基于混合神经网络的电力客户细分研究
|
摘要
随着我国电力市场改革的逐步深化,电力市场需求多元化的特性逐步凸显。制定有针对性的营销策略,满足不同客户的用电需求,实现个性化与差异化的服务,提高电网公司的核心竞争力,扩大电能在社会消费终端中的占有率,已成为电力企业一项迫切任务。为了制定电力客户的用电服务,电力客户细分就显得尤为重要。论文提出了H-LSTM(Hy?brid-Long Short-Term Memory)的混合神经网络方法来进行电力客户细分,通过用电客户的特征指标,时间之间的联系,进行对电力客户的分类。通过对电力大客户的供电营销数据进行实验,和决策树、原始LSTM神经网络等方法进行了比较,实验结果表明,H-LSTM的电力客户细分方法准确度更高,具有实际应用价值。
With the gradual deepening of China's power market reform,the characteristics of diversified demand in the powermarket have gradually emerged. Formulating targeted marketing strategies to meet different customers' electricity demand,achievingpersonalized and differentiated services,improving the core competitiveness of power grid companies,and increasing the share ofelectric energy in social consumer terminals have become a power company urgent task. In order to formulate electricity customers' power services,power customer segmentation is particularly important. In this paper,the neural network method of H-LSTM(Hybrid-Long Short-Term Memory)is proposed to subdivide the power customers,and the power customers are classified by the characteristics of customers and time. Through experiments on power supply marketing data of major power customers,comparisons aremade with decision trees and the original LSTM neural network. The experimental results show that the power customer segmentationmethod of H-LSTM is more accurate and has practical application value.
With the gradual deepening of China's power market reform,the characteristics of diversified demand in the powermarket have gradually emerged. Formulating targeted marketing strategies to meet different customers' electricity demand,achievingpersonalized and differentiated services,improving the core competitiveness of power grid companies,and increasing the share ofelectric energy in social consumer terminals have become a power company urgent task. In order to formulate electricity customers' power services,power customer segmentation is particularly important. In this paper,the neural network method of H-LSTM(Hybrid-Long Short-Term Memory)is proposed to subdivide the power customers,and the power customers are classified by the characteristics of customers and time. Through experiments on power supply marketing data of major power customers,comparisons aremade with decision trees and the original LSTM neural network. The experimental results show that the power customer segmentationmethod of H-LSTM is more accurate and has practical application value.
引文
[1]Tiefeng Zhang,Guangquan Zhang,et al. A new Indexand Classification Approach for Load Pattern Analysis ofLarge Electricity Customers[J]. IEEE Transactions onPower System,2012,27(1):153-160.
[2]周国鹏,孙志杰,傅军,等.基于模糊C均值聚类算法的电力行业负荷特性分析[J].电气应用,2016(16):86-90.ZHOU Guopeng,SUN Zhijie,FU Jun,et al. Analysis ofpower industry load characteristics based on fuzzyC-means clustering algorithm[J]. Electrical Applications,2016(16):86-90.
[3]胡晓雪,赵嵩正,吴楠.基于SOM-DB-PAM混合聚类算法的电力客户细分[J].计算机工程,2015,41(10):295-301.HU Xiaoxue,ZHAO Songzheng,WU Nan. Power Customer Segmentation Based on SOM-DB-PAM Hybrid Clustering Algorithm[J]. Computer Engineering, 2015, 41(10):295-301.
[4]胡晓雪,赵嵩正,吴楠.面向分类属性数据的一种改进相异性度量及其在客户细分中的应用[J].计算机应用研究,2016,33(2):400-403.HU Xiaoxue,ZHAO Songzheng,WU Nan. Improved dissimilarity metric for categorical attributes and its application in customer segmentatio[J]. Journal of Computer Applications,2016,33(2):400-403.
[5]黄星健,游威荣.基于数据挖掘的电力客户细分模型研究[J].自动化与仪器仪表,2017(11):146-147.HUANG Xingjian,YOU Weirong. Research on power customer segmentation model based on Data Mining[J]. Journal of Automation&Instrumentation, 2017(11):146-147.
[6]王璨,冯勤超.基于价值评价的电力用户分类研究[J].价值工程,2009,28(5):64-67.WANG Wei,FENG Qinchao. Research on Power UserClassification Based on Value Evaluation[J]. Value Engineering,2009,28(5):64-67.
[7]许青林,熊梦琪,刘扬帆.一种基于航空旅客行为的旅客细分模型[J].计算机工程与应用,2017,53(24):69-74.XU Qinglin,XIONG Mengqi,LIU Yangfan. A PassengerSubdivision Model Based on Air Passenger Behavior[J].Computer Engineering and Applications,2017,53(24):69-74.
[8]Dzobo O,Alvehag K,Gaunt C T,et al. Multi-dimensional customer segmentation model for power system reliability-worth analysis[J]. International Journal of ElectricalPower&Energy Systems,2014,62(11):532-539.
[9]Song L,Zhan W,Qian S,et al. Customer Segmentationfor Power Enterprise Based on Enhanced-FCM Algorithm[M]. Advanced Data Mining and Applications. SpringerBerlin Heidelberg,2012:127-137.
[10]徐尧强,方乐恒,赵冬华,等.基于LSTM神经网络的用电量预测[J].贵州电力技术,2017,20(8):25-29.XU Weiqiang,FANG Leheng,ZHAO Donghua,et al.Electricity Consumption Forecast Based on LSTM NeuralNetwork[J]. Guizhou Electric Power Technology,2017,20(8):25-29.
[11]任智慧,徐浩煜,封松林,等.基于LSTM网络的序列标注中文分词法[J].计算机应用研究,2017,34(5):1321-1324.REN Zhihui,XU Haozhen,FENG Songlin,et al. Chinese word segmentation based on sequence labelingbased on LSTM network[J]. Journal of Computer Applications,2017,34(5):1321-1324.
[12]Farzad A,Mashayekhi H,Hassanpour H. A comparativeperformance analysis of different activation functions inLSTM networks for classification[J]. Neural Computing&Applications,2017(5):1-15.
[13] Zhang B,Wu J L,Chang P C. A multiple time series-based recurrent neural network for short-term loadforecasting[J]. Soft Computing,2017(5):1-14.
[14]Hochreiter S,Schmidhuber J. Long short-term memory.[J]. Neural Computation,1997,9(8):1735-1780.
[15]Luo Y. Recurrent Neural Networks for Classifying Relations in Clinical Notes.[J]. Journal of Biomedical Informatics,2017:72.
[16]宗柳,李扬,王蓓蓓.计及需求响应的多维度用电特征精细挖掘[J].电力系统自动化,2012,36(20):54-58.ZONG Liu, LI Yang, WANG Wei. Fine Mining ofMulti-Dimensional Power Characteristics ConsideringDemand Response[J]. Automation of Electric Power Systems,2012,36(20):54-58.
[17]陆俊,朱炎平,彭文昊,等.智能用电用户行为分析特征优选策略[J].电力系统自动化,2017,41(5):58-63.LU Jun,ZHU Yanping,PENG Wenxi,et al. Optimization Strategy of Intelligent Behavior User Behavior Analysis Features[J]. Automation of Electric Power Systems,2017,41(5):58-63.
[18]郭崇,王征,纪建伟,等.电力用户数据中用电特征数据挖掘模型仿真[J].计算机仿真,2016,33(5):447-450.GUO Chong,WANG Zheng,JI Jianwei,et al. Simulation of Power Feature Data Mining Model in Power UserData[J]. Computer Simulation,2016,33(5):447-450.
[19]邓明斌,徐志淼,邓志飞,等.基于多特征融合的窃电识别算法研究[J].计算机与数字工程,2017,45(12):2398-2401.DENG Mingbin,XU Zhiwei,DENG Zhifei,et al. Research on Stealing Identification Algorithm Based onMulti-feature Fusion[J]. Computer and Digital Engineering,2017,45(12):2398-2401.
[20]张海燕,刘岩,马丽萌,等.决策树算法的比较与应用研究[J].华北电力技术,2017(6):42-47.ZHANG Haiyan,LIU Yan,MA Limeng,et al. Comparison and Application of Decision Tree Algorithms[J].North China Electric Power,2017(6):42-47.
[21]王文霞.数据挖掘中改进的C4.5决策树分类算法[J].吉林大学学报(理学版),2017,55(5):1274-1277.WANG Wenxia. Improved C4.5 Decision Tree Classification Algorithm in Data Mining[J]. Journal of Jilin University(Science),2017,55(5):1274-1277.
[22]张棪,曹健.面向大数据分析的决策树算法[J].计算机科学,2016,43(s1).ZHANG Yan,CAO Jian. Decision Tree Algorithm for BigData Analysis[J]. Computer Science,2016,43(s1).
[2]周国鹏,孙志杰,傅军,等.基于模糊C均值聚类算法的电力行业负荷特性分析[J].电气应用,2016(16):86-90.ZHOU Guopeng,SUN Zhijie,FU Jun,et al. Analysis ofpower industry load characteristics based on fuzzyC-means clustering algorithm[J]. Electrical Applications,2016(16):86-90.
[3]胡晓雪,赵嵩正,吴楠.基于SOM-DB-PAM混合聚类算法的电力客户细分[J].计算机工程,2015,41(10):295-301.HU Xiaoxue,ZHAO Songzheng,WU Nan. Power Customer Segmentation Based on SOM-DB-PAM Hybrid Clustering Algorithm[J]. Computer Engineering, 2015, 41(10):295-301.
[4]胡晓雪,赵嵩正,吴楠.面向分类属性数据的一种改进相异性度量及其在客户细分中的应用[J].计算机应用研究,2016,33(2):400-403.HU Xiaoxue,ZHAO Songzheng,WU Nan. Improved dissimilarity metric for categorical attributes and its application in customer segmentatio[J]. Journal of Computer Applications,2016,33(2):400-403.
[5]黄星健,游威荣.基于数据挖掘的电力客户细分模型研究[J].自动化与仪器仪表,2017(11):146-147.HUANG Xingjian,YOU Weirong. Research on power customer segmentation model based on Data Mining[J]. Journal of Automation&Instrumentation, 2017(11):146-147.
[6]王璨,冯勤超.基于价值评价的电力用户分类研究[J].价值工程,2009,28(5):64-67.WANG Wei,FENG Qinchao. Research on Power UserClassification Based on Value Evaluation[J]. Value Engineering,2009,28(5):64-67.
[7]许青林,熊梦琪,刘扬帆.一种基于航空旅客行为的旅客细分模型[J].计算机工程与应用,2017,53(24):69-74.XU Qinglin,XIONG Mengqi,LIU Yangfan. A PassengerSubdivision Model Based on Air Passenger Behavior[J].Computer Engineering and Applications,2017,53(24):69-74.
[8]Dzobo O,Alvehag K,Gaunt C T,et al. Multi-dimensional customer segmentation model for power system reliability-worth analysis[J]. International Journal of ElectricalPower&Energy Systems,2014,62(11):532-539.
[9]Song L,Zhan W,Qian S,et al. Customer Segmentationfor Power Enterprise Based on Enhanced-FCM Algorithm[M]. Advanced Data Mining and Applications. SpringerBerlin Heidelberg,2012:127-137.
[10]徐尧强,方乐恒,赵冬华,等.基于LSTM神经网络的用电量预测[J].贵州电力技术,2017,20(8):25-29.XU Weiqiang,FANG Leheng,ZHAO Donghua,et al.Electricity Consumption Forecast Based on LSTM NeuralNetwork[J]. Guizhou Electric Power Technology,2017,20(8):25-29.
[11]任智慧,徐浩煜,封松林,等.基于LSTM网络的序列标注中文分词法[J].计算机应用研究,2017,34(5):1321-1324.REN Zhihui,XU Haozhen,FENG Songlin,et al. Chinese word segmentation based on sequence labelingbased on LSTM network[J]. Journal of Computer Applications,2017,34(5):1321-1324.
[12]Farzad A,Mashayekhi H,Hassanpour H. A comparativeperformance analysis of different activation functions inLSTM networks for classification[J]. Neural Computing&Applications,2017(5):1-15.
[13] Zhang B,Wu J L,Chang P C. A multiple time series-based recurrent neural network for short-term loadforecasting[J]. Soft Computing,2017(5):1-14.
[14]Hochreiter S,Schmidhuber J. Long short-term memory.[J]. Neural Computation,1997,9(8):1735-1780.
[15]Luo Y. Recurrent Neural Networks for Classifying Relations in Clinical Notes.[J]. Journal of Biomedical Informatics,2017:72.
[16]宗柳,李扬,王蓓蓓.计及需求响应的多维度用电特征精细挖掘[J].电力系统自动化,2012,36(20):54-58.ZONG Liu, LI Yang, WANG Wei. Fine Mining ofMulti-Dimensional Power Characteristics ConsideringDemand Response[J]. Automation of Electric Power Systems,2012,36(20):54-58.
[17]陆俊,朱炎平,彭文昊,等.智能用电用户行为分析特征优选策略[J].电力系统自动化,2017,41(5):58-63.LU Jun,ZHU Yanping,PENG Wenxi,et al. Optimization Strategy of Intelligent Behavior User Behavior Analysis Features[J]. Automation of Electric Power Systems,2017,41(5):58-63.
[18]郭崇,王征,纪建伟,等.电力用户数据中用电特征数据挖掘模型仿真[J].计算机仿真,2016,33(5):447-450.GUO Chong,WANG Zheng,JI Jianwei,et al. Simulation of Power Feature Data Mining Model in Power UserData[J]. Computer Simulation,2016,33(5):447-450.
[19]邓明斌,徐志淼,邓志飞,等.基于多特征融合的窃电识别算法研究[J].计算机与数字工程,2017,45(12):2398-2401.DENG Mingbin,XU Zhiwei,DENG Zhifei,et al. Research on Stealing Identification Algorithm Based onMulti-feature Fusion[J]. Computer and Digital Engineering,2017,45(12):2398-2401.
[20]张海燕,刘岩,马丽萌,等.决策树算法的比较与应用研究[J].华北电力技术,2017(6):42-47.ZHANG Haiyan,LIU Yan,MA Limeng,et al. Comparison and Application of Decision Tree Algorithms[J].North China Electric Power,2017(6):42-47.
[21]王文霞.数据挖掘中改进的C4.5决策树分类算法[J].吉林大学学报(理学版),2017,55(5):1274-1277.WANG Wenxia. Improved C4.5 Decision Tree Classification Algorithm in Data Mining[J]. Journal of Jilin University(Science),2017,55(5):1274-1277.
[22]张棪,曹健.面向大数据分析的决策树算法[J].计算机科学,2016,43(s1).ZHANG Yan,CAO Jian. Decision Tree Algorithm for BigData Analysis[J]. Computer Science,2016,43(s1).
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |