A Data-Driven Fault Prediction Method for Power Transformers

TLDR: In this article, a data-driven fault prediction method for oil-filled transformers, combining association rules, gray prediction model and random forest algorithm to achieve fault mode and location prediction through four stages: feature parameter sets selection, fine-grained feature parameters selection, feature parameters trend prediction and fault classification.

Abstract: The safe and reliable operation of the power system is an important guarantee for the steady and rapid development of social economy. The safe operation of transformer equipment is the basis for the stability and reliability of the power system, and it is of great significance to the whole system to take effective measures to make accurate predictions of abnormal conditions or faults inside the transformer. Therefore, this paper proposes a data-driven fault prediction method for oil-filled transformers, combining association rules, gray prediction model and random forest algorithm to achieve fault mode and location prediction through four stages: feature parameter sets selection, fine-grained feature parameters selection, feature parameters trend prediction and fault classification. Based on the actual data collected, the accuracy and effectiveness of the method are verified by two groups of experiments, which is beneficial to the practical application of engineering, and to a certain extent improves the automation level of equipment maintenance.