Lithium battery health state estimation method based on adaptive migration

The invention discloses a lithium battery health state estimation method based on adaptive migration. The method comprises the following steps: giving a degradation data set of an existing complete battery circulating to failure and partial degradation data of the battery to be predicted online under an unknown working condition; health indexes having high correlation with SOH in a constant voltage charging stage are integrated, and a sliding time window technology and a normalization method are used for preprocessing the integrated data; transform is used as a main body, a plurality of adaptive modules are added to extract advanced degradation characteristics of batteries under different working conditions, and an adaptive fusion strategy of a multi-head attention mechanism is adopted to effectively combine outputs of all the adaptive modules; according to the prediction loss, weighting and integrating the output of all SOH sub-estimators as a final SOH estimated value, and then carrying out optimization tra