Abstract-The energy consumption of software plays an important role in mobile application. Firstly, on the point of measurement, the characteristics of mobile terminal software are studied, and the functional relation between software energy and software characteristic is modeled as a nonlinear function. Secondly, based on BP neural network, an energy consumption model of mobile software is built. After the analysis of software characteristics in system architecture, the training and study mechanism of BP neural network in energy consumption measures is illustrated here. Lastly, BP training to fit the functional relation between software architecture and energy consumption is verified by experimental results.