Anomaly detection is an important problem that has been well-studied within diverse research areas and application domains. The aim of this survey is two-fold, firstly we present a structured and comprehensive overview of research methods in deep learning-based anomaly detection. Furthermore, we review the adoption of these methods for anomaly across various application domains and assess their effectiveness. We have grouped state-of-the-art research techniques into different categories based on the underlying assumptions and approach adopted. Within each category we outline the basic anomaly detection technique, along with its variants and present key assumptions, to differentiate between normal and anomalous behavior. For each category, we present we also present the advantages and limitations and discuss the computational complexity of the techniques in real application domains. Finally, we outline open issues in research and challenges faced while adopting these techniques.

Deep Learning for Anomaly Detection: A Survey.

Deep learning has attracted intense interest in Prognostics and Health Management (PHM), because of its enormous representing power, automated feature learning capability and best-in-class performance in solving complex problems. This paper surveys recent advancements in PHM methodologies using deep learning with the aim of identifying research gaps and suggesting further improvements. After a brief introduction to several deep learning models, we review and analyze applications of fault detection, diagnosis and prognosis using deep learning. The survey validates the universal applicability of deep learning to various types of input in PHM, including vibration, imagery, time-series and structured data. It also reveals that deep learning provides a one-fits-all framework for the primary PHM subfields: fault detection uses either reconstruction error or stacks a binary classifier on top of the network to detect anomalies; fault diagnosis typically adds a soft-max layer to perform multi-class classification; prognosis adds a continuous regression layer to predict remaining useful life. The general framework suggests the possibility of transfer learning across PHM applications. The survey reveals some common properties and identifies the research gaps in each PHM subfield. It concludes by summarizing some major challenges and potential opportunities in the domain.

/pdf/a-review-on-deep-learning-applications-in-prognostics-and-3cw1pwlic5.pdf

A Review on Deep Learning Applications in Prognostics and Health Management

It is a great challenge to differentiate partial discharge (PD) induced by different types of insulation defects in high-voltage cables. Some types of PD signals have very similar characteristics and are specifically difficult to differentiate, even for the most experienced specialists. To overcome the challenge, a convolutional neural network (CNN)-based deep learning methodology for PD pattern recognition is presented in this paper. First, PD testing for five types of artificial defects in ethylene-propylene-rubber cables is carried out in high voltage laboratory to generate signals containing PD data. Second, 3500 sets of PD transient pulses are extracted, and then 33 kinds of PD features are established. The third stage applies a CNN to the data; typical CNN architecture and the key factors which affect the CNN-based pattern recognition accuracy are described. Factors discussed include the number of the network layers, convolutional kernel size, activation function, and pooling method. This paper presents a flowchart of the CNN-based PD pattern recognition method and an evaluation with 3500 sets of PD samples. Finally, the CNN-based pattern recognition results are shown and the proposed method is compared with two more traditional analysis methods, i.e., support vector machine (SVM) and back propagation neural network (BPNN). The results show that the proposed CNN method has higher pattern recognition accuracy than SVM and BPNN, and that the novel method is especially effective for PD type recognition in cases of signals of high similarity, which is applicable for industrial applications.

/pdf/a-convolutional-neural-network-based-deep-learning-c44kdv7j1x.pdf

A Convolutional Neural Network-Based Deep Learning Methodology for Recognition of Partial Discharge Patterns from High-Voltage Cables

This paper presents a state-of-the-art review on machine learning (ML) based intelligent diagnostics that have been applied for partial discharge (PD) detection, localization, and pattern recognition. ML techniques, particularly those developed in the last five years, are examined and classified as conventional ML or deep learning (DL). Important features of each method, such as types of input signal, sampling rate, core methodology, and accuracy, are summarized and compared in detail. Advantages and disadvantages of different ML algorithms are discussed. Moreover, technical roadblocks preventing intelligent PD diagnostics from being applied to industry are identified, such as insufficient/imbalanced dataset, data inconsistency, and difficulties in cost-effective real-time deployment. Finally, potential solutions are proposed, and future research directions are suggested.

Condition Monitoring Based on Partial Discharge Diagnostics Using Machine Learning Methods: A Comprehensive State-of-the-Art Review

Deep learning methods and applications for electrical power systems: A comprehensive review

Dissolved gas analysis (DGA) of insulating oil can provide an important basis for transformer fault diagnosis. To improve diagnosis accuracy, this paper presents a new transformer fault diagnosis method based on deep belief networks (DBN). By analyzing the relationship between the gases dissolved in transformer oil and fault types, the Non­code ratios of the gases are determined as the characterizing parameter of the DBN model. DBN adopts multi-layer and multi-dimension mapping to extract more detailed differences of fault types. In this process, the diagnosis parameters are pre-trained. A back-propagation algorithm adjusts them with the labels of the samples and optimizes the parameters. To verify the effect of the proposed method, the diagnostic DBN model is constructed and tested using various oil chromatographic datasets collected from the State Grid Corporation of China and previous publications. The performances of the DBN diagnosis model are analyzed by different characterizing parameters, different training datasets and sample datasets. In addition, the influence of discharge and overheating multiple faults on the diagnosis model is studied. The performance of the proposed approach is compared with that derived from support vector machine (SVM), back-propagation neural network (BPNN) and ratio methods respectively. The results show that the proposed method significantly improves the accuracy of power transformer fault diagnosis.

Dissolved gas analysis of insulating oil for power transformer fault diagnosis with deep belief network

This paper aims to improve recognition accuracies of partial discharge (PD) of complex data sources by employing deep convolutional neural network (DCNN). First, a dataset with complex data sources is established, which contains PD experiments data, substation live detection data and inference data. During the PD experiments, data are acquired from five types of artificial defect models on a real gas insulated switchgear (GIS) platform, using different PD detection instruments. Substation live detection data are collected from the running GIS in more than 30 substations, using two types of portable PD detection devices. Typical inference data in PD detection are also employed for algorithm validation. Second, a DCNN based PD pattern recognition method is presented. In the proposed method, all of the PD data are normalized into a uniform format of phase resolved pulse sequence (PRPS). A DCNN model is employed to automatically extract the features of a complex data set. The results are obtained by a Softmax classifier. Third, the DCNN based PD pattern recognition method is applied to the dataset with complex data sources and achieves 89.8% accuracy. The back-propagation neural network (BPNN) and support vector machine (SVM) methods with traditional statistical features are compared with the developed method. The result shows that accuracy is improved by the method proposed in this paper. With the enlargement of the data set and a more complex data sample, the improved value will further increase, thus the proposed method is more suitable for the engineering application of a big data platform.

GIS partial discharge pattern recognition via deep convolutional neural network under complex data source

Currently, the cleaning process for power equipment monitoring data is cumbersome and often leads to loss of information. To address these problems, a data cleaning method based on stacked denoising autoencoder (SDAE) networks is proposed in this paper. SDAE networks have a strong ability to denoise and restore corrupted data and have a strong feature extraction capability. The status monitoring data of equipment under normal conditions are trained by SDAE to obtain the cleaning parameters and the reconstruction errors. An upper threshold of the reconstruction errors obtained from training samples is determined through Kernel density estimation. A tolerance window width is added to achieve rapid anomaly detection. The abnormal data are classified as outliers, missing data, or fault status data according to the relationship between the reconstruction error and the threshold and between the duration of abnormal data and the tolerance window. To verify the effectiveness of the proposed method, the SDAE model is used to process the data for the dissolved gas concentration in transformer oil and the temperature of the transmission line. The results show that the proposed method can effectively identify and repair outliers and missing information. The model can perform rapid anomaly detection when the equipment is running abnormally.

Cleaning Method for Status Monitoring Data of Power Equipment Based on Stacked Denoising Autoencoders

The purpose of analyzing the dissolved gas in transformer oil is to determine the transformer’s operating status and is an important basis for fault diagnosis. Accurate prediction of the concentration of dissolved gas in oil can provide an important reference for the evaluation of the state of the transformer. A combined predicting model is proposed based on kernel principal component analysis (KPCA) and a generalized regression neural network (GRNN) using an improved fruit fly optimization algorithm (FFOA) to select the smooth factor. Firstly, based on the idea of using the dissolved gas ratio of oil to diagnose the transformer fault, gas concentration ratios are also used as characteristic parameters. Secondly, the main parameters are selected from the feature parameters using the KPCA method, and the GRNN is then used to predict the gas concentration in the transformer oil. In the training process of the network, the FFOA is used to select the smooth factor of the neural network. Through a concrete example, it is shown that the method proposed in this paper has better data fitting ability and more accurate prediction ability compared with the support vector machine (SVM) and gray model (GM) methods.

https://www.mdpi.com/1996-1073/11/1/225/pdf

Prediction of Dissolved Gas Concentrations in Transformer Oil Based on the KPCA-FFOA-GRNN Model

The state of transformer equipment is usually manifested through a variety of information. The characteristic information will change with different types of equipment defects/faults, location, severity, and other factors. For transformer operating state prediction and fault warning, the key influencing factors of the transformer panorama information are analyzed. The degree of relative deterioration is used to characterize the deterioration of the transformer state. The membership relationship between the relative deterioration degree of each indicator and the transformer state is obtained through fuzzy processing. Through the long short-term memory (LSTM) network, the evolution of the transformer status is extracted, and a data-driven state prediction model is constructed to realize preliminary warning of a potential fault of the equipment. Through the LSTM network, the quantitative index and qualitative index are organically combined in order to perceive the corresponding relationship between the characteristic parameters and the operating state of the transformer. The results of different time-scale prediction cases show that the proposed method can effectively predict the operation status of power transformers and accurately reflect their status.

Dai Jiejie

Papers

Dissolved gas analysis of insulating oil for power transformer fault diagnosis with deep belief network

GIS partial discharge pattern recognition via deep convolutional neural network under complex data source

Cleaning Method for Status Monitoring Data of Power Equipment Based on Stacked Denoising Autoencoders

Prediction of Dissolved Gas Concentrations in Transformer Oil Based on the KPCA-FFOA-GRNN Model

Power Transformer Operating State Prediction Method Based on an LSTM Network