Cooling load prediction based on correlative temporal graph convolutional network

Abstract

The efficient operation of the cooling source system depends on a reasonable control strategy, and accurate cooling load prediction provides important guidance for optimal control. As there are numerous variables that affect the prediction of cooling loads, many cooling load prediction methods try to exploit the variables in the temporal domain. However, the correlations between the variables are not reasonably utilized by many methods. To exploit the implicit information of the data and obtain an accurate cooling load prediction, the correlative temporal graph convolutional network (CTGCN) is used to predict the cooling load, which can extracted the correlation information and the temporal information. Notably, the correlations between the key variables that affect the cooling load prediction are used for the correlation graph construction, which provides guidance for correlation information extraction. Some traditional prediction methods are compared to prove the effectiveness of the proposed method in the field of cooling load prediction. The results show that the proposed model has great practical value in cooling load prediction.