Due to the limits of battery capacity of mobile devices, how to select cloud services to invoke in order to reduce energy consumption in mobile environments is becoming a critical issue. This paper addresses the problem of mobile service selection for composition in terms of energy consumption. It formally models this problem and constructs energy consumption computation models. Energy consumption aggregation rules for composite services with different structures are presented. It adopts the genetic algorithm to resolve it. A replanning mechanism is also proposed to deal with the changeable conditions and user behavior. A series of experiments are conducted to evaluate the performance of our method. The results show that our service selection method significantly outperforms traditional methods. Even if the conditions or user behavior is changeable, this method is still effective to recommend services. Moreover, the service selection method performs good scalability as the experimental scale increases. Note to Practitioners - To addresses the challenges from the prospective of service selection in mobile environment to reduce energy consumption, this paper constructs an energy consumption computation model for mobile devices and formalizes service selection for composition as an optimization problem. In order to solve the NP-hard problem, it adopts the genetic algorithm and conducts a serial of experiments to show the effectiveness and efficiency of the solution. The proposed solution can help users to select the proper services with the least energy consumption in mobile environment. It can be implemented and deployed as a cloud service to recommend services for mobile users.