Solving faults in today's complex power systems consisting of distributed generators, microgrids, and renewable resources with traditional methods is a challenging task. This paper examines recent developments in artificial intelligence (AI) studies for power system fault diagnosis. The usefulness of AI approaches in transformers, transmission lines, and rotating machines, which are the most frequently utilized high-voltage power system elements, is highlighted. A total of 60 studies published in Web of Science indexed journals after 2020 are systematically reviewed, with 20 studies analyzed for each of the three power system components. The reviewed methods include convolutional neural networks, long short-term memory, ensemble machine learning (ML), transfer learning, and hybrid deep learning (DL) architectures applied to diverse datasets including dissolved gas analysis measurements, vibration signals, thermal images, and phasor measurement unit data. The reported classification accuracies across the reviewed studies range from 85% to 100%, with DL-based approaches consistently outperforming traditional ML methods. Data editing, parameter adjustment, and model modification suggestions are also offered to improve existing AI solutions. The results indicate that automatic, fast, low-cost, and high-accuracy fault analysis will be achievable in the near future through AI-based online solutions.
Aslan et al. (2026) studied this question.