Based on the conventional building energy consumption and equipment failure prediction methods, according to the data set formed by the BIM model data and Internet of things l data in smart buildings, this paper proposes a multi task learning method based on BP neural network, which uses the same neural network model to study and predict building energy consumption and equipment failure prediction. It also expounds the design method of network structure and loss function, which not only provides a new idea for building energy consumption and equipment failure prediction, but also provides an algorithm basis for intelligent push service in intelligent building operation and maintenance management system, and truly applies artificial intelligence technology to intelligent building operation and maintenance management system.