Abstract:Due to the inherent complexity and irregularity of the cold load time series, problems such as gradient disappearance, modal aliasing and over-fitting are prone to occur in the prediction process. It is still a difficult task to predict the cold load of large public buildings. To solve this problem and improve the prediction accuracy, the VMD-GRU model is proposed. The proposed model was tested using real data from large public buildings. The prediction method process is as follows: 1) Correlation analysis of the original data, selection of highly correlated predictions; 2) Decomposition of the original data sequence into independent eigenmode functions using VMD; 3) Prediction of each component using GRU ; 4) Add the component prediction results to obtain the cold load prediction value. In order to verify the validity of the model, a large public building in Xi'an is taken as an example to analyze the energy consumption and compare it with other prediction models such as BP, GRU, EMD-BP, VMD-BP, EMD-GRU. The experimental results show that the proposed model can effectively solve the problems of gradient disappearance, modal aliasing and over-fitting, and accurately predict the cold load of large public buildings.