Tiezhou Wu

Early-stage micro-short circuits (MSCs) can remain hidden within the normal pack-voltage range because their leakage currents are much smaller than normal charging and discharging currents. This study proposes a charge self-baseline drift accumulation (C-SBDA) method to extract directional and persistent fault evidence from incomplete charging records. The framework combines data-quality control, pack-median referencing, reference-voltage-domain alignment, cell-specific historical self-baseline correction, and three complementary drift features: mean downward-drift intensity A, significant-coverage ratio P, and end-region drift E. The features are robustly standardized, fused into a segment-level score, and accumulated using an exponentially weighted moving average (EWMA) to generate a cross-segment warning score.The method was validated using 21 controlled charging segments from a six-cell battery pack and eight charging segments from a 30-cell real-vehicle battery pack. In the controlled experiment, the experimentally faulted cell first exceeded the warning threshold at Segment 12 and was continuously confirmed at Segment 13, providing an eight-segment advance relative to Segment 21. The warning score showed a strong monotonic association with independently measured leakage severity, with a Spearman correlation coefficient of 0.9121. In the vehicle dataset, the real-vehicle abnormal cell first crossed the threshold in C6, remained continuously confirmed in C7, and returned to the top rank in C8, demonstrating persistent early risk evidence under practical charging conditions. Ablation studies and 67 one-factor sensitivity tests further confirmed the roles of self-baselining, complementary drift features, and EWMA accumulation. C-SBDA requires neither SOC estimation, OCV curves, equivalent-circuit parameters, nor fault-sample training, enabling low-complexity risk screening and abnormal-cell localization.

Correctly estimating the State of Charge (SOC) is fundamental for the safe use and energy management of lithium-ion batteries. However, LiFePO4 batteries experience a prolonged voltage plateau phase that lowers SOC estimation accuracy, thereby limiting the precision of traditional estimation methods. To address these issues, this paper first constructs a Dual-Measurement Cubature Kalman Filter (DM-CKF) estimator that integrates terminal voltage and expansion force, utilizing the expansion force signal to provide additional state observation information. Then, a CNN-BiLSTM network is established that integrates multi-source information from current, voltage, temperature, and expansion force to fully extract spatiotemporal features during battery operation. Based on this, an Adaptive Weight Fusion (AWF) strategy is proposed to dynamically adjust the fusion weights according to the real-time reliability of the two estimators, thereby achieving complementary advantages between model-driven and data-driven methods. Through ablation experiments with and without expansion force, as well as comparative experiments under different dynamic operating conditions and preload conditions, the estimation accuracy and robustness of the proposed method under complex operating conditions are verified. The experimental results show that the RMSE of the proposed algorithm does not exceed 0.5% and the MAE does not exceed 0.3%, demonstrating the method’s excellent robustness and adaptability.

Accurate estimation of the State of Health (SOH) is critical for ensuring the safety and reliability of Battery Management Systems (BMS) in lithium-ion batteries. The ohmic internal resistance, a parameter in the Equivalent Circuit Model (ECM), serves as a key indicator of the aging degree during battery operation. Although data-driven methods for SOH estimation are highly flexible, their reliance on externally measured parameters often limits their ability to reflect complex internal battery dynamics, resulting in limited estimation accuracy. To address this issue, this study first identifies parameters of a first-order RC ECM using the Particle Swarm Optimization (PSO) algorithm, thereby extracting the key health feature—ohmic internal resistance. Concurrently, Differential Thermal Capacity (DTC) analysis is employed to derive thermodynamic characteristics during battery charging and discharging. An input dataset integrating these ECM parameters and DTC features is constructed to enhance the physical information density of the model inputs, thereby improving the data quality for the subsequent data-driven model. Furthermore, a hybrid model combining a Bidirectional Gated Recurrent Unit (BiGRU) neural network with the eXtreme Gradient Boosting (XGBoost) algorithm is developed to further advance the estimation accuracy. Validation experiments on multiple cells from the Oxford battery aging dataset demonstrate that the proposed method achieves a maximum Root Mean Square Error (RMSE) below 0.3% and a minimum SOH prediction error of 0.01%, confirming its high estimation precision.

Accurate prediction of the SOH(State of Health) of lithium-ion batteries is essential for ensuring the safety and efficiency of new energy vehicles. To overcome the limitations of existing methods that rely on single-category HFs(Health features), this study proposes an SOH prediction approach based on feature-type analysis and a GAPSO-GCRN neural network. Multi-dimensional HFs are extracted from voltage, IC(Incremental Capacity), and DTV(Differential Thermal Voltammetry) curves during charging and discharging, covering voltage, time, temperature, and capacity dimensions. The Pearson–Spearman mixed correlation analysis, combined with feature evolution trends during aging, identifies three optimal indicators: voltage inflection point, capacity entropy change rate, and temperature rise rate. A GCRN(Graph Convolutional Recurrent Network) model is then developed, with a GAPSO(Genetic Algorithm–Particle Swarm Optimization) hybrid strategy employed for global hyperparameter optimization. Experimental results on the Oxford Battery Degradation Dataset show that the proposed method achieves MAE(Mean Absolute Error) and RMSE(Root Mean Square Error) within 0.4% for SOH prediction, demonstrating high accuracy, robustness, and strong generalization capability.

The state of health (SOH) of lithium-ion batteries is a crucial parameter for assessing battery degradation. The aim of this study is to solve the problems of single extraction of health features (HFs) and redundancy of information between features in the SOH estimation. This article develops an SOH estimation method for lithium-ion batteries based on multifeature fusion and Bayesian optimization (BO)-bidirectional gated recurrent unit (BiGRU) model. First, a total of eight HFs in three categories, namely, time, energy, and probability, can be extracted from the charging data to accurately describe the aging mechanism of the battery. The Pearson and Spearman analysis method verified the strong correlation between HFs and SOH. Second, the multiple principal components obtained by kernel principal component analysis (KPCA) can eliminate the redundancy of information between HFs. The principal component with the highest correlation with SOH is selected by bicorrelation analysis to be defined as the fused HF. Finally, to improve SOH estimation accuracy, the BO-BiGRU model is proposed. The proposed method is validated using battery datasets from NASA. The results show that the SOH estimation accuracy of the BO-BiGRU model proposed in this article is high, while mean absolute error (MAE) is lower than 1.2%. In addition, the SOH of the lithium battery is estimated using different proportions of test sets, and the results show that the root-mean-square error (RMSE) and the mean absolute percentage error (MAPE) of the SOH remain within 3%, with high estimation accuracy and robustness.

To solve the problem of inconsistency in the use of series-connected lithium-ion battery packs, this paper proposed a topological structure of dual-layer equalization based on a flying capacitor circuit and Cuk circuit, as well as a control strategy seeking the shortest equalization path. In this structure, batteries are divided into two forms: intra-group and inter-group; the intra-group equalization is the lower-level equalization while the flying capacitor circuit is used as an equalization circuit to achieve equalization between individual battery cells; and the inter-group equalization is the upper-level equalization while Cuk circuit is used as equalization circuit to achieve equalization between battery packs; each battery pack shares a battery cell, thus to obtain more options on equalization path. The proposed strategy, with state of charge as the balancing variable, represents the topological structure of the circuit in the form of graph by adopting graph theory control, seeks the optimal equalization path via ant colony optimization algorithm with global search, thus to improve the equalization speed and efficiency. At last, the structure and the strategy proposed in this paper were simulated in matlab/simulink to compare with the maximum value equalization method in the condition of static, charging, and discharging. The result of the simulation experiments shows that the equalization method based on graph theory control reduces the equalization duration by approximately 17%, and improves the equalization efficiency by approximately 2%, which verifies the superiority and effectiveness of the structure and strategy proposed in this paper.

With the development of electric vehicles, the demand for lithium-ion batteries has been increasing annually. Accurately estimating the State of Health (SOH) of lithium-ion batteries is crucial for their efficient and reliable use. Most of the existing research on SOH estimation is based on parameters such as current, voltage, and temperature, which are prone to fluctuations. Estimating the SOH of lithium-ion batteries based on Electrochemical Impedance Spectroscopy (EIS) and data-driven approaches has been proven effective. In this paper, we explore a novel SOH estimation model for lithium batteries based on EIS and Convolutional Neural Network (CNN)-Vision Transformer (VIT). The EIS data is treated as a grayscale image, eliminating the need for manual feature extraction and simultaneously capturing both local and global features in the data. To validate the effectiveness of the proposed model, a series of simulation experiments are conducted, comparing it with various traditional machine learning models in terms of Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R²). The simulation results demonstrate that the proposed model performs best overall in the testing dataset at three different temperatures. This confirms that the model can accurately and stably estimate the SOH of lithium-ion batteries without requiring manual feature extraction and knowledge of battery aging temperature.

In this article, a grouping equalization circuit based on the single-ended primary inductor converter (SEPIC) circuit is proposed, which can transfer energy between any single cell or grouped cells. Compared with the traditional equalization circuits that transfer energy between adjacent cells, the SEPIC circuit can directly connect any two batteries that need to be equalized; the number of circuit equalization paths is calculated based on a directed graph and then used as a basis for grouping the batteries to improve the equalization efficiency. In the charging or discharging condition, the amount of charge remaining in the battery to be charged or discharged is used as the control variable for equalization, and intragroup equalization is completed before intergroup equalization starts. To ensure the equalization efficiency of the battery, the equalization current is controlled by fuzzy logic control (FLC). Taking ten single cells as an example based on the calculation of the number of equalization paths, two 5-cell groups can be confirmed as the optimal solution. Experiments were performed on matlab/simulink simulation platform, and the results show that compared with the traditional adjacent inductance equalization circuit, the equalization circuit proposed earlier reduces the time needed for equalization by 35.8%; compared with the traditional average difference method, in charging and discharging conditions, the FLC algorithm saves times by 20.5% and 31.3%, respectively, and energy loss is reduced by 9.1% and 5.5%, respectively, which verifies the feasibility of the proposed equalization scheme.

In this paper, a two-stage equalization topology based on Zero-Voltage Switching Quasi-Resonant Converter (ZVS QRC) is proposed to solve the problems of long equalization time and low energy utilization of series lithium-ion battery packs. The equalization topology is divided into two forms: intra-group and inter-group, ZVS QRCs are adopted in both intra-group and inter-group, which can equalize any single cell within a group, and the equalization circuit within each pack can equalize simultaneously, which greatly reduces the equalization time, and reduces the switching loss during the equalization process by introducing quasi-resonant circuit to improve the energy utilization. Taking the State of Charge (SOC) as the equalization target, a group equalization strategy based on the Credibility Factor (C-F) inference model is proposed. The strategy defines the SOC unbalance degree of a single battery, gets the group unbalance degree according to the C-F inference model, and gets the group equalization state from the unbalance degree, to control the group for equalization. Matlab/Simulink software is used to build the model and simulate it. The experimental results demonstrate that the proposed topology saves approximately 25% of the equalization time compared to the traditional Buck-Boost group equalization topology. Under the same static and charge/discharge conditions, the equalization time is reduced by about 21% and the energy utilization is improved by about 6% after using the C-F inference model for equalization compared to the fixed-threshold method.

This article presents an efficient and fast echelon battery equalization method based on wide voltage range bidirectional converter combined with DBSCAN (density-based spatial clustering of applications with noise) clustering control strategy. Thanks to the wide voltage range and bidirectional buck-boost characteristics of the four-switch bidirectional converter, the battery energy can be reasonably redistributed between the battery and the super capacitor. The converter operating in synchronous rectification mode can not only achieve bidirectional energy flow in a wide voltage range, but also has a high energy conversion efficiency. The DBSCAN clustering algorithm is used to divide all battery cells into groups, and each group may contain one or more adjacent and nonadjacent battery cells. First, the nonadjacent single cells are balanced and integrated into adjacent group with the closest voltage. Then, the groups with different voltages are balanced by group-to-group, thereby realizing rapid balance of the entire battery pack. In order to verify the effectiveness of the proposed method, simulations and experimental verification were carried out. The experimental results show that, comparing with traditional battery balancing methods, the proposed method achieves shorter balancing time and higher balancing efficiency.