Fault diagnosis method of insulating oil chromatography for transformer based on SSA-KELM

This article presents a method for diagnosing transformer faults utilizing machine learning. The original data consist of four types of dissolved gases in transformer insulation oil: H₂, CH₄, C₂H₆, and C₂H₄. After removing redundant data, the remaining data are used as fault sample inputs. The Kernel Extreme Learning Machine (KELM) model parameters are optimized using the Sparrow Search Algorithm (SSA). A fault diagnosis framework is developed in this paper, capable of identifying normal operating condition as well as several discharge fault conditions in transformers. Comparative experiments demonstrate that the SSA-optimized KELM-based method achieves higher diagnostic accuracy.