Search

Furan tests provide a non-intrusive and cost-effective method of estimating the degradation of paper insulation, which is critical for ensuring the reliability of power grids. However, conducting routine furan tests can be expensive and challenging, highlighting the need for alternative methods, such as machine learning algorithms, to predict furan concentrations. To establish the generalizability and robustness of the furan prediction model, this study investigates two distinct datasets from different geographical locations, Utility A and Utility B. Three scenarios are proposed: in the first scenario, a round-robin cross-validation method was used, with 75% of the data for training and the remaining 25% for testing. The second scenario involved training the model entirely on Utility A and testing it on Utility B. In the third scenario, the datasets were merged, and round-robin cross-validation was applied, similar to the first scenario. The findings reveal the effectiveness of machine learning algorithms in predicting furan concentrations, and particularly the stacked generalized ensemble method, offering a non-intrusive and cost-effective alternative to traditional testing methods. The results could significantly impact the maintenance strategies of power and distribution transformers, particularly in regions where furan testing facilities are not readily available.

Condition monitoring of high voltage apparatus is of much importance for the maintenance of electric power systems. Whether it is detecting faults or partial discharges that take place in high voltage equipment, or detecting contamination and degradation of outdoor insulators, deep learning which is a branch of machine learning has been extensively investigated. Instead of using hand-crafted manual features as an input for the traditional machine learning algorithms, deep learning algorithms use raw data as the input where the feature extraction stage is integrated in the learning stage, resulting in a more automated process. This is the main advantage of using deep learning instead of traditional machine learning techniques. This paper presents a review of the recent literature on the application of deep learning techniques in monitoring high voltage apparatus such as GIS, transformers, cables, rotating machines, and outdoor insulators.

This paper discusses the state of the art in the application of self-healing silicone-based materials for outdoor high-voltage insulation. Both the dynamic behavior of the dimethyl side groups of silicone rubber and the diffusion of a bulk siloxane to maintain low surface energy are respectively reported as intrinsic mechanisms responsible for the self-healing of silicone rubber. Localization, temporality, mobility, and the type of synthesis are the aspects defining the efficiency of the self-healing ability of silicone rubber. In addition, the deterioration of the self-healing ability with filler loaded into silicone rubber insulation housing composites is discussed. Taking the self-healing property into consideration among the other properties of silicone rubber insulators, such as tracking and erosion resistance, can be a useful design practice at the material development stage. Hydrophobicity retention, recovery, and transfer measurements are discussed as useful indicators of the self-healing ability of silicone rubber. Nevertheless, there remains a need to standardize them as design tests at the material development stage. The paper is intended to shed the light on the hydrophobicity recovery, a key material design parameter in the development of silicone rubber outdoor insulating composites, similar to the tracking and erosion resistance.

Condition monitoring of outdoor insulation systems is crucial to the integrity of distribution and transmission overhead lines and substations. The objective of this study is to use a commercial acoustic sensor along with artificial neural network (ANN), to classify different typical types of discharges in outdoor insulation systems. First, ANN was used to distinguish between five common electrical discharges that were generated under controlled conditions. Next, this approach was extended to include outdoor ceramic insulators. Three types of defects were tested under laboratory conditions, i.e. a crack in the ceramic disc, surface pollution discharge, and corona near the insulator surface. Both a single disc, and three discs connected in an insulator string were tested with respect to these defects. For both controlled samples and full insulators, a recognition rate of more than 85% was achieved.

Different types of classifiers for acoustic partial discharge (PD) pattern classification have been widely discussed in the literature. The classifier performance mainly depends on the measurement conditions (location and type of the PD, acoustic sensor position and frequency response) as well as extracted features. Recent research posits that features extracted by singular value decomposition (SVD) can exhibit the natural characteristics and energy contained in the signal. Though the technique by itself is not novel, in this paper, SVD is employed for PD classification in a revised way starting from data arrangement in Hankel form, to embedding the hypergraph-based features and finally to extracting the required set of optimal features. The algorithm is tested for various measurement conditions that include the influences of various PD locations and oil temperatures. The robustness of the algorithm is also tested using noisy PD signals. Experimental results show the proposed feature extraction method supremacy.

This paper uses a two-layered soft voting-based ensemble model to predict the interfacial tension (IFT), as one of the transformer oil test parameters. The input feature vector is composed of acidity, water content, dissipation factor, color and breakdown voltage. To test the generalization of the model, the training data was obtained from one utility company and the testing data was obtained from another utility. The model results in an optimal accuracy of 0.87 and a F1-score of 0.89. Detailed studies were also carried out to find the conditions under which the model renders optimal results.

One of the most promising techniques for condition monitoring of high voltage equipment insulation is partial discharge (PD) measurement using radio frequency (RF) antenna. Nevertheless, the accuracy of monitoring, classification, localization, or lifetime estimation could be negatively affected due to the interferences and noises measured simultaneously and contaminate the RF signals. Therefore, to achieve high accuracy of PD assessment, exploiting the denoising algorithms is inevitable. Hence, this paper seeks to introduce a new technique to suppress white noise, the most prevalent type of noise, especially for RF signals. In the proposed method, the ability of artificial neural network (ANN) in curve fitting is applied to denoising of different types of measured RF signals emitted from PD sources including ‘crack’, ‘internal void’, in the insulator discs and ‘sharp points’ from external hardware. The processes of denoising for named signals with the proposed method are carried out, and the obtained results are compared with the outputs of a wavelet transform-based method named energy conversation-based thresholding. In all tested signals, the proposed technique showed superior denoising capability.

The presented paper aims to establish a strong basis for utilizing machine learning (ML) towards the prediction of the overall insulation health condition of medium voltage distribution transformers based on their oil test results. To validate the presented approach, the ML algorithms were tested on two databases of more than 1000 medium voltage transformer oil samples of ratings in the order of tens of MVA. The oil test results were acquired from in-service transformers (during oil sampling time) of two different utility companies in the gulf region. The illustrated procedure aimed to mimic a realistic scenario of how the utility would benefit from the use of different ML tools towards understanding the insulation health index of their transformers. This objective was achieved using two procedural steps. In the first step, three different data training and testing scenarios were used with several pattern recognition tools for classifying the transformer health condition based on the full set of input test features. In the second step, the same pattern recognition tools were used along with the three training/testing scenarios for a reduced number of test features. Also, a previously developed reduced model was the basis to reduce the needed number of tests for transformer health index calculations. It was found that reducing the number of tests did not influence the accuracy of the ML prediction models, which is considered as a significant advantage in terms of transformer asset management (TAM) cost reduction.