基于霜冰优化组合模态分解及Informer的空间负荷预测OA

[Objective]This paper proposes a spatial load forecasting method based on RIME-optimized combination modal decomposition and Informer to provide accurate load data for power system planning.[Methods]First,a power geographic information system for the target area is constructed.Subsequently,the connectivity-based outlier factor method was used to detect the historical load data of the cell,and the moving average method was used to rectify the historical load data.Next,symplectic geometry mode decomposition is employed to decompose the corrected cell load time series into components with different frequencies and amplitudes.These components are reconstructed into a high-frequency component,an oscillatory component,and a trend component based on calculated permutation entropy.Then,the rime optimization algorithm optimizes key parameters of variational mode decomposition.This optimized variational mode decomposition was used to perform a secondary decomposition on the high-frequency components of the cell load,yielding high-frequency subcomponents with enhanced regularity.Finally,individual Informer forecasting models are established for each component obtained from the primary modal decomposition reconstruction and the secondary modal decomposition.The prediction results of each component are then reconstructed to obtain the load forecast values for the target year of the corresponding cell.[Results]The spatial load forecasting is completed once the load forecast values for all cells at different spatial locations within the planning area have been calculated.The results of the case analysis indicate that the method proposed in this paper significantly reduces prediction errors compared to the comparative methods,improving prediction accuracy.[Conclusions]The proposed method effectively extracts load regularities through a progressive load regularity analysis technology and achieves spatial load forecasting by establishing Informer models for individual components,obtaining improved prediction results.