Joint interval forecasting of renewable energy stocks using a secondary decomposition approach

Upon the climate change and energy transition, the renewable energy market has emerged as a new investment hotspot. Recognizing the importance of precise stock price interval forecasting for strategic planning and investment decisions, this paper introduces a novel framework that employs a two-stage decomposition and joint prediction approach for interval forecasting of multiple renewable energy stocks. By treating multiple stocks as a single Multi-Interval Time Series (MITS), our method captures inter-stock correlations and market dynamics. Initially, stocks are decomposed using the Multivariate Empirical Mode Decomposition (MEMD) algorithm, following a secondary decomposition of complex components using Multivariate Variational Mode Decomposition (MVMD) based on the complexity and noise present in the initial decomposition. Each frequency component, enriched with inter-stock information, is then predicted and integrated to yield final interval predictions. This approach leverages the inherent interdependencies among the stocks and optimizes the decomposition structure of sub-problems, resulting in superior predictive performance compared to traditional single-stock and decomposition-ensemble prediction models. In the specific empirical analysis, four representative stocks from the sustainable energy sector are selected to generate interval forecasts. The error metrics and statistical tests indicate that our proposed method outperforms other benchmark models and demonstrates strong robustness.