Long-term solar energy prediction with Robust Monte Carlo simulation method to predict solar energy using univariate and multivariate time series models

Long-term forecasting of Global Horizontal Irradiance (GHI) is a key requirement for the reliable planning and operation of renewable energy systems. This study investigates the performance of advanced deep learning and statistical models for solar irradiance prediction using both univariate and multivariate time series approaches. In the univariate setting, the Informer architecture is evaluated against classical statistical benchmarks, including Auto-Regressive Integrated Moving Average (ARIMA) and Seasonal Autoregressive Integrated Moving Average (SARIMA) models. For multivariate forecasting, meteorological variables are integrated through hybrid Transformer — RNN frameworks to capture complex temporal dependencies. The proposed models are validated using 20 years of ground-based GHI measurements and associated meteorological data under real-world operating conditions. To account for climate-induced uncertainty, Monte Carlo simulations are employed to generate probabilistic forecasts and to assess the impact of extreme climate scenarios on long-term solar resource availability. The results demonstrate that deep learning models consistently outperform traditional approaches in terms of accuracy and robustness for long-term GHI prediction. The proposed methodology provides a scalable and reliable framework for enhancing solar energy planning, supporting grid stability, and informing adaptive smart-grid strategies under future climatic variability.