Power battery state monitoring system and method based on identification theory and topology measurement

Abstract

The invention provides a power battery state monitoring system and method based on an identification theory and topology measurement, and relates to the technical field of lithium ion battery state monitoring, and the method comprises the steps: obtaining original feature data in a battery charging process, and carrying out the preprocessing of the original feature data; taking a constant-current charging response of the battery as a step response of a dynamic system, taking a low-order process model as an equivalent dynamic feature extractor, taking the preprocessed original feature data as an input fitting voltage response track of the low-order process model, and outputting an identified model parameter; constructing a dynamic transfer function, substituting the identified model parameters into the dynamic transfer function to obtain a current dynamic transfer function model, and calculating the topology gap measurement of the current model and the reference model on the frequency domain; and constructing a regression mapping model, inputting the calculated topology gap measurement into the regression mapping model, and mapping and outputting to obtain an SOH estimated value. According to the invention, high-robustness and high-precision SOH estimation can be realized.

Description

Technical Field

[0001] This disclosure relates to the field of lithium-ion battery state monitoring technology, specifically to a power battery state monitoring system and method based on identification theory and topology metrics. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] With the rapid development of new energy vehicles and energy storage power stations, lithium-ion batteries, as core energy storage components, require accurate assessment of their state of health (SOH) to ensure system safety, extend service life, and optimize charging and discharging strategies. SOH is typically defined as the ratio of the current maximum available capacity to the factory rated capacity.

[0004] However, in practical engineering applications, existing SOH estimation techniques still have the following limitations: (1) Difficulty in obtaining data: The traditional ampere-hour integration method relies on the battery to complete charge-discharge cycles from 0% to 100% to calibrate capacity. However, in actual user scenarios (such as daily commuting of electric vehicles), batteries are often in a random "fragmented" usage state, making it difficult to obtain complete capacity data, resulting in the inability to close-loop calibrate SOH estimation and large cumulative errors.

[0005] (2) Poor anti-interference capability of feature extraction: Existing voltage feature-based methods (such as direct voltage difference method and incremental capacity analysis method ICA) directly perform differentiation or differential processing on the acquired voltage sequence. This processing method amplifies current fluctuations and sensor sampling noise, resulting in the extracted feature curves (such as IC curves or internal resistance curves) being full of spikes and oscillations, making them difficult to use for high-precision state estimation.

[0006] (3) Unstable model parameter identification: The method based on the equivalent circuit model (ECM) attempts to identify micro-parameters such as ohmic internal resistance and polarization capacitance online. However, the constant current charging condition is similar to a simple step excitation, which contains limited frequency domain information, making the parameter identification process of high-order models prone to divergence (non-convergence), or the identified parameters may jump drastically between adjacent cycles, lacking physical consistency. Summary of the Invention

[0007] To address the aforementioned issues, this disclosure proposes a power battery state monitoring system and method based on identification theory and topology metrics. This system does not rely on complete cycles but utilizes only short segments of data from the constant current charging process. Noise is filtered out and features are extracted using a low-order process model, and the dynamic characteristic shift of the system is quantified using topology gap metrics, thereby achieving highly robust and accurate SOH estimation.

[0008] According to some embodiments, the present disclosure adopts the following technical solutions: A power battery state monitoring method based on identification theory and topology metrics includes: Obtain raw feature data during the battery charging process and preprocess it; The constant current charging response of the battery is regarded as the step response of the dynamic system. A low-order process model is used as the equivalent dynamic feature extractor. The preprocessed original feature data is used as the input of the low-order process model to fit the voltage response trajectory, and the identified model parameters are output. Construct a dynamic transfer function, substitute the identified model parameters into the dynamic transfer function to obtain the current dynamic transfer function model, and calculate the topological gap metric between the current model and the benchmark model in the frequency domain. A regression mapping model is constructed, and the calculated topological gap metric is input into the regression mapping model. The mapping output yields the estimated SOH value.

[0009] According to some embodiments, the present disclosure adopts the following technical solutions: The constant current segment identification and preprocessing module is used to acquire the raw feature data during the battery charging process and preprocess it. The feature extraction module treats the constant current charging response of the battery as a step response of a dynamic system. It uses a low-order process model as an equivalent dynamic feature extractor, takes the preprocessed raw feature data as input to the low-order process model, fits the voltage response trajectory, and outputs the identified model parameters. It constructs a dynamic transfer function, substitutes the identified model parameters into the dynamic transfer function to obtain the current dynamic transfer function model, and calculates the topological gap metric between the current model and the benchmark model in the frequency domain. The State Estimation module is used to construct a regression mapping model. The calculated topological gap metric is input into the regression mapping model, and the mapping output yields the SOH estimate.

[0010] Compared with the prior art, the beneficial effects of this disclosure are as follows: This disclosure presents a power battery state monitoring method based on identification theory and topology metric. This method does not rely on a complete cycle, but only uses short-segment data during constant current charging. It filters out noise and extracts features through a low-order process model (preferably first-order), and uses a topology gap metric (preferably Vinnicombe) to quantify the dynamic characteristic shift of the system, thereby achieving highly robust and accurate SOH estimation.

[0011] The power battery state monitoring method disclosed herein, based on identification theory and topology metric, forcibly filters out unstructured high-frequency measurement noise through the structural constraints of the process model, exhibiting extremely high noise resistance. This results in smooth and stable extracted topological gap feature curves, with a signal-to-noise ratio significantly superior to the traditional internal resistance method.

[0012] The power battery state monitoring method disclosed herein, based on identification theory and topological metric, has a process model corresponding to the low-frequency intercept and polarization establishment process in the battery electrochemical impedance spectrum, and the topological gap metric quantifies the degree of degradation of this impedance structure at the system level.

[0013] This disclosure presents a power battery state monitoring method based on identification theory and topological metrics. Unlike the traditional ampere-hour integration method, which relies on historical states and is prone to cumulative errors, this method employs a "single-frame independent estimation" mechanism, which can estimate the current state of energy (SOH) based solely on the characteristics of the current cycle. Experiments show that the root mean square error (RMSE) of SOH estimation over the entire life cycle is controlled within 1.5%, demonstrating extremely high engineering practical value.

[0014] The power battery state monitoring method disclosed herein, based on identification theory and topological metric, is experimentally demonstrated to exhibit a strict monotonic change trend in features as the battery ages, and has an extremely high goodness of fit with the true state of equilibrium (SOH) (R2>0.95). This method eliminates multi-valued ambiguity and significantly improves estimation accuracy. This disclosure demonstrates excellent monotonicity and accuracy. Attached Figure Description

[0015] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0016] Figure 1 This is a schematic diagram of the data preprocessing process according to an embodiment of the present disclosure; in Figure 1 (a) in the data represents the original charge / discharge data. Figure 1 (b) in the image represents the extracted constant current segment. Figure 1 (c) in the figure represents the normalized sequence after resampling and debiasing.

[0017] Figure 2 This is a graph showing the evolution of the extracted topological gap metric as a function of battery cycle count in an embodiment of this disclosure. Figure 3 This is a mapping diagram between the topological gap measurement and the actual SOH in the embodiments of this disclosure; Figure 4 The diagram shows a comparison between the method of this disclosure embodiment and the prior art (traditional internal resistance method); Figure 5 The diagram shows the SOH estimation results and error verification of the method in this embodiment over the entire battery life cycle. Figure 6 This is a diagram illustrating the overall system logic architecture of an embodiment of this disclosure.

[0018] Figure 7This is a schematic diagram of the hardware architecture of the battery management system according to an embodiment of the present disclosure. Detailed Implementation

[0019] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0020] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0021] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0022] Example 1 One embodiment of this disclosure provides a method for monitoring the state of a power battery based on identification theory and topology metrics. The method includes the following steps: Step 1: Obtain raw feature data during the battery charging process and preprocess it; Step 2: Treat the constant current charging response of the battery as the step response of the dynamic system, use a low-order process model as the equivalent dynamic feature extractor, use the preprocessed original feature data as the input of the low-order process model to fit the voltage response trajectory, and output the identified model parameters. Step 3: Construct the dynamic transfer function, substitute the identified model parameters into the dynamic transfer function to obtain the current dynamic transfer function model, and calculate the topological gap metric between the current model and the baseline model in the frequency domain. Step 4: Construct a regression mapping model. Input the calculated topological gap metric into the regression mapping model, and the mapping output will yield the SOH estimate.

[0023] As one embodiment, the power battery state monitoring method based on identification theory and topological metric disclosed herein does not rely on a complete cycle, but only utilizes short segments of data during the constant current charging process. It filters noise and extracts features using a low-order process model (preferably first-order), and quantifies the dynamic characteristic shift of the system using a topological gap metric (preferably Vinnicombe), thereby achieving highly robust and accurate SOH estimation. This solves the problems of low SOH estimation accuracy and poor feature monotonicity in existing technologies under non-full charge and high-noise conditions. The specific implementation process of this method includes: Step 1: Obtain the raw feature data during the battery charging process and preprocess it; Specifically, step 11: Obtain raw feature data during the battery charging process, including: The battery management system (BMS) is used to collect voltage and current data during the battery charging process. When a constant current charging segment is detected and the voltage is within a preset electrochemical sensitive range, the segment is extracted as the raw feature data.

[0024] Furthermore, the criteria for determining the constant current charging segment are: current I > 0.05C and current fluctuation variance ≤ 0.01C² (C is the rated capacity of the battery); the electrochemical sensitive range is determined by the battery electrochemical phase change test, and the range with the largest polarization voltage change rate is selected, wherein the preferred ranges are [3.65V, 4.15V] for ternary lithium batteries, [3.2V, 3.65V] for lithium iron phosphate batteries, and [3.7V, 4.2V] for lithium cobalt oxide batteries; Step 12: Preprocess the original feature data; Specifically, the original feature data is resampled and debiased, then mapped onto a unified time grid to obtain a standardized input-output sequence, including: (1) Detrending: Subtract the initial value from the extracted voltage and current sequences so that they change from zero.

[0025]

[0026]

[0027] in, V raw(k) , I raw(k) The processed number of the first k One voltage and current sampling point, , These are the voltage and current values ​​at the start of the segment.

[0028] (2) Resampling: Set a globally uniform sampling interval T s (Preferred duration: 5 seconds). Calculate the total duration of the segment and generate a standard uniform timeline. t query =[0,Ts,2*Ts,...]. Using linear interpolation, the original non-uniform or indeterminate length... and Mapped to t query The above yields a standardized input sequence of fixed length. u ( k(corresponding current) and output sequence y ( k (Corresponding voltage change).

[0029] Step 2: Treat the constant current charging response of the battery as a step response of a dynamic system. Use a low-order process model as an equivalent dynamic feature extractor. Use the preprocessed raw feature data as input to the low-order process model to fit the voltage response trajectory. Output the identified model parameters and construct a dynamic transfer function. Substitute the identified model parameters into the dynamic transfer function to obtain the current dynamic transfer function model. Calculate the topological gap metric between the current model and the baseline model in the frequency domain. The specific process includes: This step in the disclosure aims to extract deep electrochemical kinetic decay characteristics from seemingly stable constant-current charging-voltage curves. Specifically, it includes the following two sub-steps: Step 21: Process Model Identification; Although constant current charging appears as DC input on a macroscopic level, at the microscopic electrochemical level, the response of the battery terminal voltage is not instantaneous, but rather involves a dynamic process consisting of "ohmic voltage drop (transient)" and "polarization voltage establishment (transient)". As the battery ages, the lithium-ion diffusion rate inside the active material decreases, causing the time required for the polarization voltage to reach a steady state to increase, i.e., the "curvature" or "transient trajectory" of the voltage response curve changes.

[0030] Based on this physical fact, firstly, this disclosure regards the constant current charging process as a step excitation applied to the battery system, and innovatively adopts a first-order process model (P1) as an equivalent dynamic feature extractor. The preprocessed original feature data is used as the input of the low-order process model to fit the voltage response trajectory within the segment, and the output is the identified optimal model parameters.

[0031] Although a battery approximates an integral process under long-term constant current conditions, within a short time segment, the first-order process model can utilize its mathematical initial value response characteristics, through parameters... K p and T p The combination of these features simultaneously fits the slope characteristics of the voltage rise and the nonlinear curvature characteristics of the voltage curve.

[0032] At this point, the identified parameters K p and T pNo longer simply representing steady-state gain and time constant, these parameters serve as a set of generalized waveform structure parameters to characterize the topological shape of the voltage response trajectory under current aging conditions. As the battery's state of equilibrium (SOH) decays, the increase in internal impedance and the decrease in diffusion rate cause subtle changes in the trajectory shape, which are then precisely captured by the model parameters.

[0033] Secondly, the dynamic transfer function is established, and the transfer function G(s) is established as follows:

[0034] Where G(s) is the equivalent dynamic transfer function under the current battery state, used to describe the standardized voltage output. For current input The response relationship. K p Process gain, physically representing the steady-state impedance magnitude of a battery, is caused by aging. K p An increase corresponds to an increase in the voltage polarization amplitude. T p This is the time constant, which physically reflects how quickly the battery's polarization voltage builds up and is related to the lithium-ion diffusion rate. Although the input is a constant current, T p It keenly captures the hysteresis effect during the transition of the voltage curve from the initial moment to the steady-state rising slope. The more severe the battery aging, the slower the ion diffusion, and the more pronounced the hysteresis effect, the more effectively it can be identified. T p The larger the value.

[0035] Furthermore, using the least squares method or iterative optimization algorithms (such as the `procest` function in MATLAB), based on the obtained standardized input and output sequences, the optimal model parameters at the current time are identified with the objective of minimizing the error between the model's predicted output and the actual output sequence. K p and T p By substituting the identified parameters into the dynamic transfer function structure described above, a dynamic transfer function model for the current battery state can be constructed. P curr This model essentially acts as a physically constrained low-pass filter, extracting aging features while forcibly filtering out unstructured high-frequency noise from the measurement data.

[0036] Step 22: Gap Metric Calculation; In the first cycle of the battery's entire life cycle, a baseline model is identified using the above method, denoted as... P ref .

[0037] For any point in the subsequent operation, calculate the current dynamic transfer function model. P curr Compared with the benchmark model P ref The Vinnicombe topological gap measure between them is denoted as δ g .

[0038] Topological gap measurement δ g The definition is based on graph metric theory, and its mathematical essence is to measure the maximum distance between two linear dynamic systems under closed-loop feedback. Its calculation formula involves optimizing the maximum singular value of the system's normalized coprime factor:

[0039] in, δ g This is a dimensionless metric for topological gap, ranging from [0, 1]. 0 indicates that the dynamic characteristics of the two systems are completely identical, and 1 indicates that they are completely orthogonal. P ref and P curr Let N and M represent the transfer functions of the baseline model and the current model, respectively; (N, M) are the normalized right coprime factors of the transfer functions. w For frequency; σ max The maximum singular value is j; j is the imaginary unit. N ref , M ref As a benchmark model P ref Normalized right coprime factors satisfy , N curr , M curr For the current model P curr Normalized right coprime factors satisfy ; It is the maximum singular value operator.

[0040] Furthermore, in practical engineering implementation, this indicator directly reflects the "drift distance" of the current dynamic transfer function model of the battery relative to the new battery baseline model in the frequency domain topology space. It comprehensively reflects the overall difference between the changes in steady-state gain (impedance amplitude) and time constant (polarization response speed) caused by aging. For example... Figure 2 As shown, with the increase of the number of iterations, δ g It gradually and monotonically increases from 0, and the curve is very smooth.

[0041] Step 3: Construct a regression mapping model. Input the calculated topological gap metric into the regression mapping model, and the mapping output will yield the SOH estimate.

[0042] Specifically, a topological gap metric is established based on offline experimental data. δ g Mapping relationship with the actual SOH.

[0043] like Figure 3 As shown, experimental data δ g A significant nonlinear monotonic relationship exists between the expression and SOH. To improve accuracy, this disclosure employs a quadratic polynomial regression model for fitting, resulting in the regression mapping model:

[0044] Among them, SOH( k ) is the first k Health status estimate (percentage) at the next cycle. δ g (k) is the kth k The topological gap metric is calculated in the second iteration. p1, p2, and p3 are polynomial regression coefficients, obtained by minimizing the mean squared error (MSE) of historical data.

[0045] In actual BMS operation, the real-time calculated δ g Substituting these values ​​into the regression mapping model above, we can obtain the current estimated value of SOH.

[0046] Experimental verification To verify the effectiveness of this disclosure, actual test data of a ternary lithium battery throughout its entire life cycle (Cycle 0 to Cycle 330) were used.

[0047] 1. Comparison of anti-interference capabilities The method disclosed herein is compared with the traditional internal resistance method (calculated by ΔV / ΔI).

[0048] like Figure 4 As shown, the internal resistance method results curve exhibits violent oscillations (large noise amplitude), making it difficult to discern short-term aging trends. This is due to the differential operation amplifying sensor noise. This disclosure (Gap Metric): The results curve exhibits an extremely high signal-to-noise ratio, with a smooth and compact curve, clearly tracking subtle aging inflection points even in the middle of battery aging (approximately 150 cycles).

[0049] 2. Fitting accuracy verification The extracted features are fitted to the true SOH, such as... Figure 3 As shown in the figure, the goodness of fit R2 reached 0.9644, indicating that this feature can explain most of the SOH decay phenomenon.

[0050] 3. Validation of the accuracy of the whole life cycle estimation The constructed quadratic polynomial mapping model was applied to full lifecycle data (Cycle 0 to Cycle 330) to verify the online estimation performance of this disclosure. Figure 5 As shown in the figure above, the estimated SOH trajectory (dashed line) is compared with the actual SOH trajectory (solid line). It can be seen that the two remain highly consistent throughout the aging process, demonstrating how the estimation error changes with the number of cycles.

[0051] Statistical analysis shows that, over the entire lifecycle, the root mean square error (RMSE) of the SOH estimation using the method disclosed in this publication is only 1.22%, with the absolute error controlled within ±2% for the vast majority of cycles. This indicates that the method disclosed in this publication can achieve or even exceed the estimation accuracy of traditional full-charge-discharge tests using short segments of constant current charging data.

[0052] 4. Battery Management System (BMS) Hardware Implementation: like Figure 7 As shown, the method disclosed herein can be specifically applied to a battery management system. This system mainly includes a data acquisition unit, a core processing unit, a storage unit, and a communication unit. Specific details include: (1) Data acquisition unit: includes a high-precision ADC (analog-to-digital converter) for real-time acquisition of the battery voltage sequence V(t) and current sequence I(t) at a preset frequency (e.g., 1Hz-10Hz), and converts the analog signal into a digital signal for transmission to the core processing unit.

[0053] (2) Core processing unit: This unit can be an embedded microprocessor (such as an ARM Cortex-M series) or a digital signal processor. It integrates an algorithm logic module, specifically including: ① Constant current segment identification module: used for logical judgment of whether the constant current charging conditions and voltage range [3.65V, 4.15V] are met.

[0054] ② Preprocessing module: used to resample and debias the data in the buffer.

[0055] ③ Feature calculation module: Built-in discretization iterative algorithm of proses, used to identify process model parameters and calculate topological gap metric δg.

[0056] ④ State estimation module: Stores quadratic polynomial coefficients, used to map δg to the final SOH value.

[0057] (3) Storage unit: used to store the baseline model parameters (such as the numerator and denominator coefficients of the baseline transfer function) and the offline calibrated mapping model parameters during the early stage of battery life (BOL).

[0058] (4) Application process: When the vehicle is charging, the system automatically runs the above modules, calculates the current SOH, and sends the result to the vehicle controller or display screen via CAN bus to guide the charging cut-off strategy or display the remaining mileage.

[0059] This disclosure transforms complex frequency domain topology calculations into time domain iterative logic suitable for embedded processors through hardware and software co-design, achieving high-precision online estimation on low-cost hardware.

[0060] Example 2 One embodiment of this disclosure provides a power battery state monitoring system based on identification theory and topology metrics, including: The constant current segment identification and preprocessing module is used to acquire the raw feature data during the battery charging process and preprocess it. The feature extraction module treats the constant current charging response of the battery as a step response of a dynamic system. It uses a low-order process model as an equivalent dynamic feature extractor, takes the preprocessed raw feature data as input to the low-order process model, fits the voltage response trajectory, and outputs the identified model parameters. It constructs a dynamic transfer function, substitutes the identified model parameters into the dynamic transfer function to obtain the current dynamic transfer function model, and calculates the topological gap metric between the current model and the benchmark model in the frequency domain. The State Estimation module is used to construct a regression mapping model. The calculated topological gap metric is input into the regression mapping model, and the mapping output yields the SOH estimate.

[0061] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0062] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0063] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A power battery state monitoring method based on identification theory and topological metric, characterized in that, include: Obtain raw feature data during the battery charging process and preprocess it; The constant current charging response of the battery is regarded as the step response of the dynamic system. A low-order process model is used as the equivalent dynamic feature extractor. The preprocessed original feature data is used as the input of the low-order process model to fit the voltage response trajectory, and the identified model parameters are output. Construct a dynamic transfer function, substitute the identified model parameters into the dynamic transfer function to obtain the current dynamic transfer function model, and calculate the topological gap metric between the current model and the benchmark model in the frequency domain. A regression mapping model is constructed, and the calculated topological gap metric is input into the regression mapping model. The mapping output yields the estimated SOH value.

2. The power battery state monitoring method based on identification theory and topology metric as described in claim 1, characterized in that, The acquisition of raw feature data during the battery charging process includes: Collect voltage and current data during battery charging; Determining whether the current charging process is a constant current charging segment includes: current I > 0.05C and current fluctuation variance ≤ 0.01C², where C is the battery's rated capacity; the electrochemically sensitive range is determined by battery electrochemical phase transition testing, selecting the range with the largest polarization voltage change rate; When a constant current charging segment is detected and the voltage is within a preset electrochemically sensitive range, that segment is extracted as the raw feature data.

3. The power battery state monitoring method based on identification theory and topology metric as described in claim 1, characterized in that, Preprocessing of the raw feature data includes: The original feature data is resampled and debiased. The debiasing process involves subtracting the values ​​at the starting time from the extracted voltage and current sequences, so that they change from zero. The resampling operation includes setting a globally uniform sampling interval, calculating the total duration of the segment and generating a standard uniform time axis, and using a linear interpolation method to map the original non-uniform or variable-length voltage and current values ​​onto the uniform time axis to obtain a standardized input sequence and output sequence with fixed length.

4. The power battery state monitoring method based on identification theory and topology metric as described in claim 1, characterized in that, The constant current charging process is regarded as a step excitation applied to the battery system. A first-order process model is used as an equivalent dynamic feature extractor. The preprocessed raw feature data is used as the input of the low-order process model to fit the voltage response trajectory within the segment. The output is the identified optimal model parameters. These optimal model parameters serve as a set of generalized waveform structure parameters, characterizing the topological shape of the voltage response trajectory under the current aging state.

5. The power battery state monitoring method based on identification theory and topology metric as described in claim 1, characterized in that, Using the least squares method or iterative optimization algorithm, based on the obtained standardized input and output sequences, and with the goal of minimizing the error between the model's predicted output and the actual output sequence, the optimal model parameters at the current moment are identified, a dynamic transfer function is constructed, and the identified optimal model parameters are substituted into the dynamic transfer function structure to obtain the dynamic transfer function model of the current battery state.

6. The power battery state monitoring method based on identification theory and topology metric as described in claim 1, characterized in that, Calculate the topological gap metric between the current model and the baseline model in the frequency domain, including: A baseline model is identified during the first cycle of the battery's entire life cycle. For any point in the operation, calculate the topological gap metric between the dynamic transfer function model and the baseline model; The definition of topological gap metric is based on graph metric theory, and its mathematical essence is to measure the maximum distance between two linear dynamic systems under closed-loop feedback.

7. The power battery state monitoring method based on identification theory and topology metric as described in claim 6, characterized in that, The topology gap metric directly reflects the "drift distance" of the current dynamic transfer function model of the battery relative to the new battery reference model in the frequency domain topology space. It comprehensively reflects the overall difference between the steady-state gain change and the time constant change caused by aging. As the number of cycles increases, the topology gap metric gradually increases monotonically from 0, and the curve is very smooth.

8. The power battery state monitoring method based on identification theory and topology metric as described in claim 1, characterized in that, Based on offline experimental data, a mapping relationship between topological gap metric and true SOH was established. The topological gap metric and SOH exhibited a significant nonlinear monotonic relationship. Based on the mapping relationship, a quadratic polynomial regression model was used for fitting to obtain a regression mapping model.

9. The power battery state monitoring method based on identification theory and topology metric as described in claim 8, characterized in that, The polynomial regression coefficients are obtained by training the system to minimize the mean square error of the historical data.

10. A power battery state monitoring system based on identification theory and topological metric, characterized in that, include: The constant current segment identification and preprocessing module is used to acquire the raw feature data during the battery charging process and preprocess it. The feature extraction module treats the constant current charging response of the battery as a step response of a dynamic system. It uses a low-order process model as an equivalent dynamic feature extractor, takes the preprocessed raw feature data as input to the low-order process model, fits the voltage response trajectory, and outputs the identified model parameters. It constructs a dynamic transfer function, substitutes the identified model parameters into the dynamic transfer function to obtain the current dynamic transfer function model, and calculates the topological gap metric between the current model and the benchmark model in the frequency domain. The State Estimation module is used to construct a regression mapping model. The calculated topological gap metric is input into the regression mapping model, and the mapping output yields the SOH estimate.

Images