Transformers are utilized in generation, transmission, and distribution power system network and face an enormous number of hazards during their course of operation. Frequency response analysis (FRA) is an inexpensive, accurate, and non-destructive technique to explore the transformer mechanical integrity very fast. However, FRA results interpretation is not being automated yet. This study introduces a new setup for FRA measurement that can assist to leave the conventional FRA data interpretation techniques and obtain smart interpretation. Hence, FRA setups and interpretation techniques are studied and formulated in this work. A new measurement technique is introduced and discussed in detail. Practical studies are performed over distribution and power transformers and FRA data are recorded for inter-disk fault. The analysis of fault severity that we achieve in this work is an advantage of the proposed measurement technique. In this regard, we employ techniques from machine learning and numerical analysis to train a predictive engine for smart interpretation of FRA data. It is revealed that the proposed intelligent technique is capable of interpreting, detecting, and classifying the transformer winding inter-disk fault and its severity. The new introduced FRA measurement setup is also able to support the online FRA data assessment.

Interpreting results of transformer frequency response analysis (FRA) is quite challenging. One of the common methods to summarize FRA data is to employ Statistical Indicators (SIs) over FRA spectra. However, SI-specific boundary conditions for various operational modes of transformers are left unexplored. The lack of such boundary conditions renders interpretation of SIs difficult and subjective. Here, in an attempt to find data-driven boundary conditions, first the conventional measurement setup of FRA technique is modified to emulate inter-turn winding short circuit. Then the boundary conditions of various SIs for normal, suspicious, and critical operational modes of transformers under fault are obtained. Nevertheless, the price of moving subjective boundaries to their objective data-driven counterparts is paid in an intrinsic uncertainty introduced by the process of data collection per se. In order to capture and quantify this uncertainty, a novel solution inspired by bolstered error estimation used in pattern recognition is proposed. In particular, the proposed method allows reporting the level of confidence that an observed magnitude of SI belongs to a specific operational mode. Having this confidence level is also warranted from an operational perspective because it enables utility operators to enhance the decision-making process and estimate the severity of transformer faulty conditions.

The current interpretation procedure for the transformer frequency response analysis (FRA) is practically based on either subjective visual inspection or an observed value of a statistical indicator (SIn) programed in commercial software. However, both solutions are unable to provide very detailed information on transformer conditions in the industry. The former approach highly depends on the experience of the technical personnel, while the latter suffers from uncertainties of SIn-based decision boundaries for classification. This study has specifically focused and introduced a new comprehensive FRA data interpretation and classification method. Evaluation of the proposed technique is performed by emulation of winding short-circuit faults in transformers via a parallel connected rheostat. The method is able to categorize the operational mode of the transformer into the normal, suspicious, and critical classes based on the cumulative effect of 12 frequently used SIns and then provide a confidence level (CL) associated with classification to support operators for making a precise decision. Bootstrap sampling technique is introduced for the first time in FRA data evaluation to simulate the effect of many experiments in silico and overcome the difficulty of limited available measured data. Particularly, in a novel approach, a large set of SIn-specific decision boundaries is established by bootstrap sampling of the real experimentally obtained boundaries. In addition, a new transformer condition assessment method that employs the cumulative effect of individual SIns is developed and examined. Classification results and the CLs in FRA evaluation have a substantial practical significance and will support on-site utility managers or engineers for further diagnosis and maintenance.

Frequency response analysis (FRA) is almost certainly the most efficient and well-established method for evaluating the mechanical integrity of transformer active part. FRA has a solid background in both industrial practice and academic research. Besides conventional visual analysis, a statistical analysis has been applied to the FRA data interpretation. The existing standards and recent studies have considered different statistical indicators (SIs) on an individual basis. However, the utility of each SI in its “silo” may lead to different and possibly contradictory decisions. Inspired by the bolstered error estimation technique used in pattern recognition, this article presents a new method that can utilize multiple SIs obtained from the FRA data to classify transformer operating conditions and estimate the level of confidence in decisions made. The probabilistic method proposed herein is an attempt to bridge the gap between these decision-making silos. The practical implementation of the proposed technique on distribution and power transformers revealed a reliable interpretation and classification results. At the same time, having an estimate for the level of confidence in the decision made by the method further helps engineers and utility operators make informed decisions and have a better understanding of the level of the transformer fault severity.

Outdoor high-voltage (HV) insulators are prone to surface contamination, increasing flashover risk and threatening power transmission systems reliability. Timely inspection is essential, but conventional visual inspections are costly, and inefficient for long transmission lines. An automated inspection using unmanned aerial vehicle images can be both efficient and low-cost. Given an appropriate training sample, a deep neural network such as YOLO (You Only Look Once) can be trained to localize insulators in diverse backgrounds and classify surface contamination. However, the wide variety of YOLO architectures proposed recently makes model selection challenging due to the accuracy-complexity trade-off. Rather than proposing a new model, this study focuses on selecting the “optimal” one. To this end: 1) a “laboratory” dataset of 15,000 insulator images with various surface contaminations is collected; 2) 21 YOLO architectures (YOLOv3–v11 and their variants) are trained, followed by a domain-specific (DS) model selection based on accuracy-complexity trade-offs; and 3) the selected DS-YOLOv11m and DS-YOLOv11n are fine-tuned on a small “industry” dataset of 356 images from a local HV substation to validate DS-specific pre-training in real-world scenarios. This work presents the first DS benchmarking of YOLOv3–v11 for insulator detection and contamination classification, resulting in a lightweight model optimized for real-time edge deployment. Evaluation of the fine-tuned models shows mAP@0.5 scores of 0.983 (DS-YOLOv11m) and 0.977 (DS-YOLOv11n). Although DS-YOLOv11n achieves slightly lower accuracy, it uses significantly fewer resources, reducing FLOPs by ~10.7 times. Finally, DS-YOLOv11n is deployed on a Raspberry Pi 5 to enable real-time contamination classification in edge-computing scenarios.

Regular monitoring of outdoor insulators is crucial to ensure the reliable functioning of the power grid. With recent progress in computer vision technologies, traditional manual and expensive visual inspections can now be replaced by automated analysis using images captured by unmanned aerial vehicles (UAVs). In such applications, a practitioner might opt to choose a state‐of‐the‐art object detection and classification deep learning architecture, including You Look Only Once (YOLO). The variety of existing YOLO architectures per se makes selecting the best application‐dependent YOLO model challenging. However, selecting the best architecture solely based on performance without considering the model complexity limits its deployment on resource‐limited embedded devices. Consequently, we conduct a rigorous, systematic model selection based on the performance–complexity trade‐off across 13 YOLO architectures to determine the most effective model for detecting common mechanical faults in insulators using images captured by UAVs. A dataset comprising 15,000 images of insulators, categorized into normal condition, bird‐pecking damage, cracks, and missing caps, has been compiled for training the models. Specifically, all considered YOLO architectures are compared using model complexity and the mAP@0.5:0.95. During the model selection stage, YOLOv8l proved to be the best model in terms of mAP@0.5:0.95, while YOLOv5n was the model of choice in terms of complexity at the expense of a slight reduction in performance. Alongside YOLOv8l and YOLOv5n, an “optimal” model (OP‐YOLO) was selected using a multicriteria decision‐making approach, balancing detection accuracy and computational efficiency. In particular, in terms of test‐set performance, YOLOv8l, YOLOv5n, and OP‐YOLO achieved 0.919, 0.901, and 0.896 mAP@0.5:0.95, respectively. Although YOLOv8l reported a higher mAP@0.5:0.95, YOLOv5n requires ∼20.9 times less memory and ∼40.2 times less floating‐point operations per second (FLOPs). Also, YOLOv5n outperforms the OP‐YOLO model, still requiring ∼12 times less memory and ∼19 times less FLOPs.