Research On Short-term Forecasting Of Multi-element Loads In Integrated Energy Systems Based On Deep Learnin

In order to further achieve the dual carbon goal of “carbon peak,carbon neutralization”,improve energy utilization efficiency and reduce environmental pollution,the emerging concept of integrated energy system has been spawned.Accurate multi-load forecasting is an important prerequisite for the operation optimization of integrated energy system.Therefore,this paper takes the multi-load of integrated energy system as the research object,considers the uncertainty and coupling between multi-loads,as well as the influence of meteorological information,calendar information and other related factors,and proposes a multi-load short-term forecasting model of integrated energy system based on deep learning to achieve more accurate multi-load forecasting.Firstly,this paper analyzes the characteristics of multi-load.Aiming at the strong volatility and coupling of multi-load,it is proposed to use VMD to reduce the uncertainty of multi-load and obtain more stable multi-load modal components,and then combine the relevant influencing factors to form the input data of the prediction model.Then the TCN model is used to extract the characteristics of multivariate load time series.The GCN model mines the coupling characteristics between multiple loads and influencing factors.Using LSTM to learn the overall characteristics of the input data and ensure the output dimension of the model;a multi-model combination prediction model based on VMD and TCN-GCN-LSTM was constructed.In addition,aiming at the deficiency of subjective selection of VMD decomposition parameters,the energy difference component selection method is proposed to select the modal component parameters of various loads.At the same time,Copula entropy is used to estimate the mutual information correlation between multiple loads and influencing factors,quantitatively characterize the adjacency matrix of GCN model,and improve the solvability of prediction model for multiple load coupling characteristics.Based on this,a prediction model based on IVMD and TCN-CGCN-LSTM multi-model combination is constructed.In addition,the parallel processing method based on attention mechanism of Transformer is used to improve the acquisition of the overall characteristics of multivariate load,and a prediction model based on IVMD and TCN-CGCN-Transformer multi-model combination is constructed.Finally,through the example analysis,compared with other basic prediction models,the validity of the model proposed in this paper is verified,and the advantages and characteristics of the three prediction models are analyzed.It can select the prediction model according to its advantages,which has certain theoretical and practical significance.