基于变分自编码器的时间序列生成及异常检测模型研究

郭昌林; 周在彦; 刘春秀; 李万彬; 刘奕敏; 李龙潭; 金桂玥; 薛志伟

doi:10.11959/j.issn.2096-3750.2026.00499

您当前的位置：

首页 >

文章列表页 >

基于变分自编码器的时间序列生成及异常检测模型研究

理论与技术 | 更新时间：2026-04-16

- 基于变分自编码器的时间序列生成及异常检测模型研究
- Research on time series generation and anomaly detection model based on variational autoencoder
- 物联网学报 2026年10卷第1期页码：237-249
- 作者机构：
  
  国网山东省电力公司德州供电公司，山东德州 253000
- 作者简介：
  
  [ "郭昌林（1982‒ ），男，国网山东省电力公司德州供电公司高级工程师，主要研究方向为电力系统及其自动化。" ]
  [ "周在彦（1992‒ ），男，国网山东省电力公司德州供电公司工程师，主要研究方向为配网协同调度。" ]
  [ "刘春秀（1975‒ ），女，国网山东省电力公司德州供电公司正高级工程师，主要研究方向为电力系统及其自动化。" ]
  [ "李万彬（1974‒ ），男，国网山东省电力公司德州供电公司高级工程师，主要研究方向为电力系统及其自动化。" ]
  [ "刘奕敏（1998‒ ），女，国网山东省电力公司德州供电公司助理工程师，主要研究方向为配电网自动化。" ]
  [ "李龙潭（1987‒ ），男，国网山东省电力公司德州供电公司工程师，主要研究方向为电网自动化。" ]
  [ "金桂玥（1990‒ ），女，国网山东省电力公司德州供电公司工程师，主要研究方向为带电作业。" ]
  [ "薛志伟（1986‒ ），男，国网山东省电力公司德州供电公司高级工程师，主要研究方向为配网不停电作业。" ]
- 基金信息：
  
  国网山东省电力公司科技项目(520608240006)
- DOI：10.11959/j.issn.2096-3750.2026.00499
  中图分类号： TN915.08
- 收稿：2025-06-10，
  
  修回：2025-03-24，
  
  录用：2025-06-16，
  
  纸质出版：2026-03-30
- 稿件说明：
移动端阅览
郭昌林,周在彦,刘春秀等.基于变分自编码器的时间序列生成及异常检测模型研究[J].物联网学报,2026,10(01):237-249.

Guo Changlin,Zhou Zaiyan,Liu Chunxiu,et al.Research on time series generation and anomaly detection model based on variational autoencoder[J].Chinese Journal on Internet of Things,2026,10(01):237-249.
郭昌林,周在彦,刘春秀等.基于变分自编码器的时间序列生成及异常检测模型研究[J].物联网学报,2026,10(01):237-249. DOI： 10.11959/j.issn.2096-3750.2026.00499.

Guo Changlin,Zhou Zaiyan,Liu Chunxiu,et al.Research on time series generation and anomaly detection model based on variational autoencoder[J].Chinese Journal on Internet of Things,2026,10(01):237-249. DOI： 10.11959/j.issn.2096-3750.2026.00499.

摘要

异常检测在工业设备故障监测等能源物联网场景中具有重要的应用价值，能够帮助物联网系统实时识别时间序列中的异常模式，从而提升系统的安全性、稳定性和运维效率。然而，时序数据的稀缺性是制约模型性能的主要瓶颈之一，这主要源于高质量标注的时序数据获取成本高昂，以及工业生产等场景下的时序数据采集条件有限，难以覆盖各种可能的情境。传统的数据增强方法难以有效捕获异常事件的复杂性和多样性，进一步制约了检测模型的性能提升。为此，提出了一种基于时间序列生成的数据增强方法，以提高模型在数据稀缺条件下的异常检测能力。该方法利用变分自编码器生成模型，在稀缺数据条件下合成具有真实性和多样性的时序数据，从而有效缓解数据稀缺性对模型性能的限制，显著提升异常检测的鲁棒性和准确性。实验结果表明，该方法在智能制造等能源物联网场景中具有良好的适应性，为构建高效、智能的异常检测系统提供了有效的技术支持。

Abstract

Anomaly detection is proving to be a key capability in energy Internet of things(IoT) scenarios such as industrial equipment fault monitoring

enabling IoT systems to identify abnormal patterns in time-series data in real time

thereby improving system security

stability

and operational efficiency. However

the scarcity of time-series data is emerging as a major bottleneck limiting model performance

mainly due to the high cost of acquiring high-quality labeled data and the limited data collection conditions in industrial production

which make it difficult to cover all possible scenarios. Traditional data augmentation methods are increasingly being regarded as inadequate for capturing the complexity and diversity of abnormal events

further constraining the performance of detection models. To address these issues

a data augmentation method based on time-series generation was proposed to enhance the anomaly detection capability of models under data-scarce conditions. The method utilized a variational autoencoder generative model to synthesize realistic and diverse time-series data from limited samples

thereby mitigating the impact of data scarcity on model performance and significantly improving the robustness and accuracy of anomaly detection. Experimental results demonstrated that the proposed method exhibited good adaptability in energy IoT scenarios such as smart manufacturing

providing effective technical support for building efficient and intelligent anomaly detection systems.

关键词

Keywords

references

Wang Q Y , Farahat A , Gupta C , et al . Deep time series models for scarce data [J ] . Neurocomputing , 2021 , 456 : 504 - 518 .

Iglesias G , Talavera E , González-Prieto Á , et al . Data augmentation techniques in time series domain: a survey and taxonomy [J ] . Neural Computing and Applications , 2023 , 35 ( 14 ): 10123 - 10145 .

Navidan H , Moshiri P F , Nabati M , et al . Generative adversarial networks (gans) in networking: a comprehensive survey & evaluation [J ] . Computer Networks , 2021 , 194 : 108149 .

Rigaki M , Garcia S . Bringing a gan to a knife-fight: adapting malware communication to avoid detection [C ] // Proceedings of the 2018 IEEE Security and Privacy Workshops (SPW) . Piscataway : IEEE Press , 2018 : 70 - 75 .

Song Y , Sohl-Dickstein J , Kingma D P , et al . Score-based generative modeling through stochastic differential equations [PP ] . V2. arXiv ( 2021-02-10 )[ 2025-06-10 ] . arXiv: 2011.13456 .

Ho J , Jain A , Abbeel P . Denoising diffusion probabilistic models [J ] . Advances in neural information processing systems , 2020 , 33 : 6840 - 6851 .

Goodfellow I , Pouget-Abadie J , Mirza M , et al . Generative adversarial nets [J ] . Advances in neural information processing systems , 2014 , 27 .

Ding D Z , Zhang M , Huang Y M , et al . Towards backdoor attack on deep learning based time series classification [C ] // Proceedings of the 2022 IEEE 38th International Conference on Data Engineering (ICDE) . Piscataway : IEEE Press , 2022 : 1274 - 1287 .

Li Z J , Tang Y C , Li W , et al . Learning disentangled representation with pairwise independence [C ] // Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence and Thirty-First Innovative Applications of Artificial Intelligence Conference and Ninth AAAI Symposium on Educational Advances in Artificial Intelligence (AAAI'19/IAAI'19/EAAI'19) . Palo Alto, CA : AAAI Press , 2019 : 4245 - 4252 .

Locatello F , Bauer S , Lucic M , et al . Challenging common assumptions in the unsupervised learning of disentangled representations [C ] // Proceedings of the 36th International Conference on Machine Learning . Brookline, MA : PMLR , 2019 : 4114 - 4124 .

Locatello F , Poole B , Rätsch G , et al . Weakly-supervised disentanglement without compromises [C ] // Proceedings of the 37th International Conference on Machine Learning . Brookline, MA : JMLR , 2020 : 6348 - 6359 .

Locatello F , Tschannen M , Bauer S , et al . Disentangling factors of variations using few labels [PP ] . V2. arXiv ( 2020-02-14 )[ 2025-06-10 ] . arXiv: 1905.01258 .

Alcaraz J M L , Strodthoff N . Diffusion-based time series imputation and forecasting with structured state space models [PP ] . V1. arXiv ( 2022-08-19 )[ 2025-06-10 ] . arXiv: 2208.09399 .

Yoon J , Jarrett D , Van der Schaar M . Time-series generative adversarial networks [C ] // Proceedings of the 33rd International Conference on Neural Information Processing Systems . New York : Curran Associates Inc. , 2019 : 5508 - 5518 .

Seyfi A , Rajotte J , Ng R T . Generating multivariate time series with common source coordinated GAN (COSCI-GAN) [C ] // Proceedings of the 36th International Conference on Neural Information Processing System . New York : Curran Associates Inc. , 2022 : 32777 - 32788 .

Wang L , Zeng L , Li J . AEC-GAN: adversarial error correction GANs for auto-regressive long time-series generation [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2023 , 37 ( 8 ): 10140 - 10148 .

Desai A , Freeman C , Wang Z H . et al . TimeVAE: a variational auto-encoder for multivariate time series generation [PP ] . V1. arXiv ( 2021-11-15 )[ 2025-06-10 ] . arXiv: 2111.08095 .

Lee D , Malacarne S , Aune E . Vector quantized time series generation with a bidirectional prior model [C ] // Proceedings of The 26th International Conference on Artificial Intelligence and Statistics . Brookline, MA, USA : PMLR , 2023 : 7665 - 7693 .

Li H M , Yu S J , Principe J , et al . Causal recurrent variational autoencoder for medical time series generation [C ] // Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artificial Intelligence and Thirteenth Symposium on Educational Advances in Artificial Intelligence . Palo Alto, CA : AAAI Press , 2023 : 8562 - 8570 .

Saharia C , Chan W , Chang H W , et al . Palette: image-to-image diffusion models [C ] // Proceedings of the ACM SIGGRAPH 2022 Conference Proceedings (SIGGRAPH '22) . New York, ACM Press : 2022 : 1 - 10 .

Dhariwal P , Nichol A Q . Diffusion models beat gans on image synthesis [C ] // Proceedings of the 35th International Conference on Neural Information Processing System . New York : Curran Associates Inc. , 2021 : 8780 - 879 .

Duong H-T , Nguyen-Thi T-A . A review: preprocessing techniques and data augmentation for sentiment analysis [J ] . Computational Social Networks , 2021 , 8 : 1 .

Felix E A , Lee S P . Systematic literature review of preprocessing techniques for imbalanced data [J ] . IET Software , 2019 , 13 ( 6 ): 479 - 496 .

Mikołajczyk A , Grochowski M . Data augmentation for improving deep learning in image classification problem [C ] // Proceedings of the 2018 International Interdisciplinary PhD Workshop (IIPhDW) . Piscataway : IEEE Press , 2018 : 117 - 122 .

Blázquez-García A , Conde A , Mori U , et al . A review on outlier/anomaly detection in time series data [J ] . ACM Computing Surveys (CSUR) , 2021 , 54 ( 3 ): 1 - 33 .

Nassif A B , Talib M A , Nasir Q , et al . Machine learning for anomaly detection: a systematic review [J ] . IEEE Access , 2021 , 9 : 78658 - 78700 .

Li Z , Zhao Y , Hu X Y , et al . ECOD: unsupervised outlier detection using empirical cumulative distribution functions [J ] . IEEE Transactions on Knowledge and Data Engineering , 2023 , 35 ( 12 ): 12181 - 12193 .

Xu D , Wang Y J , Meng Y L , et al . An improved data anomaly detection method based on isolation forest [C ] // Proceedings of the 2017 10th international symposium on computational intelligence and design (ISCID) . Piscataway : IEEE Press , 2017 : 287 - 291 .

Rashid K M , Louis J . Times-series data augmentation and deep learning for construction equipment activity recognition [J ] . Advanced Engineering Informatics , 2019 , 42 : 100944 .

Wen Q S , Sun L , Yang F , et al . Time series data augmentation for deep learning: a survey [PP ] . V4. arXiv ( 2022-03-31 )[ 2025-06-10 ] . arXiv: 2022.12478 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于KNN-RF-VMD-CNN-BiLSTM的日前电价预测算法

基于数据增强的图卷积神经网络的浓度预测方法