Abstract: The prediction of the health status of electric heavy truck batteries is essential to ensure the reliability and safety of battery operation and reduce maintenance and service costs. In this study, a new electric heavy truck battery health prediction method based on Transformer is proposed. In the part of data pre-processing, the charging and discharging states of the data are distinguished, and the PCA algorithm is used to eliminate redundant feature information to minimize the calculation burden of the model. After pre-processing, the Transformer model is introduced to learn the deep correlation in the battery timing data. Experiments were carried out on the real vehicle battery data collected in the cloud, and the experimental results showed that the proposed method achieved optimal results on RE, MAE and RMSE, and the model could accurately predict the health of electric heavy truck batteries.