In accordance with the versatile characteristics of the operating state of a distribution transformer

based on the extraction and processing of multi-source data in power system

we establish a 10 kV real-time state assessment model that integrates external meteorological factors

equipment internal health and current operating conditions. Firstly

the random forest method based on Tomek links undersampling is used to predict the fault probability and the expected power outage duration under severe weather conditions. Then

the selected values for monitoring state parameters are used to calculate the real-time reliability of the distribution transformer by using Weibull distribution and maximum likelihood parameter estimation. A method for real-time failure probability caused by internal aging of the equipment is proposed by subjective and objective comprehensive weighting calculation. Finally

the model is applied to evaluate the comprehensive degradation degree of a 10 kV distribution in a certain area within 24 h. The results show that the undersampling processing based on Tomek links can significantly improve the prediction effect of power outage duration. The actual fault condition of the distribution is in accordance with the prediction result of the state evaluation model

and the validity of the model is verified. The research can provide a reference for the state evaluation of 10 kV distribution transformers.