A battery state of health estimation method based on distribution alignment adaptive adjustment

Abstract

The application discloses a battery state of health estimation method based on distribution alignment adaptive adjustment, comprising the following steps: step 1, a battery state of health estimation model comprising a feature extraction module and a state of health prediction module is constructed under a source domain working condition; the feature extraction module receives battery operation data under the source domain working condition and battery operation data under a target domain working condition respectively, and outputs a source working condition feature representation set and a feature representation set of unmarked samples of the target working condition; the state of health prediction module receives and outputs a source working condition battery state of health estimation set; step 2, a feature distribution difference measure between and is calculated, and a feature distribution alignment loss is constructed; step 3, according to the determined optimal distribution alignment weight coefficient, a joint optimization objective function is constructed, optimal parameters of the feature extraction module and optimal parameters of the state of health prediction module are obtained through convergence, and a battery state of health estimation model suitable for the target domain working condition is constructed according to and.

Images