基于模型自适应选择融合的智能电表故障多分类方法
|
摘要
智能电表故障的准确分类能大幅提高用电采集系统运维能力。融合多个分类模型的机器学习算法是解决该问题的有效手段,但现有方法无法解决输出分别为样本所属各类别概率值和类别标签的两个基分类模型融合问题。提出一种基于模型自适应选择融合的智能电表故障多分类方法。首先,分别取各基分类模型对各类样本分类准确率最大值,将其与阈值系数的乘积作为该类样本准确率阈值,实现阈值自适应调整;然后对各类样本分别计算基分类模型的准确率差值,与阈值进行比较设置样本融合标记;最后根据该标记选择参与融合的基分类模型,结合输出为概率值的基分类模型的Top-N分类标签集,得到模型融合结果。在10组KEEL公共数据集上验证了所提融合方法的有效性,且融合后准确率较基分类模型均有稳定提升,最大提升4.62%;以近年采集的智能电表故障数据为基础,对比实验表明,所提算法能够明显提高故障分类准确率。
Accurate classification of fault types on smart meters can greatly improve the operation and maintenance ability of power acquisition system. The machine learning algorithm integrating multiple classification models is an effective means to solve this problem. However, existing methods cannot solve fusion problem when the outputs of two basic classification models are probability values and labels respectively. Therefore, a multi-classification method for smart meter fault types based on model adaptive selection fusion is proposed in this paper. Firstly, the maximum classification accuracy of each type of sample is obtained from the base models and the product of the accuracy and threshold coefficient is chosen as the accuracy threshold of the sample to achieve adaptive threshold adjustment. Then, the accuracy difference between the base classification models is calculated for each sample. Comparing with the threshold, the fusion marks of each sample are set. Finally, combining the base classification of the fusion selected with the fusion marks and the output Top-N classification label set as the probability value base classification model, the output result of the fusion model is obtained. The experiment on 10 groups of KEEL public datasets proves effectiveness of the proposed fusion method with the result that maximum elevation accuracy of the syncretic model increases 4.62% compared with base classification models. Furthermore, on the basis of newly-updated fault data collected on smart meters, contrast experiment shows that the proposed fusion algorithm significantly improves the fault classification accuracy of smart meters.
Accurate classification of fault types on smart meters can greatly improve the operation and maintenance ability of power acquisition system. The machine learning algorithm integrating multiple classification models is an effective means to solve this problem. However, existing methods cannot solve fusion problem when the outputs of two basic classification models are probability values and labels respectively. Therefore, a multi-classification method for smart meter fault types based on model adaptive selection fusion is proposed in this paper. Firstly, the maximum classification accuracy of each type of sample is obtained from the base models and the product of the accuracy and threshold coefficient is chosen as the accuracy threshold of the sample to achieve adaptive threshold adjustment. Then, the accuracy difference between the base classification models is calculated for each sample. Comparing with the threshold, the fusion marks of each sample are set. Finally, combining the base classification of the fusion selected with the fusion marks and the output Top-N classification label set as the probability value base classification model, the output result of the fusion model is obtained. The experiment on 10 groups of KEEL public datasets proves effectiveness of the proposed fusion method with the result that maximum elevation accuracy of the syncretic model increases 4.62% compared with base classification models. Furthermore, on the basis of newly-updated fault data collected on smart meters, contrast experiment shows that the proposed fusion algorithm significantly improves the fault classification accuracy of smart meters.
引文
[1]王鹏,林佳颖,郭屾,等.配用电数据分析及应用[J].电网技术,2017,41(10):3333-3340.Wang Peng,Lin Jiaying,Guo Shen,et al.Distribution system data analytics and applications[J].Power System Technology,2017,41(10):3333-3340(in Chinese).
[2]彭小圣,邓迪元,程时杰,等.面向智能电网应用的电力大数据关键技术[J].中国电机工程学报,2015,35(3):503-511.Peng Xiaosheng,Deng Diyuan,Cheng Yuanjie,et al.Key technologies of electric power big data and its application prospects in smart grid[J].Proceedings of the CSEE,2015,35(3):503-511(in Chinese).
[3]Jamali S,Bahmanyar A,Bompard E.Fault location method for distribution networks using smart meters[J].Measurement,2017(102):150-157.
[4]冷喜武,陈国平,白静洁,等.智能电网监控运行大数据分析系统总体设计[J].电力系统自动化,2018,42(12):160-166.Leng Xiwu,Chen Guoping,Bai Jingjie,et al.General design of smart grid monitoring operation big data analysis system[J].Automation of Electric Power Systems,2018,42(12):160-166(in Chinese).
[5]杨金成,蒋平,陈广宇,等.智能电表故障辨识模型研究[J].电器与能效管理技术,2017(6):30-36.Yang Jincheng,Jiang Ping,Chen Guangyu,et al.Study on smart meters fault identification model[J].Electrical and Energy Efficiency Management Technology,2017(6):30-36(in Chinese).
[6]申曼丽,贺星,陈浩,等.基于数据分析的电能表故障分析系统[C]//中国电机工程学会年会,南京,2016:1-4.
[7]Yao Haowei,Wang XiaoWei,Wu Lusen,et al.Prediction method for smart meter life based on big data[J].Procedia Engineering,2018(211):1111-1114.
[8]仇红剑,李宝树,林华德.基于电力大数据挖掘的行业用电市场景气指数分析[J].电力信息与通信技术,2018,16(11):24-28.Qiu Hongjian,Li Baoshu,Lin Huade.Analysis of industry electricity market prosperity index using electric power big data mining[J].Electric Power Information and Communication Technology,2018,16(11):24-28(in Chinese).
[9]Zhang Chongsheng,Liu Changchang,Zhang Xiangliang,et al.An up-to-date comparison of state-of-the-art classification algorithms[J].Expert Systems with Applications,2017,82(C):128-150.
[10]Jiang Liangxiao,Zhang Lungan,Yu Liangjun,et al.Class-specific attribute weighted naive Bayes[J].Pattern Recognition,2019,88(C):321-330.
[11]MarkHall.A decision tree-based attribute weighting filter for naive Bayes[J].Knowledge-Based Systems,2007,20(2):120-126.
[12]Rosales-Perez A,Garcia S,Terashima-Marin H,et al.MC2ESVM:multiclass classification based on cooperative evolution of support vector machines[J].IEEE Computational Intelligence Magazine,2018,13(2):18-29.
[13]Yan He,Ye Qiaolin,Liu Ying’an,et al.The GEPSVM classifier based on L1-norm distance metric[C]//Chinese Conference on Pattern Recognition,Chengdu,China,2016:703-719.
[14]Yang Zhixia,Shao Yuanhai,Zhang Xiangsun.Multiple birth support vector machine for multi-class classification[J].Neural Computing and Applications,2013,22(1):153-161.
[15]Zhang Xiekai,Ding Shifei,Xue Yu.An improved multiple birth support machine for pattern classification[J].Neurocomputing,2016(225):119-128.
[16]Mcculloch W S,Pitts W H.A logical calculus of ideas immanent in nervous activity[J].The Bulletin of Mathematical Biophysics,1942(5):115-133.
[17]XuXinshun,Tang Zheng,Wang Jiahai.A method to improve the transiently chaotic neural network[J].Neurocomputing,2004,67(1):456-463.
[18]Sun Jianye.Local coupled feedforward neural network[J].Neural Network,2010,23(1):108-113.
[19]Ozyildirim B M,Avci M.Generalized classifier neural network[J].Neural Networks,2013(39):18-26..
[20]Wang Shitong,Chung FuLai,Wang Jun,et al.A fast learning method for feedforward neural networks[J].Neurocomputing,2015(149):295-307.
[21]López-Soto,Diana,Angel-Bello F,et al.A multi-start algorithm to design a multi-class classifier for a multi-criteria ABC inventory classification problem[J].Expert Systems with Applications,2017(81):12-21.
[22]Roy S,Mondal S,Ekbal A,et al.Dispersion ratio based decision tree model for classification[J].Expert Systems with Applications,2019,116(C):1-9.
[23]Quinlan J R.Induction of decision trees[J].Machine Learning,1986,1(1):81-106.
[24]Quinlan J R.C4.5:programs for machine learning[M].Morgan Kaufmann Publishers Inc,1993.
[25]Loh W Y.Classification and regression trees[J].Wiley Interdisciplinary Reviews Data Mining&Knowledge Discovery,2011,1(1):14-23.
[26]Rooney N,Hui Wang,Taylor P S.An investigation into the application of ensemble learning for entailment classification[J].Information Processing and Management,2014(1):87-103.
[27]Chaudhary A,Kolhe S,Kamal R.An improve random forest classifier for multi-class classification[J].Information Processing in Agriculture,2016,3(4):215-222.
[28]Chen T,Guestrin C.XGBoost:A scalable tree boosting System[C]//ACM SIGKDD International Conference on Knowledge Discovery and Data Mining,San Francisco,CA,USA,2016:785-794.
[29]Genuer R,Poggi J M,Tuleau-Malot C,et al.Random forests for big data[J].Big Data Research,2017(9):28-46.
[30]Ke Guolin,Meng Qi,Finley T,et al.LightGBM:a highly efficient gradient boosting decision tree[C]//Advances in Neural Information Processing Systems,San Mateo,CA,USA,2017:3148-3156.
[31]Xu Yong,Lu Yuwu.Adaptive weighted fusion:A novel fusion approach for image classification[J].Neurocomputing,2015(168):566-574.
[32]Mazher A,Li P,Moughal T A,et al.A decision fusion method using an algorithm for fusion of correlated probabilities[J].International Journal of Remote Sensing,2016,37(1):14-25.
[33]Alcalá-Fdez J,Fernandez A,Luengo J,et al.KEEL data-mining software tool:data set repository,integration of algorithms and experimental analysis framework[J].Journal of Multiple-Valued Logic and Soft Computing,2011,17(2-3):255-287.KEEL数据集[DB/OL].[2018-05-06].http://www.keel.es/.
[2]彭小圣,邓迪元,程时杰,等.面向智能电网应用的电力大数据关键技术[J].中国电机工程学报,2015,35(3):503-511.Peng Xiaosheng,Deng Diyuan,Cheng Yuanjie,et al.Key technologies of electric power big data and its application prospects in smart grid[J].Proceedings of the CSEE,2015,35(3):503-511(in Chinese).
[3]Jamali S,Bahmanyar A,Bompard E.Fault location method for distribution networks using smart meters[J].Measurement,2017(102):150-157.
[4]冷喜武,陈国平,白静洁,等.智能电网监控运行大数据分析系统总体设计[J].电力系统自动化,2018,42(12):160-166.Leng Xiwu,Chen Guoping,Bai Jingjie,et al.General design of smart grid monitoring operation big data analysis system[J].Automation of Electric Power Systems,2018,42(12):160-166(in Chinese).
[5]杨金成,蒋平,陈广宇,等.智能电表故障辨识模型研究[J].电器与能效管理技术,2017(6):30-36.Yang Jincheng,Jiang Ping,Chen Guangyu,et al.Study on smart meters fault identification model[J].Electrical and Energy Efficiency Management Technology,2017(6):30-36(in Chinese).
[6]申曼丽,贺星,陈浩,等.基于数据分析的电能表故障分析系统[C]//中国电机工程学会年会,南京,2016:1-4.
[7]Yao Haowei,Wang XiaoWei,Wu Lusen,et al.Prediction method for smart meter life based on big data[J].Procedia Engineering,2018(211):1111-1114.
[8]仇红剑,李宝树,林华德.基于电力大数据挖掘的行业用电市场景气指数分析[J].电力信息与通信技术,2018,16(11):24-28.Qiu Hongjian,Li Baoshu,Lin Huade.Analysis of industry electricity market prosperity index using electric power big data mining[J].Electric Power Information and Communication Technology,2018,16(11):24-28(in Chinese).
[9]Zhang Chongsheng,Liu Changchang,Zhang Xiangliang,et al.An up-to-date comparison of state-of-the-art classification algorithms[J].Expert Systems with Applications,2017,82(C):128-150.
[10]Jiang Liangxiao,Zhang Lungan,Yu Liangjun,et al.Class-specific attribute weighted naive Bayes[J].Pattern Recognition,2019,88(C):321-330.
[11]MarkHall.A decision tree-based attribute weighting filter for naive Bayes[J].Knowledge-Based Systems,2007,20(2):120-126.
[12]Rosales-Perez A,Garcia S,Terashima-Marin H,et al.MC2ESVM:multiclass classification based on cooperative evolution of support vector machines[J].IEEE Computational Intelligence Magazine,2018,13(2):18-29.
[13]Yan He,Ye Qiaolin,Liu Ying’an,et al.The GEPSVM classifier based on L1-norm distance metric[C]//Chinese Conference on Pattern Recognition,Chengdu,China,2016:703-719.
[14]Yang Zhixia,Shao Yuanhai,Zhang Xiangsun.Multiple birth support vector machine for multi-class classification[J].Neural Computing and Applications,2013,22(1):153-161.
[15]Zhang Xiekai,Ding Shifei,Xue Yu.An improved multiple birth support machine for pattern classification[J].Neurocomputing,2016(225):119-128.
[16]Mcculloch W S,Pitts W H.A logical calculus of ideas immanent in nervous activity[J].The Bulletin of Mathematical Biophysics,1942(5):115-133.
[17]XuXinshun,Tang Zheng,Wang Jiahai.A method to improve the transiently chaotic neural network[J].Neurocomputing,2004,67(1):456-463.
[18]Sun Jianye.Local coupled feedforward neural network[J].Neural Network,2010,23(1):108-113.
[19]Ozyildirim B M,Avci M.Generalized classifier neural network[J].Neural Networks,2013(39):18-26..
[20]Wang Shitong,Chung FuLai,Wang Jun,et al.A fast learning method for feedforward neural networks[J].Neurocomputing,2015(149):295-307.
[21]López-Soto,Diana,Angel-Bello F,et al.A multi-start algorithm to design a multi-class classifier for a multi-criteria ABC inventory classification problem[J].Expert Systems with Applications,2017(81):12-21.
[22]Roy S,Mondal S,Ekbal A,et al.Dispersion ratio based decision tree model for classification[J].Expert Systems with Applications,2019,116(C):1-9.
[23]Quinlan J R.Induction of decision trees[J].Machine Learning,1986,1(1):81-106.
[24]Quinlan J R.C4.5:programs for machine learning[M].Morgan Kaufmann Publishers Inc,1993.
[25]Loh W Y.Classification and regression trees[J].Wiley Interdisciplinary Reviews Data Mining&Knowledge Discovery,2011,1(1):14-23.
[26]Rooney N,Hui Wang,Taylor P S.An investigation into the application of ensemble learning for entailment classification[J].Information Processing and Management,2014(1):87-103.
[27]Chaudhary A,Kolhe S,Kamal R.An improve random forest classifier for multi-class classification[J].Information Processing in Agriculture,2016,3(4):215-222.
[28]Chen T,Guestrin C.XGBoost:A scalable tree boosting System[C]//ACM SIGKDD International Conference on Knowledge Discovery and Data Mining,San Francisco,CA,USA,2016:785-794.
[29]Genuer R,Poggi J M,Tuleau-Malot C,et al.Random forests for big data[J].Big Data Research,2017(9):28-46.
[30]Ke Guolin,Meng Qi,Finley T,et al.LightGBM:a highly efficient gradient boosting decision tree[C]//Advances in Neural Information Processing Systems,San Mateo,CA,USA,2017:3148-3156.
[31]Xu Yong,Lu Yuwu.Adaptive weighted fusion:A novel fusion approach for image classification[J].Neurocomputing,2015(168):566-574.
[32]Mazher A,Li P,Moughal T A,et al.A decision fusion method using an algorithm for fusion of correlated probabilities[J].International Journal of Remote Sensing,2016,37(1):14-25.
[33]Alcalá-Fdez J,Fernandez A,Luengo J,et al.KEEL data-mining software tool:data set repository,integration of algorithms and experimental analysis framework[J].Journal of Multiple-Valued Logic and Soft Computing,2011,17(2-3):255-287.KEEL数据集[DB/OL].[2018-05-06].http://www.keel.es/.