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 be 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. Firstly, PD testing for five types of artificial defects in Ethylene-Propylene-Rubber (EPR) cables was carried out in the High Voltage (HV) laboratory to generate signals containing PD data. Secondly, 3500 sets of PD transient pulses were extracted and then 33 kinds of PD features were 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. The 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.

The paper shows that open-loop noise reduction techniques can be effective in separating partial discharge (PD) signals from noise. Wavelet-based denoising through the use of thresholding of wavelet coefficients provides near optimal noise reduction for many classes of signals including PD signals in noise. The effectiveness of wavelet transform-based noise reduction depends on selection of an appropriate wavelet basis function as well as on careful selection of threshold function and levels.

Optimal selection of features of Partial Discharge (PD) signals recorded from defects in High Voltage (HV) cables will contribute not only to the improvement of PD pattern recognition accuracy and efficiency but also to PD parameter visualization in HV cable condition monitoring and diagnostics. This paper presents a novel Random Forest (RF)-based feature selection algorithm for PD pattern recognition of HV cables. Firstly, the experimental data acquisition and feature extraction processes are introduced. PD signals were captured from faults created in a cable to obtain the raw PD data, then a set of 3500 transient PD pulses and a set of 3500 typical interference pulses were extracted, based on which 34 PD features were extracted for further processing. Furthermore, 119 two-dimensional features and 1082 three-dimensional features were generated. The paper then discusses the basic principle of the RF algorithm. Finally, RF-based feature selection was implemented to determine the optimal features for PD pattern recognition. The results were obtained and evaluated with the Back Propagation Neural Network (BPNN) and Support Vector Machine (SVM). Results show that the proposed RF-based method is effective for PD feature selection of HV cables with the potential for application to additional HV power apparatus.

On-line Partial Discharge (PD) monitoring is being increasingly adopted in an effort to improve asset management of the vast network of MV and HV power cables. This paper presents a novel method for autonomous recognition of PD patterns recorded under conditions in which a phase-reference voltage waveform from the HV conductors is not available, as is often the case in on-line PD based insulation condition monitoring. The paper begins with an analysis of two significant challenges for automatic PD pattern recognition. A methodology is then proposed for applying the K-Means method to the task of recognizing PD patterns without phase reference information. Results are presented to show that the proposed methodology is capable of recognising patterns of PD activity in on-line monitoring applications for both single-phase and three-phase cables and is also effective technique for rejecting interference signals.

Partial discharge (PD) testing of high voltage (HV) cables and cable accessories has been implemented predominantly using high frequency current transformers (HFCTs) as PD sensors. PD currents initiating at PD sources are coupled onto cable conductors and travel away from the PD sources and will be detected by HFCTs installed at cable terminations. In this paper, based on combining finite-difference time-domain (FDTD) modeling and transfer function theory, a hybrid modeling approach is proposed to investigate the processes of PD coupling and detection involved in HFCT-based PD testing of HV cables. This approach allows exciting a PD event anywhere in FDTD models of the cables and predicting output from HFCTs some distance away. Implementation of the method is illustrated using an 11 kV XLPE cable. Moreover, a “direct measurement” method to obtain original PD pulses as the excitation source waveform is presented. The modeling approach introduced here will facilitate studies on the relationship between measured PD signals and those excited at PD sources, which can potentially give useful insight into the basic mechanisms behind PD detection in cables.

Partial discharge (PD) diagnostics is the most widely used tool to assess the insulation condition of insulated power cables which facilitates informed maintenance planning leading to extended service life of ageing assets. Time domain reflectometry (TDR) using a single ended or double ended approach is the most widely used method for locating PD sources. The success of the single ended method is dependent upon cable network design. However, by monitoring PDs at both ends of the cable, i.e. double-ended PD monitoring, higher accuracy of PD location can be achieved with a higher success rate. The double ended method is not widely used due to its complex system design, time synchronization and communication requirement between measurement units. This paper proposes a double ended PD location system which triggers on the predefined one pulse per second (1PPS) signal obtained from a global positioning systems (GPS) using novel time based triggering logic (TBTL) implemented in field programmable gate arrays (FPGA). This system ignores non-useable (not synchronized) data caused by flywheel 1PPS from GPS receiver due to any short-term loss of satellite signals which eventually reduces the PD location accuracy. Furthermore TBTL also ignores spurious triggering pulses radiated from noise sources within the substation. With the use of a communication link between two ends of the cable provided via mobile broadband together with TBTL, eliminated the acquisition of non-useable(not synchronized) data. Based on laboratory tests and on-site measurements PD location accuracy of less than ± 10 m can be achieved. The system design, laboratory tests and on-site measurements are discussed.