Reverse decomposition method and verification of air conditioning load for an underground railway station

 The random forest method is proposed to decompose the cooling load of the air conditioning system in an underground railway station in order to obtain the subitem cooling load, and provide some reference for the cooling load calculation, system design and system operation of the HVAC system. Based on the analysis of the characteristics of these subitem cooling loads, outdoor parameters, total cooling load and time are selected for model training. The results show that the RMSEs of the training set are 0 to 5.1 kW with average error of 0 to 7.2%. The RMSEs of the test set are 0 to 15.7 kW with average error of 0 to 22.6%. Among these four subitem cooling loads, the lighting equipment load has the highest accuracy in decomposition, while the occupant load and outdoor air load have relatively large errors.