Abstract: Targeting the energy consumption problem of metro VAC (ventilation and air-conditioning) water system equipment, a water system energy consumption early warning model is established, using cooling performance coefficient as the core, based on historical normal data of the equipment, and combining data statistics, evidence theory and neural network genetic algorithm. Through the early warning model, it is possible to allocate abnormal energy consumption of cooling equipment or pump in water system in time. By detecting systematic problems such as collection abnormality or equipment failure and dealing with them in time, energy waste caused by this can be reduced, as well as urban rail transit operation cost.