models perform better than conventional ML techniques when it comes to identifying the intricate, non-linear correlations present in solar irradiance data. By identifying the most effective models, this research advances reliable solar forecasting tools for enabling continuous grid integration and supporting the global transition to sustainable energy.