一种基于协同过滤的电动汽车充电推荐方法  

Recommendation method for electric vehicle charging based on collaborative filtering

作　　者：卜凡鹏[1] 田世明[1] 高晶晶[2] 齐林海 

机构地区：[1]中国电力科学研究院,北京100192 [2]华北电力大学控制与计算机工程学院,北京102206

出　　处：《科技导报》2017年第21期61-67,共7页Science & Technology Review

基　　金：国家高技术研究发展计划(863计划)项目(2015AA050203);国家电网公司科技项目(52094017002U)

摘　　要：基于大量的充电行为数据,建立电动汽车用户的充电兴趣模型,将用户感兴趣但未发现的最佳充电选择推荐给用户,实现充电行为有序引导是一个重要问题。本文针对电动汽车充电提出一种基于协同过滤算法的推荐模型,得出最佳推荐模型参数指标,充电10次以下的新用户采用基于用户的协同过滤算法,充电10次以上的老用户采用基于物品的协同过滤算法;基于用户的协同过滤算法的最佳邻居数和推荐列表长度均为3;基于物品的协同过滤算法的最佳推荐列表长度为4。指出负荷聚合商可以结合参与需求响应计划的情况,对推荐列表进行再优化,将与需求响应冲突的推荐信息过滤掉,从而实现有序充电控制。Based on a large number of charging behavior data,a charging interest model of electric vehicle users is established.The best charging options which users are interested in but have not found are recommended and charging behavior is orderly guided.In this paper,a recommended model based on a collaborative filtering algorithm is proposed for electric vehicle charging,and the best model parameter index is obtained through test and evaluation.New users with 10 times or less may use the user-based collaborative filtering algorithm while old users who have charged more than 10 times may adopt the collaborative filtering algorithm based on items.The optimal neighbor and recommended list length is 3 for the user-based collaborative filtering algorithm,and 4 for the algorithm based on items.The paper points out that the load aggregator can re-optimize the recommendation list in combination with the demand response plan and filter out the recommended information that conflicts with the demand response to realize the orderly charging control.

关 键 词：有序充电 电动汽车 协同过滤 智能推荐 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]