A real time monitoring method of energy consumption based on data mining techniques is proposed to compensate the deficiency of common energy consumption methods in real time and intelligence. The new method can identify energy consumption patterns by clustering analysis of historical energy consumption data, get the decision tree of energy consumption pattern by classifying the energy consumption data, match the real time energy consumption data with the energy consumption patterns, make outlier analysis with historical data of the same pattern, and then determine whether the current energy consumption is abnormal. The experiment with energy consumption data from the comprehensive building proves that the new method is effective in detecting the abnormal data of energy consumption real timely.