基于改进高斯混合模型的热工过程异常值检测

doi:10.16182/j.issn1004731x.joss.22-0047

系统仿真学报 ›› 2023, Vol. 35 ›› Issue (5): 1020-1033.doi: 10.16182/j.issn1004731x.joss.22-0047

基于改进高斯混合模型的热工过程异常值检测

吴铮¹^,²(), 张悦¹^,², 董泽¹^,²()

^1.华北电力大学控制与计算机工程学院, 北京 102206
^2.河北省发电过程仿真与优化控制技术创新中心(华北电力大学), 河北保定 071003

收稿日期:2022-01-16 修回日期:2022-02-18 出版日期:2023-05-30 发布日期:2023-05-22
通讯作者: 董泽 E-mail:wuzhengncepu@ncepu.edu.cn;dongze@ncepu.edu.cn
作者简介:吴铮(1996-)，男，满族，博士生，研究方向为火电机组建模理论与方法。E-mail：wuzhengncepu@ncepu.edu.cn
基金资助:
中央高校基本科研业务费专项资金(2018QN096);河北省自然科学基金(E2018502111)

Outlier Detection During Thermal Processes Based on Improved Gaussian Mixture Model

Zheng Wu¹^,²(), Yue Zhang¹^,², Ze Dong¹^,²()

^1.School of Control and Computer Engineering, North China Electric Power University, Beijing 102206, China
^2.Hebei Technology Innovation Center of Simulation & Optimized Control for Power Generation, North China Electric Power University, Baoding 071003, China

Received:2022-01-16 Revised:2022-02-18 Online:2023-05-30 Published:2023-05-22
Contact: Ze Dong E-mail:wuzhengncepu@ncepu.edu.cn;dongze@ncepu.edu.cn

摘要/Abstract

摘要：

热工过程异常数据检测是进行系统建模、控制、优化的基础，也是数据处理的重要组成部分。提出一种基于改进高斯混合模型的无监督热工过程异常值检测算法，利用每一维高斯分量捕获一类特定工况数据集群，通过添加惩罚约束因子修正传统模型的后验概率密度，对误检、漏检项加以惩罚，并根据与集群的相关性差异实现异常数据的识别。仿真实验结果表明，模型在多种误差条件下均可准确定位异常数据位置，具有很强的泛化性能，同时相较于传统高斯混合模型，误检、漏检点的检测效果总体提升了37.8%和15%，反映出模型改进的有效性。

关键词: 异常值检测, 高斯混合模型, 惩罚约束, 热工过程, 无监督

Abstract:

Abnormal data detection during thermal processes is the basis for performing system modeling, control, and optimization and constitutes an important part of data processing. In this paper, an unsupervised outlier detection algorithm during thermal processes based on an improved Gaussian mixture model is proposed. The algorithm captures a class of data clusters under specific working conditions by using Gaussian components in each dimension, modifies the posterior probability density of the traditional model by adding penalty constraint factors to penalize the false detection and missed detection items, and identifies abnormal data according to the correlation differences with the clusters. The simulation experimental results show that the model can accurately locate the abnormal data location under a variety of error conditions with strong generalization performance, and the overall detection effects of false detection and missed detection items are improved by 37.8% and 15% compared with the traditional Gaussian mixture model, which proves the effectiveness of the model improvement.

Key words: outlier detection, Gaussian mixture model, penalty constraint, thermal processes, unsupervised

中图分类号:

TH81

吴铮, 张悦, 董泽. 基于改进高斯混合模型的热工过程异常值检测[J]. 系统仿真学报, 2023, 35(5): 1020-1033.

Zheng Wu, Yue Zhang, Ze Dong. Outlier Detection During Thermal Processes Based on Improved Gaussian Mixture Model[J]. Journal of System Simulation, 2023, 35(5): 1020-1033.

图/表 13

表1

评价指标描述

变量名称	数学公式
准确率	$准确率 = 正确个数真实异常值个数$
漏检率	$漏检率 = 漏检个数真实异常值个数$
误判率	$误判率 = 误判个数真实异常值个数$

表1

图1

表2

图2

图3

图4

图5

图6

图7

表3

图8

表4

图9

参考文献 27

1	董泽, 贾昊. 基于EWT-LOF的热工过程数据异常值检测方法[J]. 仪器仪表学报, 2020, 41(2): 126-134.
	Dong Ze, Jia Hao. Detection Method for Abnormal Value of Thermal Process Data Based on EWT-LOF[J]. Chinese Journal of Scientific Instrument, 2020, 41(2): 126-134.
2	郜传厚, 渐令, 陈积明, 等. 复杂高炉炼铁过程的数据驱动建模及预测算法[J]. 自动化学报, 2009, 35(6): 725-730.
	Gao Chuanhou, Ling Gradually, Chen Jiming, et al. Data-Driven Modeling and Prediction Algorithm for Complex Blast Furnace Ironmaking Process[J]. Acta Automatica Sinica, 2009, 35(6): 725-730.
3	徐飞阳, 薛安成, 常乃超, 等. 电力系统同步相量异常数据检测与修复研究现状与展望[J]. 中国电机工程学报, 2021, 41(20): 6869-6886.
	Xu Feiyang, Xue Ancheng, Chang Naichao, et al. Research Status and Prospect of Synchrophasor Anomaly Data Detection and Restoration in Power Systems[J]. Chinese Journal of Electrical Engineering, 2021, 41(20): 6869-6886.
4	杨茂, 杨琼琼. 基于云分段最优熵算法的风电机组异常数据识别研究[J]. 中国电机工程学报, 2018, 38(8): 2294-2301, 2539.
	Yang Mao, Yang Qiongqiong. Research on Abnormal Data Recognition of Wind Turbines Based on Cloud Segmentation Optimal Entropy Algorithm[J]. Proceedings of the Chinese Society for Electrical Engineering, 2018, 38(8): 2294-2301, 2539.
5	闫媞锦, 夏元清, 张宏伟, 等. 一种非规则采样航空时序数据异常检测方法[J]. 航空学报, 2021, 42(4): 558-568.
	Yan Tijin, Xia Yuanqing, Zhang Hongwei, et al. Anomaly Detection Method of Irregularly Sampled Aviation Time Series Data[J]. Acta Aeronautica Sinica, 2021, 42(4): 558-568.
6	范涛. 基于钻孔瞬变电磁的煤层气压裂效果检测方法[J]. 煤炭学报, 2020, 45(9): 3195-3207.
	Fan Tao. Detection Method of Coal Seam Gas Pressure Cracking Effect Based on Borehole Transient Electromagnetic[J]. Journal of China Coal Society, 2020, 45(9): 3195-3207.
7	Zhao J, Kai L, Wei W, et al. Adaptive Fuzzy Clustering Based Anomaly Data Detection in Energy System of Steel Industry[J]. Information Sciences (S0020-0255), 2014, 259(3): 335-345.
8	窦珊, 张广宇, 熊智华. 基于LSTM时间序列重建的生产装置异常检测[J]. 化工学报, 2019, 70(2): 481-486.
	Dou Shan, Zhang Guangyu, Xiong Zhihua. Anomaly Detection of Production Equipment Based on LSTM Time Series Reconstruction[J]. CIESC Journal, 2019, 70(2): 481-486.
9	彭宇, 庞景月, 刘大同, 等. 大数据: 内涵、技术体系与展望[J]. 电子测量与仪器学报, 2015, 29(4): 469-482.
	Peng Yu, Pang Jingyue, Liu Datong, et al. Big Data: Connotation, Technical System and Prospect[J]. Journal of Electronic Measurement and Instrument, 2015, 29(4): 469-482.
10	D M auth Hawkins. Identification of Outliers[M]. South Africa: Chapman and Hall, Springer, 1980.
11	于冰洁. 基于高斯模型的异常检测算法[D]. 徐州: 中国矿业大学, 2017.
	Yu Bingjie. Anomaly Detection Algorithm Based on Gaussian Model[D]. Xuzhou: China University of Mining and Technology, 2017.
12	谢敬东, 卢浩哲, 陆池鑫, 等. 基于分阶段离群点检测的电力市场异常辨识[J]. 科学技术与工程, 2021, 21(9): 3633-3641.
	Xie Jingdong, Lu Haozhe, Lu Chixin, et al. Electricity Market Anomaly Identification Based on Staged Outlier Detection[J]. Science Technology and Engineering, 2021, 21(9): 3633-3641.
13	Ye Q, Zhao H, Li Z, et al. L1-Norm Distance Minimization-Based Fast Robust Twin Support Vector k-Plane Clustering[J]. IEEE Transactions on Neural Networks and Learning Systems (S2162-237X), 2018, 29(9): 4494-4503.
14	徐瑶, 李栓, 韩英华. 基于CNN-GS-SVM的用户异常用电行为检测[J]. 控制工程, 2021, 28(10): 1989-1997.
	Xu Yao, Li Shuan, Han Yinghua. Detection of Abnormal Power Consumption Behavior of Users Based on CNN-GS-SVM[J]. Control Engineering, 2021, 28(10): 1989-1997.
15	韩旭东. 基于数据驱动的火电机组高压加热系统异常检测研究[J]. 华电技术, 2021, 43(8): 67-73.
	Han Xudong. Research on Abnormal Detection of High-Voltage Heating System of Thermal Power Unit Based on Data Drive[J]. Huadian Technology, 2021, 43(8): 67-73.
16	翟怡萌, 李晶, 程强, 等. 基于OCSVM的火电厂一次风机异常检测[J]. 工业控制计算机, 2019, 32(6): 102-104.
	Zhai Yimeng, Li Jing, Cheng Qiang, et al. Anomaly Detection of Primary Fan in Thermal Power Plant Based on OCSVM[J]. Industrial Control Computer, 2019, 32(6): 102-104.
17	Cho J H, Lee J M, Sang W C, et al. Fault Identification for Process Monitoring Using Kernel Principal Component Analysis[J]. Chemical Engineering Science (S0009-2509), 2005, 60(1): 279-288.
18	王小邦, 贺凯迅, 苏照阳. 融合互信息与慢特性的火电机组异常监控[J]. 控制工程, 2021, 10(1): 1-6.
	Wang Xiaobang, He Kaixun, Su Zhaoyang. Anomaly Monitoring of Thermal Power Units Integrating Mutual Information and Slow Characteristics[J]. Control Engineering, 2021, 10(1): 1-6.
19	Yc A, Ning C A, Mao C A, et al. A Spatially Constrained Asymmetric Gaussian Mixture Model for Image Segmentation[J]. Information Sciences (S0020-0255), 2021, 575(1): 41-45.
20	Zhang B, Shin Y C. An Adaptive Gaussian Mixture Method for Nonlinear Uncertainty Propagation in Neural Networks[J]. Neurocomputing (S0925-2312), 2021, 458(2): 170-183.
21	Shahin I, Nassif A B, Nemmour N, et al. Novel Hybrid DNN Approaches for Speaker Verification in Emotional and Stressful Talking Environments[J]. Neural Computing and Applications (S0941-0643), 2021, 33(1): 1-23.
22	燕彩蓉, 张青龙, 赵雪, 等. 基于广义高斯分布的贝叶斯概率矩阵分解方法[J]. 计算机研究与发展, 2016, 53(12): 2793-2800.
	Yan Cairong, Zhang Qinglong, Zhao Xue, et al. Bayesian Probability Matrix Factorization Method Based on Generalized Gaussian Distribution[J]. Computer Research and Development, 2016, 53(12): 2793-2800.
23	周帆, 郑常宝, 胡存刚, 等. 基于高斯混合模型的光伏发电功率概率区间预测[J]. 科学技术与工程, 2021, 21(24): 10284-10290.
	Zhou Fan, Zheng Changbao, Hu Cungang, et al. Probability Interval Prediction of Photovoltaic Power Generation Power Based on Gaussian Mixture Model[J]. Science Technology and Engineering, 2021, 21(24): 10284-10290.
24	Yu J, Chen K, Mori J, et al. A Gaussian Mixture Copula Model Based Localized Gaussian Process Regression Approach for Long-Term Wind Speed Prediction[J]. Energy (S0360-5442), 2013, 61(1): 673-686.
25	刘岩, 李文文, 周丽霞, 等. 基于高斯混合模型的光伏发电出力中高比例异常数据检测方法研究[J]. 电测与仪表, 2021, 58(9): 14-21.
	Liu Yan, Li Wenwen, Zhou Lixia, et al. Research on Abnormal Data Detection Method of Photovoltaic Power Generation Output Based on Gaussian Mixture Model[J]. Electrical Measurement and Instrumentation, 2021, 58(9): 14-21.
26	Reynolds, Gaussian Mixture Models D.A. : Encycl. Biometric Recognition[J]. Springer (S2193-1801), 2009, 10(5): 659-663.
27	Schwarz G E. Estimating the Dimension of a Model[J]. The Annals of Statistics (S0090-5364), 1978, 6(2): 461-464.

类别	数据集1	数据集2	数据集3	数据集4
1001-1050	+2	+5	+8	+10
1101-1150	-2	-5	-8	-10
3001-3050	+2	+5	+8	+10
3101-3150	-2	-5	-8	-10
6001-6050	+2	+5	+8	+10
6101-6150	-2	-5	-8	-10
9001-9050	+2	+5	+8	+10
9101-9150	-2	-5	-8	-10

误差/%	参数	5×10^-4	1×10^-4	5×10^-5	1×10^-5
±2	误判率	22/400	17/400	2/400	0
±2	漏检率	15/400	24/400	42/400	50/400
±5	误判率	15/400	2/400	0	0
±5	漏检率	0	0	0	0
±8	误判率	15/400	2/400	0	0
±8	漏检率	0	0	0	0
±10	误判率	15/400	2/400	0	0
±10	漏检率	0	0	0	0

算法	准确率	漏检率	误判率
PC-GMM	0.94	0.06	0.04
GMM	0.79	0.21	0.42
孤立森林	0.76	0.23	0.31
DBSCAN	0.71	0.29	0.12

[1]	何一帆, 何玉林, 蔡湧达, 黄哲学. 基于夹角几何的I-niceMO增强算法[J]. 系统仿真学报, 2023, 35(4): 797-808.
[2]	周培培, 侯幸林. 一种用于图像融合的无监督深度神经网络[J]. 系统仿真学报, 2022, 34(6): 1267-1274.
[3]	祁成, 熊伟丽. 基于BGMM的即时学习软测量建模方法[J]. 系统仿真学报, 2019, 31(8): 1555-1561.
[4]	陈芳芳, 计忠平. 人脸肤色检测与编辑[J]. 系统仿真学报, 2019, 31(7): 1377-1386.
[5]	赵帅, 史旭东, 熊伟丽. 一种分层集成的高斯过程回归软测量建模方法[J]. 系统仿真学报, 2019, 31(10): 2042-2051.
[6]	詹永杰, 龙飞, 卜轶坤. 基于独立子空间分析特征学习的表情识别[J]. 系统仿真学报, 2015, 27(10): 2316-2319.

基于改进高斯混合模型的热工过程异常值检测

Outlier Detection During Thermal Processes Based on Improved Gaussian Mixture Model

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 27

相关文章 6

编辑推荐

Metrics

本文评价