The defect development stages of power equipment follow different laws

and the measured data of defect development process before the fault is scarce. The artificial intelligence prediction method based on historical data series can not solve the prediction problem in the new stage. The statistical analysis method based on traditional reliability theory has good practicability. Based on the performance degradation track theory in reliability theory

combined with the data of dissolved gas in oil before the overheating failure of several on-site transformers

this paper builds a nonlinear Wiener degradation model for the development process of gas volume fraction by combining the degraded track model as a time-scale conversion function with the linear Wiener model. Moerover

the steps and methods for solving the model parameters are also given. The analysis results of the field data show that the developments of the main characteristic gas volume fraction before transformer overheating faults obey the exponential law

and the nonlinear Wiener degradation model can be constructed by using the exponential function as the time scale transformation function. The model has low requirements on the length and time interval of data series

can reflect the nonlinear trend and randomness of gas volume fraction development process at the same time

and has the interpretability based on reliability theory. Finally

this paper statistically analyzes the probability density distribution functions of each parameter in the Wiener degradation model of methane

ethylene and total hydrocarbon volume fraction in transformer oil with overheating fault. The research results provide a new method with good interpretation

high accuracy and low data requirement for gas concentration prediction and transformer overheating fault probability calculation.