Model training method and battery remaining life prediction method

Abstract

The invention provides a model training method and a battery remaining life prediction method. The training method comprises the following steps: acquiring first sample working condition data of a first sample vehicle; the first sample working condition data comprises multiple indexes, vehicle-mounted data corresponding to the multiple indexes and generation time of the vehicle-mounted data; the various indexes comprise a battery state index, an environment index and a driving habit index; constructing a first sample working condition matrix according to the first sample working condition data; the first sample working condition matrix represents the distribution condition of the vehicle-mounted data under the combined influence of multiple indexes; training the initial multi-branch auto-encoder model by using the first sample working condition matrix to obtain a target multi-branch auto-encoder model; the target multi-branch auto-encoder model comprises a target encoder; and taking the target encoder as a feature vector extraction model. Based on the above method, the feature vector extraction model can be constructed, the feature vector of the input data can be accurately extracted, and a data basis is provided for subsequent residual life prediction.

Images