Building energy consumption prediction provides the possibility for regulating running condition of equipments in advance. Then the equipments will keep good movement and building energy consumption will reduce obviously. This paper built an energy consumption prediction evaluation model according to Matlab Artificial Neural Network Toolbox. The model was trained and simulated by operation data in June-September of 2008 and 2009 of a case building. Then it can be used to predict this building energy consumption by special data, such as meteorological characteristics of prediction year, operation load, operation time and energy consumption of last year. With more building samples, the model will be used in wide range of building energy consumption prediction.