基于Fisher主元分析和核极限学习机的非侵入式电力负荷辨识模型

doi:10.12204/j.issn.1000-7229.2021.02.011

电力建设 ›› 2021, Vol. 42 ›› Issue (2): 85-92.doi: 10.12204/j.issn.1000-7229.2021.02.011

基于Fisher主元分析和核极限学习机的非侵入式电力负荷辨识模型

仝瑞宁¹, 李鹏¹, 郎恂¹, 沈鑫², 曹敏²

1.云南大学信息学院,昆明市 650504
2.云南电网有限责任公司电力科学研究院,昆明市 650217

收稿日期:2020-04-29 出版日期:2021-02-01 发布日期:2021-02-09
通讯作者: 李鹏
作者简介:仝瑞宁(1997),男,硕士研究生,主要研究方向为电力大数据建模、电力物联网与能源互联网、智能配用电与新能源发电功率预测技术;|郎恂(1994),男,博士,博士后,主要研究方向为工业大数据分析与建模;|沈鑫(1981),男,博士,教授级高级工程师,主要研究方向为智能配电网、计量自动化;|曹敏(1961),男,学士,教授级高级工程师,主要研究方向为智能电网、电力物联网。
基金资助:
国家自然科学基金项目(61763049);云南省应用基础研究重点课题项目(2018FA032)

Non-intrusive Power Load Identification Model Based on FPCA and KELM

TONG Ruining¹, LI Peng¹, LANG Xun¹, SHEN Xin², CAO Min²

1. School of Information, Yunnan University, Kunming 650504, China
2. Electric Power Research Institute, Yunnan Power Grid Co., Ltd., Kunming 650217, China

Received:2020-04-29 Online:2021-02-01 Published:2021-02-09
Contact: LI Peng
Supported by:
National Natural Science Foundation of China(61763049);the Key Project of Applied Basic Research in Yunnan Province(2018FA032)

摘要/Abstract

摘要：

非侵入式电力负荷监测与辨识是实现泛在电力物联网客户侧智能感知的关键技术。针对现有辨识模型存在的特征冗余度高、辨识准确率差、计算效率低等问题,提出了一种基于Fisher主元分析(Fisher principal component analysis,FPCA)和核极限学习机(kernel extreme learning machine,KELM)的非侵入式电力负荷辨识模型。首先,选取电流、功率、谐波含有率等稳态特征作为原始输入变量,运用Fisher得分和主成分分析相融合的Fisher主元分析法剔除可分性较差的无效特征,同时降低有效特征之间的相关性;然后,引入径向基核函数搭建网络结构,并采用遗传算法(genetic algorithm,GA)对惩罚系数等模型参数进行寻优,从而建立核极限学习机分类模型进行负荷识别;最后,通过公开的TIPDM负荷数据集进行算例分析。仿真结果表明,所提模型相比于传统负荷辨识模型具有更好的辨识准确率和计算效率,运用该模型可对常见家用负荷进行有效识别。

关键词: 非侵入式负荷辨识, Fisher得分, 主成分分析, 遗传算法(GA), 核极限学习机(KELM)

Abstract:

Non-intrusive power load monitoring and identification is the key technology to realize customer-side intelligent sensing in the ubiquitous power Internet of things. Aiming at the problems of high feature redundancy, low identification accuracy and low calculation efficiency in the existing identification model, a non-intrusive power load identification model based on Fisher principal component analysis and kernel extreme learning machine is proposed. Firstly, the steady-state characteristics such as current, power and harmonic contents are selected as the original input variables, and the Fisher principal component analysis (FPCA), which combines Fisher score and principal component analysis, is used to eliminate the invalid features with poor separability and to eliminate the correlation among the effective features at the same time. Then, radial basis function is introduced to build the network structure, and genetic algorithm (GA) is used to optimize the model parameters such as penalty coefficient, so as to build the kernel extreme learning machine(KELM) classification model for load identification. Finally, the open TIPDM load data set is used for example analysis. The simulation results show that the proposed model has better identification accuracy and calculation efficiency than the traditional load identification models, and it can effectively identify common household loads.

Key words: non-intrusive load identification, Fisher score, principal component analysis, genetic algorithm(GA), kernel extreme learning machine(KELM)

中图分类号:

TM73

仝瑞宁, 李鹏, 郎恂, 沈鑫, 曹敏. 基于Fisher主元分析和核极限学习机的非侵入式电力负荷辨识模型[J]. 电力建设, 2021, 42(2): 85-92.

TONG Ruining, LI Peng, LANG Xun, SHEN Xin, CAO Min. Non-intrusive Power Load Identification Model Based on FPCA and KELM[J]. ELECTRIC POWER CONSTRUCTION, 2021, 42(2): 85-92.

图/表 9

图1 部分电器前25次的电流谐波含有率

Fig.1 Contents of the 2nd to 25th harmonic currents of some electrical appliances

图2 基于Fisher主元分析和KELM的非侵入式电力负荷辨识建模流程

Fig.2 Flow chart of non-intrusive power load identification based on FPCA and KELM

表1 遗传算法优化参数

Table 1 Parameters of GA

图3 基于FPCA-GA-KELM的网络结构

Fig.3 Network structure of FPCA-GA-KELM

表2 17种稳态特征的Fisher得分

Table 2 Fisher scores of 17 steady-state characteristics

表3 8个变量的协方差矩阵特征值

Table 3 Eigenvalues of the covariance matrix of 8 variables

表4 不同特征下测试样本分类准确率

Table 4 Classification accuracy of the test samples under different characteristics

表5 4种不同的辨识模型测试分类准确率

Table 5 Test classification accuracy of four different identification models

图4 模型计算消耗时间随样本个数增加变化曲线

Fig.4 Time-consuming of the model calculation increasing with the number of samples

参考文献 21

[1]	CORBETT J, WARDLE K, CHEN C. Toward a sustainable modern electricity grid:The effects of smart metering and program investments on demand-side management performance in the US electricity sector 2009-2012[J]. IEEE Transactions on Engineering Management, 2018,65(2):252-263.
[2]	王继业, 季知祥, 史梦洁, 等. 智能配用电大数据需求分析与应用研究[J]. 中国电机工程学报, 2015,35(8):1829-1836.
	WANG Jiye, JI Zhixiang, SHI Mengjie, et al. Scenario analysis and application research on big data in smart power distribution and consumption systems[J]. Proceedings of the CSEE, 2015,35(8):1829-1836.
[3]	王守相, 葛磊蛟, 王凯. 智能配电系统的内涵及其关键技术[J]. 电力自动化设备, 2016,36(6):1-6.
	WANG Shouxiang, GE Leijiao, WANG Kai. Main contents and key technologies of smart distribution system[J]. Electric Power Automation Equipment, 2016,36(6):1-6.
[4]	江帆, 杨洪耕. 基于选择性贝叶斯分类的非侵入式负荷识别方法[J]. 电力建设, 2019,40(2):94-99.
	JIANG Fan, YANG Honggeng. Non-intrusive load identification method based on selective Bayesian classification[J]. Electric Power Construction, 2019,40(2):94-99.
[5]	GUZEL T, USTUNEL E. Principal components null space analysis based non-intrusive load monitoring[C]∥ IEEE Electrical Power and Energy Conference. Ontario: IEEE, 2015: 420-423.
[6]	LIU B, LUAN W P, YU Y X. Dynamic time warping based non-intrusive load transient identification[J]. Applied Energy, 2017,195:634-645.
[7]	何恒靖, 王灏, 肖勇, 等. 基于Bi-LSTM算法的非侵入式负荷监测模型[J]. 南方电网技术, 2019,13(2):20-26.
	HE Hengjing, WANG Hao, XIAO Yong, et al. Non-intrusive load monitoring model based on Bi-LSTM algorithm[J]. Southern Power System Technology, 2019,13(2):20-26.
[8]	祁兵, 程媛, 武昕. 基于Fisher 有监督判别的非侵入式居民负荷辨识方法[J]. 电网技术, 2016,40(8):2484-2491.
	QI Bing, CHENG Yuan, WU Xin. Non-intrusive household appliance load identification method based on Fisher supervised discriminant[J]. Power System Technology, 2016,40(8):2484-2491.
[9]	宋旭帆, 周明, 涂京, 等. 基于k-NN结合核Fisher判别的非侵入式负荷监测方法[J]. 电力系统自动化, 2018,42(6):73-80.
	SONG Xufan, ZHOU Ming, TU Jing, et al. Non-intrusive load monitoring method based on k-NN and kernel Fisher discriminant[J]. Automation of Electric Power Systems, 2018,42(6):73-80.
[10]	DUFOURL, GENOUD D, JARA A, et al. A non-intrusive model to predict the exible energy in a residential building[C]∥ Wireless Communications and Networking Conference Workshops(WCNCW).IEEE, 2015: 69-74.
[11]	李如意, 王晓换, 胡美璇, 等. RPROP神经网络在非侵入式负荷分解中的应用[J]. 电力系统保护与控制, 2016,44(7):55-61.
	LI Ruyi, WANG Xiaohuan, HU Meixuan, et al. Application of RPROP neural network in nonintrusive load decomposition[J]. Power System Protection and Control, 2016,44(7):55-61.
[12]	江帆, 杨洪耕. 基于广义回归神经网络的非侵入式负荷识别方法[J]. 电测与仪表, 2020,57(3):1-6, 18.
	JIANG Fan, YANG Honggeng. Non-intrusive load identification method based on generalized regression neural network[J]. Electrical Measurement & Instrumentation, 2020,57(3):1-6, 18.
[13]	程祥, 李林芝, 吴浩, 等. 非侵入式负荷监测与分解研究综述[J]. 电网技术, 2016,40(10):3108-3117.
	CHENG Xiang, LI Linzhi, WU Hao, et al. A survey of the research on non-intrusive load monitoring and disaggregation[J]. Power System Technology, 2016,40(10):3108-3117.
[14]	LIANG J, NG S K, KENDALL G, et al. Load signature study,part I:Basic concept, structure, and methodology[J]. IEEE Transactions on Power Delivery, 2010,25(2):551-560.
[15]	李郅琴, 杜建强, 聂斌, 等. 特征选择方法综述[J]. 计算机工程与应用, 2019,55(24):10-19.
	LI Zhiqin, DU Jianqiang, NIE Bin, et al. Summary of feature selection methods[J]. Computer Engineering and Application, 2019,55(24):10-19.
[16]	贺川双, 杜修明, 严英杰, 等. 基于数据挖掘和主成分分析的电力设备状态评价[J]. 高压电器, 2017,53(12):34-41.
	HE Chuanshuang, DU Xiuming, YAN Yingjie, et al. State evaluation of power equipment based on data mining and principal component analysis[J]. High Voltage Apparatus, 2017,53(12):34-41.
[17]	孙朝江, 汤健, 魏忠军, 等. 基于选择性更新的在线核极限学习机建模[J]. 控制工程, 2013,20(4):659-662.
	SUN Chaojiang, TANG Jian, WEI Zhongjun, et al. Modeling of online nuclear limit learning machine based on selective update[J]. Control Engineering, 2013,20(4):659-662.
[18]	周明, 宋旭帆, 涂京, 等. 基于非侵入式负荷监测的居民用电行为分析[J]. 电网技术, 2018,42(10):3268-3276.
	ZHOU Ming, SONG Xufan, TU Jing, et al. Residential electricity consumption behavior analysis based on non-intrusive load monitoring[J]. Power System Technology, 2018,42(10):3268-3276.
[19]	徐青山, 娄藕蝶, 郑爱霞, 等. 基于近邻传播聚类和遗传优化的非侵入式负荷分解方法[J]. 电工技术学报, 2018,33(16):3868-3878.
	XU Qingshan, LOU Oudie, ZHENG Aixia, et al. Non-intrusive load decomposition method based on near neighbor propagation clustering and genetic optimization[J]. Transactions of China Electrotechnical Society, 2018,33(16):3868-3878.
[20]	DEB K, PRATAP A, AGARWAL S, et al. A fast and elitist multi-objective genetic algorithm: NSGA-II[J]. IEEE Transactions on Evolutionary Computation, 2002,6(2):182-197.
[21]	Dreis. Topic a of the sixth Teddy cup data mining challenge: Power data mining based on non-invasive load monitoring and decomposition [EB/OL]. [2018-11-01]. http://www.tipdm.org/bdrace/index.html.

基于Fisher主元分析和核极限学习机的非侵入式电力负荷辨识模型

Non-intrusive Power Load Identification Model Based on FPCA and KELM

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 21

相关文章 11

编辑推荐

Metrics

本文评价

[1]	刘赛, 刘宇, 高山, 郭昊旻, 宋天成, 蒋伟毅, 李正, 王俊成, 徐逸民. 基于多特征目标函数PCA-ILP的非侵入式负荷分解方法[J]. 电力建设, 2020, 41(8): 1-8.
[2]	艾欣，杨子豪，胡寰宇，王智冬，彭冬，赵朗. 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.
[3]	周宏宇，薛友，刘祚宇，沈洁，俞敏，文福拴. 特高压变电站设备运维检修费用影响因素度量与分析[J]. 电力建设, 2018, 39(1): 19-.
[4]	杜蜻蜓,刘继春,王冬,杨常. 计入柔性负荷的售电公司参与市场竞价两阶段优化模型 [J]. 电力建设, 2017, 38(6): 142-.
[5]	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-.
[6]	和敬涵，卢育梓，陆金耀，胡波，杨方，何博. 基于“进化”主成分分析法的用户分类及其应用[J]. 电力建设, 2017, 38(3): 101-.
[7]	孟金岭，胡嘉骅，文福拴，林国营，党三磊. 利用电参数的工业用户能效模糊综合评估[J]. 电力建设, 2016, 37(11): 16-.
[8]	崔巍，都秀文，杨海峰. 基于主成分分析法的电力需求影响因素分析[J]. 电力建设, 2013, 34(8): 34-39.
[9]	郑燕赵彪. 优化BP神经网络在垃圾发电厂选址中的应用[J]. 电力建设, 2011, 32(6): 67-69.
[10]	刘吉成，王素花. 基于PCA和DP的电力施工企业动态联盟[J]. 电力建设, 2011, 32(3): 113-116.
[11]	刘克文张贲王宇飞. 基于复合分类器的网络入侵检测模型[J]. 电力建设, 2011, 32(11): 40-44.