Fault Diagnosis Method of Power Transformer Based on Historical Case

Power transformer is an indispensable key component in power systems. The accurate and reliable fault diagnosis is crucial to ensure the safe and stable operation of the power grid. Traditional diagnosis methods suffer from issues such as rule conflicts, high data quality requirements, and a lack of historical case analysis, leading to suboptimal performance in practical applications. Therefore, this paper proposes a historical case-enhanced oil chromatography diagnostic method. Firstly, building upon multiple improved ratio methods and incorporating fuzzy processing, the method calculates the fault pattern membership degrees. Secondly, a neural network is employed to analyze gas content data and historical case texts, determining the weights for the ratio methods. Lastly, based on the weights and the membership degrees of various ratio methods, the final fault probability is computed. Experimental validation demonstrates that the proposed method effectively provides equipment diagnostic results, showing an improvement of 8.21% to 24.25% in F1 values compared to traditional methods.