Multi-space collaboration framework based optimal model selection for power load forecasting. (15th May 2022)

• Record Type: Journal Article
• Authors: Xian, Huafeng; Che, Jinxing
• Is Part Of: Applied energy. Volume 314(2022)
• Publication Date: 2022-05-15
• Subjects: Optimal model selection -- Multi-space collaboration -- Meta-heuristic algorithm -- Power load forecasting; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2022.118937
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: A multi-space collaboration framework for optimal model selection is proposed. The model selection capability of the MSC framework is verified through simulation study. The MSC framework is applied to select the optimal SVR model for the power load forecasting. The MSC framework has strong adaptability to the candidate size of the parameter domain. Abstract: In recent years, power load forecasting has become a hot and open issue in the field of energy. However, the optimal model selection for power load forecasting is a tricky problem. In this paper, we propose a multi-space collaboration (MSC) framework for optimal model selection. Specifically, our framework adopts space separation strategy to do the model selection on the subspace, which increases the probability of selecting the optimal model; A subspace elimination strategy is introduced, and the subspace with low development potential is gradually eliminated as iteration progresses, making the framework pay more attention to better parameter domain. We conduct a simulation study and a real-world case study of experimental analysis to verify the effectiveness of the proposed framework. On several test functions of known optimal situation, the model selection ability of the MSC framework is better than the ordinary meta -heuristic algorithms, and it has excellent robustness. In addition, the results of the real-world case study show that the optimal SVR model selected by our framework is absolutely superior to various comparison models, and our framework has strong adaptability to the candidate size of the parameter domain.