[1]廖彬生,聂益民,曾江蛟,等.基于VMD 与LSTM-VAE密度聚类的火电机组主蒸汽压力异常检测方法[J].机械与电子,2026,44(01):103-110.
　LIAO Binsheng,NIE Yimin,ZENG Jiangjiao,et al. Anomaly Detection of Main Steam Pressure in Thermal Power Units Based on VMD and LSTM-VAE Density Clustering[J].Machinery & Electronics,2026,44(01):103-110.
点击复制

基于VMD 与LSTM-VAE密度聚类的火电机组主蒸汽压力异常检测方法()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
44
期数:
2026年01期
页码:
103-110
栏目:
电力控制
出版日期:
2026-01-27

文章信息/Info

Title:
 Anomaly Detection of Main Steam Pressure in Thermal Power Units Based on VMD and LSTM-VAE Density Clustering
文章编号:
1001-2257(2026)01-0103-08
作者:
 廖彬生1; 聂益民1; 曾江蛟1; 袁宇龙2; 何 钧3; 袁小翠2
 (1.国能九江发电有限公司,江西 九江 332504;2.江西水利电力大学电气工程学院,江西 南昌 330099;3.国网江西省电力有限公司电力科学研究院,江西 南昌 330096)
Author(s):
 LIAO Binsheng1; NIE Yimin1; ZENG Jiangjiao1; YUAN Yulong2; HE Jun3; YUAN Xiaocui2
关键词:
火电机组; 异常检测; 变分模态分解; LSTM VAE; DBSCAN
Keywords:
 ; (1.Guoneng Jiujiang Power Generation Co.; Ltd.; Jiujiang 332504; China; 2.School of Electrical Engineering; Jiangxi University of Water Resources and Electric Power; Nanchang 330099; China; 3.Electric Power Research Institute; State Grid Jiangxi Electric Power Co.; Ltd.; Nanchang 330096; China)
分类号:
TM621.6;TP183
文献标志码:
A
摘要:
为提升火电机组主控参数异常检测能力,以主控参数主蒸汽压力为对象,提出基于VMD 与LSTM-VAE密度聚类的火电机组主蒸汽压力异常检测方法。采用变分模态分解方法对主蒸汽压力时序信号进行多分量分解,提取主导模态并重建信号以实现降噪;构建长短期记忆网络变分自编码器模型,通过无监督学习提取正常工况下的时序数据分布特征,计算每个滑窗的重构误差;将重构误差作为聚类特征,并应用自适应密度聚类算法对重构误差特征分类,从而实现异常检测。以某电厂650 MW 火电机组主蒸汽压力数据为算例样本,分析结果表明,异常检测精确率达到0.832,召回率达到1.000,F1分数达到0.908,且在标准差0.05的高斯噪声干扰下F1分数仍达到0.887,验证了所提方法的检测性能和应用价值。
Abstract:
To enhance the anomaly detection capability for critical control parameters in thermal power units,a novel method based on the variational mode decomposition (VMD) and density clustering using the long short term memory variational autoencoder (LSTM VAE) is proposed,focusing on the main steam pressure parameter.The VMD approach is first employed to decompose the raw steam pressure time series signal into multiple components.The dominant modes are extracted and used to reconstruct the signal with effective noise reduction.Subsequently,an LSTM VAE model is constructed to learn the distribution characteristics of time series data under normal operating conditions in an unsupervised manner,and the reconstruction error for each sliding window is computed.These reconstruction errors are then utilized as clustering features.An adaptive density based clustering algorithm is applied to classify the features of reconstruction error,enabling the identification of anomalies.The proposed method is validated by a case study that using main steam pressure data from a 650 MW thermal power unit.The results demonstrate that the method achieves an anomaly detection precision of 0.832,a recall of 1.000,and an F1 score of 0.908.Furthermore,it maintains a high F1 score of 0.887 under the interference of Gaussian noise with a standard deviation of 0.05,confirming its robust detection performance and practical application value.

参考文献/References:

 [1] 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.
[2] 赵悦,方彦军,董政呈.基于状态识别的经验模态分解法火电厂运行数据预处理[J].热力发电,2019,48(1):49-54.
[3] 苏安龙,孙志鑫,何晓洋,等.基于多元经验模式分解的电力系统低频振荡模式辨识[J].电力系统保护与控制,2019,47(22):113-125.
[4] 邢鼎皇,杨光,叶娟,等.基于DBSCAN算法的燃气流量数据异常检测[J].煤气与热力,2024,44(6):24-29.
[5] ZHANG H,LIN J,HUA J D,et al.Data anomaly detection for bridge SHM based on CNN combined with statistic features[J].Journal of nondestructive evaluation,2022,41:28.
[6] ULLAH I,MAHMOUD Q H.Design and development of RNN anomaly detection model for IoT networks[J].IEEE Access,2022,10:62722-62750.
[7] LINDEMANN B,MASCHLER B,SAHLAB N,et al.A survey on anomaly detection for technical systems using LSTM networks[J].Computers in industry,2021,131:103498.
[8] BERGAMIN L,CARRARO T,POLATO M,et al.Novel applications for vae based anomaly detection systems[C]∥2022 International Joint Conference on Neural Networks (IJCNN).New York:IEEE,2022:1-8.
[9] LIN S Y,CLARK R,BIRKE R,et al.Anomaly detection for time series using VAE LSTM hybrid model[C]∥ICASSP 2020 2020 IEEE International Conference on Acoustics,Speech and Signal Processing (ICASSP).New York:IEEE,2020:4322-4326.
[10] HU D,ZHANG C,YANG T,et al.Anomaly detection of power plant equipment using long short term memory based autoencoder neural network[J].Sensors,2020,20(21):6164.
[11] KIM H,KO J U,NA K,et al.Opt TCAE:optimal temporal convolutional auto encoder for boiler tube leakage detection in a thermal power plant using multi sensor data[J].Expert systems with applications,2023,215:119377.
[12] XU W M,ZHANG P.Steam turbine anomaly detection:an unsupervised learning approach using enhanced long short term memory variational autoencoder[J].Applied thermal engineering,2025,278:127138.
[13] LI G L,LI Y J,LI S T,et al.Research on anomaly detection of steam power system based on the coupling of thermoeconomics and autoencoder[J].Energy,2025,318:134819.
[14] YAN X A,JIA M P.Application of CSA VMD and optimal scale morphological slice bispectrum in enhancing outer race fault detection of rolling element bearings[J].Mechanical systems and signal processing,2019,122:56-86.
[15] JAIN M,SAIHJPAL V,SINGH N,et al.An overview of variants and advancements of PSO algorithm[J].Applied sciences,2022,12(17):8392.
[16] MARINI F,WALCZAK B.Particle swarm optimization(PSO).A tutorial[J].Chemometrics and intelligent laboratory systems,2015,149:153-165.
[17] TAO Z F,XU Q H,LIU X,et al.An integrated approach implementing sliding window and DTW distance for time series forecasting tasks[J].Applied intelligence,2023,53(17):20614-20625.
[18] 朱乔木,李弘毅,王子琪,等.基于长短期记忆网络的风电场发电功率超短期预测[J].电网技术,2017,41(12):3797-3802.
[19] 翟正利,梁振明,周炜,等.变分自编码器模型综述[J].计算机工程与应用,2019,55(3):1-9.
[20] 方知,贾晓芬,赵佰亭.复杂场景下的双编码器裂缝分割方法[J/OL].激光与光电子学进展,1-18[2025-06-12].https:∥link.cnki.net/urlid/31.1690.TN.20250611.1641.058.

相似文献/References:

[1]孙兆光.基于支持向量机的农田水利灌溉分流机械运行异常检测[J].机械与电子,2022,(07):43.
　SUN Zhaoguang.Detection of Abnormal Operation of Farmland Irrigation and Diversion Machinery Based on Support Vector Machine[J].Machinery & Electronics,2022,(01):43.

备注/Memo

备注/Memo:
收稿日期:2025-06-13
基金项目:国家能源集团科技研发项目(GNJ 24 KJ 03);国网公司重点研发项目(52182025000C)
作者简介:廖彬生 (1971-),男,江西宁都人,工程师,研究方向为电厂热能动力工程;何 钧 (1979-),男,湖南永州人,正高级工程师,研究方向为热工自动化、网源协调控制,通信作者,E-mail:13870653216@163.com。
更新日期/Last Update: 2026-03-09