改进最小二乘变点识别法在负荷分解的应用
|
摘要
随着全世界正进行的大规模智能电表的推广安装,使用非侵入式负荷监测分解方法,总电能消耗分解为单独设备的消耗,成为最近的研究热点;而变点识别是负荷分解方法中的第一步;精确的变点检测为后续提取特征以及识别负荷,打下了坚实的基础;提出了一种基于均值变点模型的识别算法,通过滑动窗口,利用最小二乘法计算目标函数,以确定变点个数;最后,提出假设检验,来验证变点检测的准确性;它能根据相关信号准确检测到负荷投切等引起的电气量变化、发生时刻等重要信息,并记录下来,然后为后续的负荷识别和分解提供保障;最后以某商业写字楼为例,通过测量该商业部分用电负荷数据,从而验证了该算法的可行性。
With the promotion and installation of large-scale smart meters in the world,the use of Non-Intrusive Load Monitoring(NILM)to decompose the total power consumption into the consumption of individual devices has become a new research hotspot.Changepoint recognition is the first step in the NILM method.Accurate change point detection lays a solid foundation for subsequent extraction of features and identification of loads.In this paper,a recognition algorithm based on mean change point model is proposed.By using the sliding window,the objective function is calculated by least squares method to determine the number of change points.Finally,a hypothesis test is proposed to verify the accuracy of the changepoint detection.It can accurately detect important changes such as electrical quantity changes and occurrence times caused by load switching according to relevant signals,and record them,and then ensure subsequent load identification and decomposition.At the end of this paper,a commercial office building is taken as an example to verify the feasibility of the algorithm by measuring the electrical load data of the commercial part.
With the promotion and installation of large-scale smart meters in the world,the use of Non-Intrusive Load Monitoring(NILM)to decompose the total power consumption into the consumption of individual devices has become a new research hotspot.Changepoint recognition is the first step in the NILM method.Accurate change point detection lays a solid foundation for subsequent extraction of features and identification of loads.In this paper,a recognition algorithm based on mean change point model is proposed.By using the sliding window,the objective function is calculated by least squares method to determine the number of change points.Finally,a hypothesis test is proposed to verify the accuracy of the changepoint detection.It can accurately detect important changes such as electrical quantity changes and occurrence times caused by load switching according to relevant signals,and record them,and then ensure subsequent load identification and decomposition.At the end of this paper,a commercial office building is taken as an example to verify the feasibility of the algorithm by measuring the electrical load data of the commercial part.
引文
[1]Hart G W.Non-intrusive appliance load monitoring[J].Proceedings of IEEE,1992:1870-1891.
[2]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(7):55-61.
[3]EPRI Commercial nonintrusive load monitoring system beta test result[J].电力系统保护与控制,1999(3):50-55.
[4]龙勇,苏振宇,盖晓平.成分分解方法预测月度电力负荷[J].电力系统及其自动化学报,2015,29(5):36-40.
[5]Steven S R,Leeb S B,et al.Nonintrusive load monitoring and diagnostics in power systems[J].IEEE Trans on Instrumentation and Measurement,2008:1445-1454.
[6]Shaw S R,Abler C B,Lepard R F,et al.Instrumentation for high performance nonintrusive electrical load monitoring[J].IEEE Journal of Solar Energy Engineering,1998:224-229.
[7]Zhu Z C,Wei Z Q,Yin B,et al.A novel approach for event detection in nonintrusive load monitoring[A].2017,IEEE CS[C].2017:54-61.
[8]Amirach N,Xerri B,Borloz B.A new approach for event detection and feature extraction for NILM[C].2014,IEEE.
[9]Hsuehy Y L,Liny N H,Chang C C.Abnormal event detection using Bayesian networks at a smart home[A].2015,IEEE UMEDIA[C].2015:273-277.
[10]Kolter J Z,Johnson M J.The reference energy disaggregation data set[EB/OL].http://redd.csail.mit.edu/,2011.
[11]Kolter J Z,Batra S,Ng A Y.Energy disaggregation via discriminative sparse coding[Z].Neural Information Processing Systems,2010.
[12]陈希孺.变点统计分析简介(Ⅱ)最小二乘法[J].数理统计与管理,1991(2):52-59.
[13]Pattem S.Unsupervised disaggregation for non-intrusive load monitoring[A]. 11th International Conference Machine Learning Applications(ICMLA)[C].2012(2):515-520.
[2]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(7):55-61.
[3]EPRI Commercial nonintrusive load monitoring system beta test result[J].电力系统保护与控制,1999(3):50-55.
[4]龙勇,苏振宇,盖晓平.成分分解方法预测月度电力负荷[J].电力系统及其自动化学报,2015,29(5):36-40.
[5]Steven S R,Leeb S B,et al.Nonintrusive load monitoring and diagnostics in power systems[J].IEEE Trans on Instrumentation and Measurement,2008:1445-1454.
[6]Shaw S R,Abler C B,Lepard R F,et al.Instrumentation for high performance nonintrusive electrical load monitoring[J].IEEE Journal of Solar Energy Engineering,1998:224-229.
[7]Zhu Z C,Wei Z Q,Yin B,et al.A novel approach for event detection in nonintrusive load monitoring[A].2017,IEEE CS[C].2017:54-61.
[8]Amirach N,Xerri B,Borloz B.A new approach for event detection and feature extraction for NILM[C].2014,IEEE.
[9]Hsuehy Y L,Liny N H,Chang C C.Abnormal event detection using Bayesian networks at a smart home[A].2015,IEEE UMEDIA[C].2015:273-277.
[10]Kolter J Z,Johnson M J.The reference energy disaggregation data set[EB/OL].http://redd.csail.mit.edu/,2011.
[11]Kolter J Z,Batra S,Ng A Y.Energy disaggregation via discriminative sparse coding[Z].Neural Information Processing Systems,2010.
[12]陈希孺.变点统计分析简介(Ⅱ)最小二乘法[J].数理统计与管理,1991(2):52-59.
[13]Pattem S.Unsupervised disaggregation for non-intrusive load monitoring[A]. 11th International Conference Machine Learning Applications(ICMLA)[C].2012(2):515-520.