基于SAE-ELM的电动汽车充电站负荷预测模型

Model of Load Forecasting of Electric Vehicle Charging Station Based on SAE-ELM

电动汽车(electric vehicle,EV)用户充电行为在时间和空间上的随机性增加了EV充电站负荷预测的难度,为此以提高负荷预测的准确度为目的,通过改进深度学习中的栈式自编码器提出栈式自编码器-极限学习机(SAEELM)的混合模型,并深入研究EV与电网的交互模式;综合考虑影响充电站负荷量的关键因素,如历史负荷、环境、日类型等,对某地充电站进行短期负荷预测并验证;最后与SAE-BP、ELM算法做对比实验,实验结果表明SAEELM对充电站的短期负荷预测更加有效准确,更有利于电网稳定运行。

The randomness of charging behavior in time and space increases the difficulty of load forecasting of EV charging station.In this paper,the stacked auto encoder neural network-extreme learning machine(SAE-ELM) hybrid model is proposed by improving the stack auto encoder of deep-learning to realize short-term load forecasting of charging stations.The interactive mode of electric vehicle and power grid is introduced and the key factors affecting the charging station load,such as historical load,environment,typical day type,etc.,are also considered.Finally,the short-term load forecasting of a practical charging station is realized and compared with SAE-BP and ELM algorithm.The result shows the proposed approach can provide more accurate forecasting result,which benefits the stable operation of power grid.