Electric vehicle battery health assessment method and system based on multi-source data fusion

Abstract

The invention discloses an electric vehicle battery health assessment method and system based on multi-source data fusion, and relates to the technical field of new energy vehicle battery application, and the method comprises the steps: collecting a Nyquist impedance curve, terminal voltage relaxation time sequence data and equalization current time sequence data based on a constant current stage cut-off event of a battery pack of an electric vehicle; extracting an impedance form embedding vector corresponding to the Nyquist impedance curve based on an auto-encoder; performing wavelet packet decomposition on the terminal voltage relaxation time sequence data to obtain a corresponding multi-scale voltage relaxation energy feature vector; calculating a dynamic equalization imbalance index based on the terminal voltage relaxation time sequence data and the equalization current time sequence data; and inputting the impedance form embedded vector, the multi-scale voltage relaxation energy feature vector and the dynamic equilibrium imbalance index into a battery SOH prediction model to output an SOH estimated value of the battery pack. Therefore, the accuracy of the SOH evaluation result is improved through the joint modeling of the coupling characteristics of the multi-physical process of the battery.

Description

Technical Field

[0001] The present application relates to the field of new energy vehicle battery application technology, and in particular to an electric vehicle battery health assessment method and system based on multi-source data fusion. Background Art

[0002] As an important component of new energy transportation, the safety, reliability, and lifespan of electric vehicle (EV) power battery systems have become a focus of current research and industry attention. State of Health (SOH) assessment of power batteries is a key technology for enabling intelligent battery management systems (BMS). Accurately assessing battery SOH helps optimize charge and discharge control strategies, extend service life, and reduce operating and maintenance costs.

[0003] Current mainstream battery health assessment methods typically rely on a single data source or indicator for modeling and judgment. For example, some methods construct criteria based on frequency-domain measurement information, reflecting the battery's internal state by capturing its response characteristics under specific operating conditions. Other methods utilize static or dynamic modeling using voltage and current curves during the charge and discharge process to infer capacity fade and performance degradation.

[0004] However, current battery health assessment methods primarily focus on a specific physical process or signal dimension. The extracted features are often limited to single-value quantities at fixed frequencies or parameter estimates based on simplified fitting models, making it difficult to fully reflect the complex time-frequency coupled evolutionary processes within the battery. Furthermore, as battery modules age, consistency between cells often degrades. Modeling fails to effectively account for the dynamic differences and regulatory behavior between cells, resulting in reduced accuracy in SOH assessment results when faced with early imbalances. Summary of the Invention

[0005] The present application provides an electric vehicle battery health assessment method, system, storage medium, computer program product and electronic device based on multi-source data fusion, which is used to at least solve the problem that the current related technology relies on a single signal dimension and cannot accurately reveal the evolution mechanism of the mutual coupling of multiple physical fields during battery aging, resulting in deviations in SOH assessment results.

[0006] In a first aspect, an embodiment of the present application provides an electric vehicle battery health assessment method based on multi-source data fusion, comprising: synchronously triggering an electrochemical impedance spectroscopy scan, terminal voltage relaxation recording, and battery pack balancing operation based on a constant current phase cutoff event of a battery pack of an electric vehicle to obtain corresponding Nyquist impedance curves, terminal voltage relaxation time series data, and balancing current time series data, respectively; the terminal voltage relaxation time series data includes multiple sampling moments and corresponding terminal voltage values, and the balancing current time series data includes multiple sampling moments and corresponding balancing current values; extracting an impedance morphology embedding vector corresponding to the Nyquist impedance curve based on an autoencoder; performing wavelet packet decomposition on the terminal voltage relaxation time series data to obtain a corresponding multi-scale voltage relaxation energy feature vector; and calculating a dynamic balancing imbalance index based on the terminal voltage relaxation time series data and the balancing current time series data:

[0007]

[0008] Where M bal (t) is the dynamic balance imbalance index; Δt represents the sampling time interval, σ V (t) is the standard deviation of all monomer terminal voltages at sampling time t, is the regularization constant, ΔI bal (t) is the balancing current increment, which represents the difference in balancing current between two adjacent sampling times t-Δt and t;

[0009] The impedance morphology embedding vector, the multi-scale voltage relaxation energy feature vector, and the dynamic balance imbalance index are input into a battery SOH prediction model to output an estimated SOH value of the battery pack.

[0010] In a second aspect, an embodiment of the present application provides an electric vehicle battery health assessment system based on multi-source data fusion, including: a multi-source data acquisition unit for synchronously triggering electrochemical impedance spectroscopy scanning, terminal voltage relaxation recording, and battery pack balancing operation based on a constant current phase cutoff event for an electric vehicle battery pack, so as to obtain corresponding Nyquist impedance curves, terminal voltage relaxation time series data, and balancing current time series data, respectively; the terminal voltage relaxation time series data includes multiple sampling moments and corresponding terminal voltage values, and the balancing current time series data includes multiple sampling moments and corresponding balancing current values; an impedance morphology encoding unit for extracting an impedance morphology embedding vector corresponding to the Nyquist impedance curve based on an autoencoder; a voltage relaxation multi-scale extraction unit for performing wavelet packet decomposition on the terminal voltage relaxation time series data to obtain a corresponding multi-scale voltage relaxation energy feature vector; a dynamic balance imbalance index calculation unit for calculating a dynamic balance imbalance index based on the terminal voltage relaxation time series data and the balancing current time series data:

[0011]

[0012] Where M bal (t) is the dynamic balance imbalance index; Δt represents the sampling time interval, σ V (t) is the standard deviation of all monomer terminal voltages at sampling time t, is the regularization constant, ΔI bal (t) is the balancing current increment, which represents the difference in balancing current between two adjacent sampling times t-Δt and t;

[0013] A battery pack SOH prediction unit is configured to input the impedance morphology embedding vector, the multi-scale voltage relaxation energy feature vector, and the dynamic balance imbalance index into a battery SOH prediction model to output an estimated SOH value of the battery pack.

[0014] In a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the steps of the electric vehicle battery health assessment method based on multi-source data fusion of any embodiment of the present application.

[0015] In a fourth aspect, an embodiment of the present application provides a storage medium on which a computer program is stored, characterized in that when the program is executed by a processor, the steps of the electric vehicle battery health assessment method based on multi-source data fusion of any embodiment of the present application are implemented.

[0016] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the electric vehicle battery health assessment method based on multi-source data fusion of any embodiment of the present application.

[0017] The electric vehicle battery health assessment method and system based on multi-source data fusion provided in this application can produce at least the following technical effects:

[0018] (1) By synchronously triggering electrochemical impedance spectroscopy scanning, terminal voltage relaxation, and equilibrium current measurement, the internal state of the battery is comprehensively characterized from three dimensions: time-frequency coupling, energy distribution, and cell consistency. On the one hand, the Nyquist impedance morphology embedding vector obtained by the autoencoder can capture the subtle changes in the charge transfer and diffusion process at the battery interface, making the characterization of the internal electrochemical process more discriminative. On the other hand, the multi-scale voltage relaxation energy characteristics extracted by wavelet packet decomposition can reflect the polarization and recovery dynamics of the battery at different time scales, improving the model's sensitivity to the initial signals of capacity decay. Combined with the real-time quantification of the voltage imbalance and equilibrium behavior between cells by the dynamic equilibrium imbalance index, it can accurately identify the early imbalance risk caused by consistency degradation. The fusion of multi-source features enables the SOH prediction model to maintain high accuracy and robustness at different degradation stages, significantly improving the early decline diagnosis capability and the assessment accuracy under imbalanced conditions, thereby providing a more reliable decision-making basis for the intelligent BMS.

[0019] (2) By introducing the dynamic balancing imbalance index, the standard deviation of the terminal voltage and the balancing current increment of all cells in the battery pack are quantified in real time, quantitatively describing the differential changes between cells during the balancing process. This can effectively identify the degradation of the internal consistency of the module and accurately identify the imbalance state and balancing behavior between cells, especially the identification and adjustment of the imbalance state in the early stage of battery pack use, further improving the accuracy and reliability of the overall SOH assessment.

[0020] (3) The three types of time-frequency coupling, multi-energy distribution and single-cell consistency assessment features, namely impedance morphology embedding vector, multi-scale voltage relaxation energy characteristics and dynamic balance imbalance index, are fused and input into the SOH prediction model, so that the assessment results maintain high accuracy and robustness in each degradation stage, thereby improving the accuracy of SOH assessment under unbalanced battery pack conditions.

[0021] Through this technical solution, by synchronously collecting electrochemical impedance spectra, terminal voltage relaxation data and battery pack balancing current data, integrating frequency domain response characteristics, time domain relaxation characteristics and module consistency status, the joint modeling of the coupling characteristics of battery multi-physical processes is achieved, which can more comprehensively and accurately characterize the battery health status. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0023] Figure 1A flowchart of an example of an electric vehicle battery health assessment method based on multi-source data fusion according to an embodiment of the present application is shown;

[0024] Figure 2 A schematic diagram showing an experimental comparison simulation effect of an example of dynamic SOH estimation and static SOH estimation based on a dynamic balance imbalance index according to an embodiment of the present application is shown;

[0025] Figure 3 An operational flow chart of an example of synchronously triggering multi-source data acquisition based on a cutoff event in a constant current phase according to an embodiment of the present application is shown;

[0026] Figure 4 A schematic structural connection diagram of an example of an encoder structure of an autoencoder according to an embodiment of the present application is shown;

[0027] Figure 5 A schematic diagram of the structural connection of an example of a battery SOH prediction model according to an embodiment of the present application is shown;

[0028] Figure 6 A structural block diagram of an example of an electric vehicle battery health assessment system based on multi-source data fusion according to an embodiment of the present application is shown;

[0029] Figure 7 This is a schematic structural diagram of an embodiment of an electronic device of the present application. DETAILED DESCRIPTION

[0030] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0031] It should be noted that the current battery health assessment methods mainly include analysis methods based on electrochemical impedance spectroscopy and assessment methods based on static fitting of charge and discharge curves. In the analysis method based on electrochemical impedance spectroscopy, the AC impedance of the battery at different frequencies is measured to reflect the state of its internal electrochemical process. Generally, the impedance modulus or phase angle at a specific frequency point of 1Hz or 10Hz is selected as a characteristic indicator for SOH estimation. However, this method lacks comprehensive consideration of the overall morphological changes of the spectrum curve, and it is difficult to distinguish subtle curve changes caused by different degradation mechanisms. At the same time, the single-point feature has poor stability in the presence of measurement noise, which can easily lead to evaluation deviations.

[0032] In the method based on charge and discharge curve analysis, the voltage relaxation curve after the charge and discharge cutoff is fitted, and the time constant is extracted using an exponential function or a biexponential function to estimate the SOH. However, the voltage relaxation process is essentially a coupling effect of multiple physical processes (such as charge transfer, concentration polarization, and diffusion processes), and its decay characteristics have significant multi-time scale characteristics. The simple function model currently used cannot effectively decouple the decay components of different time scales, resulting in insufficient sensitivity for detecting early minor degradation. In addition, the static fitting method has difficulty capturing the nonlinear characteristics in the dynamic changes of voltage, which limits its adaptability under complex working conditions.

[0033] Furthermore, battery modules consist of multiple cells connected in series and parallel. Due to manufacturing variations and inconsistent aging rates, cell-to-cell voltage inconsistencies often occur during use. BMS systems typically adjust cell voltages through active or passive balancing strategies. However, existing health assessment methods often ignore the dynamic information of balancing current and cell voltage differences, making it difficult to identify early signs of internal consistency degradation, affecting the sensitivity and accuracy of SOH assessments.

[0034] It should be understood that the purpose of the above description of the current related art is only to facilitate the public to better understand the inventive spirit and motivation of this application, and is not to be construed as limiting this application. In addition, the technical solutions described in the above-mentioned current related art are not prior art and may also be undisclosed technical solutions, such as solutions under research or in the laboratory stage.

[0035] In the technical solutions of this application, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved shall comply with the provisions of relevant laws and regulations and shall not violate public order and good morals.

[0036] Figure 1 A flowchart of an example of an electric vehicle battery health assessment method based on multi-source data fusion according to an embodiment of the present application is shown.

[0037] It should be noted that static characteristics only reflect the overall state of the battery pack at the end of a charge-discharge cycle and cannot capture the details of the instantaneous voltage differences between cells during the cycle and the lag in the balancing current response. When the capacity or internal resistance of certain cells begin to deviate due to manufacturing variations or local damage, this deviation is partially smoothed out by the balancing circuit after the cycle ends, causing the static characteristics to appear to be within the normal range, but it masks the risks that have begun to accumulate internally. Furthermore, as the number of cycles increases, the differences in cell capacity and internal resistance within the battery pack gradually accumulate, and the real-time compensation speed of the balancing circuit varies. This dynamic imbalance is a precursor to potential early failure, but it is often overlooked.

[0038] Regarding the execution subject of the method of the embodiment of the present application, it can be any controller or processor with computing or processing capabilities. Specifically, it can be implemented by a BMS controller or SOH assessment component, using a data-driven and physical feature fusion approach for feature extraction and modeling. It is the first time to propose integrating the dynamic balance imbalance index into the SOH assessment, incorporating the inconsistency of cells within the group and their real-time balance response into the model. The constructed health assessment model has good generalization and anti-interference capabilities, and is applicable to power battery systems of different models and different operating conditions.

[0039] In some examples, it can be integrated into an electronic device or terminal through software, hardware, or a combination of software and hardware, and the type of terminal or electronic device can be diverse, such as a mobile phone, tablet computer, or desktop computer, etc.

[0040] like Figure 1 As shown, in step S110, based on the constant current phase cut-off event of the battery pack of the electric vehicle, the electrochemical impedance spectroscopy scan, the terminal voltage relaxation recording and the battery pack balancing operation are synchronously triggered to obtain the corresponding Nyquist impedance curve, the terminal voltage relaxation time series data and the balancing current time series data respectively.

[0041] Here, the terminal voltage relaxation timing data includes multiple sampling moments and corresponding terminal voltage values, and the balanced current timing data includes multiple sampling moments and corresponding balanced current values, for example, synchronously triggering the acquisition of terminal voltage values and balanced current values based on the same sampling timestamp.

[0042] A constant current phase cutoff event is a state transition event triggered by the system during constant current charge and discharge of a battery pack when a preset cutoff condition is met (such as a cell voltage reaching an upper threshold, total charge capacity reaching a threshold, or cumulative time exceeding a limit). This event marks the end of the constant current phase and triggers subsequent operations (such as switching to the constant voltage phase and starting data acquisition).

[0043] For example, the BMS system sends a constant current phase end trigger signal when any of the following preset conditions is met:

[0044] Termination voltage condition: the battery pack terminal voltage rises to or drops to the preset charge and discharge termination voltage;

[0045] Capacity threshold condition: The cumulative charge and discharge capacity reaches the set rated capacity threshold;

[0046] Time limit condition: the duration of constant current charge and discharge reaches the pre-specified maximum time;

[0047] SOC / SOH safety threshold condition: The battery pack's state of charge (SOC) or SOH reaches a safety threshold that requires switching the control strategy.

[0048] When any of the above conditions is met, the BMS system immediately interrupts the constant current control mode, switches to the terminal voltage relaxation measurement and equilibrium operation mode, and synchronously triggers the electrochemical impedance spectroscopy scan to ensure the time alignment of various signals.

[0049] It should be noted that after the constant current phase ends, the charge distribution within the battery has reached the specified bias level; it then enters the constant voltage or open-circuit relaxation state, at which point the system's dynamic response significantly slows and the electrochemical process becomes controllable. Using this demarcation point as the triggering moment, a multi-source signal can be obtained that not only captures the end-of-charge and discharge characteristics (polarization effects at high currents) but also captures the intrinsic recovery dynamics during the subsequent static relaxation phase.

[0050] More specifically, electrochemical impedance spectroscopy (EIS) measurements must be performed under steady-state or known bias current conditions to accurately reflect battery interface and diffusion processes. The bias current must reach the set constant current value around the end of the constant current measurement. Applying a sweeping perturbation voltage immediately after triggering ensures a clear and consistent initial baseline for the EIS measurement, avoiding data distortion caused by initiating EIS during the dynamic charge and discharge process.

[0051] The terminal voltage relaxation curve reflects the recovery kinetics from constant current cutoff to electrochemical equilibrium. Starting from the end of the constant current phase, the system records the entire relaxation process from the maximum polarization value. Subsequent multi-scale analysis extracts the energy distribution characteristics at different time constants, greatly improving sensitivity to early capacity decay and polarization evolution.

[0052] Furthermore, BMS systems typically balance and compensate individual cells after constant current charging or discharging, or after entering constant voltage mode. Switching to the balancing circuit at this time best reflects the differences in cell status within the group. Using this event to trigger balancing current measurement quantifies the BMS system's response to imbalance in real time, directly comparing it to the degree of voltage imbalance and generating a highly correlated dynamic imbalance index.

[0053] In some embodiments, in the BMS system, the current state of the battery pack is detected in real time. When it is detected that the constant current charge and discharge process is close to termination (for example, reaching the set termination voltage condition or capacity threshold condition), or through the constant current mode switching interrupt signal (for example, generated by the charge and discharge controller), and this moment is identified as the "constant current stage cutoff event", the corresponding absolute timestamp t0 is recorded. Furthermore, when the electric vehicle battery pack is at the end of the constant current charge and discharge stage, the BMS control unit is used to synchronously issue a trigger instruction: on the one hand, the electrochemical impedance spectroscopy (EIS) scanner is started, a small amplitude sinusoidal AC disturbance voltage is applied and the current response is recorded; on the other hand, the terminal voltage relaxation measurement is started at the same time, and the balancing action sampling of the battery pack balancing circuit is started at the same time. To ensure the time alignment of the three signals, the BMS controller can pre-calibrate the clocks of each measurement subsystem and issue a "synchronous sampling" command through a hardware interrupt or a dedicated bus (such as CAN or LIN).

[0054] Specifically, the EIS scanner applies a ±10mV peak-to-peak sinusoidal AC voltage with frequencies covering 10mHz, 100mHz, 1Hz, 10Hz, 100Hz, and 1kHz. The sampling time for each frequency point is approximately 100ms, and the total scanning time is less than 1s. Specifically, the scanning hardware is integrated into the BMS daughterboard and can be run in parallel with subsequent relaxation recordings. The scanning results are marked with t0+Δt in real time. i (Δt i The timestamp of the start delay of each frequency point.

[0055] The voltage relaxation measurement module records all cell terminal voltages according to the sampling frequency. Specifically, starting from time t0, the terminal voltage sequence V(t) is recorded at a sampling rate of 100 Hz. The default recording duration is 300 s and can be adjusted according to the diffusion time constant of the battery material.

[0056] The balancing current measurement unit triggers the balancing module switch, applies a constant voltage bypass or a bidirectional balancing circuit, and records the balancing current waveform through a shunt resistor and a high-precision ADC with the same sampling timing as the relaxation measurement.

[0057] As a result, multi-dimensional data is collected under unified event drive with high time alignment, realizing a cross-domain synchronous data collection mechanism for the three types of data, and providing a basis for accurate feature fusion modeling.

[0058] In step S120, an impedance morphology embedding vector corresponding to the Nyquist impedance curve is extracted based on the autoencoder.

[0059] Here, the impedance shape embedding vector compresses the "shape information" of a Nyquist impedance curve into a fixed-length, low-dimensional numerical representation that can be used by subsequent models. The impedance shape embedding vector reflects the electrochemical impedance characteristics of the battery pack in its current state.

[0060] It should be noted that on the Nyquist plot, different degradation mechanisms will cause the curve to exhibit different features such as arc semicircles, depressions, double semicircle superpositions, or the tail end of the Warburg line. We collectively refer to these geometric structures as "morphology." Through the encoder part of the autoencoder, the morphological information scattered on the curve in different frequency bands is "converged" to obtain the impedance morphology embedding vector. Here, each component of the impedance morphology embedding vector does not directly correspond to the impedance value at a specific frequency point, but rather integrates the "geometric" features learned at different scales by multiple layers of convolution, such as inflection points, depression depth, semicircle diameter, tail slope, etc., to more effectively provide morphological discriminant features.

[0061] Furthermore, the autoencoder consists of an encoder and a decoder. The encoder maps the Nyquist curve to a low-dimensional latent space (i.e., the impedance morphology embedding vector), while the decoder attempts to reconstruct the original curve to ensure that the extracted features have sufficient information fidelity. It should be noted that the decoder is used to optimize the encoder and is only applied during the training phase of the autoencoder. It can also be set to have a mirror-symmetric structure with the encoder.

[0062] It should be noted that in traditional methods, using only single-point impedance values or equivalent circuit parameters easily misses information about the overall morphological changes of the curve, resulting in insufficient ability to distinguish different degradation mechanisms. In contrast, in the embodiment of the present application, the autoencoder automatically extracts the morphological comprehensive discriminant representation of key physical mechanisms such as interfacial charge transfer, electrolyte diffusion, and electrode aging in the frequency domain during the process of compressing and reconstructing the impedance curve, giving the embedded vector both denoising and high-dimensional discrimination capabilities.

[0063] In step S130 , wavelet packet decomposition is performed on the terminal voltage relaxation time series data to obtain a corresponding multi-scale voltage relaxation energy feature vector.

[0064] The terminal voltage relaxation curve typically encompasses a dynamic process lasting several seconds to minutes from the end of constant current to the equilibrium stage. To capture characteristics at different kinetic scales, we introduce wavelet packet decomposition technology into the terminal voltage relaxation time series data to extract multi-scale voltage relaxation energy eigenvectors, which reflect the dynamic changes of the battery at different time scales.

[0065] In some implementations, a terminal voltage sequence sampled at uniform intervals can be used as the input signal, ensuring that the sampling frequency covers both rapid and slow voltage variations. Furthermore, based on the nonstationary nature of the voltage curve, a tightly supported wavelet basis (such as the Daubechies wavelet) is selected, and a 3-5-layer wavelet packet decomposition structure is established to fully capture both high- and low-frequency information. By calculating the energy values in each subband as eigenvector elements, a multi-scale energy distribution map is generated to describe the voltage relaxation process.

[0066] In the examples of this application, by employing multi-scale wavelet packet decomposition, the energy distribution of both the short-term polarization effect and the long-term diffusion recovery process can be simultaneously reflected. The dynamic characteristics at different time constants are quantified separately, allowing the model to separately perceive the evolutionary trajectories of three mechanisms: early charge accumulation, surface polarization, and deep diffusion. Compared to a single time window or Fourier transform, the wavelet packet method excels in time-frequency localization, significantly enhancing the ability to capture differential signals in the capacity decay curve at various degradation stages.

[0067] In step S140 , a dynamic balance imbalance index is calculated based on the terminal voltage relaxation time series data and the balance current time series data.

[0068] In the embodiment of the present application, a dynamic balance imbalance index is introduced as a metric for the degradation of the internal consistency of the module, aiming to infer the trend of uneven distribution of health status from the differences and balance behaviors between battery cells.

[0069] In some implementations, the standard deviation of multiple cell voltage values is first calculated to reflect the degree of imbalance within the battery pack. The incremental change between two balancing current samples is then calculated to characterize the strength of the balancing action. Finally, a ratio calculation is performed on the standard deviation and the current increment, and regularization is added to suppress the influence of extreme values. Furthermore, continuous updates are performed using a sliding window, and a reasonable calculation cycle (e.g., 100ms) and time window width (e.g., 1s) are set in the controller, allowing the exponential curve to smoothly display the dynamic balance process of voltage dispersion and current adjustment. The calculated dynamic balancing imbalance index can be fed back to the BMS decision module in real time to determine whether to trigger balancing side compensation or early warning prompts.

[0070] More specifically, the dynamic equilibrium imbalance index is calculated by the following formula:

[0071]

[0072] Where M bal (t) is the dynamic balance imbalance index, which represents the internal inconsistency index of the battery pack calculated at the sampling time t, and is used to characterize the ratio of the single-terminal voltage difference to the balance current fluctuation. Δt represents the sampling time interval, σ V(t) is the standard deviation of all monomer terminal voltages at sampling time t, is the regularization constant, ΔI bal (t) is the balancing current increment, which represents the difference in balancing current between two adjacent sampling times t-Δt and t.

[0073] Here, the dynamic balance imbalance index combines the dual information of single-cell voltage dispersion and balance current regulation strength, which can not only reflect the degree of imbalance, but also measure the BMS's response efficiency to the imbalance. In the early stage of battery consistency degradation, the index change is more discriminative than a single voltage fluctuation or current change, and can warn of potential imbalance risks several cycles in advance.

[0074] Figure 2 A schematic diagram of experimental comparison simulation effects of an example of dynamic SOH estimation and static SOH estimation based on a dynamic balance imbalance index according to an embodiment of the present application is shown.

[0075] like Figure 2 As shown in the figure, both the static SOH estimation curve and the dynamic SOH estimation curve show a downward trend, but in the critical range (50-100 cycles), the dynamic curve is significantly ahead of the static curve, reflecting the accelerated decline of health.

[0076] exist Figure 2 At the 50th cycle at point A, the dynamic curve has dropped to 95%, while the static curve remains at 98%. At this point, the dynamic equilibrium imbalance index rises due to the increasing inconsistency of the monomers, allowing the fusion model to capture weak degradation signals in the early cycles, while static methods relying solely on impedance or relaxation characteristics have not yet shown any changes.

[0077] exist Figure 2 At the 100th cycle at point B, the dynamic SOH has further dropped to 90%, while the static curve only begins to decline rapidly from 95%. The dynamic model remains sensitive to inconsistencies, while the static model does not respond significantly to degradation until the end of the cycle. This is because it only evaluates "steady-state" characteristics at the end of the cycle and is unable to promptly reflect the accumulated fluctuations during the cycle.

[0078] Further, refer to Figure 2 In the shaded interval (i.e., cycles 50-100), the gap in the SOH estimates of the two curves gradually widens, reaching a maximum of approximately 5%. This is precisely the result of the dynamic model continuously adjusting the SOH prediction slope through the interaction of the real-time equilibrium imbalance index and static characteristics. The dynamic SOH continues to lead in this interval, providing a longer lead time for maintenance and equilibrium strategy optimization. Therefore, through the embodiments of the present application, the dynamic equilibrium imbalance index is deeply integrated with the static impedance and relaxation characteristics, retaining static macro-decay information while supplementing the real-time feedback of dynamic consistency changes.

[0079] In some examples of the present invention, the dynamic balance imbalance index also helps optimize the balancing strategy. More specifically, when the index shows a significant upward trend, compensation is increased, and after it stabilizes, the balancing switching frequency can be reduced, thereby reducing system power consumption and balancing cell lifespan.

[0080] Exemplarily, the dynamic balance imbalance index is compared with a preset balance imbalance threshold, and when the dynamic balance imbalance index exceeds the balance imbalance threshold, an alarm operation for a battery cell inconsistency fault is triggered.

[0081] It should be noted that when the dynamic balance imbalance index M bal (t) exceeds the preset threshold M th , it indicates that the voltage difference between the individual cells in the battery pack is too large, and the balancing current response rate is not fast enough to quickly bridge this difference. Battery cell inconsistency failures can include a variety of fault conditions, such as uneven cell capacity decay, abnormal internal resistance increase, balancing circuit failure or performance degradation, etc. When the alarm is triggered, the system also writes the cycle number, timestamp, and related exponential curve data of the fault to the log for subsequent fault location and analysis. Therefore, with the help of the comprehensive quantified dynamic balancing imbalance index and the strict threshold comparison, an alarm can be issued when the battery cell inconsistency just exceeds the normal fluctuation range, which is more sensitive and accurate than the traditional method that relies only on voltage difference or temperature difference limit.

[0082] In step S150 , the impedance morphology embedding vector, the multi-scale voltage relaxation energy feature vector, and the dynamic balance imbalance index are input into a battery SOH prediction model to output an estimated SOH value of the battery pack.

[0083] In some implementations, a fusion neural network model (such as a Transformer or LSTM-CNN hybrid) can be used to jointly learn static and dynamic features and output the current SOH estimate. Specifically, the three features described above—impedance morphology embedding, multi-scale relaxation energy vector, and dynamic misalignment index—are concatenated in a sequential or parallel manner and then input into the SOH prediction model to comprehensively capture nonlinear and temporal dependencies.

[0084] To train a battery SOH prediction model, a library of battery pack experimental samples can be constructed, encompassing different cycle times, temperatures, and load conditions. SOH values are annotated based on actual capacity decay or authoritative test data. Cross-validation, hyperparameter search, and regularization techniques are used during training to avoid overfitting. During online inference, the model first normalizes the input features and then outputs the SOH value. The BMS automatically adjusts the charge and discharge strategy or issues maintenance prompts based on the prediction results.

[0085] By jointly learning multiple sources of features representing different physical mechanisms within the same model, a comprehensive understanding of battery SOH can be achieved. The battery SOH prediction model not only identifies the aging state of the electrode and electrolyte based on morphological impedance characteristics, but also determines the capacity decay process by combining the relaxation energy spectrum. It also relies on the imbalance index to compensate for consistency deviations, reducing single-signal noise interference and improving the model's generalization ability for complex degradation patterns. This allows SOH estimation to maintain high accuracy (error less than ±2%) and high robustness (significantly reduced standard deviation) under extreme operating conditions and imbalanced states, providing reliable decision support for intelligent BMS.

[0086] Figure 3 An operational flow chart of an example of synchronously triggering multi-source data acquisition based on a cutoff event in a constant current phase according to an embodiment of the present application is shown.

[0087] In order to ensure that the data collected by subsequent EIS scans, terminal voltage relaxation and balancing operations all reflect the same "battery pack completes constant current charge and discharge and enters open circuit static" state, an agreed time point is required for unified triggering. This time point is not necessarily the sampling moment itself, and can also be the event moment after confirmation by the current judgment logic.

[0088] like Figure 3 As shown, in step S310, the charge and discharge current is sampled in real time at the shunt resistor of the battery pack, and the sampled signal is processed by first-order low-pass filtering and sliding average processing to obtain a smooth current curve

[0089] Here, the high-frequency noise caused by electromagnetic interference and current pulsation is effectively suppressed through first-order low-pass filtering and sliding average processing.

[0090] In step S320, the current curve is smoothed With the preset threshold I thr For comparison, when And the duration is not less than the debounce time t deb When , it is determined that the constant current stage is terminated.

[0091] In some embodiments, I thr It is recommended to set it to 2% of the rated current or a fixed value of 0.5A to balance false alarms and response speed. It can also be dynamically adjusted through production calibration or online learning. The debounce time can be selected between 10-50ms to filter out abnormal glitches of more than ten milliseconds while ensuring that roll-off trends within tens of milliseconds are captured. When all the above conditions are met, the physical state corresponding to the judgment period is determined to be a constant current phase termination event, "constant current phase terminated and entered open circuit static state."

[0092] In step S330, when it is determined that the constant current stage is terminated, a hardware trigger message is generated and broadcast through the vehicle communication bus to synchronously trigger the electrochemical impedance spectroscopy scan, terminal voltage relaxation recording and battery pack balancing operation. The hardware trigger message includes the cycle number and the timestamp of the sampling time.

[0093] Once an event is confirmed, a hardware trigger message is generated and broadcast via the vehicle communication bus, notifying downstream modules to initiate EIS scanning, terminal voltage relaxation timing data recording, and battery pack balancing in parallel. The "cycle number N" and "sampling time timestamp t0" carried in the hardware trigger message serve as key indexes, ensuring that the three parallel acquisition channels can be aligned one-to-one at the same cycle and state point on the backend, eliminating data alignment ambiguity.

[0094] Here, by combining the dual-channel trigger mechanism of hardware interrupt signals and vehicle bus message broadcasts, not only can the trigger moment alignment error be controlled below the millisecond level, but the cycle number and precise timestamp can also be synchronously transmitted to each acquisition module, ensuring that multi-source data is consistently collected under the same physical state.

[0095] The signal optimization and dual-channel trigger fusion scheme proposed in the embodiment of the present application can significantly improve the accuracy and stability of the constant current stage cutoff event identification compared to the traditional method of directly judging by the original current threshold.

[0096] Figure 4 A structural connection diagram of an example of an encoder structure of an autoencoder according to an embodiment of the present application is shown.

[0097] It should be noted that in battery health assessment, the Nyquist impedance spectrum contains multiple mechanism information such as electrochemical interface, diffusion process and material internal resistance, but traditional single-point impedance value or equivalent circuit parameters are difficult to fully characterize its overall morphological changes. To this end, in the embodiment of this application, an encoder structure of an autoencoder is introduced to map the geometric morphology of the impedance curve to a low-dimensional embedding vector, providing high-dimensional, distinguishable morphological features for multi-source data fusion.

[0098] like Figure 4 As shown, the encoder structure 400 of the autoencoder includes a cascaded input preprocessing layer 410, a multi-scale dilated convolution layer 420, a channel attention pooling layer 430 and a fully connected dimensionality reduction layer 440.

[0099] The input pre-processing layer 410 is used to perform the following operations:

[0100] The obtained discrete impedance spectrum point set The Nyquist curve is obtained by interpolation and arc length parameterization, so that the arc length of the curve is:

[0101]

[0102] The Nyquist curve is divided into the resampling point sequence P according to the equidistant arc length interval L / N j =(x j ,y j ), L is the total arc length of the Nyquist curve, N is the total number of resampling points, j = 1, ..., N; where ω i represents the frequency of the i-th sinusoidal excitation signal, and are the measured excitation frequencies ω i The real and imaginary values of the electrochemical impedance of the battery pack are shown in Figure 2. K represents the total number of discrete frequency points collected for the Nyquist curve, τ is the curve parameter, and s(t) represents the geometric arc length from the starting point of the curve to the parameter value t; P j represents the Nyquist plane coordinates of the jth resampling point, x j Corresponding to the real part, y j Corresponding to the imaginary part.

[0103] It should be noted that the discrete impedance spectrum point set is obtained through EIS scanning. For example, at the i-th excitation frequency ω i In this case, a small voltage excitation is applied to the battery port, and then the corresponding AC voltage or current response is measured. After phase-locked amplification or fast Fourier transform, the real part of the impedance is obtained. and the imaginary part A total of K such frequency pairs are collected to form a discrete impedance spectrum point set.

[0104] The impedance spectrum points obtained from the original EIS scan are unevenly distributed across the frequency or impedance plane, and the curve is sparse due to the number of test points and the scanning strategy. In the input preprocessing layer 410, the discrete points are padded to a geometrically evenly spaced sequence through curve arc length parameterization and equidistant resampling. Normalization is then performed to eliminate scale differences in the original data.

[0105] After stacking each resampling point row by row, we get the input matrix H (0) , each row of the matrix is (x j ,y j ).

[0106] The multi-scale dilated convolutional layer 420 is used to perform the following operations:

[0107] The input matrix H (0) Input three layers of one-dimensional dilated convolution, each layer is

[0108]

[0109] Iteration

[0110] Among them, H (l) represents the intermediate features of the lth layer, Indicates that the expansion rate d is 2 l One-dimensional convolution operation, b (l) is the bias of the lth layer, and C is the number of channels.

[0111] It should be noted that the different mechanism features on the Nyquist curve are manifested as arc semicircles and depressions of different scales. Single-scale convolution is difficult to take into account both tiny grooves and large semicircles at the same time, so dilated convolution is used to form a "hole" receptive field with different dilation rates, taking into account both local and global morphology. Specifically, the first layer has a dilation rate of 1 to capture the subtlest single-point depression, the second layer has a dilation rate of 2 to connect wider features between two sampling points, and the third layer has a dilation rate of 4 to expand to a larger range (such as covering the entire semicircle structure). In this way, a multi-scale receptive field is established to automatically learn microscopic and macroscopic morphological features.

[0112] The channel attention pooling layer 430 is used to perform the following operations:

[0113] For H (3) Each row vector of Calculating attention weights

[0114]

[0115] Where, α j represents the attention weight corresponding to the j-th resampling point, represents the attention mapping matrix, is the attention weight vector. Represents the normalization term, which is used to obtain α through softmax normalization j .

[0116] It should be noted that different sampling locations contribute differently to health status—low-frequency semicircular depressions are often more effective in distinguishing aging types, but standard global pooling cannot. The channel-based attention mechanism learns position weights, allowing the network to focus on the most discriminative local features.

[0117] Compute the global morphological eigenvector:

[0118]

[0119] Therefore, by dynamically adjusting feature aggregation, irrelevant noise is suppressed and robustness is achieved against sudden measurement errors. By visualizing the attention weights, it is possible to explain which frequency bands and curve areas contribute most to the SOH evaluation, making the fused vector z more discriminative.

[0120] The fully connected dimension reduction layer 440 is used to perform the following operations:

[0121] The global morphological feature vector z is mapped through two-level full connection to output impedance morphological embedding vector

[0122]

[0123] Where, Represents the first layer of dimensionality reduction mapping matrix, represents the first layer bias vector, Represents the second-layer dimensionality reduction mapping matrix, represents the second layer bias vector, D is the dimension of the impedance morphology embedding vector, and E represents the impedance morphology embedding vector.

[0124] Here, the high-dimensional aggregate feature z is mapped to the preset embedding space D to facilitate subsequent splicing and unified fusion with other modal features, and two-level mapping is used to achieve feature fusion and dimensionality compression respectively, thereby retaining sufficient nonlinear expression capabilities to ensure good distinction between different degradation modes.

[0125] Through the embodiments of the present application, uniform arc length resampling is performed on the Nyquist curve constructed from a discrete impedance spectrum point set, and multi-scale dilated convolution is combined with attention pooling to enable the embedding vector to accurately capture the morphological differences of various degradation mechanisms on the Nyquist curve, thereby capturing more discriminative impedance morphological features.

[0126] It should be noted that the decoder structure in the autoencoder can be diverse and can be mirror-symmetric with the encoder structure. For example, the decoder structure consists of three layers of one-dimensional transposed convolution that are symmetrical with the encoder, and the number of channels and convolution kernel size of each layer are mirrored with the corresponding encoder layer. In addition, the activation function (ReLU) and batch normalization used are the same as those of the encoder, and the last layer is linearly mapped back to the original N×2 Nyquist sequence. When training the autoencoder, the reconstruction error (MSE) plus the curvature loss can be used as the overall loss to optimize the encoding performance.

[0127] Regarding the explanation of the battery terminal voltage relaxation operation, the battery terminal voltage relaxation is the superposition result of a multi-stage, multi-physical process. After the discharge or charge is completed, the voltage first undergoes a rapid reconstruction of the double-layer capacitance (typical time constant is in the millisecond level), followed by slower processes such as liquid phase diffusion (in the second level) and solid phase (solid state) diffusion (tens to hundreds of seconds). Traditional single exponential or double exponential fitting models can only roughly separate the two time constants, cannot distinguish more mechanisms, and are difficult to capture weak differences in the early decay stage.

[0128] Compared with simple analysis in the time domain or frequency domain, in the embodiment of the present application, by adopting wavelet packet decomposition (WPD), it is possible to provide localized multi-scale decomposition on the time-frequency plane, that is, retaining time positioning information and obtaining frequency component distinction at different resolutions, which helps to decompose voltage relaxation sequences containing multiple time scale characteristics in the signal.

[0129] More specifically, the obtained terminal voltage is relaxed into a discrete sequence Applying the bandpass filter operator BP{·} and subtracting the final steady-state voltage V[M-1] yields the correction sequence:

[0130] ΔV[n]=BP{V[n]}-V[M-1], formula (7)

[0131] Where ΔV[n] is the corrected voltage attenuation value, representing the terminal voltage relaxation after removing the steady-state baseline at the nth sampling moment; V[n] is the terminal voltage value measured at the nth sampling moment; BP{·} is a bandpass filter operator with a preset cutoff frequency; M is the total number of discrete moments sampled in the relaxation record; V[M-1] is the steady-state terminal voltage, representing the terminal voltage baseline value corresponding to the last sampling point in the sequence.

[0132] Here, the digital bandpass filter BP{·} can typically use a second-order Butterworth or Chebyshev II filter, with a low cutoff frequency set to 0.05 Hz (to filter out DC drift) and a high cutoff frequency set to 5 Hz (to suppress high-frequency measurement noise). The filter design uses bidirectional zero-phase filtering to avoid phase distortion.

[0133] Furthermore, using the final sample V[M-1] as the steady-state voltage baseline, the DC component is precisely subtracted through ΔV[n], ensuring that ΔV[n] contains only the relaxation decay component, and subsequent decomposition is unaffected by drift. This eliminates DC drift, avoids energy statistical shifts, suppresses high-frequency mutations, and enhances decomposition stability.

[0134] Set ΔV[n] as the initial approximation coefficient:

[0135] a (0) [n]=ΔV[n], n=0,...,M-1, Formula (8)

[0136] to a (0) [n] Perform U-layer wavelet packet decomposition and recursively obtain the approximation coefficient of the v-th subband of the r+1th layer and detail coefficient

[0137]

[0138] Where r is the decomposition layer index, r=0,...,U-1, U is the total number of decomposition layers, v=1,...,2 r is the subband index of the rth layer; h[p] and g[p] are the low-pass filter coefficient and high-pass filter coefficient of the Db4 mother wavelet respectively, and p is the filter coefficient index;

[0139] For all approximation coefficient sets of the Uth layer and detail coefficient set Calculate the time domain energy separately:

[0140]

[0141] Where N v is the sub-band coefficient length of the v-th sub-band coefficient of the U-th layer.

[0142] Here, by statistically analyzing the energy of each decomposed subband coefficient, a set of multi-scale energy values is obtained, which directly reflects the relative intensity and time evolution characteristics of different physical processes in the entire relaxation process, providing quantitative and highly differentiated input.

[0143] More specifically, DaubechiesDb4 was chosen, as it has a reasonable fourth-order support length, sufficient orthogonality, and time-frequency localization, making it suitable for oscillatory decaying signals. The filter coefficients h[p] and g[p] are derived from the Daubechies definition and can be precomputed and hard-coded.

[0144] In addition, the number of decomposition layers is selected according to the total duration of the signal and the target frequency resolution, and can be selected as 4 or 6 layers. The higher the number of layers, the finer the frequency band division, and the more degradation mechanisms can be distinguished, but the amount of computation and feature dimension increase exponentially. Each layer simultaneously filters and downsamples the subband coefficients (approximation and details) to obtain 2 U sub-bands, each sub-band corresponds to a narrow frequency band.

[0145] Furthermore, the approximation coefficient and detail coefficient of the U-th layer decomposition output are totaled 2 U Each subband is independently counted, and the square sum of all coefficients within each subband is performed. This can lead to large energy differences that can affect model convergence. Therefore, energy values directly correspond to signal strengths in different frequency bands, making them easy to compare and sort. High-frequency subbands naturally have low noise energy, while low-frequency subbands carry the majority of attenuation information, so energy statistics inherently suppress noise.

[0146] Approximating each energy {Q a,v} and detail energy {Q q,v}Arrange in ascending order by subband index to form a multi-scale energy feature vector:

[0147]

[0148] Where, e is the multi-scale voltage relaxation energy eigenvector, which is used to characterize the relaxation attenuation process at different time scales. U is the total number of subbands after decomposition.

[0149] Here, by arranging the multi-scale energy values into a vector of fixed dimension, it can be directly used as the input of the deep fusion network or time series regression model, which helps to simplify the data preprocessing process.

[0150] Compared with the traditional exponential fitting which can only distinguish two attenuation stages, the embodiment of the present application adopts wavelet packet decomposition to analyze the attenuation process up to two stages. L The time-frequency decomposition is refined to the energy of each frequency band. By performing refined time-frequency decomposition and energy quantization on the terminal voltage relaxation signal, various complex coupling mechanisms such as double layer, electrolyte diffusion, and solid phase diffusion can be accurately distinguished, thereby improving the distinguishability and robustness of multi-physical degradation processes.

[0151] Figure 5 A structural connection diagram of an example of a battery SOH prediction model according to an embodiment of the present application is shown.

[0152] It should be noted that while impedance morphology embedding, multi-scale decomposition of relaxation energy, and the dynamic equilibrium imbalance index together constitute a three-dimensional depiction of the multi-mechanism degradation within the battery, they are essentially a collection of high-dimensional or time-series features and cannot be directly used to guide a single numerical output for SOH. Therefore, the battery SOH prediction model analyzes these multimodal, high-dimensional comprehensive features to obtain the final SOH indicator output.

[0153] like Figure 5 As shown, the battery SOH prediction model 500 includes a static feature encoding module 510 , a modulation coefficient generation module 520 , a dynamic feature time series extraction module 530 , a time series pooling module 540 and a regression prediction module 550 .

[0154] The static feature encoding module 510 is used to concatenate the impedance morphology embedding vector E and the multi-scale voltage relaxation energy feature vector e to form a static joint feature S raw =[E;e]; Subsequently, the static feature encoding module uses a multi-layer fully connected network and self-attention mechanism to encode S raw Perform nonlinear mapping and reweighting, and output the context vector S.

[0155] Here, the static feature encoding module is responsible for fusing two essentially independent, high-dimensional static information types—electrochemical morphology and relaxation energy—into a contextual representation that is discriminative of battery degradation. The impedance morphology embedding vector E captures the geometric features of the EIS curve, such as the size of the semicircle, the low-frequency dip, and the Warburg line, while the multi-scale voltage relaxation energy feature vector e separates the energy distributions of the double layer, liquid phase, and solid phase diffusion processes in the time-frequency domain. Because directly concatenating these two components can easily lead to excessive dimensionality and make it difficult to balance their weights, a nonlinear mapping and attention mechanism are used to automatically learn the synergistic and mutually exclusive relationships between different components, compressing and reweighting them into a fixed-dimensional context vector S. This effectively compresses the high-dimensional static features, and the self-attention module automatically assigns higher weights to key channels, improving sensitivity to subtle differences.

[0156] The modulation coefficient generation module 520 is used to input the context vector S into the parameter mapping network, and generate the feature modulation coefficient set {γ f ,β f |f=1,...,G}; where G represents the total number of dilated convolutional layers, γ f and β f Represent the channel scaling coefficient vector and channel offset coefficient vector of the f-th layer respectively, and each pair of γ f ,β f Used to modulate the channel activations of the f-th dilated convolutional layer.

[0157] Here, inspired by the concept of feature-by-feature linear modulation (FiLM), the static context S and the dynamic temporal network are bridged through the modulation coefficient generation module, so that the static features can "guide" each layer of dynamic feature extraction, that is, through γ f ,β f The channel activations of the f-th layer of temporal convolution are linearly scaled and offset element by element, thereby achieving cross-modal information transfer and modulation. In this way, the dynamic temporal network no longer "blindly calculates" the equilibrium imbalance index sequence, but instead dynamically adjusts the attention of each convolutional layer to changes at different time scales based on the current context of electrochemical morphology and diffusion characteristics.

[0158] Therefore, the impedance morphology embedding vector and multi-scale voltage relaxation energy features jointly construct a high-dimensional contextual representation of the electrochemical interface and diffusion process in the static encoding module. By organically integrating the influence of static information on the dynamic extraction of each layer of features, the model can fundamentally distinguish different degradation mechanisms.

[0159] The dynamic feature timing extraction module 530 is used to receive the dynamic balance imbalance index sequence M bal(t), and extract the temporal features layer by layer through the G-layer dilated convolution unit, and calculate each layer

[0160]

[0161] and follow the characteristic modulation

[0162]

[0163] Among them, h (f) (t) is the activation output of the fth layer, which represents the feature vector after convolution and activation of the layer at sampling time t, h (0) (t) = M bal (t); Indicates that the expansion rate d is 2 f-1 One-dimensional convolution operation; ⊙ represents element-wise product operation, is the activation after modulation of the fth layer, which represents the output feature vector of the layer after context-driven channel-level modulation at sampling time t.

[0164] It should be noted that the dynamic balance imbalance index M bal (t) is a high-frequency time series signal whose degradation information is distributed across short-term mutations, periodic fluctuations, and long-term trends. A dilated convolutional network (TCN) is chosen over a recurrent neural network (RNN) to enable parallel computation and capture diverse temporal dependencies through exponential dilation. Combined with a modulation coefficient module, a context-driven linear modulation of channel activations is performed after each convolution layer, enabling the model to emphasize responses at different timescales in different recurrent nodes.

[0165] Therefore, based on the context-driven channel-level modulation mechanism, the responses of each layer of the dynamic equilibrium imbalance index sequence in the temporal convolutional network are precisely amplified or suppressed, ensuring that the model is sensitive to key inconsistency fluctuations rather than being smoothed out.

[0166] The time series pooling module 540 is used to extract the modulated feature sequence output by the dynamic feature time series extraction module. Global average pooling is performed along the time dimension to generate a pooled vector P.

[0167] It should be understood that the output of a TCN is still a sequence that evolves over time. To integrate it with the static context and feed it into the regression network, the temporal features need to be aggregated into a fixed-length vector. Global average pooling (or attention pooling) can highlight the overall trend of the sequence while ensuring time invariance.

[0168] The regression prediction module 550 is used to concatenate the pooling vector P and the context vector S to form a fusion feature

[0169] [P; S], and output the estimated SOH value of the battery pack through a multi-layer fully connected regression network.

[0170] The regression prediction module 550 takes the fused static context S and the temporal pooling vector P as input, and maps the multi-source feature space to the SOH estimation scalar through deep nonlinear mapping. Through multi-layer nonlinear regression mapping, while retaining the richness of the above-mentioned multi-source information, it effectively suppresses overfitting and improves generalization ability through hierarchical feature fusion and regularization.

[0171] Through the embodiments of the present application, deep coding and fusion of multi-source heterogeneous features are introduced to achieve multi-angle, multi-scale and dynamic adaptive analysis of the battery health status, effectively enhancing the accuracy and reliability of the SOH assessment results.

[0172] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of combined actions, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application. In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0173] Figure 6 A structural block diagram of an example of an electric vehicle battery health assessment system based on multi-source data fusion according to an embodiment of the present application is shown.

[0174] like Figure 6 As shown, the electric vehicle battery health assessment system 600 based on multi-source data fusion includes a multi-source data acquisition unit 610, an impedance morphology encoding unit 620, a voltage relaxation multi-scale extraction unit 630, a dynamic balance imbalance index calculation unit 640 and a battery pack SOH prediction unit 650.

[0175] The multi-source data acquisition unit 610 is used to synchronously trigger electrochemical impedance spectroscopy scanning, terminal voltage relaxation recording and battery pack balancing operation based on the constant current phase cut-off event of the battery pack of the electric vehicle, so as to obtain corresponding Nyquist impedance curves, terminal voltage relaxation time series data and balancing current time series data respectively; the terminal voltage relaxation time series data includes multiple sampling moments and corresponding terminal voltage values, and the balancing current time series data includes multiple sampling moments and corresponding balancing current values.

[0176] The impedance morphology encoding unit 620 is configured to extract an impedance morphology embedding vector corresponding to the Nyquist impedance curve based on an autoencoder.

[0177] The voltage relaxation multi-scale extraction unit 630 is configured to perform wavelet packet decomposition on the terminal voltage relaxation time series data to obtain a corresponding multi-scale voltage relaxation energy feature vector.

[0178] The dynamic balance imbalance index calculation unit 640 is configured to calculate a dynamic balance imbalance index based on the terminal voltage relaxation time series data and the balance current time series data:

[0179]

[0180] Where M bal (t) is the dynamic balance imbalance index; Δt represents the sampling time interval, σ V (t) is the standard deviation of all monomer terminal voltages at sampling time t, is the regularization constant, ΔI bal (t) is the balancing current increment, which represents the difference in balancing current between two adjacent sampling times t-Δt and t.

[0181] The battery pack SOH prediction unit 650 is configured to input the impedance morphology embedding vector, the multi-scale voltage relaxation energy feature vector, and the dynamic balance imbalance index into a battery SOH prediction model to output an estimated SOH value of the battery pack.

[0182] In some embodiments, an embodiment of the present application provides a non-volatile computer-readable storage medium, which stores one or more programs including execution instructions, and the execution instructions can be read and executed by an electronic device (including but not limited to a computer, server, or network device, etc.) to execute any of the steps of the above-mentioned electric vehicle battery health assessment method based on multi-source data fusion in the present application.

[0183] In some embodiments, the embodiments of the present application also provide a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer performs any step of the above-mentioned electric vehicle battery health assessment method based on multi-source data fusion.

[0184] In some embodiments, an embodiment of the present application also provides an electronic device comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the steps of the electric vehicle battery health assessment method based on multi-source data fusion.

[0185] Figure 7 This is a hardware structure diagram of an electronic device for executing an electric vehicle battery health assessment method based on multi-source data fusion provided by another embodiment of the present application, such as Figure 7 As shown, the device includes:

[0186] One or more processors 710 and memory 720, Figure 7 A processor 710 is taken as an example.

[0187] The device for executing the electric vehicle battery health assessment method based on multi-source data fusion may further include: an input device 730 and an output device 740 .

[0188] The processor 710, the memory 720, the input device 730 and the output device 740 may be connected via a bus or other means. Figure 7 The bus connection is taken as an example.

[0189] Memory 720, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the electric vehicle battery health assessment method based on multi-source data fusion in the embodiments of this application. Processor 710 executes the non-volatile software programs, instructions, and modules stored in memory 720 to execute various server functional applications and data processing, thereby implementing the electric vehicle battery health assessment method based on multi-source data fusion in the aforementioned method embodiment.

[0190] The memory 720 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device, etc. In addition, the memory 720 may include a high-speed random access memory and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 720 may optionally include a memory remotely located relative to the processor 710, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0191] The input device 730 may receive input digital or character information and generate signals related to user settings and function control of the electronic device. The output device 740 may include a display device such as a display screen.

[0192] The one or more modules are stored in the memory 720 and, when executed by the one or more processors 710 , perform the electric vehicle battery health assessment method based on multi-source data fusion in any of the above method embodiments.

[0193] The above-mentioned product can execute the method provided in the embodiment of this application, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not fully described in this embodiment, please refer to the method provided in the embodiment of this application.

[0194] The electronic devices of the embodiments of the present application exist in various forms, including but not limited to:

[0195] (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and their primary purpose is to provide voice and data communications. These terminals include smartphones, multimedia phones, feature phones, and low-end phones.

[0196] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers and have computing and processing capabilities, and generally also have mobile Internet access. These terminals include PDAs, MIDs, and UMPCs.

[0197] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players, handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0198] (4) Other onboard electronic devices with data interaction functions, such as onboard computer devices installed in vehicles.

[0199] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0200] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a general hardware platform, or of course, by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the relevant technology, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiment.

[0201] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for evaluating the health of electric vehicle batteries based on multi-source data fusion, comprising: Based on the cutoff event of the constant current phase of the battery pack for electric vehicles, the electrochemical impedance spectroscopy scan, terminal voltage relaxation recording and battery pack balancing operation are synchronously triggered to obtain the corresponding Nyquist impedance curve, terminal voltage relaxation time series data and balancing current time series data respectively; The terminal voltage relaxation time series data includes a plurality of sampling moments and corresponding terminal voltage values, and the balancing current time series data includes a plurality of sampling moments and corresponding balancing current values; Extracting an impedance morphology embedding vector corresponding to the Nyquist impedance curve based on an autoencoder; performing wavelet packet decomposition on the terminal voltage relaxation time series data to obtain a corresponding multi-scale voltage relaxation energy feature vector; Based on the terminal voltage relaxation time series data and the balancing current time series data, a dynamic balancing imbalance index is calculated: Where M bal (t) is the dynamic balance imbalance index; Δt represents the sampling time interval, σ V (t) is the standard deviation of all monomer terminal voltages at sampling time t, is the regularization constant, ΔI bal (t) is the balancing current increment, which represents the difference in balancing current between two adjacent sampling times t-Δt and t; The impedance morphology embedding vector, the multi-scale voltage relaxation energy feature vector, and the dynamic balance imbalance index are input into a battery SOH prediction model to output an estimated SOH value of the battery pack.

2. The method according to claim 1, wherein The method for synchronously triggering electrochemical impedance spectroscopy scanning, terminal voltage relaxation recording, and battery pack balancing operations based on a constant current phase cutoff event for an electric vehicle battery pack includes: The charge and discharge current is sampled in real time at the shunt resistor of the battery pack, and the sampling signal is processed by first-order low-pass filtering and sliding average to obtain a smooth current curve Will With the preset threshold I thr For comparison, when And the duration is not less than the debounce time t deb When , it is determined that the constant current stage is cut off; When it is determined that the constant current stage is terminated, a hardware trigger message is generated and broadcasted via the vehicle communication bus to synchronously trigger electrochemical impedance spectroscopy scanning, terminal voltage relaxation recording, and battery pack balancing operations; the hardware trigger message includes a cycle number and a timestamp of the sampling time.

3. The method according to claim 1, wherein The encoder structure of the autoencoder includes a cascaded input preprocessing layer, a multi-scale dilated convolution layer, a channel attention pooling layer, and a fully connected dimensionality reduction layer; The input preprocessing layer is used to perform the following operations: The obtained discrete impedance spectrum point set The Nyquist curve is obtained by interpolation and arc length parameterization, so that the arc length of the curve is: The Nyquist curve is divided into a resampling point sequence P according to an equidistant arc length interval L / N. j =(x j ,y j ), L is the total arc length of the Nyquist curve, N is the total number of resampling points, j = 1, ..., N; where ω i represents the frequency of the i-th sinusoidal excitation signal, and are the measured excitation frequencies ω i The real and imaginary values of the electrochemical impedance of the battery pack are shown in Figure 2. K represents the total number of discrete frequency points collected for the Nyquist curve, τ is the curve parameter, and s(t) represents the geometric arc length from the starting point of the curve to the parameter value t; P j represents the Nyquist plane coordinates of the jth resampling point, x j Corresponding to the real part, y j Corresponding to the imaginary part; After stacking each resampling point row by row, we get the input matrix H (0) , each row of the matrix is (x j ,y j ); The multi-scale dilated convolutional layer is used to perform the following operations: The input matrix H (0) Input three layers of one-dimensional dilated convolution, each layer is Iteration Among them, H (l) represents the intermediate features of the lth layer, Indicates that the expansion rate d is 2 l One-dimensional convolution operation, b (l) is the bias of the lth layer, C is the number of channels; The channel attention pooling layer is used to perform the following operations: For H (3) Each row vector of Calculating attention weights Where, α j represents the attention weight corresponding to the j-th resampling point, represents the attention mapping matrix, is the attention weight vector; Compute the global morphological eigenvector: The fully connected dimension reduction layer is used to perform the following operations: The global morphological feature vector z is mapped through two-level full connection to output impedance morphological embedding vector Where, Represents the first layer of dimensionality reduction mapping matrix, represents the first layer bias vector, Represents the second-layer dimensionality reduction mapping matrix, represents the second layer bias vector, D is the dimension of the impedance morphology embedding vector, and E represents the impedance morphology embedding vector.

4. The method according to claim 1, wherein The performing wavelet packet decomposition on the terminal voltage relaxation time series data to obtain a corresponding multi-scale voltage relaxation energy feature vector includes: Relax the discrete sequence of terminal voltages obtained Applying the bandpass filter operator BP{·} and subtracting the final steady-state voltage V[M-1] yields the correction sequence: ΔV[n]=BP{V[n]}-V[M-1], Where ΔV[n] is the corrected voltage attenuation value, representing the terminal voltage relaxation amount after removing the steady-state baseline at the nth sampling moment; V[n] is the terminal voltage value measured at the nth sampling moment; BP{·} is a bandpass filter operator with a preset cutoff frequency; M is the total number of discrete moments sampled in the relaxation record; V[M-1] is the steady-state terminal voltage, representing the terminal voltage baseline value corresponding to the last sampling point in the sequence; Set ΔV[n] as the initial approximation coefficient: a (0) [n]=ΔV[n],n=0,...,M-1, to a (0) [n] Perform U-layer wavelet packet decomposition and recursively obtain the approximation coefficient of the v-th subband of the r+1th layer and detail coefficient Where r is the decomposition layer index, r=0,...,U-1, U is the total number of decomposition layers, v=1,...,2 r is the subband index of the rth layer; h[p] and g[p] are the low-pass filter coefficient and high-pass filter coefficient of the Db4 mother wavelet respectively, and p is the filter coefficient index; For all approximation coefficient sets of the Uth layer and detail coefficient set Calculate the time domain energy separately: Where N v is the sub-band coefficient length of the v-th sub-band coefficient of the U-th layer; Approximating each energy {Q a,v } and detail energy {Q q,v }Arrange in ascending order by subband index to form a multi-scale energy feature vector: Where, e is the multi-scale voltage relaxation energy eigenvector, which is used to characterize the relaxation attenuation process at different time scales. U is the total number of subbands after decomposition.

5. The method according to claim 1, wherein After calculating the dynamic balancing imbalance index based on the terminal voltage relaxation time series data and the balancing current time series data, the method further includes: The dynamic balance imbalance index is compared with a preset balance imbalance threshold, and when the dynamic balance imbalance index exceeds the balance imbalance threshold, an alarm operation for a battery cell inconsistency fault is triggered.

6. The method according to claim 1, wherein The battery SOH prediction model includes a static feature encoding module, a modulation coefficient generation module, a dynamic feature time series extraction module, a time series pooling module and a regression prediction module; The static feature encoding module is used to concatenate the impedance morphology embedding vector E and the multi-scale voltage relaxation energy feature vector e to form a static joint feature S raw =[E;e]; Then, the static feature encoding module uses a multi-layer fully connected network and self-attention mechanism to encode S raw Perform nonlinear mapping and reweighting to output the context vector S; The modulation coefficient generation module is used to input the context vector S into the parameter mapping network, and generate the feature modulation coefficient set {γ corresponding to each dilated convolution layer of the dynamic feature timing extraction module through at least two layers of full connection transformation. f ,β f |f=1,...,G}; where G represents the total number of dilated convolutional layers, γ f and β f Represent the channel scaling coefficient vector and channel offset coefficient vector of the f-th layer respectively, and each pair of γ f ,β f Used to modulate the channel activation of the f-th dilated convolutional layer; The dynamic feature timing extraction module is used to receive the dynamic balance imbalance index sequence M bal (t), and extract the temporal features layer by layer through the G-layer dilated convolution unit, and calculate each layer and follow the characteristic modulation Among them, h (f) (t) is the activation output of the fth layer, which represents the feature vector after convolution and activation of the layer at sampling time t, h (0) (t) = M bal (t); Indicates that the expansion rate d is 2 f-1 One-dimensional convolution operation; ⊙ represents element-wise product operation, is the modulated activation of the fth layer, which represents the output feature vector of the layer after context-driven channel-level modulation at sampling time t; The time series pooling module is used to process the modulated feature sequence output by the dynamic feature time series extraction module. Perform global average pooling along the time dimension to generate a pooled vector P; The regression prediction module is used to splice the pooling vector P with the context vector S to form a fusion feature [P; S], and output the estimated SOH value of the battery pack through a multi-layer fully connected regression network.

7. An electric vehicle battery health assessment system based on multi-source data fusion, comprising: A multi-source data acquisition unit is used to synchronously trigger electrochemical impedance spectroscopy scanning, terminal voltage relaxation recording, and battery pack balancing operations based on the constant current phase cutoff event of the battery pack of the electric vehicle, so as to obtain the corresponding Nyquist impedance curve, terminal voltage relaxation time series data, and balancing current time series data respectively; The terminal voltage relaxation time series data includes a plurality of sampling moments and corresponding terminal voltage values, and the balancing current time series data includes a plurality of sampling moments and corresponding balancing current values; An impedance morphology encoding unit, configured to extract an impedance morphology embedding vector corresponding to the Nyquist impedance curve based on an autoencoder; a voltage relaxation multi-scale extraction unit, configured to perform wavelet packet decomposition on the terminal voltage relaxation time series data to obtain a corresponding multi-scale voltage relaxation energy feature vector; A dynamic balance imbalance index calculation unit is used to calculate a dynamic balance imbalance index based on the terminal voltage relaxation time series data and the balance current time series data: Where M bal (t) is the dynamic balance imbalance index; Δt represents the sampling time interval, σ V (t) is the standard deviation of all monomer terminal voltages at sampling time t, is the regularization constant, ΔI bal (t) is the balancing current increment, which represents the difference in balancing current between two adjacent sampling times t-Δt and t; A battery pack SOH prediction unit is configured to input the impedance morphology embedding vector, the multi-scale voltage relaxation energy feature vector, and the dynamic balance imbalance index into a battery SOH prediction model to output an estimated SOH value of the battery pack.

Images