The application of artificial intelligence methods for the interpretation of dissolved gas analysis results in transformer oil

Abstract

The growing complexity of power grids driven by the integration of renewable energy sources, electric vehicles, and climate change has increased the importance of reliable maintenance of power transformers (PT). Transformer failures can severely impact electricity generation, transmission, and distribution. Accurate condition assessment, especially through Dissolved Gas Analysis (DGA) of transformer oil, is crucial for early fault detection. This paper presents an overview of recent developments in the application of artificial intelligence (AI) methods for interpreting DGA results to improve diagnostic accuracy. Various supervised, unsupervised, and hybrid approaches are analyzed, with a focus on their capabilities, advantages, and limitations. The paper also discusses two representative case studies from the literature to illustrate the practical implementation and effectiveness of AI-based diagnostics. AI significantly enhances the reliability and automation of transformer condition monitoring, particularly in borderline or uncertain cases, thereby promoting the adoption of web-based and e-commerce preventive maintenance solutions.