基于GLDSC-ConvAutoformer模型的区域电动汽车短期充电负荷预测  

SHORT-TERM CHARGING LOAD PREDICTION OF REGIONAL ELECTRIC VEHICLES BASED ON GLDSC-CONVAUTOFORMER MODEL

作　　者：李练兵[1] 郭兴辰 曾四鸣 梁纪峰 Li Lianbing;Guo Xingchen;Zeng Siming;Liang Jifeng(State Key Laboratory of Electrical Equipment Reliability and Intelligence(Hebei University of Technology),Tianjin 300131,China;School of Electrical Engineering,Hebei University of Technology,Tianjin 300131,China;Electricity Academy of State Grid Hebei Electric Power Co.,Shijiazhuang 050000,China)

机构地区：[1]省部共建电工装备可靠性与智能化国家重点实验室(河北工业大学),天津300131 [2]河北工业大学电气工程学院,天津300131 [3]国网河北省电力有限公司电科院,石家庄050000

出　　处：《太阳能学报》2025年第2期90-98,共9页Acta Energiae Solaris Sinica

基　　金：河北省省级科技计划(20312102D)。

摘　　要：针对大规模电动汽车并网过程中对电网负荷产生波动的问题,电动汽车短期负荷预测可为电动汽车的优化调度提供决策依据。为更好地保证电网的稳定性与可靠性,提出一种电动汽车短期充电负荷预测方法,以提高负荷预测精度。首先,根据每个充电桩上电动汽车充电的时空差异,构建基于受限动态时间弯曲距离算法的灰关联度模型,将关联度矩阵作为谱聚类算法的度矩阵,构建灰色受限动态谱聚类算法,对所有电动汽车日充电负荷曲线进行聚类,使聚类数据有更好的周期性;其次,对聚类数据分别进行双重卷积化处理,将提取的数据特征分别输入到Autoformer,构建ConvAutoformer负荷预测模型,分别对所聚类结果进行负荷预测;最后,采用实际电动汽车充电桩充电负荷数据进行算例分析。实验结果表明,所提方法能有效提高电动汽车短期充电负荷预测准确度。Aiming at the problem of load fluctuation caused by large-scale electric vehicles in the grid-connected process,short-term load prediction of electric vehicles provides decision-making basis for optimal scheduling of electric vehicles.In order to better ensure the stability and reliability of the power grid,a short-term charging load prediction method of electric vehicles is proposed to improve the load prediction accuracy.Firstly,according to the spatiotemporal differences of EV charging on each charging pile,a grey relational degree model based on dynamic time warping under limited warping path length algorithm is constructed.The correlation degree matrix was used as the degree matrix of spectral clustering algorithm,and the gray limited dynamic spectrum clustering algorithm model was constructed.Cluster the daily charging load curves of all electric vehicles to make the clustering data have better periodicity.Secondly,the cluster data were processed by double convolution,and the extracted data features were input into Autoformer respectively to build ConvAutoformer load prediction model,and load prediction was carried out on the cluster results respectively.Finally,the actual charging load data of electric vehicle charging pile was used for example analysis.Experimental results show that the proposed method can effectively improve the accuracy of short-term charging load prediction for electric vehicles.

关 键 词：电动汽车 特征提取 预测 受限动态时间弯曲距离 灰色受限动态谱聚类 ConvAutoformer 

分 类 号：TM734[电气工程—电力系统及自动化]