基于优化模态分解与DGRUK的综合能源系统负荷预测OA

[Objective]To further explore the potential structure of load sequence data in integrated energy systems(IES)and enhance the overall prediction accuracy and reliability of IES load forecasting models,this paper proposes a novel load forecasting method for IES based on optimized modal decomposition and the DGRUK network.[Methods]Firstly,for the multi-energy load sequence decomposition stage,an improved ivy algorithm is employed to optimize the parameters of the improved complete ensemble empirical mode decomposition.Decomposes cooling,heating,electricity,and other multi-energy load sequences into intrinsic mode function components,thereby reducing the non-stationarity and complex coupling of the original sequences.Secondly,during the feature extraction phase,the discrete cosine transform is integrated into the channel attention mechanism to efficiently capture global correlations among different channels and enhance the representation of key features.Finally,a DGRUK network is constructed by leveraging the advantages of Kolmogorov-Arnold networks in nonlinear mapping.This step compensates for the limitations of traditional fully connected layers in handling complex nonlinear relationships,thereby improving the model's capability to process high-dimensional,non-stationary load data.[Results]The proposed method achieves mean absolute percentage errors(MAPE)of 2.045%,2.379%,and 1.234%for cooling,heating,and electrical load forecasting,respectively.All error metrics are lower than those of other commonly used methods,verifying the effectiveness of the proposed approach.[Conclusions]The proposed method effectively addresses the issues of non-stationarity and complex coupling in multi-energy load sequences of integrated energy systems.It provides scientific support for the optimal scheduling and operational management of integrated energy systems.