Dissolved gas analysis (DGA) is crucial for identifying incipient failures in transformers by analyzing gas concentrations due to degradation. However, its high cost and time-consuming nature limit practical use. To address this, a metal-oxide semiconductor based electronic nose (E-nose) is utilized in this study to detect gases in transformer oil, including hydrogen (H ₂ ), methane (CH ₄ ), ethane (C ₂ H ₆ ), ethylene (C ₂ H ₄ ), and acetylene (C ₂ H ₂ ). Machine learning techniques are integrated with the E-nose system to enhance classification performance. Experimental results using artificially contaminated mineral oil samples demonstrate promising accuracy in gas classification. Initially, without feature reduction, the F1 score was 0.2972. Feature ranking increased the F1 score to 0.7956, and after implementing dimensionality reduction, it further improved to 0.9313. Subsequently, the combination of support vector machine and genetic algorithm was employed for sensor selection, achieving an F1 score of 0.9869. Among the combinations of 2, 3, and 4 sensors, MQ 8 and TGS 2612 consistently showed the best F1 scores, with TGS 813 and TGS 2611 also contributing significantly. This innovative approach suggests a potential solution for transformer oil condition monitoring, offering a rapid, simple, and cost-effective alternative to traditional DGA analyses. By combining E-nose technology with machine learning, this method holds promise for facilitating routine measurements and ensuring the reliability and efficiency of transformer operations.

Acoustic emission detection is a non-intrusive technique for monitoring transformer conditions by capturing emissions from partial discharges (PD), hotspots, and noise. Machine learning (ML) has been widely used in PD diagnostics in power transformers. While effective in managing complex data characteristics, traditional ML algorithms cannot quantify uncertainty in classification. This study uses Bayesian networks integrated with a classification algorithm to quantify uncertainty in acoustic signal classification. Bayesian deep learning (BDL) and ensemble models are the two commonly used techniques for uncertainty quantification. BDL models with multiple Bayesian layers are more prone to convergence issues with difficulty interpreting the sources of uncertainty. The performance of the ensemble method is based on model diversity with a higher risk of overfitting and does not offer insight into aleatoric uncertainty. The present work proposes architectures with a single Bayesian layer to quantify both aleatoric and epistemic uncertainties in acoustic signal classification. A Bayesian CNN layer combined with an ensemble architecture demonstrates comparatively higher performance among the considered Bayesian architectures. Raw acoustic signals are transformed into spectrogram images to enhance feature representation, capturing time-frequency characteristics. The proposed method is assessed using laboratory and field-measured data, demonstrating significant improvements in estimating uncertainty in classification.

With the increase in transmission voltage level, and the guidelines of the International Commission on Non‐Ionising Radiation Protection (ICNIRP), the effects of non‐ionising radiation on biological elements at high‐voltage substations have gained significant importance. The electric field ( E ‐field) distribution in an upcoming 765 kV substation in the Indian subcontinent is computed using the charge simulation method (CSM). CSM is used to model the 765 kV bays, transmission lines, buses, and ground wires in the substation. The three‐dimensional (3D) E ‐field is calculated through the superimposition of E ‐fields obtained in two orthogonal planes using infinite‐line charges. This proposed method of using infinite‐line charges gives realistic results. The simplistic model using infinite‐line charges greatly reduces the complexity of the CSM‐based model (due to the reduced number of charges) apart from increasing the CSM‐based model accuracies. This fact has been demonstrated by comparing these results with those of CSM‐3D‐model of a detailed bay model (including major equipment and associated support structures). The complex‐charge‐based CSM helps in computing the root mean square value of the E ‐field at a point, directly, as per ICNIRP guidelines. This RMS value of the E ‐field is compared with the occupational exposure reference value prescribed in the ICNIRP guidelines.

Condition monitoring of the power transformer is important in extending the lifetime of the transformer through timely maintenance. Acoustic-emission (AE)-based partial discharge (PD) monitoring in a transformer is a commonly used technique for assessing the PD levels inside the transformer. The classification of the type of PD or localization of PD generally involves using multiple acoustic sensors, and recently, deep learning techniques have been widely used with the acquired PD data for classification. In the present work, a single-channel synchrosqueezing transform (SST)-based FastIVA blind-source-separation (BSS) technique in the time-frequency (TF) domain is proposed for the separation of multiple PD signals. SST, along with singular value decomposition (SVD), is used to identify the number of sources. The SST-FastIVA-based method is used for PD signal separation for the first time and it performs better than other separation techniques in noisy conditions. An SVM-based machine learning model is developed with original unmixed source signals for a classification task. The separated signals are tested using the developed model, and it attains an F1 score of 78.70% on the separated signals. Furthermore, the proposed methodology is used with the experimental data, and the F1 score obtained with the separated signal is 84.44%. The proposed single-channel (sensor) technique with BSS is a cost-effective method that can be used for the classification of multiple PD sources.

Numerical methods plays a vital role in computation of electric fields for the design of high voltage systems. For electric fields studies related to high voltage applications, the preferred numerical methods are charge simulation method (CSM) and method of moments (MoM). In the present study, the performance of these two numerical methods are analyzed by simulating sphere-plane gap arrangement (which has an analytical solution to compare). Through this example, CSM is found to be more accurate compared to MoM. Further, the effect of change in sphere-plane gap geometry on the electric field computed is reported and discussed.

An attempt is made to reduce the maximum value of rms E-field via transposition of conductors in a 765 kV substation with reference to the occupational exposure. The method is applied to an EHV substation which is under construction. The effect of transposing the bay conductors (alone) and the bus conductors (alone) on the E-field distribution (at 2 m above the ground) over the area of the substation is discussed. Also, the topology of substation conductors which results in the minimum value for the maximum of rms E-fields is obtained via simultaneous transpositions of bus and bay conductors. The area of the substation where the rms E-field exceeds the reference value of 10 kV/m (ICNIRP occupational exposure limit) is also reported for the existing topology. For this topology, the computed maximum value of the rms E-field over the area of the substation is 13.06 kV/m. This E-field is reduced to 7.999 kV/m with the best configuration of substation conductors obtained via transposition. Charge simulation method based generalized algorithm for obtaining the best topology (among all permutations) of substation conductors, which results in the minimum value for the maximum of rms E-field magnitudes over the substation area is also reported.