To address the issues of low accuracy and poor robustness in traditional SOC and SOH estimation models, an improved model, EKF-HInformer, is proposed based on the Extended Kalman Filter (EKF) and the deep learning model Informer. This model enables real-time and accurate estimation of the State of Charge (SOC) and State of Health (SOH) of electric vehicle batteries. First, the EKF algorithm is used to normalize the real-time battery data, and the adaptive gain factor is adjusted to reduce noise fluctuations, enhancing the performance of EKF data filtering. Then, the Informer network model is used to intelligently estimate the normalized battery data. To reduce the bias in attention weights caused by outliers or abnormal values, the Hampel algorithm is applied to optimize the Informer model, improving the feature learning ability of the multi-head probabilistic sparse self-attention mechanism. Finally, the filtered data is fed into the HInformer network to estimate real-time SOC and SOH. Experiments are conducted using battery datasets from the University of Oxford and the University of Maryland. The results show that the estimation accuracy for SOC and SOH exceeds 99.5%, with RMSE less than 1% and MAXE less than 0.5%. Compared to traditional Informer, Transformer, and LSTM models, this model is faster and more accurate in estimating SOC and SOH, demonstrating superior robustness and generalization ability.