Accurate estimation of battery state including state of charge (SOC) and state of health (SOH) plays an important role in improving the performance and lifespan of battery packs in electric vehicles. In this paper, a method based on a simplified battery EM, which is optimized by BP neural network, was proposed to estimate the battery SOH and SOC. The simplified battery EM called single particle model (SPM) is employed to obtain the number of recyclable lithium ions at different aging degrees and the lithium-ion changes of solid particles with different loaded currents. The parameters of the SPM are identified by an algorithm hybridizing Coyote Optimization Algorithm (COA) and Grey Wolf Optimizer (GWO), abbreviated to HCOAG for its great balance in operational speed and optimization accuracy. To improving the speed and accuracy of model parameter identification, the rang of the number of recyclable lithium ions is narrowed by battery capacity which is estimated by the BP neural network. The accuracy and robustness of the proposed method are systematically evaluated, and with the mean absolute percentage errors of promising battery capacity and SOC prediction results are less than 2% and 1.8%, respectively.

Accurate battery power capability prediction can contribute to reliable and sufficient utilization of the battery to absorb or deliver a certain amount of power within its safe operating area. The power capability of a battery is a finite quantity that is limited by the electrochemical reaction properties occurring inside the battery. Note that the instantaneous available power of the battery is strongly related to the surface lithium concentration of solid particles in battery electrodes, but their relationship has not been explored sufficiently yet. This paper proposes a novel method for battery instantaneous available power prediction using a practical physical limit (i.e. lithium concentration limit) rather than the limits of macroscopically observed variables, such as the cell terminal voltage and current, thus providing a direct insight into electrochemical processes inside batteries. The surface lithium concentration of the solid particle is derived from a simplified battery electrochemical model, and a relationship between battery instantaneous available power and surface lithium concentration is quantified for the power capability prediction. Promising results with small forecast errors can be achieved for battery charging and discharging at different cell aging levels and ambient temperatures, which highlights the superior accuracy and robustness of the proposed method.

In this study, we introduce a novel denoising transformer-based neural network (DTNN) model for predicting the remaining useful life (RUL) of lithium-ion batteries. The proposed DTNN model significantly outperforms traditional machine learning models and other deep learning architectures in terms of accuracy and reliability. Specifically, the DTNN achieved an $R^2$ value of 0.991, a mean absolute percentage error (MAPE) of 0.632%, and an absolute RUL error of 3.2, which are superior to other models such as Random Forest (RF), Decision Trees (DT), Multilayer Perceptron (MLP), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Dual-LSTM, and DeTransformer. These results highlight the efficacy of the DTNN model in providing precise and reliable predictions for battery RUL, making it a promising tool for battery management systems in various applications.

This paper presents a model-predictive sliding-mode control (MPSMC) scheme for a three-phase ac/dc converter to achieve better stability and dynamic performances. In the conventional model-predictive control method, a proportional-integral (PI) controller is used to generate the active power reference. This traditional model-predictive PI control (MPPIC) scheme, however, produces a large overshoot/undershoot, a long settling time, and a large steady-state error under disturbances. To overcome these deficiencies, a sliding-mode controller is employed to replace the PI controller. Since the control law and the controller are designed based on the system model, the proposed MPSMC scheme can reduce the effects of unexpected disturbances, such as the output voltage demand and the resistance load variations. Both methods have been simulated in MATLAB/Simulink during various disturbances. Compared with the performances of MPPIC, the results obtained from MPSMC show that the settling time of the dc voltage can be minimized by about 91%, and the overshoot can be eliminated from 9.13% during the steady-state progress. The active and reactive power from MPSMC can also be controlled to the desired values, respectively, with a much smaller overshoot/undershoot and a faster response speed. Similar dynamic improvements can be achieved with MPSMC when the dc voltage demand varies. The simulation results are validated by experimental results.

Safe and efficient battery operations necessitate the technique of accurate state of power (SOP) capability prediction for lithium-ion batteries. From the perspective of electrochemical mechanisms, the battery SOP depends strongly on the status of battery electrochemical reaction processes and associated variables, especially the lithium-ion concentrations in solid phases. Battery electrochemical models (EMs) are capable of depicting dynamics inside the battery with high fidelity, but most EMs fail to capture the temperature dependences of model parameters, thereby limiting the model accuracy and their application. This paper proposes an EM-based SOP prediction method considering temperature effect, in which battery peak currents and power capabilities are mainly determined by the intercalated or de-intercalated number of lithium-ions from solid particles, and the temperature dependences of battery EM parameters are investigated to enhance the robustness of the model. With the physical limits of lithium-ion concentrations in solid particles, the grey wolf optimizer (GWO) algorithm is applied to seek the peak currents and power capabilities within different predictive time horizons. The accuracy and robustness of the proposed method are evaluated systematically, and promising SOP prediction results with most of the mean absolute percentage errors of less than 4.0% can be achieved under various temperatures.

Lithium-ion batteries are generally regarded as a leading candidate for energy storage systems. The safe and reliable operation of lithium-ion batteries depends largely on accurate estimation of the state of charge (SOC), which requires an accurate battery model. Bearing strong mechanisms, the electrochemical model (EM) can mimic the battery dynamics with high fidelity, and thus the EM-based methods can produce more reliable SOC estimates. This paper proposes a novel EM-based SOC estimation method for lithium-ion batteries from the electrochemical mechanism perspective. Firstly, a single particle model is employed to gain a direct insight into the electrochemical reactions inside the battery, and it is found that the model output voltage and SOC are strongly related to the lithium-ion concentrations of solid phases. A simple negative voltage feedback module is then applied to observe the voltage error between the cell referenced terminal voltage and the model output voltage. To eliminate the voltage error and achieve a precise estimate, a quantitative relationship between the voltage error and corrected amount of lithium-ion concentrations is deduced based on the Nernst equation. The performance of proposed method has been systemically evaluated under different operating conditions, including various charging and discharging current rates, erroneous initial SOCs, and cell aging levels. Although an erroneous initial SOC of 50% is applied to the proposed algorithm, promising estimates with the mean absolute errors of 0.22% and 1.35% can be still achieved under the constant and dynamic loading conditions, respectively.