IGBT (Insulated Gate Bipolar Translator) life prediction method of low data sample

The invention discloses an IGBT life prediction method of a low data sample, and the method comprises the steps: carrying out the fitting of a degradation model of selected input data, fitting a degradation function, and constructing a state equation and an observation equation; according to the method, coefficient adjustment of an observation equation is completed by replacing prediction data with measured data at any time, a proposed target error function is utilized to optimize a particle filter, the weight of the particle filter is adjusted through optimal parameter searching to improve prediction precision, the convergence speed is high, the problem that IGBT life prediction needs huge data as a sample support is solved, and the prediction precision is improved. The method solves the problem that particle filter precision is reduced due to particle degeneration, can complete IGBT service life prediction for one module only through a small amount of data, can be suitable for prediction of various differen