Machine Learning-Based Solar Irradiance Forecasting Model Using GPS

Abstract

Accurate solar irradiance forecasting is critical for optimizing photovoltaic (PV) systems, enhancing grid stability, and enabling effective energy management. This study explores the integration of machine learning (ML) techniques with Global Positioning System (GPS) data to improve the accuracy of solar irradiance prediction. By incorporating Total Electron Content (TEC) and Integrated Water Vapor (IWV) derived from GPS data, alongside meteorological variables such as pressure and temperature, a robust forecasting model was developed. Among the three backpropagation algorithms tested—Levenberg-Marquardt, Bayesian Regularization, and Scaled Conjugate Gradient—the Bayesian Regularization algorithm with a 10-layer neural network achieved the best performance, with the lowest Mean Square Error (MSE) and highest Correlation Coefficient (R). The model's predictions closely aligned with measured solar irradiance, demonstrating its reliability. Despite challenges such as data availability and computational complexity, the study highlights the potential of integrating GPS-derived data into ML-based solar irradiance forecasting. This approach offers a promising solution for advancing renewable energy management and supporting the transition to sustainable energy systems.