Bms battery state ai intelligent estimation method and system

Abstract

The application discloses a BMS battery state AI intelligent estimation method and system, relates to the technical field of battery management, and comprises the following steps: carrying out multi-level feature decomposition on a voltage-time response curve, generating a pulse impedance spectrum feature vector, and establishing an internal state anchor point parameter representing the internal health state of a battery based on the pulse impedance spectrum feature vector; based on the internal state anchor point parameter and battery operation data, evolving and deducing the battery state, taking the internal state anchor point parameter as a state reference benchmark, and obtaining a continuous evolution trajectory of the battery state; when obtaining a new internal state anchor point parameter by means of a current pulse again in the battery operation process, carrying out deviation comparison and identification on the predicted state of the continuous evolution trajectory at the corresponding moment and the internal state anchor point parameter, smoothing and calibrating the continuous evolution trajectory based on the deviation, and obtaining a continuous battery full state vector; and providing high-precision state estimation for real-time energy management and full-life-cycle health prediction of the battery.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, and in particular to a method and system for intelligent estimation of battery status using AI in battery management systems (BMS). Background Technology

[0002] With the rise of new energy vehicles and the energy storage industry, the requirements for estimating the state of charge (SOC) and state of health (SOH) of batteries by battery management systems (BMS) are constantly increasing. Currently, model-based state estimation methods are widely used, especially those combining equivalent circuit models with Kalman filtering. This method establishes a mathematical model of the battery's external electrical behavior and uses real-time collected voltage and current data for state observation and filtering to achieve online SOC estimation. To further enhance the model's adaptability and nonlinear characterization performance, academia and industry have begun to introduce data-driven artificial intelligence technologies, such as deep learning models using long short-term memory networks, to directly learn the complex mapping rules of battery state from operational data, thus promoting the development of intelligent estimation technology that integrates mechanisms and data.

[0003] However, existing state estimation methods based on equivalent circuit models and filtering algorithms, as well as pure data-driven estimation methods based on deep learning, mainly rely on external macroscopic electrical signals to infer the internal state of the battery. Their estimation results lack directly observable internal state references, making it difficult to effectively calibrate the state evolution process. When the battery undergoes slow-evolving health degradation processes such as the decline in lithium-ion diffusion capacity and the increase in interface reaction impedance, the influence of internal mechanism changes on external voltage and current characteristics often exhibits weak, hysteretic, and highly coupled characteristics. This makes it difficult for existing methods to obtain high-confidence internal state parameters that can be used to constrain state evolution over a long period of time. Consequently, estimation errors and state drift are prone to occur during long-term operation, making it difficult to achieve continuous and reliable tracking of the battery's internal health state. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an AI-based intelligent estimation method for BMS battery state to solve the problem of decreased accuracy and error drift in long-term state estimation caused by relying on reverse inference from external signals.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an AI-based intelligent estimation method for BMS battery state, comprising: During battery operation, a current pulse is applied to the battery, and the voltage at the battery terminals changes over time to obtain a voltage-time response curve. Multi-level feature deconstruction is performed on the voltage-time response curve to generate pulse impedance spectrum feature vector, and internal state anchor point parameters characterizing the internal health state of the battery are established based on the pulse impedance spectrum feature vector. Based on the internal state anchor point parameters and battery operation data, the battery state is evolved and deduced, and the internal state anchor point parameters are used as the state reference benchmark to obtain the continuous evolution trajectory of the battery state. When new internal state anchor parameters are obtained again through current pulse during battery operation, the predicted state of the continuous evolution trajectory at the corresponding time is compared and identified with the internal state anchor parameters, and the continuous evolution trajectory is smoothed and calibrated based on the deviation to obtain the continuous battery full state vector. The continuous battery full-state vector is analyzed into battery state of charge and battery health state, and used for battery energy scheduling control and health early warning.

[0007] Preferably, the method for obtaining the voltage-time response curve includes: During battery operation, the changes in battery terminal voltage and current are continuously monitored; when the rate of change of voltage and the rate of change of current are both lower than the change threshold within a preset time window, the battery is determined to be in a stable working state. After the battery enters a stable operating state, a current pulse with a constant amplitude and fixed duration is applied to the battery; Before, during, and after the application of the current pulse, the battery terminal voltage is collected to obtain voltage data; the voltage data is then time-stamped and sequentially arranged to construct a voltage-time response curve.

[0008] Preferably, the method for generating the pulse impedance spectrum eigenvector includes: Based on the timing of the current pulse application and the voltage relaxation process after the pulse ends, the voltage-time response curve is divided into a transient response segment and a relaxation response segment. Multi-level feature extraction is performed on the voltage data in the transient response and relaxation response regions to obtain the internal electrochemical dynamic characteristic parameters of the battery, and a unified scale mapping is performed on the internal electrochemical dynamic characteristic parameters of the battery. The internal electrochemical dynamic characteristic parameters of the battery are arranged and combined in the order of their formation to form the pulse impedance spectrum characteristic vector of the battery's internal response characteristics under current pulse excitation.

[0009] Preferably, the method for establishing internal state anchor point parameters characterizing the internal health state of the battery includes: Based on the time distribution and amplitude variation characteristics of the pulse impedance spectrum eigenvector under the action of current pulse, state correlation analysis is performed to distinguish the characteristic parameters of fast and slow electrochemical response processes. Based on the characteristic parameters of fast and slow electrochemical response processes, determine the set of state characterization parameters for the internal electrochemical kinetic state of the battery; The set of state representation parameters is quantified to obtain internal state parameters; these internal state parameters are then used as internal state anchor parameters for the current internal health state of the battery.

[0010] Preferably, the method for obtaining the continuous evolution trajectory of the battery state includes: Based on the internal state anchor point parameters and the battery operation data continuously collected during battery operation, a battery state evolution time benchmark with time as the continuous independent variable is established, and the internal state anchor point parameters are used as the state reference benchmark to divide the continuous evolution interval of the battery state between adjacent anchor point parameter acquisition times. Based on battery operation data within the continuous evolution interval, the battery state is progressively evolved over a continuous time scale to obtain continuous state change results; and the continuous state change results are then connected in time sequence to form a continuous evolution trajectory of the battery state.

[0011] Preferably, the method for obtaining the continuous battery full-state vector includes: During battery operation, when a current pulse is applied to the battery again and new internal state anchor parameters are obtained, the timestamp corresponding to the internal state anchor parameters is determined. Based on the timestamps corresponding to the internal state anchor point parameters, the predicted battery state at the corresponding moment is extracted from the continuous evolution trajectory of the battery state and used as the anchor point to align the predicted state. A comparison analysis of the anchor point alignment prediction state and the newly acquired internal state anchor point parameters is performed to obtain the state deviation vector of the degree of inconsistency between the continuously extrapolated state and the actual internal state of the battery. Based on the state deviation vector, the continuous evolution trajectory of the battery state is smoothed and calibrated in the timestamp neighborhood, so that it gradually converges to the internal state anchor parameters; the smoothed and calibrated continuous evolution trajectory of the battery state is used as the continuous battery full state vector.

[0012] Preferably, the method for resolving the continuous battery full-state vector into battery state of charge and battery state of health includes: Based on the time-varying characteristics of each state component in the full state vector of a continuous battery, the state dimension of the full state vector of a continuous battery is analyzed to separate the state of charge parameters and the health state parameters. The state of charge parameter is used as a characterization index of the current energy availability of the battery for energy scheduling control during the battery charging and discharging process; The health status parameters are compared and analyzed with preset health status criteria. When the health status parameters meet the abnormal or degradation conditions, corresponding battery health warning information is generated.

[0013] Preferably, the method for performing state association analysis includes: Based on the time distribution characteristics of each dimension of the characteristic parameters in the pulse impedance spectrum feature vector before, during and after the current pulse is applied, the correspondence between each dimension of the characteristic parameters and the timing of the internal electrochemical response of the battery is established. Based on the correspondence, the characteristic parameters of the time distribution feature that are concentrated in the initial stage of current pulse application, and whose amplitude changes occur within a short response time window and whose amplitude change rate is higher than the first preset amplitude threshold are classified as characteristic parameters of the fast electrochemical response process. The characteristic parameters of the slow electrochemical response process are those whose response time distribution is concentrated in the voltage relaxation stage after the current pulse ends, whose amplitude changes are distributed within the long response time window, and whose amplitude change rate is lower than the second preset amplitude threshold.

[0014] Preferably, the method for performing smooth calibration on the continuous evolution trajectory of the battery state within the timestamp neighborhood includes: Centered on the timestamp corresponding to the internal state anchor point parameter, the timestamp neighborhood covering the time range before and after the battery state continuous evolution trajectory is determined, and the continuous inferred state within the timestamp neighborhood is extracted.

[0015] Based on the magnitude of the state deviation vector and the time distance between the continuously derived state and the timestamp of the internal state anchor point parameter, a correlation-based smooth convergence adjustment is performed on the continuously derived state in the timestamp neighborhood.

[0016] The continuously extrapolated state after smooth convergence adjustment is updated to the continuous evolution trajectory of the battery state in chronological order, so that it gradually converges to the internal state anchor point parameters.

[0017] Secondly, this invention provides a BMS battery state AI intelligent estimation system, including: The acquisition module is used to apply current pulses to the battery during battery operation and acquire the voltage at the battery terminals as a function of time to obtain a voltage-time response curve. The feature extraction module is used to perform multi-level feature deconstruction on the voltage-time response curve, generate pulse impedance spectrum feature vector, and establish internal state anchor point parameters characterizing the internal health state of the battery based on the pulse impedance spectrum feature vector. The evolution module is used to extrapolate the evolution of the battery state based on the internal state anchor parameters and battery operation data, and to obtain the continuous evolution trajectory of the battery state by using the internal state anchor parameters as the state reference benchmark. The calibration module is used to compare and identify the deviation between the predicted state of the continuous evolution trajectory at the corresponding moment and the internal state anchor parameters when new internal state anchor parameters are obtained again through current pulses during battery operation, and to perform smooth calibration on the continuous evolution trajectory based on the deviation to obtain the continuous battery full state vector. The output module is used to resolve the continuous battery full state vector into battery state of charge and battery health state, and is used for battery energy scheduling control and health warning.

[0018] The beneficial effects of this invention are as follows: By actively stimulating and extracting internal state anchors, and by continuously evolving and periodically smoothing calibration based on the anchors, a new intelligent estimation paradigm of "active detection-closed-loop calibration" is constructed; by directly obtaining a high-confidence internal electrochemical state benchmark through physical stimulation, a new approach is provided to solve the uncertainty at the source of estimation, and the ability to perceive the microscopic health state in its early stages is enhanced; furthermore, by using this benchmark to perform periodically smoothing correction on the continuous evolution model, long-term state drift is suppressed, and continuous and seamless monitoring of the state throughout the entire life cycle is achieved, which can provide a long-term stable and mechanism-clear high-precision state estimate for real-time energy management and full life cycle health prediction of batteries. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of the AI-powered intelligent estimation method for BMS battery state in this invention; Figure 2 This is a schematic diagram of the BMS battery state AI intelligent estimation system in this invention; Figure 3 This is a flowchart illustrating the output of internal state anchor point parameters in this invention; Figure 4 This is a flowchart of the process for outputting the continuous battery state vector in this invention. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Reference Figure 1 , Figure 2 , Figure 3 and Figure 4 This is one embodiment of the present invention, which provides a BMS battery state AI intelligent estimation method, including the following steps: Methods for obtaining voltage-time response curves include: During battery operation, the changes in battery terminal voltage and current are continuously monitored; when the rate of change of voltage and the rate of change of current are both lower than the change threshold within a preset time window, the battery is determined to be in a stable working state. It should be noted that the preset time window is used to limit the continuous observation interval of the battery terminal voltage change rate and current change rate. Its value is determined based on the natural fluctuation time scale of the battery under constant load or slow load change conditions. For example, the value is a continuous time range of 1 second to 10 seconds to suppress the influence of instantaneous noise. The change threshold is used to characterize the upper limit of the allowable fluctuation of the rate of change of battery terminal voltage and the rate of change of current when the battery is in a stable operating state. The setting process is based on the statistical results of the corresponding rate of change of the battery in the stable operating range. The example value is the upper limit of the low amplitude fluctuation of the rate of change distribution in the stable range. During the monitoring process, when the rate of change of battery terminal voltage and the rate of change of current are continuously lower than the change threshold throughout the entire preset time window, it is determined that the battery is in a stable operating state.

[0025] After the battery enters a stable operating state, a current pulse with a constant amplitude and a fixed duration is applied to the battery.

[0026] Specifically, after the battery is determined to be in a stable operating state, the battery charging and discharging control process is time-controlled to ensure that the battery current switches from a stable value to a target current value and remains constant at the stable operating state, forming a current pulse with constant amplitude and fixed duration. Here, a current pulse refers to a current process in which the battery terminal current undergoes a step change in a short period of time and remains unchanged for a fixed period of time, such as switching from the original stable current to a current level of 5% to 20% of the rated operating current. Constant amplitude means that the battery terminal current does not change with time throughout the entire pulse duration, and its amplitude is selected based on the battery's rated operating current range, such as 5% to 20% of the rated operating current. Fixed duration means that the current maintains the target amplitude for a consistent length of time, and the battery terminal current returns to the stable level before the pulse is applied after the pulse ends.

[0027] Before, during, and after the application of the current pulse, the battery terminal voltage is collected to obtain voltage data; the voltage data is then time-stamped and sequentially arranged to construct a voltage-time response curve.

[0028] Specifically, the battery terminal voltage is continuously acquired before, during, and after the current pulse is applied to the battery to obtain the initial voltage change under stable operating conditions, the transient voltage response under current step action, and the voltage stabilization process after current recovery, forming a complete voltage change process covering the period before, during, and after the pulse. During the acquisition process, the battery terminal voltage acquired at each moment is time-stamped and arranged in chronological order, and a voltage-time response curve is constructed based on the correspondence between the time stamp and the battery terminal voltage.

[0029] Current BMS state estimation methods largely rely on current integration or external characteristic modeling, treating the battery as a "black box" for reverse inference, making it difficult to directly perceive the internal electrochemical dynamics. Due to fluctuations in operating conditions and environmental influences, traditional algorithms are prone to cumulative errors and state drift, and lack physical benchmark calibration, leading to distortion of estimation results over long-term operation. Furthermore, single-point estimation lacks temporal continuity, making it difficult to provide robust state trajectories during complex charge and discharge processes. Therefore, this solution establishes internal state anchor points through current pulses, aiming to build a solid physical observation benchmark for AI estimation, as detailed below: Methods for generating pulse impedance spectrum eigenvectors include: Based on the timing of the current pulse application and the voltage relaxation process after the pulse ends, the voltage-time response curve is divided into transient response segment and relaxation response segment.

[0030] Specifically, based on the voltage-time response curve, the moment when the current pulse switches from a stable value to the target current value and the moment when the current pulse ends and returns to a stable current level are determined, with the moment of current pulse application serving as the starting boundary for segment division. Within the time range from the application of the current pulse to its end, the battery terminal voltage undergoes rapid changes and short-term adjustments with the current step change, and the corresponding voltage-time response curve is divided into the transient response segment, used to characterize the instantaneous response characteristics of the battery terminal voltage to the current pulse. After the current pulse ends and returns to a stable current level, the battery terminal voltage gradually decreases and tends to stabilize, and the corresponding voltage-time response curve is divided into the relaxation response segment, used to characterize the recovery process of the battery terminal voltage after the current disturbance disappears, thus completing the division of the transient response segment and the relaxation response segment.

[0031] Multi-level feature extraction is performed on the voltage data in the transient response and relaxation response regions to obtain the internal electrochemical dynamic characteristic parameters of the battery, and a unified scale mapping is performed on the internal electrochemical dynamic characteristic parameters of the battery.

[0032] Specifically, within the transient response region, based on the voltage change process after the current step, the voltage mutation amount, initial change slope, short-term adjustment amplitude, and change characteristics before reaching a local steady state are extracted sequentially to characterize the dynamic response characteristics of the battery's internal electrochemical process under rapid disturbance conditions. Within the relaxation response region, based on the voltage stabilization process after the current recovers to a stable current level, the voltage drop amplitude, stabilization rate, and voltage difference characteristics before and after stabilization are extracted to characterize the slow recovery characteristics of the battery's internal electrochemical process after the disturbance disappears. The battery's internal electrochemical dynamic characteristic parameters extracted from the transient response region and the relaxation response region are collected and normalized to unify the scale of the battery's internal electrochemical dynamic characteristic parameters, forming a set of battery internal electrochemical dynamic characteristic parameters.

[0033] The internal electrochemical dynamic characteristic parameters of the battery are arranged and combined in the order of their formation to form the pulse impedance spectrum characteristic vector of the battery's internal response characteristics under current pulse excitation.

[0034] Specifically, the internal electrochemical dynamic characteristic parameters of the battery are arranged in chronological order of their formation. First, the internal electrochemical dynamic characteristic parameters of the battery that characterize the rapid response characteristics under current step action in the transient response segment are arranged in order. Then, the internal electrochemical dynamic characteristic parameters of the battery that characterize the recovery and stabilization characteristics after current recovery in the relaxation response segment are arranged in order. The internal electrochemical dynamic characteristic parameters of the battery corresponding to the transient response segment and the internal electrochemical dynamic characteristic parameters of the battery corresponding to the relaxation response segment are sequentially spliced ​​to form the pulse impedance spectrum feature vector.

[0035] Methods for establishing internal state anchor point parameters characterizing the internal health state of a battery include: Based on the time distribution and amplitude variation characteristics of the pulse impedance spectrum eigenvectors under current pulse, state correlation analysis is performed to distinguish the characteristic parameters of fast and slow electrochemical response processes.

[0036] Based on the characteristic parameters of fast and slow electrochemical response processes, a set of state characterization parameters for the internal electrochemical kinetic state of the battery is determined.

[0037] Specifically, the characteristic parameters of the fast electrochemical response process are used as characterization factors of the internal polarization and interfacial reaction activity of the battery, while the characteristic parameters of the slow electrochemical response process are used as characterization factors of the internal diffusion and concentration recovery characteristics of the battery. Under the same current pulse excitation condition, the two types of characterization factors are jointly organized and combined to form a set of state characterization parameters that include both fast response and slow recovery characteristics.

[0038] The set of state representation parameters is quantified to obtain internal state parameters; these internal state parameters are then used as internal state anchor parameters for the current internal health state of the battery.

[0039] Specifically, after forming the set of state characterization parameters, based on the voltage-time response results corresponding to each characterization factor under the same current pulse excitation condition, each characterization factor in the set of state characterization parameters is quantitatively characterized. Among them, the characteristic parameters of the fast electrochemical response process are taken as their corresponding stable statistical values ​​or representative amplitudes in the transient response segment, which are used to reflect the actual response level of the battery's internal polarization and interface reaction under the current operating conditions; the characteristic parameters of the slow electrochemical response process are taken as their corresponding recovery amplitude, change trend, or stable value in the relaxation response segment, which are used to reflect the actual performance of the battery's internal diffusion and concentration recovery process; the numerical combination results of each characterization factor in the set of state characterization parameters are used as internal state parameters, and the internal state parameters are determined as internal state anchor point parameters characterizing the current internal health state of the battery. By establishing anchor parameters through pulse impedance spectrum feature deconstruction, this scheme achieves in-depth analysis of polarization response and diffusion characteristics. This physically-excited "anchor point" provides a high-confidence starting point for state evolution, effectively suppressing random disturbances. Time-series advancement and smoothing adjustments are performed within the evolution trajectory, ensuring the physical consistency of state parameters across continuous time scales. Ultimately, this closed-loop mechanism of "anchor point support and trajectory evolution" enables the BMS to output high-precision and clearly defined continuous full-state vectors, providing robust data support for energy scheduling and health early warning.

[0040] Methods for obtaining the continuous evolution trajectory of battery states include: Based on the internal state anchor point parameters and the battery operation data continuously collected during battery operation, a battery state evolution time benchmark with time as the continuous independent variable is established, and the internal state anchor point parameters are used as the state reference benchmark to divide the continuous evolution interval of the battery state between adjacent anchor point parameter acquisition times.

[0041] Specifically, during battery operation, battery terminal voltage, battery terminal current, and corresponding time markers are continuously collected to form battery operation data arranged in chronological order. When a current pulse excitation is completed and internal state anchor parameters are obtained, the internal state anchor parameters are bound to the corresponding time markers as a reference benchmark for the battery's internal health state at that moment. Using time as a continuous independent variable, the time positions corresponding to each internal state anchor parameter are arranged on the same time axis to establish a battery state evolution time benchmark with time as a continuous independent variable and internal state anchor parameters as the state reference benchmark. Time is obtained by using the parameters of two adjacent internal state anchor points. and As the interval boundary, the state anchor point parameters within the adjacent internal region and Between these points, combined with continuously collected battery operation data, the changes in the battery's internal state parameters over time are continuously described, thereby dividing the continuous evolution interval of the battery state between adjacent anchor point parameter acquisition times; in any Internally, by continuously mapping or interpolating the internal state parameters in chronological order between adjacent anchor points, the continuous evolution of the battery's internal health state during operation is described. This continuous mapping or interpolation can be expressed as (this expression is a standard linear interpolation formula): ; in, Is the battery at all times? The internal state parameters, It is in the Secondary current pulse excitation completion time The obtained internal state anchor point parameters, It is located in and Any consecutive time points between, It is the first The time when the internal state anchor point parameters are obtained. It is the first The time when the internal state anchor point parameters are obtained. It is in the Secondary current pulse excitation completion time The obtained internal state anchor point parameters.

[0042] Based on battery operation data within the continuous evolution interval, the battery state is progressively evolved over a continuous time scale to obtain continuous state change results; and the continuous state change results are then connected in time sequence to form a continuous evolution trajectory of the battery state.

[0043] Specifically, within the continuous evolution interval, based on the battery state evolution time reference determined by the internal state anchor point parameters, and combined with the battery terminal voltage, battery terminal current, and corresponding time markers continuously collected during battery operation, the internal state change process of the battery on a continuous time scale is calculated time-by-time in chronological order. Specifically, using the time markers within the continuous evolution interval as the advancement order, the internal state parameters corresponding to each time position are continuously updated, so that the battery internal state parameters maintain temporal continuity and state smoothness between adjacent time positions, obtaining continuous state change results covering the entire continuous evolution interval. Furthermore, the continuous state change results obtained in each continuous evolution interval are sequentially connected in chronological order, so that the internal state parameters of adjacent continuous evolution intervals remain consistent at the time boundaries, forming a continuous evolution trajectory of the battery state where the internal health state of the battery changes continuously with operating time.

[0044] Existing technologies still struggle to completely avoid nonlinear deviations between model predictions and the actual physical state of the battery when handling state extrapolation across time periods. Traditional calibration methods often perform "jump" corrections only at observation points, which not only causes discontinuous step changes in the battery state curve at the moment of calibration but also fails to eliminate accumulated prediction biases that occurred before the observation point. To address the mismatch between the extrapolated trajectory and the actual state, this solution designs a difference comparison mechanism based on timestamp indexes, as follows: Methods for obtaining the full state vector of a continuous battery include: During battery operation, when a current pulse is applied to the battery again and new internal state anchor parameters are obtained, the timestamp corresponding to the internal state anchor parameters is determined.

[0045] It should be noted that, based on the continuous evolution trajectory of the battery state, when the current pulse excitation is triggered again during battery operation, the current pulse excitation process is fully recorded according to the method of current pulse application and voltage-time response acquisition; when the current pulse excitation is completed and the new internal state anchor point parameters are calculated based on the pulse impedance spectrum eigenvector, the time stamp corresponding to the end of the current pulse excitation is read synchronously, and the time stamp is determined as the timestamp corresponding to the newly acquired internal state anchor point parameters.

[0046] Based on the timestamps corresponding to the internal state anchor parameters, the predicted battery state at the corresponding moment is extracted from the continuous evolution trajectory of the battery state and used as the anchor point to align the predicted state.

[0047] Specifically, after obtaining the timestamp corresponding to the new internal state anchor point parameter, the timestamp is used as a time index to locate the time position that is consistent with or closest to the timestamp in the continuous evolution trajectory of the battery state; by reading the continuous state change results corresponding to the time position, the predicted battery state obtained by the continuous evolution of the battery at that moment is obtained, and the predicted battery state is determined as the anchor point aligned prediction state for alignment comparison with the internal state anchor point parameter.

[0048] By comparing and analyzing the differences between the anchor point alignment prediction state and the newly acquired internal state anchor point parameters, a state deviation vector is obtained to determine the degree of inconsistency between the continuously extrapolated state and the actual internal state of the battery.

[0049] Specifically, at the same timestamp Next, align the anchor point with the predicted state. With the newly acquired internal state anchor parameters A dimension-by-dimensional comparison is performed. Based on the correspondence of each component dimension of the internal state parameters, the numerical differences between the two in each dimension are calculated. The results of the differences in each dimension are then combined according to the original arrangement order of the internal state parameters to form a state deviation vector. It can be exemplarily represented as: ;in, It is the time stamp corresponding to the internal state anchor point parameter.

[0050] Based on the state deviation vector, the continuous evolution trajectory of the battery state is smoothed and calibrated in the timestamp neighborhood, so that it gradually converges to the internal state anchor parameters; the smoothed and calibrated continuous evolution trajectory of the battery state is used as the continuous battery full state vector.

[0051] By performing this neighborhood smoothing calibration based on the state deviation vector, this scheme achieves a leap from "point-to-point correction" to "temporally continuous convergence." Dissimilarity comparison analysis accurately quantifies the degree of drift in the continuous evolution trajectory, while subsequent smoothing ensures that the trajectory maintains the continuity and stability of physical evolution as it approaches the new anchor point. This design suppresses data mutations caused by algorithm updates, resulting in a final output continuous battery state vector that not only possesses extremely high instantaneous accuracy but also exhibits strong robustness in the time dimension.

[0052] Methods for resolving the continuous battery state vector into battery state of charge and battery health state include: Based on the time-varying characteristics of each state component in the full state vector of a continuous battery, the state dimension of the full state vector of the continuous battery is analyzed to separate the charge state parameters and health state parameters.

[0053] Specifically, after the continuous battery full state vector, the change characteristics of each internal state parameter in the continuous battery full state vector on the running time axis are analyzed one dimension at a time. Among them, based on the characteristics that the internal state parameters change synchronously with the battery charging and discharging behavior in a short time scale and have a direct correspondence with the changes in the direction and amplitude of the battery terminal current, the state components reflecting the battery energy access behavior are identified, and the identified state components are analyzed into charge state parameters. Based on the characteristics of the slow evolution of internal state parameters over a long time scale with the cumulative changes of battery cycle number and operating conditions, and the weak correlation with instantaneous charge and discharge behavior, state components reflecting the degradation trend of internal electrochemical performance of the battery are identified, and the identified state components are analyzed as health state parameters.

[0054] The state of charge (SOC) parameter is used as a characterization index of the battery's current energy availability for energy scheduling control during battery charging and discharging.

[0055] It should be explained that a correspondence is established between the numerical variation range of the state of charge (SCC) parameter and the available energy level of the battery at the current operating moment. By using the SCC parameter as a state indicator reflecting the remaining available energy and charging / discharging capacity of the battery, the allowable charging / discharging direction, intensity, and duration range of the battery are determined based on the changing trend and current value of the SCC parameter during the battery charging / discharging process. This allows for the constraint and adjustment of the energy scheduling behavior during the battery charging / discharging process without exceeding the battery's available energy boundary.

[0056] The health status parameters are compared and analyzed with preset health status criteria. When the health status parameters meet the abnormal or degradation conditions, corresponding battery health warning information is generated.

[0057] Specifically, the current values ​​and trends of the health status parameters over time are compared and analyzed item by item with preset health status criteria. The preset health status criteria are used to characterize the range of internal electrochemical state changes that the battery is allowed to undergo under normal operating conditions. When the current value or evolution trend of the health status parameter exceeds the normal range defined by the health status criteria, or shows a continuous deviation from the normal range, it is determined that there is an abnormality or degradation risk in the battery's internal health status. After the comparison and judgment are completed, battery health warning information is generated to indicate abnormal or degraded battery health status.

[0058] Methods for performing state association analysis include: Based on the time distribution characteristics of each dimension of the characteristic parameters in the pulse impedance spectrum feature vector before, during and after the current pulse is applied, the correspondence between each dimension of the characteristic parameters and the timing of the internal electrochemical response of the battery is established.

[0059] Specifically, the internal electrochemical dynamic characteristic parameters of the battery, derived from the transient response region, are mapped to the time range after the current switches from a stable value to the target current value. This is used to characterize the response behavior of the internal electrochemical process of the battery during the initial and maintenance phases of the current pulse. The internal electrochemical dynamic characteristic parameters of the battery, derived from the relaxation response region, are mapped to the time range after the current pulse ends and the current returns to a stable level. This is used to characterize the recovery behavior of the internal electrochemical process of the battery after the current disturbance disappears. Through time position matching, a one-to-one correspondence is established between the characteristic parameters of each dimension in the pulse impedance spectrum feature vector and the actual occurrence sequence of the internal electrochemical response of the battery.

[0060] Based on the correspondence, the characteristic parameters of the fast electrochemical response process are distinguished as those whose response time distribution is concentrated in the initial stage of current pulse application, whose amplitude change occurs within a short response time window, and whose amplitude change rate is higher than the first preset amplitude threshold.

[0061] It should be noted that the short-time response window is used to limit the rapid response interval in the initial stage of the current pulse application, and its exemplary value can be a time range of several milliseconds to hundreds of milliseconds after the current pulse is applied; the first preset amplitude threshold is used to distinguish between rapid changes and slow changes, and its exemplary value can be the lower limit of the high change interval in the statistical distribution of the amplitude change rate of the corresponding characteristic parameter under stable working conditions.

[0062] The characteristic parameters of the slow electrochemical response process are those whose response time distribution is concentrated in the voltage relaxation stage after the current pulse ends, whose amplitude changes are distributed within the long response time window, and whose amplitude change rate is lower than the second preset amplitude threshold.

[0063] It should be noted that the long-time response window is used to limit the slow recovery interval during the voltage relaxation process, and its exemplary value can be a time range of several seconds to tens of seconds after the current pulse ends; the exemplary value of the second preset amplitude threshold can be the upper limit of the low change interval in the statistical distribution of the amplitude change rate of the corresponding characteristic parameter under stable operating conditions.

[0064] Methods for performing smooth calibration of the continuous evolution trajectory of battery state within the timestamp neighborhood include: Centered on the timestamp corresponding to the internal state anchor point parameter, the timestamp neighborhood covering the time range before and after the battery state continuous evolution trajectory is determined, and the continuous inferred state within the timestamp neighborhood is extracted.

[0065] It should be noted that after determining the timestamp corresponding to the internal state anchor point parameter, this timestamp is used as the time center point, and a certain continuous time range is extended in the direction before and after the time axis in the continuous evolution trajectory of the battery state to form a timestamp neighborhood covering the state change process before and after the acquisition time of the internal state anchor point parameter; within the timestamp neighborhood, the corresponding continuous deduced state in the continuous evolution trajectory of the battery state is extracted in chronological order.

[0066] Based on the magnitude of the state deviation vector and the time distance between the continuously derived state and the timestamp of the internal state anchor point parameter, a correlation-based smooth convergence adjustment is performed on the continuously derived state in the timestamp neighborhood.

[0067] It should be noted that after obtaining the continuous deduced states within the timestamp neighborhood, the state deviation vector calculated at the anchor point time is used as the deviation benchmark for the continuous deduced states at the anchor point time. Combined with the time distance between the timestamps corresponding to each continuous deduced state within the timestamp neighborhood and the timestamps corresponding to the internal state anchor point parameters, different degrees of deviation correction are applied to the continuous deduced states, causing the state deviation vector to gradually decrease along the time axis. At any consecutive time point within the timestamp neighborhood... At this point, the corresponding weighted bias correction expression is: ; in, At consecutive time points The continuous deduction state is after weighted bias correction. At consecutive time points The continuously deduced state obtained at that point, It is related to continuous time points The corresponding deviation correction weight; Continuous simulation states with a shorter time interval (exemplary time interval less than 1 second) correspond to a larger deviation correction weight (exemplary weight coefficient close to 1), while continuous simulation states with a longer time interval (exemplary time interval greater than 1 second) correspond to a smaller deviation correction weight (exemplary weight coefficient close to 0). This allows the state deviation vector to gradually decay along the time axis and act on the continuous simulation states, thereby achieving a smooth transitional state correction while ensuring time continuity.

[0068] The continuously extrapolated state after smooth convergence adjustment is updated to the continuous evolution trajectory of the battery state in chronological order, so that it gradually converges to the internal state anchor point parameters.

[0069] It should be noted that after the weighted bias correction of the continuous extrapolated states within the timestamp neighborhood is completed, the bias-corrected continuous extrapolated states are written back to the corresponding time positions in the continuous evolution trajectory of the battery state in chronological order according to the time stamps. During the update process, the continuity of the state between adjacent time positions is maintained, so that the continuous evolution trajectory of the battery state is consistent with the internal state anchor parameters at the corresponding time, and achieves a smooth transition within the time range before and after them, so that the continuous evolution trajectory of the battery state gradually converges towards the internal state anchor parameters as a whole. The smoothed and calibrated continuous evolution trajectory of the battery state is used as the continuous battery full state vector.

[0070] This embodiment also provides a BMS battery state AI intelligent estimation system, including: The acquisition module is used to apply current pulses to the battery during battery operation and acquire the voltage at the battery terminals as a function of time to obtain a voltage-time response curve. The feature extraction module is used to perform multi-level feature deconstruction on the voltage-time response curve, generate pulse impedance spectrum feature vector, and establish internal state anchor point parameters characterizing the internal health state of the battery based on the pulse impedance spectrum feature vector. The evolution module is used to extrapolate the evolution of the battery state based on the internal state anchor parameters and battery operation data, and to obtain the continuous evolution trajectory of the battery state by using the internal state anchor parameters as the state reference benchmark. The calibration module is used to compare and identify the deviation between the predicted state of the continuous evolution trajectory at the corresponding moment and the internal state anchor parameters when new internal state anchor parameters are obtained again through current pulses during battery operation, and to perform smooth calibration on the continuous evolution trajectory based on the deviation to obtain the continuous battery full state vector. The output module is used to resolve the continuous battery full state vector into battery state of charge and battery health state, and is used for battery energy scheduling control and health warning.

[0071] This embodiment also provides a computer device applicable to the BMS battery state AI intelligent estimation method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the BMS battery state AI intelligent estimation method proposed in the above embodiment.

[0072] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0073] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the AI-based intelligent estimation method for BMS battery status as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0074] In summary, this invention constructs a new intelligent estimation paradigm of "active detection-closed-loop calibration" through active pulse excitation and internal state anchor point extraction, as well as continuous evolution and periodic smoothing calibration based on the anchor points. It directly obtains a high-confidence internal electrochemical state benchmark through physical excitation, providing a new approach to addressing estimation source uncertainties and enhancing the early perception capability of microscopic health states. Furthermore, by using this benchmark to perform periodic smoothing correction on the continuous evolution model, long-term state drift is suppressed, achieving continuous and seamless monitoring of the entire life cycle state. This provides long-term stable and mechanistically clear high-precision state estimation for real-time energy management and full life cycle health prediction of batteries.

[0075] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A BMS battery state AI intelligent estimation method, characterized in that, include: During battery operation, a current pulse is applied to the battery, and the voltage at the battery terminals changes over time to obtain a voltage-time response curve. Multi-level feature deconstruction is performed on the voltage-time response curve to generate pulse impedance spectrum feature vector, and internal state anchor point parameters characterizing the internal health state of the battery are established based on the pulse impedance spectrum feature vector. Based on the internal state anchor point parameters and battery operation data, the battery state is evolved and deduced, and the internal state anchor point parameters are used as the state reference benchmark to obtain the continuous evolution trajectory of the battery state. When new internal state anchor parameters are obtained again through current pulse during battery operation, the predicted state of the continuous evolution trajectory at the corresponding time is compared and identified with the internal state anchor parameters, and the continuous evolution trajectory is smoothed and calibrated based on the deviation to obtain the continuous battery full state vector. The continuous battery full-state vector is analyzed into battery state of charge and battery health state, and used for battery energy scheduling control and health early warning.

2. The BMS battery state AI intelligent estimation method as described in claim 1, characterized in that, The method for obtaining the voltage-time response curve includes: During battery operation, the changes in battery terminal voltage and current are continuously monitored; when the rate of change of voltage and the rate of change of current are both lower than the change threshold within a preset time window, the battery is determined to be in a stable working state. After the battery enters a stable operating state, a current pulse with a constant amplitude and fixed duration is applied to the battery; Before, during, and after the application of the current pulse, the battery terminal voltage is collected to obtain voltage data; the voltage data is then time-stamped and sequentially arranged to construct a voltage-time response curve.

3. The BMS battery state AI intelligent estimation method as described in claim 2, characterized in that, The method for generating pulse impedance spectrum eigenvectors includes: Based on the timing of the current pulse application and the voltage relaxation process after the pulse ends, the voltage-time response curve is divided into a transient response segment and a relaxation response segment. Multi-level feature extraction is performed on the voltage data in the transient response and relaxation response regions to obtain the internal electrochemical dynamic characteristic parameters of the battery, and a unified scale mapping is performed on the internal electrochemical dynamic characteristic parameters of the battery. The internal electrochemical dynamic characteristic parameters of the battery are arranged and combined in the order of their formation to form the pulse impedance spectrum characteristic vector of the battery's internal response characteristics under current pulse excitation.

4. The BMS battery state AI intelligent estimation method as described in claim 3, characterized in that, The method for establishing internal state anchor point parameters characterizing the internal health state of a battery includes: Based on the time distribution and amplitude variation characteristics of the pulse impedance spectrum eigenvector under the action of current pulse, state correlation analysis is performed to distinguish the characteristic parameters of fast and slow electrochemical response processes. Based on the characteristic parameters of fast and slow electrochemical response processes, determine the set of state characterization parameters for the internal electrochemical kinetic state of the battery; The set of state representation parameters is quantified to obtain internal state parameters; these internal state parameters are then used as internal state anchor parameters for the current internal health state of the battery.

5. The BMS battery state AI intelligent estimation method as described in claim 4, characterized in that, The method for obtaining the continuous evolution trajectory of the battery state includes: Based on the internal state anchor point parameters and the battery operation data continuously collected during battery operation, a battery state evolution time benchmark with time as the continuous independent variable is established, and the internal state anchor point parameters are used as the state reference benchmark to divide the continuous evolution interval of the battery state between adjacent anchor point parameter acquisition times. Based on battery operation data within the continuous evolution interval, the battery state is progressively evolved over a continuous time scale to obtain continuous state change results; and the continuous state change results are then connected in time sequence to form a continuous evolution trajectory of the battery state.

6. The BMS battery state AI intelligent estimation method as described in claim 5, characterized in that, The method for obtaining the full state vector of a continuous battery includes: During battery operation, when a current pulse is applied to the battery again and new internal state anchor parameters are obtained, the timestamp corresponding to the internal state anchor parameters is determined. Based on the timestamps corresponding to the internal state anchor point parameters, the predicted battery state at the corresponding moment is extracted from the continuous evolution trajectory of the battery state and used as the anchor point to align the predicted state. A comparison analysis of the anchor point alignment prediction state and the newly acquired internal state anchor point parameters is performed to obtain the state deviation vector of the degree of inconsistency between the continuously extrapolated state and the actual internal state of the battery. Based on the state deviation vector, the continuous evolution trajectory of the battery state is smoothed and calibrated in the timestamp neighborhood, so that it gradually converges to the internal state anchor parameters; the smoothed and calibrated continuous evolution trajectory of the battery state is used as the continuous battery full state vector.

7. The BMS battery state AI intelligent estimation method as described in claim 6, characterized in that, The method for resolving the continuous battery full-state vector into battery state of charge and battery health state includes: Based on the time-varying characteristics of each state component in the full state vector of a continuous battery, the state dimension of the full state vector of a continuous battery is analyzed to separate the state of charge parameters and the health state parameters. The state of charge parameter is used as a characterization index of the current energy availability of the battery for energy scheduling control during the battery charging and discharging process; The health status parameters are compared and analyzed with preset health status criteria. When the health status parameters meet the abnormal or degradation conditions, corresponding battery health warning information is generated.

8. The BMS battery state AI intelligent estimation method as described in claim 4, characterized in that, The method for performing state association analysis includes: Based on the time distribution characteristics of each dimension of the characteristic parameters in the pulse impedance spectrum feature vector before, during and after the current pulse is applied, the correspondence between each dimension of the characteristic parameters and the timing of the internal electrochemical response of the battery is established. Based on the correspondence, the characteristic parameters of the time distribution feature that are concentrated in the initial stage of current pulse application, and whose amplitude changes occur within a short response time window and whose amplitude change rate is higher than the first preset amplitude threshold are classified as characteristic parameters of the fast electrochemical response process. The characteristic parameters of the slow electrochemical response process are those whose response time distribution is concentrated in the voltage relaxation stage after the current pulse ends, whose amplitude changes are distributed within the long response time window, and whose amplitude change rate is lower than the second preset amplitude threshold.

9. The BMS battery state AI intelligent estimation method as described in claim 6, characterized in that, The method for performing smooth calibration on the continuous evolution trajectory of the battery state within the timestamp neighborhood includes: Centered on the timestamp corresponding to the internal state anchor point parameter, the timestamp neighborhood covering the time range before and after the battery state continuous evolution trajectory is determined, and the continuous inferred state within the timestamp neighborhood is extracted. Based on the magnitude of the state deviation vector and the time distance between the continuously extrapolated state and the timestamp of the internal state anchor point parameter, a correlation-based smooth convergence adjustment is performed on the continuously extrapolated state in the timestamp neighborhood. The continuously extrapolated state after smooth convergence adjustment is updated to the continuous evolution trajectory of the battery state in chronological order, so that it gradually converges to the internal state anchor point parameters.

10. A BMS battery state AI intelligent estimation system, based on the BMS battery state AI intelligent estimation method according to any one of claims 1 to 9, characterized in that, include: The acquisition module is used to apply current pulses to the battery during battery operation and acquire the voltage at the battery terminals as a function of time to obtain a voltage-time response curve. The feature extraction module is used to perform multi-level feature deconstruction on the voltage-time response curve, generate pulse impedance spectrum feature vector, and establish internal state anchor point parameters characterizing the internal health state of the battery based on the pulse impedance spectrum feature vector. The evolution module is used to extrapolate the evolution of the battery state based on the internal state anchor parameters and battery operation data, and to obtain the continuous evolution trajectory of the battery state by using the internal state anchor parameters as the state reference benchmark. The calibration module is used to compare and identify the deviation between the predicted state of the continuous evolution trajectory at the corresponding moment and the internal state anchor parameters when new internal state anchor parameters are obtained again through current pulses during battery operation, and to perform smooth calibration on the continuous evolution trajectory based on the deviation to obtain the continuous battery full state vector. The output module is used to resolve the continuous battery full state vector into battery state of charge and battery health state, and is used for battery energy scheduling control and health warning.

Images