Comparison of solar irradiance forecasting performance with K- means++ clustering combined with hybrid deep learning models

Abstract

Solar irradiance forecasting plays a crucial role in renewable energy, weather prediction, and climate modeling. Accurate forecasts are essential for optimizing solar power efficiency, grid integration, and energy planning. Traditional forecasting methods, based on physical and statistical models, struggle to capture the complex, nonlinear relationships inherent in solar irradiance. To address these challenges, this chapter presents a comparative analysis of advanced machine learning (ML) and deep learning (DL) models. Techniques like CNNs, RNNs, and hybrid models have demonstrated strong capabilities in extracting temporal and spatial patterns from Solar data. The integration of K- means++ clustering with DL frameworks further enhances model robustness, generalization, and interpretability. This chapter evaluates hybrid models, such as Temporal CNN- LSTM, CNN- GRU using metrics based on Solcast data. Results reveal that the TCNN- GRU model outperforms other state- of- the- art approaches, underscoring the value of clustering- enhanced DL frameworks for accurate solar irradiance forecasting. © 2025, IGI Global Scientific Publishing. All rights reserved.