GTAD: Graph and Temporal Neural Network for Multivariate Time Series Anomaly Detection

Guan, Siwei; Zhao, Baowei; Dong, Zhekang; Gao, Mingyu; He, Zhiwei

doi:10.3390/e24060759

Cited by 22 publications

(11 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to verify the optimization performance of the model for the CVT internal insulation abnormality detection in this paper, in the comparison method based on the autoencoder, this paper compared the LSTM-AE model [20], based on the autoencoder, with the MTAD-GAT model [18] and the GTAD model [23], based on the prediction and reconstruction model. The Precision, Recall and F1-score of these three indicators were compared.…”

Section: Compare and Verifymentioning

confidence: 99%

“…The literature [17] separates the CVT secondary side voltage into a primary voltage component and residual component and achieves the evaluation of measurement error by evaluating the residual component. In recent years, deep learning methods for anomaly detection in various fields have been studied in depth, such as MTAD-GAT [18], TCN-AE [19], LATM-AE [20], USAD [21], TranAD [22] and GTAD [23], achieving good results in the field of anomaly detection.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Online Evaluation Method of CVT Internal Insulation Abnormality Based on Self-Supervised Learning

He,

Zhou,

Tong

et al. 2023

Energies

View full text Add to dashboard Cite

A capacitive voltage transformer (CVT) is one of the electrical quantities measurement devices, and the state of its internal insulation is the key factor for ensuring the accuracy of its measurement of electrical energy. In view of the fact that the traditional real-time evaluation method of a CVT internal insulation anomaly mainly relies on empirical rules and prior knowledge and lacks the ability to independently mine effective features, an online evaluation method of a CVT internal insulation anomaly based on self-supervised learning is proposed. Firstly, an autoencoder is constructed to extract the residual sequence of the CVT secondary voltage and eliminate the influence of primary voltage fluctuation and power system voltage regulation. Without any prior knowledge, the complex dependence of the residual sequences in time and feature dimensions is learned by using a parallel graph attention layer (GATv2). Finally, a joint optimization based on the prediction and reconstruction model is introduced to obtain the abnormal inference score at each timestamp and realize the evaluation of the CVT internal insulation status. Experimental analysis shows that this method can effectively eliminate the influence of primary voltage fluctuation and power system voltage regulation on the online evaluation of the CVT internal insulation status and independently excavate the abnormal characteristics of the CVT secondary voltage to realize real-time monitoring and early warning of the CVT internal insulation status.

show abstract

Section: Compare and Verifymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Online Evaluation Method of CVT Internal Insulation Abnormality Based on Self-Supervised Learning

He,

Zhou,

Tong

et al. 2023

Energies

View full text Add to dashboard Cite

show abstract

“…The problem of anomaly detection has been an important subject of study in several research communities, such as statistics, signal processing, machine learning, information theory, and data mining, either specifically for an application domain or as a generic method. To name a few, an SVM classification approach for anomaly detection was proposed in [ 10 ]; Bayesian methods were developed for social networks [ 11 ], partially observed traffic networks [ 12 ], and streaming environmental data [ 13 ]; deep neural network models were proposed for detecting anomalies multivariate time series [ 14 , 15 , 16 , 17 , 18 ]; several information theoretic measures were proposed in [ 19 ] for the intrusion detection problem; and two new information metrics for DDoS attack detection was introduced in [ 3 ]. Due to the challenging nature of the problem and considering the challenges posed by today’s technological advances such as big data problems, there is still a need for studying the anomaly detection problem.…”

Section: Related Workmentioning

confidence: 99%

Online Multivariate Anomaly Detection and Localization for High-Dimensional Settings

Mozaffari

Doshi

Yılmaz

2022

Sensors

View full text Add to dashboard Cite

This paper considers the real-time detection of abrupt and persistent anomalies in high-dimensional data streams. The goal is to detect anomalies quickly and accurately so that the appropriate countermeasures could be taken in time before the system possibly gets harmed. We propose a sequential and multivariate anomaly detection method that scales well to high-dimensional datasets. The proposed method follows a nonparametric, i.e., data-driven, and semi-supervised approach, i.e., trains only on nominal data. Thus, it is applicable to a wide range of applications and data types. Thanks to its multivariate nature, it can quickly and accurately detect challenging anomalies, such as changes in the correlation structure. Its asymptotic optimality and computational complexity are comprehensively analyzed. In conjunction with the detection method, an effective technique for localizing the anomalous data dimensions is also proposed. The practical use of proposed algorithms are demonstrated using synthetic and real data, and in variety of applications including seizure detection, DDoS attack detection, and video surveillance.

show abstract

“…Multivariate time series are formed in smart building energy monitoring systems with multiple sensing measurements, such as room temperature and humidity sensors, water detection sensors, flow meters, etc. As can be seen, univariate time series anomaly detection algorithms are able to detect anomalies in individual metrics but do not represent the overall state of the smart building system well [ 18 , 22 ]. It can be seen that in practical application scenarios, building energy consumption anomalies are not only related to time series but also depend on the correlation between multiple characteristic variables.…”

Section: Introductionmentioning

confidence: 99%

Anomaly detection method for building energy consumption in multivariate time series based on graph attention mechanism

et al. 2023

View full text Add to dashboard Cite

A critical issue in intelligent building control is detecting energy consumption anomalies based on intelligent device status data. The building field is plagued by energy consumption anomalies caused by a number of factors, many of which are associated with one another in apparent temporal relationships. For the detection of abnormalities, most traditional detection methods rely solely on a single variable of energy consumption data and its time series changes. Therefore, they are unable to examine the correlation between the multiple characteristic factors that affect energy consumption anomalies and their relationship in time. The outcomes of anomaly detection are one-sided. To address the above problems, this paper proposes an anomaly detection method based on multivariate time series. Firstly, in order to extract the correlation between different feature variables affecting energy consumption, this paper introduces a graph convolutional network to build an anomaly detection framework. Secondly, as different feature variables have different influences on each other, the framework is enhanced by a graph attention mechanism so that time series features with higher influence on energy consumption are given more attention weights, resulting in better anomaly detection of building energy consumption. Finally, the effectiveness of this paper’s method and existing methods for detecting energy consumption anomalies in smart buildings are compared using standard data sets. The experimental results show that the model has better detection accuracy.

show abstract

GTAD: Graph and Temporal Neural Network for Multivariate Time Series Anomaly Detection

Cited by 22 publications

References 45 publications

Online Evaluation Method of CVT Internal Insulation Abnormality Based on Self-Supervised Learning

Online Evaluation Method of CVT Internal Insulation Abnormality Based on Self-Supervised Learning

Online Multivariate Anomaly Detection and Localization for High-Dimensional Settings

Anomaly detection method for building energy consumption in multivariate time series based on graph attention mechanism

Contact Info

Product

Resources

About