融合外部注意力机制的序列到点非侵入式负荷分解

doi:10.16183/j.cnki.jsjtu.2022.534

摘要/Abstract

摘要：

非侵入式负荷分解可以深度挖掘用户电力消耗数据蕴含的信息价值,为电力设备故障监测、需求响应等决策分析提供重要参考.为有效解决非侵入式负荷分解算法训练时间成本与分解精度间的冲突,提出一种融合外部注意力机制的序列到点非侵入式负荷分解算法.首先,将总负荷功率消耗序列进行数据清理、标准化等预处理,以固定窗口长度构建训练输入数据,输入数据通过编码层自动提取设备特征;然后,设计外部注意力机制增强重要特征权值;最终,输入到解码层得到负荷分解结果.利用REDD与UK-DALE两种公开数据集进行模型仿真计算,在信号聚合误差、平均绝对误差、标准化分解误差指标、模型分解曲线、特征图和用户耗能等方面进行对比分析,本文模型克服了卷积层注意力分散的缺点,增强了对有效信息的提取与利用能力,在未增加训练时间成本的前提下具有更高的分解精度.

关键词: 非侵入式负荷分解, 外部注意力机制, 神经网络, 序列到点

Abstract:

Non-intrusive load disaggregation (NILD) can deeply explore the value of customer power consumption data, providing an important reference for decision analysis such as power equipment fault monitoring and demand response. Aimed at the conflict between the training time and the accuracy of non-intrusive load disaggregation, a non-intrusive load disaggregation algorithm using sequence-to-point integrating external attention (EA) mechanism is proposed. First, the original data is pre-processed by data purification, normalization, and some other operations, and the train data is built with a same length window. The equipment feature is extracted through the encoder layer. Then, the feature weights of important parts are enhanced by introducing an external attention mechanism. Finally, the results are yielded through the decoder layer. Simulation calculation of the proposed model and the current mainstream model is performed using the publicly available datasets, REDD and UK-DALE, while the indicators of signal aggregate error, mean absolute error, normalized disaggregation error, model disaggregation curves, feature map, and user energe consumption are compared and analyzed. The proposed model overcomes the shortcomings of attention scattering in the convolutional layer, enhances the ability to extract and utilize effective information, and has a more accurate decomposition accuracy without increasing the training time cost.

Key words: non-intrusive load disaggregation (NILD), external attention (EA) mechanism, neural networks, sequence-to-point (Seq2Point)

中图分类号:

TM721

李利娟, 刘海, 刘红良, 张青松, 陈永东. 融合外部注意力机制的序列到点非侵入式负荷分解[J]. 上海交通大学学报, 2024, 58(6): 846-854.

LI Lijuan, LIU Hai, LIU Hongliang, ZHANG Qingsong, CHEN Yongdong. Non-Intrusive Load Disaggregation Using Sequence-to-Point Integrating External Attention Mechanism[J]. Journal of Shanghai Jiao Tong University, 2024, 58(6): 846-854.

图/表 14

图1

图2

图3

表1

图4

表2

表3

图5

图6

图7

图8

图9

图10

表4

参考文献 29

[1]	刘博, 栾文鹏. 基于负荷分解的用电数据云架构方案及应用场景[J]. 电网技术, 2016, 40(3): 791-796.
	LIU Bo, LUAN Wenpeng. Conceptual cloud solution architecture and application scenarios of power consumption data based on load disaggregation[J]. Power System Technology, 2016, 40(3): 791-796.
[2]	郭红霞, 陆进威, 杨苹, 等. 非侵入式负荷监测关键技术问题研究综述[J]. 电力自动化设备, 2021, 41(1): 135-146.
	GUO Hongxia, LU Jinwei, YANG Ping, et al. Review on key techniques of non-intrusive load monito-ring[J]. Electric Power Automation Equipment, 2021, 41(1): 135-146.
[3]	HART G W. Nonintrusive appliance load monitoring[J]. Proceedings of the IEEE, 1992, 80(12): 1870-1891.
[4]	邓晓平, 张桂青, 魏庆来, 等. 非侵入式负荷监测综述[J]. 自动化学报, 2022, 48(3): 644-663.
	DENG Xiaoping, ZHANG Guiqing, WEI Qinglai, et al. A survey on the non-intrusive load monitoring[J]. Acta Automatica Sinica, 2022, 48(3): 644-663.
[5]	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.
[6]	WICHAKOOL W, REMSCRIM Z, ORJI U A, et al. Smart metering of variable power loads[J]. IEEE Transactions on Smart Grid, 2015, 6(1): 189-198.
[7]	PATEL S N, ROBERTSON T, KIENTZ J A, et al. At the flick of a switch: Detecting and classifying unique electrical events on the residential power line (nominated for the best paper award)[C]//International Conference on Ubiquitous Computing. Berlin, Germany: Springer, 2007: 271-288.
[8]	刘然. 结合改进最近邻法与支持向量机的住宅用电负荷识别研究[D]. 重庆: 重庆大学, 2014.
	LIU Ran. Research on household load identification combining improved nearest neighbour method and support vector machine[D]. Chongqing: Chongqing University, 2014.
[9]	FIGUEIREDO M, RIBEIRO B, DE ALMEIDA A. Electrical signal source separation via nonnegative tensor factorization using on site measurements in a smart home[J]. IEEE Transactions on Instrumentation & Measurement, 2014, 63(2): 364-373.
[10]	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.
[11]	CHOU P A, CHANG R I. Unsupervised adaptive non-intrusive load monitoring system[C]//2013 IEEE International Conference on Systems, Man, and Cybernetics. Manchester, UK: IEEE, 2013: 3180-3185.
[12]	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 & Control, 2016, 44(7): 55-61.
[13]	BATRA N, KELLY J, PARSON O, et al. NILMTK: An open source toolkit for non-intrusive load monitoring[C]//Proceedings of the 5th international conference on Future energy systems. Cambridge, UK: ACM, 2014: 265-276.
[14]	FIGUEIREDO M, DE ALMEIDA A, RIBEIRO B. Home electrical signal disaggregation for non-intrusive load monitoring (NILM) systems[J]. Neurocomputing, 2012, 96: 66-73.
[15]	涂京, 周明, 宋旭帆, 等. 基于监督学习的非侵入式负荷监测算法比较[J]. 电力自动化设备, 2018, 38(12): 128-134.
	TU Jing, ZHOU Ming, SONG Xufan, et al. Comparison of supervised learning-based non-intrusive load monitoring algorithms[J]. Electric Power Automation Equipment, 2018, 38(12): 128-134.
[16]	KARMAKAR P, TENG S W, LU G J. Thank you for attention: A survey on attention-based artificial neural networks for automatic speech recognition[DB/OL]. (2021-02-14)[2022-05-01]. https://arxiv.org/abs/2102.07259.
[17]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach, USA: ACM, 2017: 6000-6010.
[18]	MIKOLOV T, CHEN K, CORRADO G, et al. Efficient estimation of word representations in vector space[DB/OL]. (2013-01-16)[2022-05-01]. https://arxiv.org/abs/1301.3781.
[19]	KELLY J, KNOTTENBELT W. Neural NILM: Deep neural networks applied to energy disaggregation[C]//Proceedings of the 2nd ACM International Conference on Embedded Systems for Energy-Efficient Built Environments. Seoul, South Korea: ACM, 2015: 55-64.
[20]	ZHANG C Y, ZHONG M J, WANG Z Z, et al. Sequence-to-point learning with neural networks for non-intrusive load monitoring[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Palo Alto, California, USA: AAAI, 2018: 2604-2611.
[21]	PAN Y G, LIU K, SHEN Z Y, et al. Sequence-to-subsequence learning with conditional Gan for power disaggregation[C]//ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing. Barcelona, Spain:IEEE, 2020: 3202-3206.
[22]	YANG M Z, LIU Y E, LIU Q L. Nonintrusive residential electricity load decomposition based on transfer learning[J]. Sustainability, 2021, 13(12): 6546.
[23]	KAZEMI S M, GOEL R, EGHBALI S, et al. Time2Vec: Learning a vector representation of time[DB/OL]. (2019-07-11)[2022-05-01]. https://arxiv.org/abs/1907.05321.
[24]	徐晓会, 赵书涛, 崔克彬. 基于卷积块注意力模型的非侵入式负荷分解算法[J]. 电网技术, 2021, 45(9): 3700-3706.
	XU Xiaohui, ZHAO Shutao, CUI Kebin. Non-intrusive load disaggregate algorithm based on convolutional block attention module[J]. Power System Technology, 2021, 45(9): 3700-3706.
[25]	KOLTER J Z, JOHNSON M J. REDD: A public data set for energy disaggregation research[EB/OL]. (2011-01-01)[2022-05-01]. https://people.csail.mit.edu/mattjj/papers/kddsust2011.pdf.
[26]	KELLY J, KNOTTENBELT W. The UK-DALE dataset, domestic appliance-level electricity demand and whole-house demand from five UK homes[J]. Scientific Data, 2015, 2: 150007.
[27]	GUO M H, LIU Z N, MU T J, et al. Beyond self-attention: External attention using two linear layers for visual tasks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2023, 45(5): 5436-5447.
[28]	GUO M H, CAI J X, LIU Z N, et al. PCT: Point cloud transformer[J]. Computational Visual Media, 2021, 7(2): 187-199.
[29]	PRECHELT L. Early stopping-but when?[M]//Lecture Notes in Computer Science. Berlin, Germany: Springer Berlin Heidelberg, 1998: 55-69.

设备	平均值/W	标准差/W
主电源	522	814
微波炉	500	800
冰箱	200	400
洗碗机	700	1000
洗衣机	400	700

设备	E_SAE				E_MAE				E_NDE
设备	CNN	T2V	CBAM	本文	CNN	T2V	CBAM	本文	CNN	T2V	CBAM	本文
微波炉	0.889	0.485	0.373	0.323	19.681	20.066	20.712	22.481	0.996	0.751	0.689	0.722
冰箱	0.219	0.209	0.159	0.092	35.69	31.076	37.603	28.494	0.338	0.279	0.345	0.231
洗碗机	0.478	0.643	0.395	0.358	19.142	29.610	19.476	18.365	0.505	0.872	0.437	0.431
洗衣机	0.053	0.095	0.089	0.045	15.106	12.427	19.065	18.622	0.199	0.139	0.172	0.174
均值	0.410	0.358	0.254	0.205	23.450	23.295	23.421	21.991	0.510	0.510	0.411	0.390

设备	E_SAE				E_MAE				E_NDE
设备	CNN	T2V	CBAM	本文	CNN	T2V	CBAM	本文	CNN	T2V	CBAM	本文
微波炉	0.100	0.877	0.719	0.028	5.732	10.318	7.447	6.042	0.613	0.984	0.606	0.594
冰箱	0.143	0.178	0.072	0.057	23.342	31.686	22.823	20.549	0.410	0.526	0.403	0.384
洗碗机	0.244	0.737	0.274	0.153	7.334	14.953	7.227	5.290	0.060	0.457	0.074	0.074
洗衣机	0.004	0.223	0.004	0.005	8.927	13.923	8.081	7.960	0.089	0.274	0.072	0.070
均值	0.122	0.504	0.267	0.061	11.333	17.720	11.394	9.960	0.293	0.560	0.288	0.280

模型	训练参数量	训练时间/s
CNN	30,708,249	123
CBAM	30,709,748	351
本文	29,229,199	120