The state-of-charge and health prognosis are important factors for electric vehicles. The long short-term memory (LSTM) is used to estimate battery states, and it attracts a lot of attention. However, the traditional LSTM network has limited ability of feature extraction and battery states prediction in long time series. To solve this problem, an integrated attention mechanism is proposed to improve the performance of bidirectional LSTM networks, and multiple-dimensional temperature compensation is proposed to enhance prediction under changing temperatures. In experiments, the maximum errors of the proposed method are less than 1%, and the accuracy is improved by 9.39% and 22.36% at two current conditions. In the battery health prognosis, the proposed method improves the accuracy by 21.45% compared with that of bidirectional LSTMs.

For power batteries used in the electric vehicle, accurate state-of-charge (SOC) and state-of-health (SOH) are important. Meanwhile, the current rate in actual working conditions often varies dramatically. However, the influence of current rates on voltage and battery states estimation has been neglected for a long time. To solve this problem, in this article, a hierarchical estimation model considering the current rate is proposed. The fractional-order model is used in the battery modeling, the data-driven method is used to identify parameters, and a multiscale dual extended Kalman filter (DEKF) is used in estimation of battery states. For SOC estimation, the proposed method improves the accuracy by 35.8% and 36.5% compared with traditional DEKF under two conditions. For SOH estimation, the proposed method improves the accuracy by 34.8% and 43.1% under two current conditions.

Accurate estimation of state-of-charge (SOC) and state-of-health (SOH) is extremely important for the state diagnosis of power batteries, which is related to the energy efficiency and safety of electric vehicles. However, in order to represent the signal noises of sensors, the most commonly used method based on Kalman filter introduces the random Gaussian noise into the estimation, which causes the uncertainty of the estimation results. In this article, the multithread dynamic optimization method is proposed to solve the problem. In addition, the fractional-order model and the unscented Kalman filter are used in SOC estimation. The Gaussian linear models based on parameters of six commonly used open-circuit-voltage models are proposed to estimate SOH. Finally, the dynamic stress test current condition and four lithium-ion batteries are implemented to verify the effectiveness of the proposed method in the experiment. For SOC estimation, root-mean-square error (RMSE) of the proposed method is 0.098 and the average value of the six models is 0.123, thus the proposed method improves the estimation accuracy by 20.32%. For SOH estimation, we compare the smallest RMSE of the six models and that of the proposed method for four experimental batteries, thus the average improvement of accuracy is 25.44%.

Due to the complexity of the actual operating conditions of lithium-ion batteries, accurately estimating the state-of-health (SOH) of them often requires a significant amount of battery data, but most of the current SOH estimation methods lack generalisability. To address this issue, this article proposes a meta-learning SOH estimation method, which combines the meta-learning model with the CNN-LSTM model to improve the generalization of lithium-ion battery SOH estimation. It not only possesses better generalization ability, but also has higher estimation accuracy. In addition, regardless of the four different types of CALCE datasets or lithium-ion battery datasets in the laboratory, the maximum root mean square error and mean absolute error of the proposed method is 2.31% and 2.03%, which indicates the good performance of the proposed method for SOH estimation. Compared with two prevalent deep learning methods, this method enhances the estimation accuracy by an average of 25% across different battery data.

For lithium-ion batteries used in the electric vehicles, accurate prediction of capacity and remaining useful life online is extremely important. However, most of the researches focus on the prediction accuracy but neglect the complexity of the test environment, which makes many methods show poor robustness in application. To solve the problem, in this paper, we firstly introduce the simultaneous input and state estimation (SISE) algorithm into the online prediction of state-of-charge (SOC) and capacity, and combine the Gauss-Hermite extended particle filter (GH-EPF) to predict the remaining useful life. By setting different gradients of state noises in experiments, the proposed algorithm demonstrates the best accuracy and robustness in comparison with other algorithms. Through the two-factor authentication in simulations, the maximum error of capacity estimation is 35 mAh. For the prediction of remaining useful life, the minimum relative error of the proposed method is 0.4%. Therefore, the high accuracy and strong robustness of the proposed algorithm are verified.

For power batteries used in the electric vehicle, accurate state-of-charge estimation is important. However, as one of the most commonly used estimation method, the least square support vector regression is hard to balance the accuracy and efficiency. To solve this problem, the fuzzy information granulation based on asymmetric Gaussian membership function is proposed to improve the utilization efficiency of the effective data and enhance the prediction accuracy of battery states. In addition, the performance of the proposed method is compared with that of other commonly used membership functions. In experiments, the dynamic stress test condition and the urban dynamometer driving schedule condition are used to verify the effectiveness. Compared with the most commonly used triangle membership function, the proposed method improves the accuracy of estimation by 6.47% and 2.18% under two current conditions, respectively.