The accurate energy consumption perdition for building is critical to improve the energy efficient of the operation of the operation of large-scale central air conditioning system in summer. Firstly, the influencing factors of cooling load were identified to determine the inputs of cooling load predition model. Then, the indirect measurement method was proposed to obtain the shopper rate based on the supply frequencies of new wind-8units to identify the custom number in summer. Last, an AFC-HCMAC neural network algorithm is proposed to for dynamic cooling load prediction. The results show that compared with the traditional HCMAC algorithm, the proposed AFC-HCMAC algorithm can effectively reduce the neural network nodes and improve the prediction accuracy. The shoppers rate plays an important role in the cooling load prediction for shopping mall. Increasing shopper rate in the inputs of prediction model can significantly improve the prediction accuracy of dynamical cooling load forecasting for shopping mall.