Seasonal waste, geotherm, nuclear, wood net power generations forecasting using a novel hybrid grey model with seasonally buffered and time-varying effect

Seasonal volatility data is often disturbed by uncertain external shocks, making accurate forecasting particularly strenuous. This paper proposes a progressive adaptive prediction framework of data preprocessing, feature recognition, and seasonal prediction, namely SAWBO-TNGBM (1,1) model. Specifically, the seasonal full information variable weight weakening buffering operator is employed to effectively smooth the nonlinear fluctuation data. Furthermore, the grey Bernoulli model is extended by considering the time-varying effect, and Grey Wolf Optimization algorithm improves the overall prediction efficiency. Necessarily, the Convertibility, Unbiasedness, and Recursiveness are fully derived and proven, which undoubtedly improves the reliability and the ability to capture seasonal information. Empirically, from a data-driven perspective, US seasonal clean energy net generations with diverse fluctuating characteristics are utilized to validate the predicted performance, including quarterly series (waste, geotherm) and monthly series (nuclear, wood). Results obtained from comprehensive experimental comparative analyses show that the fitting ability of the SAWBO-TNGBM (1,1) model exceeds that of other models, demonstrating its flexibility, universality, and high precision. Lastly, innovative robustness testing and extended analysis ensure that the novel model provides an effective tool for seasonal forecasting in clean energy generation.