基于Hadoop的电动汽车充电站负荷预测
2018-07-10刘晓悦孙玉容
刘晓悦 孙玉容
摘 要: 为了应对大规模电动汽车充电站负荷的调度管理,提出一种基于Hadoop的模糊灰色GA?BP神经网络负荷预测模型。在云平台上,运用模糊聚类及灰色关联度分析选取相似日,将相似日负荷代入MapReduce架构下的GA?BP神经网络预测模型进行学习,获得待测日的预测负荷。以城市辖区快换式充电站实测数据进行实验,实验结果证明,此方法在快换式充电站的负荷预测上兼具高效性与精确性。
关键词: Hadoop架构; 模糊聚类; 灰色关联分析; 负荷预测; BP神经网络; 快换式电动汽车充电站
中图分类号: TN711?34; TM734 文献标识码: A 文章编号: 1004?373X(2018)13?0074?04
Abstract: A fuzzy gray GA?BP neural network load forecasting model based on Hadoop is proposed to deal with the load dispatch management of large?scale electric vehicle charging station. The fuzzy clustering and gray relational analysis are used in the cloud platform to select the similar days, and then the similar daily loads are brought into the GA?BP neural network prediction model under MapReduce architecture for learning, so as to obtain the forecasting load of the testing day. The experiment was performed for the measured data of quick?change charging station in city area. The experimental results show this method both has efficiency and accuracy for the load forecasting of the quick?change charging station.
Keywords: Hadoop architecture; fuzzy clustering; gray relational analysis; load forecasting; BP neural network; quick?change electric vehicle charging station
0 引 言
电动汽车行业迅速发展,电动汽车负荷数据逐年增加,导致充电站负荷数据量急剧增涨,对电网系统及基础设施产生重大的影响[1?3]。众多数据和实例表明,精准的负荷预测保障了快换式电动汽车的平稳运行和能源的合理调度,对国家电网调度、能源管理有着重要意义[4?6]。
目前,针对电动汽车充电站负荷预测的方法还存在着局限性,文献[7]提出基于模糊聚类分析和BP神经网络的电力系统负荷预测,通过模糊聚类分析选择相似日,建立BP网络模型。但是通过模糊聚类得到的只是与预测日影响因素相似的负荷数据,不具备良好的负荷曲线相似性[8],并且训练样本复杂,容易陷入局部极小值。文献[9]运用模糊聚类灰色关联分析法选择相似日,根据负荷特点分时建立网络模型预测,但BP网络存在收敛速度慢、权重初始化随机等问题。……
