Base station is a major node for communication network's energy consumption. The accurate calculation of the energy-saving amount for the base station under EPC model is a technology bottleneck in this field. This paper proposed a modeling method of energy consumption of the base station based on particle swarm optimization (PSO) and least squares support vector machine (LSSVM) of sliding window, oriented at three kinds of typical scenarios base station. In this approach, a sliding window was established by selecting configuration parameters of base station and real-time data for pretreatment, and then the dynamic energy consumption model was obtained for the base station, which varied in accordance with that of the sliding window by means of the parameters for PSO training model and LSSVM regression training model. Compared with the simulation and test results from the sample base station, the proposed energy consumption model shows high prediction accuracy and generalization ability, and is applicable for the evaluation of energy-saving engineering of the base station.