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 Non-code 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.
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.
Polypropylene (PP) has special advantages in replacing conventional crosslinked polyethylene (XLPE) to be insulation material of power cable. In order to reveal the effects of nano-filler addition on ...PP, five kinds of experiments, namely TEM, DSC, breakdown strengths (BDs), space charge and dielectric measurement were investigated to evaluate the insulating properties of PP and its nanocomposites with surface-treated nano-MgO of different concentration. It is revealed that, with the addition of nano-MgO, favorable dispersibility of nano-filler and significant increase of crystallinity of polymer are observed. BDs under dc voltage increase apparently with loading of nano MgO, but when the nano concentration reaches higher than 1 wt%, the BDs have a slight decline compared with 1 wt%. It is clarified that space charge and electric field distortion are well restricted with the addition of nano-MgO, while this effect is not obvious with the nano concentration reaches 6 wt%. Permittivity has trends of increase with the rise of temperature at first and then decrease when the temperature reaches about 333 K, and the addition of nano-filler MgO could also decrease permittivity. When the nano-filler concentration reaches high, dielectric loss increases to a high level. Low nano-filler concentration in MgO/PP shows better electrical insulation properties compared with PP and high nano-filler concentration composites.
Power equipment operation and maintenance (O&M) requires plenty of domain knowledge to improve equipment security and power grid reliability. However, most knowledge is implicitly represented by the ...semantic text, which is hard to be comprehended by algorithms and restricts the intelligence level of equipment O&M. Therefore, this article proposes an event knowledge graph system to automatically extract event knowledge from O&M reports and represent the knowledge by an event knowledge graph (KG). First, a bidirectional long short‐term memory network (BiLSTM) combined with conditional random field (CRF) is employed to recognize the entities. Next, an attention‐based sequence‐to‐sequence model is proposed to detect multiple events and extract document‐level representation vectors. Then, a tree‐based table filling strategy is utilized to complete the artificially designed event tables. Finally, the proposed system renders the event tables into an event KG. Verified by the experiments, the proposed system outperforms the baselines with the F‐measure of 83.5% for event extraction and 94.2% for event detection. When dealing with coexisting events and scattering arguments, the F‐measure of the proposed system achieves 74.8% and 79.0%, respectively. Moreover, the proposed system has positive effects on intelligent retrieval, fault diagnosis, and condition assessment, verifying its potential value for equipment O&M.
Faults with line breaks occur more and more frequently and are difficult to detect by relay protection device in small-current grounding distribution networks. However, faults with line breaks are ...more likely to cause security accidents such as fires, overvoltages and electric shocks. Considering the low cost and high reliability of collecting data on the low-voltage (LV) side in distribution networks, this paper proposes a method for distinguishing between faults with line breaks and short-circuit faults on the medium-voltage (MV) side. Voltage amplitude and phase on the LV side of distribution transformer (DT) are analyzed through symmetrical component method. The amplitudes are used to identify two-phase faults with line breaks (TPFs-LBs) and two-phase short-circuit faults (TPFs-SCs); while the phases are selected to distinguish single-line-to-ground faults with line breaks (SLGFs-LBs) from single-line-to-ground faults (SLGFs). The effects of network parameters, operation conditions and external factors on the proposed method have been discussed. Simulation results verified the effectiveness of the proposed method.
Existing interturn fault detecting methods rely on winding impedance, winding current, and dissolved gases. They are effective only when the insulation is severely damaged. This paper proposes a ...novel detection method based on fusion analysis of electrothermal characteristics including winding currents, temperatures of four areas on the tank wall, top oil and ambient temperatures, which can identify the interturn fault at an early stage. When an incipient interturn fault occurs, the heat generated by the faulty turns is transferred to the oil and tank wall, leading to an increase in top oil and tank wall temperatures. Thus, the incipient fault can be detected by analysing these electrothermal characteristic parameters. Borrowing the idea of digital twin (DT), this method establishes a high‐fidelity simulation model to simulate the transformer electrothermal characteristics under different operating conditions. Afterward, an intelligent neural network is adopted to extract the quantitative relationship between the eight feature attributions and fault conditions. Finally, this neural network is utilized to detect the incipient interturn fault for the transformer entity. Case studies are conducted on a 100 kVA transformer with oil natural air natural (ONAN) cooling mode. The detection accuracy is improved by 68.5% compared to the winding current‐based method.
The manuscript proposes a novel method based on fusion analysis of electrothermal characteristics for detecting incipient interturn fault of ONAN transformers. This method relies on winding currents, ambient temperature, top oil temperature, and the temperatures in four areas of the tank wall, which can identify the early faults before interturn short circuit and arc discharge faults. The work in this paper provides a new idea for the early interturn fault diagnosis of transformers.
A novel algorithm based on L-shaped antenna array signal processing is proposed in this paper for the localization of partial-discharge (PD) sources in substations. The principle of estimation of ...signal parameter via rotational invariance techniques has been used for finding the direction of arrival (DOA) of signals. Third-order cumulants of signals are used in this algorithm, by which Gaussian white noises and periodic narrowband interference mixed in observed signals can be efficiently suppressed. Planar location of PD sources can be obtained by solving the intersecting point of two lines in DOAs. Therefore, solving nonlinear equations can be avoided. Besides, it is convenient to replace the observed signals with their envelopes in this algorithm. The proposed algorithm is used to process mixed signals with simulated ultra-high frequency (UHF) signals by electromagnetic-wave simulation software, Gaussian white noises of different signal-to-noise ratios, and fixed-frequency noises. The planar location of PD sources is obtained approximately. UHF signals collected in substations and their envelopes have proven to be suitable to locate PD sources effectively by the proposed algorithm as well. Therefore, the accuracy and feasibility of the proposed algorithm are proved.
Partial-discharging (PD) test is one of the most effective methods for insulation diagnosis. Because of the advantages of ultra-high frequency (UHF) electromagnetic wave, UHF signals radiated by PD ...have been widely used to locate faults on power equipment. Time delay estimation is not only on the basis of the PD localization algorithm, but is also the key in determining the localization accuracy. Currently, common methods are the threshold value method, energy accumulation method, and the cross-correlation analysis method, etc. However, field signals always include a variety of unknown noises, so common methods cannot be effective enough. This paper proposes time delay estimation algorithms based on four-order cumulant and bispectrum, and gives their computational realization. These algorithms have a prominent advantage of being insensitive to Gaussian noises with unknown correlation properties. The time delay of simulated UHF signals with Gaussian noises and fixed-frequency interference is estimated by the given algorithms. Numerical robustness of the algorithms is verified. Finally, time delays of field UHF signals received by UHF antennas are estimated, and then substituting the sequence of time delay estimations into the localization equations based on the time difference and 3-D coordinate of the PD source can be calculated accurately. The effectiveness and practicability of this algorithm are verified.
To monitor the insulation deterioration of power equipment and realize prompt fault warning systems in air-insulated substations, in this study, we propose a multiple signal classification (MUSIC) ...algorithm-based partial discharge (PD) localization method with an angle of arrival (AOA) and ultrahigh frequency (UHF)-received signal strength indicator (RSSI). Compared with traditional UHF time-difference-based techniques, this RSSI-based AOA localization method is a more economical solution. In addition, by comparing the measured RSSI vector to a prebuilt reliable reference data set, the MUSIC method can effectively locate the direction of the PD source with high accuracy. Compared with the method that directly determines the smallest RSSI values by several sensors, this method can accomplish localization by fewer sensors without impeding accuracy. Furthermore, the interpolation method was adopted to improve the precision of the relationship curve of AOA/RSSI, which it did with a limited number of sensors. Laboratory tests were conducted to verify the accuracy of the proposed method, and most of the localization errors were less than 1°, which indicates its potential application in the prompt identification of faults regarding the insulation deterioration of power equipment in substations.
Pin plays the role of fixing power equipment on the overhead line. Once it is missing, it will lay a serious hidden danger for the normal operation of the overhead line. In order to improve the ...efficiency of UAV patrol transmission line and improve the detection rate of pin defect of transmission line, this paper proposes a pin defect detection method based on cascaded convolution network. In view of the complex background of inspection image and the small size of pin, the overall detection method is divided into two parts: positioning and diagnosis. Firstly, all fastener positions including pins are located by the improved Faster-RCNN network, and then RetinaNet network is cascaded to diagnose the defects of the fastener. The difficulty of learning is decreased and the generalization ability is improved in this way. Finally the experiment shows that this method can effectively detect the pin defects in the UAV patrol image.