Photovoltaic power prediction system based on dual-layer decomposition strategy and a novel dynamic grouping multi-objective coati optimization algorithm
Xiaole Tang, Hao Lu, Yanting Kang, Wenjun Zhao
Abstract
The substantial volatility of photovoltaic (PV) power output presents challenges to the stable operation of power grids. To improve the accuracy and stability of PV power prediction, this study proposes a PV power prediction system based on a dual-layer decomposition strategy and a dynamic grouping multi-objective Coati optimization algorithm (DGMOCOA). This system includes three core modules: data preprocessing, optimization, and prediction. The dual-layer decomposition strategy effectively utilizes high-frequency signal information, overcoming the limitations of single-layer methods. Additionally, the DGMOCOA employs a dynamic grouping strategy to balance exploration and exploitation, enhancing the system’s accuracy and stability. To validate the proposed strategy and hybrid model, extensive experiments were conducted. The results show that, compared to 11 benchmark models, the proposed model achieves higher prediction accuracy and greater stability. In interval prediction, it also demonstrates higher coverage and narrower intervals.