基于谱聚类和LSTM神经网络的电动公交车充电负荷预测方法

doi:10.12204/j.issn.1000-7229.2021.06.006

电力建设 ›› 2021, Vol. 42 ›› Issue (6): 58-66.doi: 10.12204/j.issn.1000-7229.2021.06.006

• 电动汽车参与电网调度的关键技术·栏目主持傅质馨副教授· • 上一篇下一篇

基于谱聚类和LSTM神经网络的电动公交车充电负荷预测方法

王哲¹, 万宝², 凌天晗², 董晓红³, 穆云飞³, 邓友均³, 唐舒懿³

1.国网天津市电力公司,天津市 300010
2.国网天津市电力公司滨海供电分公司,天津市 300450
3.智能电网教育部重点实验室(天津大学),天津市 300072

收稿日期:2020-10-20 出版日期:2021-06-01 发布日期:2021-05-28
作者简介:王哲(1984),男,硕士,高级工程师,主要研究方向为配电网规划与电网智能化;|万宝(1988),男,学士,工程师,主要研究方向为配电网规划;|凌天晗(1989),男,学士,工程师,主要研究方向为配电网规划;|董晓红(1989),女,博士研究生,主要研究方向为电动汽车与电网互动控制,配电网规划;|穆云飞(1984),男,博士,教授,主要研究方向为电力系统安全与稳定、综合能源集成与应用;|邓友均(1990),男,博士研究生,主要研究方向为电动汽车充放电管理、配电网规划;|唐舒懿(1996),女,博士研究生,主要研究方向为电动汽车负荷预测及电动汽车有序充电控制、配电网规划。
基金资助:
国网天津市电力公司科技项目(KJ20-1-38)

Electric Bus Charging Load Forecasting Method Based on Spectral Clustering and LSTM Neural Network

WANG Zhe¹, WAN Bao², LING Tianhan², DONG Xiaohong³, MU Yunfei³, DENG Youjun³, TANG Shuyi³

1. State Grid Tianjin Electric Power Company, Tianjin 300010, China
2. State Grid Tianjin Binhai Electric Power Company, Tianjin 300450, China
3. Key Laboratory of Smart Grid of Ministry of Education (Tianjin University), Tianjin 300072, China

Received:2020-10-20 Online:2021-06-01 Published:2021-05-28
Supported by:
Science and Technology Program of State Grid Tianjin Electric Power Company(KJ20-1-38)

摘要/Abstract

摘要：

目前电动公交车的渗透率较大,且充电频率和充电量较高,故而其充电负荷对电网运行与调度产生着不可忽略的影响。因此,电动公交车的充电负荷预测研究具有重要的理论与现实意义,但由于公交车间歇性与随机性的充电行为在时间上给充电负荷预测增加了难度。为此,提出基于谱聚类和长短期记忆(long short-term memory,LSTM)神经网络的电动公交车充电负荷预测方法。首先,利用考虑距离与形态的谱聚类,对充电负荷曲线进行聚类;其次,综合考虑影响充电负荷的关键因素,如温度、日类型等,利用不同簇的总充电负荷,分别训练LSTM神经网络的模型参数,并预测每簇的充电负荷;接着,对不同簇的预测结果求和即可得到预测日的总充电负荷;最后,通过利用某市实际数据,验证本文所提方法。结果表明,所提方法充电负荷预测结果的平均绝对百分误差(mean absolute percentage error, MAPE)在11%以下,预测准确度有所提升。

关键词: 谱聚类, 长短期记忆网络(LSTM), 电动公交车, 负荷预测

Abstract:

At present, the penetration rate, charging frequency and charging capacity of electric buses are relatively high, so the charging load has a non-negligible impact on the operation and dispatch of the power grid. So, the charging load forecasting research has important theoretical and practical significance, but the intermittent and random charging behavior increase the spatial forecasting difficulty. Therefore, the charging load forecasting method of electric buses is proposed on the basis of spectral clustering and long short-term memory (LSTM) neural network. First of all, the charging load curve is clustered according to spectral clustering considering the distance and the shape. And then, considering the key factors that affect the charging load, such as historical load, temperature and day type, the model parameter of LSTM neural network is trained using each cluster charging load, and the charging load of each cluster is predicted. Then, the total charging load of the forecasting day is to sum the forecasting results of different clusters. Finally, on the basis of the historical real data in a certain city, the proposed method is verified. The result shows the mean absolute percentage error (MAPE) of charging load prediction result of the proposed method is below 11%, and the accuracy of load forecasting is improved.

Key words: spectral clustering, long short-term memory network (LSTM), electric bus, load forecasting

中图分类号:

TM715

王哲, 万宝, 凌天晗, 董晓红, 穆云飞, 邓友均, 唐舒懿. 基于谱聚类和LSTM神经网络的电动公交车充电负荷预测方法[J]. 电力建设, 2021, 42(6): 58-66.

WANG Zhe, WAN Bao, LING Tianhan, DONG Xiaohong, MU Yunfei, DENG Youjun, TANG Shuyi. Electric Bus Charging Load Forecasting Method Based on Spectral Clustering and LSTM Neural Network[J]. ELECTRIC POWER CONSTRUCTION, 2021, 42(6): 58-66.

图/表 11

图1 充电负荷预测方法框架

Fig.1 Framework of charging load forecasting method

图2 电动公交车日负荷曲线

Fig.2 Daily load curve of the electric bus

图3 LSTM网络单元结构

Fig.3 Structure of LSTM network unit

图4 SC指数和DBI指数随聚类数K的变化趋势

Fig.4 Change of SC and DBI indices with the amount of the amount K of the clusters

图5 聚类中心线

Fig.5 Cluster center line

表1 每类的LSTM结构参数与预测性能

Table 1 Structural parameters and forecasting performance of LSTM networks for each cluster

图6 类1训练和预测值

Fig.6 Training and predictive values of Class 1

图7 类2训练和预测值

Fig.7 Training and predictive values of Class 2

图8 类1和类2训练时的损失值变化

Fig.8 Changes in loss value during the training for Class 1 and Class 2

图9 负荷预测模型对比

Fig.9 Comparison of load forecasting model

表2 预测模型MAPE误差对比

Table 2 MAPE error comparison of the prediction model

参考文献 24

[1]	罗卓伟, 胡泽春, 宋永华, 等. 电动汽车充电负荷计算方法[J]. 电力系统自动化, 2011,35(14):36-42.
	LUO Zhuowei, HU Zechun, SONG Yonghua, et al. Study on plug-in electric vehicles charging load calculating[J]. Automation of Electric Power Systems, 2011,35(14):36-42.
[2]	WANG H L, ZHANG Y J, MAO H P. Load forecasting method of EVs based on time charging probability[C]// 2018 International Conference on Power System Technology (POWERCON). November 6-8, 2018, Guangzhou, China. IEEE, 2018:1731-1735.
[3]	袁正平, 周伟, 王文斌. 电动汽车充电负荷预测方法研究[J]. 华东电力, 2013,41(12):2567-2572.
	YUAN Zhengping, ZHOU Wei, WANG Wenbin. Charging load forecasting method for electric vehicles[J]. East China Electric Power, 2013,41(12):2567-2572.
[4]	WANG Y, XIA Q, KANG C Q. Secondary forecasting based on deviation analysis for short-term load forecasting[J]. IEEE Transactions on Power Systems, 2011,26(2):500-507. doi: 10.1109/TPWRS.2010.2052638 URL
[5]	ZHU J C, YANG Z L, CHANG Y, et al. A novel LSTM based deep learning approach for multi-time scale electric vehicles charging load prediction[C]// 2019 IEEE Innovative Smart Grid Technologies-Asia (ISGT Asia). May 21-24, 2019, Chengdu, China. IEEE, 2019: 3531-3536.
[6]	牛东晓, 马天男, 王海潮, 等. 基于KPCA和NSGAⅡ优化CNN参数的电动汽车充电站短期负荷预测[J]. 电力建设, 2017,38(3):85-92.
	NIU Dongxiao, MA Tiannan, WANG Haichao, et al. Short-term load forecasting of electric vehicle charging station based on KPCA and CNN parameters optimized by NSGAⅡ[J]. Electric Power Construction, 2017,38(3):85-92.
[7]	王琨, 高敬更, 张勇红, 等. 基于LSTM神经网络的复合变量电动汽车充电负荷预测方法技术研究[J]. 工业仪表与自动化装置, 2019(1):27-31.
	WANG Kun, GAO Jinggeng, ZHANG Yonghong, et al. Study on forecasting method of charging load of hybrid variable electric vehicle based on LSTM neural network[J]. Industrial Instrumentation & Automation, 2019(1):27-31.
[8]	王潇笛, 刘俊勇, 刘友波, 等. 采用自适应分段聚合近似的典型负荷曲线形态聚类算法[J]. 电力系统自动化, 2019,43(1):110-118.
	WANG Xiaodi, LIU Junyong, LIU Youbo, et al. Shape clustering algorithm of typical load curves based on adaptive piecewise aggregate approximation[J]. Automation of Electric Power Systems, 2019,43(1):110-118.
[9]	DONG X S, QIAN L J, HUANG L. Short-term load forecasting in smart grid: A combined CNN and K-means clustering approach[C]// 2017 IEEE International Conference on Big Data and Smart Computing (BigComp). February 13-16, 2017, Jeju, Korea (South). IEEE, 2017: 119-125.
[10]	庞传军, 余建明, 冯长有, 等. 基于LSTM自动编码器的电力负荷聚类建模及特性分析[J]. 电力系统自动化, 2020,44(23):57-63.
	PANG Chuanjun, YU Jianming, FENG Changyou, et al. Clustering modeling and characteristic analysis of power load based on long-short-term-[J]. Automation of Electric Power Systems, 2020,44(23):57-63.
[11]	SUN R Q, XIAO X Q, ZHOU F, et al. Research of power user load classification method based on K-means and FSVM[C]// 2019 IEEE 3rd Information Technology, Networking, Electronic and Automation Control Conference (ITNEC). March 15-17, 2019, Chengdu, China. IEEE, 2019: 2138-2142.
[12]	金之榆, 王毛毛, 史会磊. 基于DBSCAN和改进K-means聚类算法的电力负荷聚类研究[J]. 东北电力技术, 2019,40(6):10-14.
	JIN Zhiyu, WANG Maomao, SHI Huilei. Research on power load clustering based on DBSCAN and improved K-means clustering algorithm[J]. Northeast Electric Power Technology, 2019,40(6):10-14.
[13]	YU T M, YANG J H, LU W. Dynamic background subtraction using histograms based on fuzzy C-means clustering and fuzzy nearness degree[J]. IEEE Access, 2019,7:14671-14679. doi: 10.1109/ACCESS.2019.2893771 URL
[14]	DINESH C, MAKONIN S, BAJIć I V. Residential power forecasting based on affinity aggregation spectral clustering [J]. IEEE Access, 2020,8:99431-99444. doi: 10.1109/Access.6287639 URL
[15]	艾欣, 杨子豪, 胡寰宇, 等. 基于改进k-means算法的VPP负荷曲线聚类方法及应用[J]. 电力建设, 2020,41(5):28-36.
	AI Xin, YANG Zihao, HU Huanyu, et al. A load curve clustering method based on improved K-means algorithm for virtual power plant and its application[J]. Electric Power Construction, 2020,41(5):28-36.
[16]	林顺富, 田二伟, 符杨, 等. 基于信息熵分段聚合近似和谱聚类的负荷分类方法[J]. 中国电机工程学报, 2017,37(8):2242-2253.
	LIN Shunfu, TIAN Erwei, FU Yang, et al. Power load classification method based on information entropy piecewise aggregate approximation and spectral clustering[J]. Proceedings of the CSEE, 2017,37(8):2242-2253.
[17]	端祝超. 基于数据简化拟合的电动公交车充电负荷预测[J]. 电工电气, 2020(3):23-27.
	DUAN Zhuchao. Electric bus charging load forecast based on data simplification fitting[J]. Electrotechnics Electric, 2020(3):23-27.
[18]	陆继翔, 张琪培, 杨志宏, 等. 基于CNN-LSTM混合神经网络模型的短期负荷预测方法[J]. 电力系统自动化, 2019,43(8):131-137.
	LU Jixiang, ZHANG Qipei, YANG Zhihong, et al. Short-term load forecasting method based on CNN-LSTM hybrid neural network model[J]. Automation of Electric Power Systems, 2019,43(8):131-137.
[19]	SUN Y P, GAO Y J, YANG W H, et al. Research on user optimal aggregation based on demand response potential spectrum clustering analysis[J]. The Journal of Engineering, 2017,2017(13):2152-2157. doi: 10.1049/tje2.v2017.13 URL
[20]	彭勃, 张逸, 熊军, 等. 结合负荷形态指标的电力负荷曲线两步聚类算法[J]. 电力建设, 2016,37(6):96-102.
	PENG Bo, ZHANG Yi, XIONG Jun, et al. A two-step clustering algorithm combined with load shape index for power load curve[J]. Electric Power Construction, 2016,37(6):96-102.
[21]	杨甲甲, 刘国龙, 赵俊华, 等. 采用长短期记忆深度学习模型的工业负荷短期预测方法[J]. 电力建设, 2018,39(10):20-27.
	YANG Jiajia, LIU Guolong, ZHAO Junhua, et al. A long short term memory based deep learning method for industrial load forecasting[J]. Electric Power Construction, 2018,39(10):20-27.
[22]	陈吕鹏, 殷林飞, 余涛, 等. 基于深度森林算法的电力系统短期负荷预测[J]. 电力建设, 2018,39(11):42-50.
	CHEN Lüpeng, YIN Linfei, YU Tao, et al. Short-term power load forecasting based on deep forest algorithm[J]. Electric Power Construction, 2018,39(11):42-50.
[23]	MUZAFFAR S, AFSHARI A. Short-term load forecasts using LSTM networks[J]. Energy Procedia, 2019,158:2922-2927. doi: 10.1016/j.egypro.2019.01.952 URL
[24]	中国气象数据网. 中国地面国际交换站气候资料日值数据集(V3.0). 2019[2020-12-30]. https://data.cma.cn.

基于谱聚类和LSTM神经网络的电动公交车充电负荷预测方法

Electric Bus Charging Load Forecasting Method Based on Spectral Clustering and LSTM Neural Network

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 24

相关文章 15

编辑推荐

Metrics

本文评价

[1]	刘敦楠, 张悦, 彭晓峰, 刘明光, 王文, 加鹤萍, 秦光宇, 王峻, 杨烨. 计及相似日与气象因素的电动汽车充电负荷聚类预测[J]. 电力建设, 2021, 42(2): 43-49.
[2]	宋建，束洪春，董俊，梁雨婷，李雨龙，杨博. 基于GM(1，1)与BP神经网络的综合负荷预测[J]. 电力建设, 2020, 41(5): 75-80.
[3]	钱科军，秦萌，宋远军，周振凯，刘乙，谢鹰，陈丽娟. 基于模糊隶属度的充电站多目标优化调度[J]. 电力建设, 2020, 41(2): 118-124.
[4]	朱刘柱, 王绪利, 马静, 陈庆会, 齐先军. 基于小波包分解与循环神经网络的综合能源系统短期负荷预测[J]. 电力建设, 2020, 41(12): 133-140.
[5]	庄家懿, 杨国华, 郑豪丰, 王煜东, 胡瑞琨, 丁旭. 并行多模型融合的混合神经网络超短期负荷预测[J]. 电力建设, 2020, 41(10): 1-8.
[6]	陈倩，齐林海，王红. 基于LSTM网络的谐波多标签分类[J]. 电力建设, 2019, 40(7): 101-106.
[7]	徐杉杉，朱俊澎，袁越，吴涵. 基于马尔可夫模型的光伏出力聚类与模拟[J]. 电力建设, 2019, 40(2): 54-62.
[8]	李睿雪，胡泽春，段小宇. 电动公交车充电站代理商的月度交易模型与日运行优化策略[J]. 电力建设, 2019, 40(1): 27-34.
[9]	张涌新，沈弘，马静. 综合能源系统负荷特性分析及应用研究[J]. 电力建设, 2018, 39(9): 18-29.
[10]	陈吕鹏，殷林飞，余涛，王克英. 基于深度森林算法的电力系统短期负荷预测[J]. 电力建设, 2018, 39(11): 42-50.
[11]	梁荣，杨波，马润泽，吴健，吴奎华，林振智，文福拴. 利用多源信息和深度置信神经网络的配电系统空间负荷预测[J]. 电力建设, 2018, 39(10): 12-19.
[12]	杨甲甲, 刘国龙, 赵俊华, 文福拴, 董朝阳. 采用长短期记忆深度学习模型的工业负荷短期预测方法[J]. 电力建设, 2018, 39(10): 20-27.
[13]	刘扬，严正，马洪艳，徐潇源. 风电接入下基于半监督谱聚类的备用动态分区方法[J]. 电力建设, 2018, 39(1): 90-.
[14]	李力，傅子昊，孙磊，杨银国，谭嫣，林振智，文福拴. 基于半监督谱聚类的黑启动分区策略[J]. 电力建设, 2017, 38(4): 9-.
[15]	牛东晓，马天男，王海潮，刘鸿飞，黄雅莉. 基于KPCA和NSGAII优化CNN参数的电动汽车充电站短期负荷预测[J]. 电力建设, 2017, 38(3): 85-.