Long range wide area networks (LoRaWAN) is one of the novel specification for the applications of Internet-of-Things. Recent research focused on the propagation model and coverage discussion based on experimental measurement. In this paper, power- class analytical model is firstly proposed to model the operations of LoRaWAN end-devices for throughput and power consumption. Based on this model, the energy efficient class-selection algorithm (EECA) is proposed to achieve the optimal energy efficiency with the consideration of power consumption on end-devices by the suitable class selection scheme. A system level simulation has been conducted for performance evaluation. The simulation results provide the guideline for system parameters configuration and class selection on each end-devices under different scenarios.