Long Term Solar Energy Prediction Using Deep Learning Models with Monte Carlo Simulation

Abstract

Accurate long-term forecasting of Global Horizontal Irradiance (GHI) is essential for resilient solar energy planning in the face of climate uncertainty. This study presents a robust hybrid framework that integrates the Informer and Prophet models, supported by data preprocessing and feature selection using temperature, humidity, and the clear sky index. Prophet is used to analyze long-term trends and seasonality, while Informer effectively models complex temporal dependencies for extended forecasts. To account for uncertainty and improve the reliability of the forecast, Monte Carlo simulations generate probabilistic predictions with 95% confidence intervals over 30 years, based on historical data from 2004 to 2024. The results demonstrate the effectiveness of Informer in producing climate-resilient solar irradiance forecasts, contributing to the design of sustainable and adaptive energy systems.