Battery state of health prediction method, management system, energy storage device, and storage medium

Abstract

The embodiment of the application provides a battery health state prediction method, a management system, a storage device and a storage medium. The method comprises the following steps: based on a real-time collected state of charge sequence of a battery, a cycle parameter of at least one charge-discharge cycle is extracted; the damage contribution of each charge-discharge cycle to the battery aging is calculated respectively according to each cycle parameter; a total damage is obtained by accumulating a plurality of damage contributions; and the health state of the battery is determined according to the total damage. The embodiment of the application can realize real-time and high-precision prediction of the battery health state under a high-frequency micro-cycle working condition.

Description

Technical Field

[0001] This invention relates to the field of battery state monitoring technology, specifically to a battery health state prediction method, a battery management system, a battery module, an energy storage device, and a computer-readable storage medium. Background Technology

[0002] With the rapid development of cloud computing, artificial intelligence, and big data applications, data centers have become critical infrastructure. To ensure power supply reliability and mitigate the impact of load fluctuations on the power grid, data center energy storage systems need to alternate between uninterruptible power supply (UPS) backup mode and frequency regulation mode. In UPS backup mode, the battery is float-charged at a high state of charge for an extended period; in frequency regulation mode, the battery's state of charge fluctuates frequently around a midpoint with slight amplitude.

[0003] Under such high-frequency micro-cycle conditions, battery life assessment faces unique challenges. Most battery state of health (SOH) prediction methods are based on relatively regular charge-discharge cycle conditions, such as cycle tests with a fixed depth of discharge. However, for data center energy storage scenarios, due to the smaller fluctuations in state of charge and the infrequent occurrence of deep charge-discharge cycles, methods using equivalent full cycle counts have significant errors.

[0004] Furthermore, the rainflow counting methods used in related technologies are mostly offline batch processing methods, which require complete historical load data to perform counting. In high-frequency sampling scenarios (second-level or even millisecond-level) in data centers, offline processing cannot meet the needs of real-time evaluation and is prone to misidentifying spurious peaks and valleys caused by noise as valid cycles, leading to overestimation of damage.

[0005] On the other hand, some existing lifetime prediction models mainly consider the stress in the dimension of depth of discharge. However, under the floating charge condition of high charge state in data centers, calendar aging and cyclic aging exhibit strong nonlinear coupling. Ignoring multi-dimensional stress factors such as average state of charge and charge / discharge rate will lead to systematic deviations in the prediction results. Summary of the Invention

[0006] The embodiments of the present invention provide a battery health status prediction method, a battery management system, a battery module, an energy storage device, an electronic device, and a computer-readable storage medium, which can improve the technical problems of real-time performance and accuracy of battery health status prediction under high-frequency micro-circulation conditions in data centers.

[0007] In a first aspect, embodiments of the present invention provide a battery health state prediction method, comprising: Based on the real-time acquired state of charge sequence of the battery, at least one charge-discharge cycle parameter is extracted from the state of charge sequence. Based on each cycle parameter, calculate the damage contribution of each charge-discharge cycle to battery aging. Accumulate multiple damage contributions to obtain the total damage; The health status of the battery is determined based on the total damage.

[0008] In the embodiments of the present invention, cyclic extraction and damage accumulation are performed simultaneously during the real-time acquisition of the state of charge sequence, which can output the battery health status in real time and meet the online evaluation requirements in high-frequency sampling scenarios in data centers. At the same time, by decomposing the complex operating conditions experienced by the battery into a series of charge-discharge cycles for damage accumulation, health status prediction based on cycle-level damage is realized, improving the accuracy and real-time performance of the prediction.

[0009] In one embodiment, extracting cycle parameters for at least one charge-discharge cycle from the state-of-charge sequence includes: The peak and valley extraction thresholds are dynamically adjusted according to the battery's operating mode. Based on the adjusted peak-valley extraction threshold, a sequence of reversal points arranged in chronological order is obtained from the charged state sequence, wherein the reversal points are local maxima or local minima in the charged state sequence. The reversal point sequence is counted online to obtain the loop parameters.

[0010] This embodiment effectively filters out measurement noise under different operating conditions by adaptively adjusting the peak and valley extraction threshold, avoiding misidentification of noise as valid cycles, while retaining real shallow fluctuation information, thereby improving the accuracy of cycle identification.

[0011] In one embodiment, the online counting process of the reversal point sequence to obtain the loop parameters includes: The reversal points are added to the stack structure in sequence, and each time a new reversal point is added to the stack structure, it is determined whether there is a charge-discharge cycle in the stack structure. When a charge / discharge cycle exists, the cycle parameters of the charge / discharge cycle are output, and the corresponding inversion point is removed from the stack structure.

[0012] This embodiment employs an online counting algorithm based on a stack structure, which can complete loop identification and parameter output in real time while data is being acquired, significantly reducing processing latency and storage requirements, and meeting the real-time requirements of high-frequency sampling scenarios.

[0013] In one embodiment, the step of sequentially adding the reversal points to the stack structure, and determining whether a charge-discharge cycle exists in the stack structure each time a new reversal point is added to the stack structure, includes: Obtain four consecutive reversal points at the top of the stack structure, wherein the four reversal points include a first reversal point, a second reversal point, a third reversal point, and a fourth reversal point; Calculate the absolute value of the first difference between the first reversal point and the second reversal point, the absolute value of the second difference between the second reversal point and the third reversal point, and the absolute value of the third difference between the third reversal point and the fourth reversal point; When the absolute value of the first difference is not greater than the absolute value of the second difference, and the absolute value of the second difference is not greater than the absolute value of the third difference, the interval corresponding to the second reversal point and the third reversal point is determined to form the charge-discharge cycle.

[0014] This embodiment uses a four-point method to identify closed charge-discharge cycles from the reversal point sequence in a simple and efficient way, thus improving the accuracy of online counting.

[0015] In one embodiment, dynamically adjusting the peak-valley extraction threshold according to the battery's operating mode includes: Calculate the standard deviation of the state of charge within the sliding time window; When the standard deviation is less than the preset threshold, it is determined to be in floating mode, and the peak-valley extraction threshold is set to the first threshold. When the standard deviation is greater than or equal to the preset threshold, it is determined to be a frequency modulation mode, and the peak-valley extraction threshold is set to the second threshold. Wherein, the first threshold is greater than the second threshold.

[0016] This embodiment distinguishes between float charging mode and frequency modulation mode by calculating the standard deviation of the state of charge, and dynamically adjusts the peak and valley extraction threshold accordingly, thereby realizing the adaptive configuration of the peak and valley extraction threshold according to the operating conditions.

[0017] This embodiment provides an optimal parameter range suitable for data center energy storage scenarios, which can reasonably configure peak and valley extraction thresholds in this scenario, ensuring effective identification of real cycles while avoiding misidentification of noise.

[0018] In one embodiment, the cyclic parameters include cyclic amplitude and average state of charge; The accumulation of multiple damage contributions to obtain total damage includes: Based on the average state of charge of each charge-discharge cycle, determine the nonlinear index corresponding to the charge-discharge cycle; The damage contribution of each charge-discharge cycle is accumulated according to the corresponding nonlinear exponent to obtain the total damage.

[0019] This embodiment uses a nonlinear exponential accumulation method to dynamically adjust the weight of each cycle damage in the total damage according to the average state of charge. This can correct the systematic error of the traditional linear accumulation method under high charge floating charging conditions and improve the prediction accuracy.

[0020] In one embodiment, the nonlinear exponent deviates from 1 when the average state of charge deviates from a preset reference value.

[0021] This embodiment further defines the relationship between the nonlinear exponent and the average state of charge. By setting a nonlinear exponent of less than 1 under high charge state, the damage contribution is amplified, simulating the strong nonlinear coupling effect of calendar aging and cyclic aging.

[0022] In one embodiment, the cycling parameters include cycling amplitude, average state of charge, and charge / discharge rate, wherein the charge / discharge rate is determined based on the ratio of the collected current value to the battery's rated capacity. The step of calculating the damage contribution of each charge-discharge cycle to battery aging based on each of the cycle parameters includes: Based on the pre-calibrated multidimensional stress damage function, the damage contribution of each charge-discharge cycle to battery aging is calculated, wherein the independent variable of the multidimensional stress damage function includes the cycle parameter.

[0023] This embodiment considers three stress dimensions simultaneously: cycle amplitude, average state of charge, and charge / discharge rate, which can more comprehensively reflect the aging pattern of the battery under actual operating conditions and avoid the systematic bias of single-dimensional models.

[0024] In a second aspect, embodiments of the present invention provide a battery management system, including: Sensor components are used to collect voltage, current, and temperature data from energy storage batteries; The state of charge estimation module is used to output a state of charge sequence in real time based on the voltage, current and temperature data; The control module, connected to the state of charge estimation module, is used to execute the battery health state prediction method described above.

[0025] This battery management system integrates sensor acquisition, state of charge estimation, and health status prediction functions. It can monitor battery operating data online and output health status in real time, providing decision-making basis for the operation and maintenance management of energy storage systems.

[0026] Thirdly, embodiments of the present invention provide a battery module, comprising: One or more batteries; A battery management system as described above that is connected to the one or more batteries.

[0027] This battery module integrates individual battery cells with a battery management system that has a health status prediction function, making it easy to install, maintain and replace, and improving the reliability and safety of the battery system.

[0028] Fourthly, embodiments of the present invention provide an energy storage device, comprising: Battery array; The battery management system described above is connected to the battery array.

[0029] This energy storage system can monitor the health status of the battery array in real time, detect abnormal battery cells in a timely manner, and avoid power outages or safety accidents caused by battery aging. It is suitable for data center backup power and power grid frequency regulation services.

[0030] Fifthly, embodiments of the present invention provide an electronic device, comprising: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the battery health state prediction method as described above.

[0031] This electronic device can function as a standalone battery monitoring device or be integrated into a battery management system to enable real-time prediction of battery health status.

[0032] In a sixth aspect, embodiments of the present invention provide a computer-readable storage medium storing executable instructions for inducing a processor to execute and implement the battery health state prediction method as described above.

[0033] The beneficial effects of the embodiments of the present invention are as follows: In the embodiments of this invention, charge-discharge cycle parameters are extracted online based on real-time acquired state-of-charge sequences. The damage contribution of each cycle is calculated and accumulated to obtain the total damage. Then, the battery health status is determined based on the total damage, achieving real-time, high-precision prediction of battery health status under high-frequency micro-cycle conditions in data centers. Compared with related technologies that require complete load history for offline processing, the embodiments of this invention can meet the real-time evaluation requirements in second-level or even millisecond-level sampling scenarios. By adaptively adjusting the peak and valley extraction threshold according to the operating conditions, spurious peak and valley values ​​caused by measurement noise are effectively filtered out, avoiding overestimation of damage. By introducing a multidimensional stress damage function and a nonlinear accumulation rule, the aging contribution of three stress dimensions—cycle amplitude, average state of charge, and charge-discharge rate—is considered simultaneously. The systematic error of the linear accumulation rule in the strongly nonlinear coupling scenario of high-charge float charging and cyclic aging is corrected, significantly improving the prediction accuracy. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0035] Figure 1 This is a schematic flowchart of the battery health status prediction method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a battery health status prediction device provided in an embodiment of the present invention; Figure 3 This is another structural schematic diagram of the battery health status prediction device provided in the embodiments of the present invention; Figure 4 This is another structural schematic diagram of the battery health status prediction device provided in the embodiments of the present invention; Figure 5 This is a schematic diagram of the battery management system provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the battery module provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the energy storage device provided in an embodiment of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Furthermore, it should be understood that the specific embodiments described herein are only for illustration and explanation of the present invention and are not intended to limit the present invention. In the present invention, unless otherwise stated, directional terms such as "upper" and "lower" generally refer to the upper and lower positions of the device in actual use or operation, specifically the drawing directions in the accompanying drawings; while "inner" and "outer" refer to the outline of the device.

[0037] Based on the problems mentioned in the background, the battery life prediction methods in related technologies mainly suffer from the following issues: Rainflow counting methods are mostly offline batch processing methods, requiring complete state-of-charge (POC) change curves before cycle identification and damage calculation can be performed. This fails to meet the real-time evaluation requirements of high-frequency sampling scenarios in data centers, and high-frequency noise can easily lead to false peaks and valleys being misidentified as valid cycles, resulting in overestimation of damage. Damage functions typically only consider the depth of discharge, ignoring the influence of average POC and charge / discharge rate. However, in data centers where high POC float charging and shallow micro-cycles alternate, there is a strong nonlinear coupling between calendar aging and cycle aging. The higher the average POC and the greater the rate, the more significant the actual damage caused by the same discharge depth, inevitably leading to systematic bias in the single-dimensional model. Damage accumulation commonly uses the Miner linear rule, which assumes that damage from each cycle is independently additive. However, in the aforementioned nonlinear coupling scenario, the linear accumulation assumption deviates significantly from the actual aging law, requiring the introduction of nonlinear correction.

[0038] Based on the above analysis, embodiments of the present invention provide a battery health status prediction method, apparatus, and computer-readable storage medium, which can improve the technical problems of real-time performance and accuracy of battery health status prediction under high-frequency micro-circulation conditions in data centers.

[0039] This application provides a method for predicting battery health status. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating the battery health state prediction method provided in this application embodiment, as shown below. Figure 1 As shown, the method includes: Step S101: Based on the real-time collected battery operating data, obtain the state of charge sequence, and extract the cycle parameters of at least one charge-discharge cycle from the state of charge sequence.

[0040] Specifically, before step S101, which involves extracting cycle parameters for at least one charge-discharge cycle from the real-time acquired state-of-charge sequence of the battery, the method further includes: Collect battery voltage, current, and temperature data; The state of charge is estimated based on voltage, current, and temperature data to form a sequence of states of charge.

[0041] During battery operation, the battery management system (BMS) estimates and outputs the data in real time, using algorithms such as the ampere-hour integration method, extended Kalman filtering method, or open-circuit voltage correction method, based on collected voltage, current, and temperature data. Those skilled in the art can choose the appropriate estimation method according to the actual application scenario, which will not be elaborated here.

[0042] In this embodiment, the sampling frequency of the state of charge sequence can be set according to the actual application scenario. For example, in the data center energy storage scenario, the sampling frequency is not less than 1 Hz to meet the monitoring requirements of high-frequency micro-circulation conditions.

[0043] As an optional embodiment, before step S101, the charged state sequence is further subjected to denoising processing. The purpose of denoising processing is to suppress the interference of high-frequency measurement noise on subsequent extreme point extraction and cycle identification, and to avoid misidentifying noise-induced pseudo-peaks and valleys as valid cycles. Denoising processing can be implemented in various ways, such as discrete wavelet transform denoising. Specifically, the original charged state sequence can be decomposed into wavelet components, and the parts of the detailed components obtained from the decomposition with amplitudes less than a preset threshold can be regarded as noise and filtered out. Then, wavelet reconstruction is performed based on the low-frequency components to obtain the denoised charged state sequence. It should be noted that the specific implementation method of denoising processing can be flexibly selected according to the noise level in the actual scenario, and those skilled in the art will understand that it is not limited to discrete wavelet transform.

[0044] A charge-discharge cycle refers to a battery undergoing a complete discharge and charge process, or the process of the state of charge changing from one extreme point to another and back. Cycle parameters can be extracted in various ways. One optional implementation is to identify charge-discharge cycles by detecting local maxima and minima (peaks and valleys) in the state of charge sequence. The change between every two adjacent extreme points constitutes a half-cycle, and these are paired to form a complete cycle. Another optional implementation is to use rainflow counting to perform cyclic statistics on the state of charge sequence and output the cycle parameters for each cycle. Yet another optional implementation is to use a sliding window method based on a time window to detect the fluctuation amplitude of the state of charge in real time; when the fluctuation amplitude exceeds a preset threshold, it is identified as a charge-discharge cycle.

[0045] It should be noted that the above extraction process is performed synchronously during the real-time acquisition of the state of charge sequence, that is, there is no need to wait for the complete state of charge sequence to be acquired before the extraction operation is performed, thus meeting the needs of real-time prediction.

[0046] Step S102: Calculate the damage contribution of each charge-discharge cycle to battery aging based on each cycle parameter.

[0047] The method for calculating damage contribution can be determined based on a pre-defined damage model. For example, damage coefficients under different combinations of cycling parameters can be pre-calibrated experimentally, a mapping table can be established, and then the cycling parameters of the current cycle can be matched with the mapping table to obtain the corresponding damage contribution value. Alternatively, a damage function with cycling parameters as independent variables can be constructed, and the damage contribution can be calculated by substituting the cycling parameters into the function. The damage function can be a linear function, an exponential function, or a polynomial function, depending on the battery type and application scenario. For high-frequency micro-cycle conditions in data centers, the damage function can consider the influence of multiple stress factors such as cycle amplitude, average state of charge, and charge / discharge rate on aging to more accurately reflect the aging pattern of the battery under actual operating conditions.

[0048] Step S103: Accumulate multiple damage contributions to obtain the total damage.

[0049] The accumulation method can be linear accumulation, which directly adds the damage contributions of each cycle; or it can be weighted accumulation, for example, assigning different weights based on the order in which the cycles occur or the magnitude of the cycle amplitude. During the accumulation process, whenever a new charge-discharge cycle is identified and its damage contribution is calculated, that damage contribution is immediately added to the total damage, thereby achieving real-time updates of the damage.

[0050] Step S104: Determine the health status of the battery based on the total damage.

[0051] Battery health status is typically expressed as a percentage of remaining capacity. For example, a correlation can be set between total damage and health status: when total damage is 0, the health status is 100% (new battery); when total damage reaches a preset threshold (e.g., 1), the health status drops to the end-of-life threshold (e.g., 80%). By comparing the current total damage with this correlation, the current battery health status can be obtained.

[0052] In addition, the remaining battery life can be estimated based on the total rate of damage accumulation and the recent average damage contribution per cycle.

[0053] This embodiment extracts cyclic parameters of charge-discharge cycles online from a real-time acquired state-of-charge (POC) sequence and accumulates the total damage based on the damage contribution of each cycle. This decomposes the complex operating conditions experienced by the battery into a series of charge-discharge cycles, enabling health status prediction based on cycle accumulation. Compared to related technologies that require a complete load history for offline analysis, this embodiment performs cyclic extraction and damage calculation simultaneously during the real-time acquisition of the POC sequence, providing real-time output of the battery health status and meeting the online evaluation requirements of high-frequency sampling scenarios. Furthermore, by rationally designing the damage contribution calculation method, such as considering factors like average POC and charge-discharge rate, it can more accurately reflect the aging pattern of the battery under actual operating conditions, improving prediction accuracy. Therefore, this embodiment can effectively improve the technical issues of real-time performance and accuracy in predicting battery health status under high-frequency micro-circulation conditions in data centers.

[0054] In some embodiments, extracting cycle parameters for at least one charge-discharge cycle from the state-of-charge sequence includes: The peak and valley extraction thresholds are dynamically adjusted according to the battery's operating mode. Based on the adjusted peak-valley extraction threshold, a sequence of reversal points arranged in chronological order is obtained from the charged state sequence, where the reversal points are local maxima or local minima in the charged state sequence. The reversal point sequence is counted online to obtain the loop parameters.

[0055] The battery's operating modes can include float charging mode and frequency regulation mode. Float charging mode is suitable for scenarios where the battery is in a high state of charge (SOC) for extended periods with minimal SOC fluctuations, while frequency regulation mode is suitable for scenarios where the SOC fluctuates frequently around a midpoint. Based on the determined operating mode, the peak-valley extraction threshold is dynamically adjusted. For example, in float charging mode, noise in the SOC sequence is relatively significant, requiring a larger peak-valley extraction threshold to filter out minor fluctuations; in frequency regulation mode, SOC fluctuations are more frequent, requiring a smaller peak-valley extraction threshold to capture the true charge-discharge cycle.

[0056] In this context, reversal points are local maxima or minima in the state-of-charge (SOC) sequence. Specifically, by iterating through each sampling point in the SOC sequence, a sampling point is identified as an extreme point when the difference between its SOC value and that of its neighboring sampling points meets a preset condition. A peak-valley extraction threshold controls the sensitivity of extreme point extraction: a larger threshold results in fewer extracted extreme points (only large fluctuations are identified); a smaller threshold results in more extracted extreme points (small fluctuations are also identified). Arranging the extracted local maxima and minima in chronological order yields the reversal point sequence.

[0057] Online counting processing refers to the simultaneous cyclic identification during the real-time acquisition of the state of charge (SOC) sequence. This means that the processing results are updated in real time as soon as a new inversion point is acquired, without waiting for the entire inversion point sequence to be acquired before processing. Specifically, during the acquisition of the inversion point sequence, the amplitude relationship between each inversion point is determined in real time, identifying pairs of inversion points that can form a charge-discharge cycle, and calculating the cycle parameters for each charge-discharge cycle. The cycle parameters include at least the cycle amplitude and the average SOC.

[0058] There are several methods to identify charge-discharge cycles. For example, adjacent pairs of maxima and minima can be considered as a half-cycle, forming a complete cycle through pairing. Alternatively, rainflow counting can be used to process the sequence of extreme points and output the cycle parameters for each closed cycle. The cycle parameters include at least the cycle amplitude (depth of discharge) and the average state of charge.

[0059] By following the steps above, the cyclic parameters of each charge-discharge cycle can be extracted from the real-time acquired state of charge sequence, providing basic data for subsequent damage contribution calculation and health status prediction.

[0060] This embodiment adaptively adjusts the peak-valley extraction threshold based on operating conditions, enabling reasonable configuration of the reversal point extraction sensitivity according to different operating modes. This avoids misidentifying noise as a reversal point in float charging mode, while retaining true shallow fluctuation information in frequency modulation mode. Through online counting processing, it can output cyclic parameters in real time while acquiring data, meeting the real-time requirements of high-frequency sampling scenarios.

[0061] In some embodiments, the reversal point sequence is subjected to online counting to obtain the loop parameters, including: Add the reversal points to the stack structure one by one, and each time a new reversal point is added to the stack structure, check whether there is a charging and discharging cycle in the stack structure; When a charge / discharge cycle exists, output the cycle parameters of the charge / discharge cycle and remove the corresponding inversion point from the stack structure.

[0062] Reversal points, also known as local extrema, reflect the turning points in the direction of change of the state of charge, such as points where the charge state changes from decreasing to increasing (local minimum) or from increasing to decreasing (local maximum). The sequence of reversal points is the basis for subsequent cyclic identification.

[0063] A stack is a Last-In-First-Out (LIFO) data structure used to temporarily store unclosed extreme points. Initially, the stack is empty. As the charged state sequence is acquired in real time, reversal points are extracted sequentially. Each time a new reversal point is added to the stack, it is called a push operation on the stack.

[0064] Each time a new inversion point is added to the stack structure, it is determined whether an outputtable charge-discharge cycle exists in the stack structure. This determination can be made using known online rainflow counting rules such as the four-point method or the three-point method. When the cycle closure condition is met, an outputtable charge-discharge cycle is confirmed. This charge-discharge cycle corresponds to the interval between two inversion points in the stack, and its cycle parameters (such as cycle amplitude and average state of charge) can be calculated based on the values ​​of these two inversion points.

[0065] When the determination result indicates the existence of an outputtable charge-discharge cycle, the cycle parameters of that cycle are output, and the corresponding reversal point is removed from the stack structure. The removal operation typically pops the two reversal points forming a closed loop from the stack, while retaining other reversal points relevant to subsequent determinations. After outputting the cycle parameters, the cycle can be immediately used for damage calculation and accumulation. Afterwards, the process continues to wait for new reversal points to be added, repeating the above determination and output process.

[0066] Using the above method, during the real-time acquisition of the state of charge sequence, each time a new reversal point is extracted, the cycle identification and parameter output can be completed simultaneously without waiting for the complete state of charge sequence to be acquired, thus realizing online and real-time charge-discharge cycle extraction.

[0067] This embodiment discretizes the continuously changing state of charge sequence into a series of reversal points reflecting the turning points of the change direction by extracting extreme points. Utilizing the last-in-first-out (LIFO) characteristic of a stack structure, it maintains the sequence of extreme points that are not yet closed. Each time a new reversal point is added, it determines whether an outputtable charge-discharge cycle exists according to a pre-set closure rule; if so, it is immediately output and the corresponding point is removed. Compared to the reversal point pairing method that requires batch processing after completing the complete load history, the online counting algorithm in this embodiment can complete cycle identification simultaneously with data acquisition, significantly reducing latency and storage requirements, and is suitable for high-frequency sampling scenarios such as data center energy storage.

[0068] In some embodiments, reversal points are sequentially added to a stack structure, and each time a new reversal point is added to the stack structure, it is determined whether a charging / discharging cycle exists in the stack structure, including: Obtain the four consecutive reversal points at the top of the stack structure, where the four reversal points include the first reversal point, the second reversal point, the third reversal point, and the fourth reversal point; Calculate the absolute value of the first difference between the first reversal point and the second reversal point, calculate the absolute value of the second difference between the second reversal point and the third reversal point, and calculate the absolute value of the third difference between the third reversal point and the fourth reversal point; When the absolute value of the first difference is not greater than the absolute value of the second difference, and the absolute value of the second difference is not greater than the absolute value of the third difference, the interval corresponding to the second reversal point and the third reversal point is determined to form the charging and discharging cycle.

[0069] In processing the reversal point sequence using an online counting algorithm based on a stack structure, after each new reversal point is added to the stack, it is necessary to determine whether there is an outputtable charge-discharge cycle in the current stack. This embodiment provides a judgment rule based on the four-point method.

[0070] Specifically, obtain the four consecutive inversion points at the top of the current stack structure. Due to the Last-In-First-Out (LIFO) property of the stack structure, the top of the stack is the most recently added inversion point. For ease of description, these four inversion points are denoted as the first inversion point P1, the second inversion point P2, the third inversion point P3, and the fourth inversion point P4 in the order from the bottom of the stack to the top (i.e., in chronological order), where the fourth inversion point P4 is the most recently added inversion point.

[0071] For the four reversal points obtained, calculate the absolute value of the difference between any two adjacent points. Specifically, calculate the first absolute value of the difference between the first reversal point P1 and the second reversal point P2, |P2-P1|, which is the absolute value of the difference in the numerical values ​​of the charge state of the two reversal points; calculate the second absolute value of the difference between the second reversal point P2 and the third reversal point P3, |P3-P2|; and calculate the third absolute value of the difference between the third reversal point P3 and the fourth reversal point P4, |P4-P3|.

[0072] Then, the absolute values ​​of the three differences are compared. When the absolute value of the first difference is not greater than the absolute value of the second difference, and the absolute value of the second difference is not greater than the absolute value of the third difference, |P2-P1| ≤ |P3-P2| ≤ |P4-P3|, that is, when the absolute values ​​of the three differences show a monotonically non-decreasing relationship, the interval corresponding to the second reversal point P2 and the third reversal point P3 is determined to form an output charge-discharge cycle.

[0073] In a reversal point sequence, if the fluctuation range between the two middle reversal points is "small" enough relative to the fluctuation range on both sides, then the interval enclosed by the two middle reversal points can be regarded as a closed and complete charge-discharge cycle, which can be separated from the sequence.

[0074] Once an outputtable charge-discharge cycle is determined, the corresponding cycle parameters (e.g., cycle magnitude ΔDOD and average state of charge SOC) are determined. mean The cyclic amplitude ΔDOD can be calculated based on the values ​​of the second and third reversal points. For example, the cyclic amplitude ΔDOD... i = |P2 P3∣ / 2, Average State of Charge (SOC) mean,i= (P2+P3) / 2. Then, the second and third reversal points P2 and P3 corresponding to this loop are removed from the stack structure, while the first and fourth reversal points P1 and P4 are retained to continue participating in subsequent judgments. If the above inequality condition is not met, the system waits for the next reversal point to be added, and then takes the new top four points of the stack for another judgment.

[0075] This embodiment provides a simple and efficient online cycle judgment rule through the four-point method, which can identify closed charge-discharge cycles in real time without relying on the complete load history.

[0076] In some embodiments, dynamically adjusting the peak and valley extraction thresholds according to the battery's operating mode includes: Calculate the standard deviation of the state of charge within the sliding time window; When the standard deviation is less than the preset threshold, it is determined to be in floating mode, and the peak and valley extraction threshold is set to the first threshold. When the standard deviation is greater than or equal to the preset threshold, it is determined to be in frequency modulation mode, and the peak-valley extraction threshold is set to the second threshold. The first threshold is greater than the second threshold.

[0077] A sliding time window is a data window that slides forward along the time axis with a fixed window length, used to capture a sequence of states of charge over a continuous period of time for analysis. As an optional example, the window length can be set to 60 seconds.

[0078] For the current time point, obtain the sequence of charged states within the sliding time window, and calculate the standard deviation σ of this sequence. SOC The formula for calculating the standard deviation is:

[0079] Where, N w SOC is the number of sampling points within the sliding time window. i SO is the state of charge value of the i-th sampling point within the window. This represents the average state of charge within the window. The standard deviation characterizes the dispersion of the state of charge within that time window, i.e., the magnitude of the fluctuation.

[0080] The calculated standard deviation σ SOC With preset threshold σ th The comparison is performed. The preset threshold σ... th It is used to distinguish between float charging mode and frequency regulation mode, and can be pre-calibrated according to battery characteristics and operating conditions.

[0081] When σ SOC <σ thAt this point, it is determined that the battery is currently in float charging mode. In float charging mode, the battery maintains a high state of charge for an extended period, resulting in relatively small fluctuations in state of charge and more noticeable measurement noise. Therefore, the peak-valley extraction threshold δ is set as the first threshold δ. float .

[0082] When σ SOC ≥ σ th At this point, it is determined that the current state is in frequency modulation mode. In frequency modulation mode, the battery state of charge fluctuates frequently around the median value, but the actual amplitude of the fluctuation is relatively small. Therefore, the peak-valley extraction threshold δ is set as the second threshold δ. fr .

[0083] Wherein, the first threshold is greater than the second threshold, i.e., δ float >δ fr This setting method uses a larger threshold in float charging mode to filter out noise and a smaller threshold in frequency modulation mode to capture real fluctuations, thus achieving adaptive adjustment of the peak and valley extraction threshold according to the operating mode.

[0084] In practical applications, the standard deviation within the sliding time window can be calculated in real time at each sampling moment or at certain time intervals, and the peak and valley extraction threshold can be dynamically adjusted accordingly to achieve a rapid response to changes in operating conditions.

[0085] In some embodiments, the preset threshold ranges from 0.3% to 0.5%, the first threshold is 2.0%, and the second threshold is 0.5%.

[0086] In data center energy storage scenarios, a set of preferred parameter value ranges can be obtained by statistically analyzing and experimentally verifying the state of charge sequences under different operating conditions.

[0087] Preset threshold σ th The value range is 0.3% to 0.5%. When the standard deviation of the state of charge within the sliding time window is less than this range, it is determined to be in float charging mode; when the standard deviation is greater than or equal to this range, it is determined to be in frequency regulation mode. This value range is obtained based on the stable fluctuation range of the state of charge in the data center energy storage system under float charging mode and the fluctuation amplitude under frequency regulation mode.

[0088] First threshold δ float The peak-valley extraction threshold used in float charge mode is 2.0%. In float charge mode, there is usually measurement noise of about 1% to 2% in the state-of-charge sequence. Setting the first threshold to 2.0% can effectively filter out most of the noise and avoid misidentifying noise as valid cycles.

[0089] Second threshold δ frThe peak-valley extraction threshold used in frequency modulation mode is 0.5%. In frequency modulation mode, the fluctuation range of the state of charge during a real charge-discharge cycle is usually between 1% and 5%. Setting the second threshold to 0.5% can ensure that these real shallow fluctuations are captured, while filtering out residual noise with very small amplitude.

[0090] The above parameter values ​​are preferred values ​​for data center energy storage scenarios. In practical applications, they can be appropriately adjusted according to factors such as specific battery type, sampling frequency, and noise level. This embodiment is not limited to the specific values ​​mentioned above.

[0091] By adopting the above parameter values, it is possible to reasonably configure the peak and valley extraction threshold under high-frequency micro-circulation conditions in the data center, which not only ensures the effective identification of the real cycle, but also avoids the misidentification of noise, thereby improving the accuracy of subsequent damage calculation and health status prediction.

[0092] In some embodiments, the cyclic parameters include the cyclic amplitude and the average state of charge.

[0093] Accumulate multiple damage contributions to obtain the total damage, including: The nonlinear index corresponding to each charge-discharge cycle is determined based on the average state of charge for each charge-discharge cycle. The damage contribution of each charge-discharge cycle is summed according to the corresponding nonlinear exponent to obtain the total damage.

[0094] The cycle amplitude is used to characterize the depth of discharge of the cycle and can be calculated by the difference between the maximum and minimum values ​​of the cycle. For example, for a cycle formed by the state of charge decreasing from a high point to a low point and then rising back to a high point, its amplitude is the difference between the high point and the low point.

[0095] The average state of charge (SOC) is used to characterize the charge level at the time the cycle occurs. It can be calculated by averaging the maximum and minimum values ​​of the cycle, or by averaging the SOC values ​​of each sampling point over the duration of the cycle.

[0096] After calculating the damage contribution (i.e., the damage amount per cycle) for each charge-discharge cycle, the damage contributions from multiple cycles need to be accumulated to obtain the total damage. This embodiment uses a non-linear exponential accumulation method.

[0097] Specifically, let d be the damage contribution of the i-th charge-discharge cycle. i The average state of charge corresponding to this cycle is SOC. mean,i A nonlinear exponent λ is determined based on the average state of charge. i , where λ i It is a positive number. Then, the damage contribution of the cycle is transformed using this exponent, i.e., d is calculated. i λi The power of 1. Total damage D total It equals the sum of the damage contributions after all the transformations in the loops, as shown in the following formula:

[0098] Wherein, the nonlinear exponent λ i With average state of charge (SOC) mean,i The correspondence between them can be predefined. For example, a mapping function can be obtained by fitting experimental data, or it can be determined by looking up a table.

[0099] In some embodiments, when the average state of charge deviates from a preset reference value, the nonlinear exponent deviates from 1.

[0100] As a specific embodiment of this application, a preset reference value can be set, for example, 50%. This reference value corresponds to the midpoint of the battery's nominal state of charge. Near this reference value, the transformed damage contribution is equal to the original damage contribution, and the damage contribution can be approximately considered to accumulate linearly. Therefore, λ is set... i = 1.

[0101] When the average state of charge (SOC) exceeds the preset benchmark value (e.g., 80%, 90%, or higher), the battery remains in a high SOC for an extended period, leading to intensified side reactions in the electrode materials, accelerated SEI film growth, and significantly enhanced calendar aging. Simultaneously, during shallow charge-discharge cycles at a high SOC, calendar aging and cycle aging mutually reinforce each other, resulting in actual damage exceeding that caused by either cycle alone. To reflect this phenomenon, the nonlinear exponent λ is used... i Set to a positive number less than 1. Due to the damage contribution d i It is usually a positive number less than 1. For numbers less than 1, calculate its power less than 1, i.e., d. i ^ {λ i}, λ i A value less than 1 will increase the value, thereby amplifying the damage contribution of each cycle under high charge conditions.

[0102] For example, when the average state of charge is 80%, λ can be calibrated. i = 0.8; when the average state of charge is 95%, λ can be calibrated. i = 0.6. As the average state of charge increases, λ_i gradually decreases, and the amplification effect gradually strengthens.

[0103] When the average state of charge is lower than the preset benchmark value, λ can be set according to the actual aging pattern. i A value greater than 1 compresses the damage contribution to reflect the characteristic of a slower aging rate under low charge conditions.

[0104] By using the aforementioned nonlinear exponential accumulation method, the weight of the cycle damage in the total damage can be dynamically adjusted according to the charge level of different cycles, thereby more accurately simulating the aging law of the battery under actual working conditions. In particular, under the complex working conditions of alternating high charge state float charging and shallow micro-cycle in data centers, the nonlinear coupling effect of calendar aging and cycle aging can be effectively corrected.

[0105] This embodiment employs a nonlinear exponent with the average state of charge as the independent variable to exponentially transform the damage contribution of each cycle before summing them, rather than directly adding the damage contributions of each cycle linearly as in traditional methods. When the battery is in a high state of charge for a long period, the actual damage of subsequent microcycles will be more severe than when it occurs alone. By setting a nonlinear exponent less than 1, the damage contribution of high-charge cycles is amplified, which can reflect this phenomenon. Conversely, in a low-charge state, a nonlinear exponent greater than 1 is set to compress the damage contribution, making the accumulated total damage more consistent with the actual aging process of the battery. Therefore, this embodiment can improve the systematic error of the traditional linear accumulation method in the coupled scenario of high-charge float charging and cycle aging, and improve the accuracy of health status prediction.

[0106] In some embodiments, the cycling parameters include the cycling amplitude, the average state of charge, and the charge / discharge rate.

[0107] Based on the parameters of each cycle, the contribution of each charge-discharge cycle to battery aging damage is calculated, including: Based on the pre-calibrated multidimensional stress damage function, the damage contribution of each charge-discharge cycle to battery aging is calculated. The independent variables of the multidimensional stress damage function include the aforementioned cycle parameters.

[0108] Cyclic amplitude is used to characterize the depth of discharge in the cycle, and can be calculated from the difference between the maximum and minimum values ​​in the cycle; average state of charge is used to characterize the charge level when the cycle occurs, and can be calculated from the average of the maximum and minimum values ​​in the cycle; charge / discharge rate is used to characterize the current intensity during the cycle.

[0109] The multidimensional stress-damage function is obtained through pre-calibration via experiments. Specifically, cyclic aging tests can be performed on the battery under different stress conditions, covering different cycle amplitudes ΔDOD and average state of charge (SOC). mean and charge / discharge rate C rate The system combines measurements of battery capacity decay under various conditions to establish a mapping relationship between independent variables and damage amounts. This mapping relationship can be stored using multidimensional surface fitting, lookup tables, or other methods. In actual prediction, the loop parameters of the current cycle are input into this pre-calibrated damage function to obtain the corresponding single-cycle damage amount.

[0110] The above method can simultaneously consider the contributions of three stress dimensions—cycle amplitude, average state of charge, and charge / discharge rate—to battery aging. Compared to damage functions that only consider a single dimension, it can more comprehensively reflect the aging pattern of batteries under actual operating conditions.

[0111] In some embodiments, the charge / discharge rate is determined based on the ratio of the collected current value to the battery's rated capacity.

[0112] Specifically, for the i-th charge-discharge cycle, the current value I during that cycle is obtained. i (Usually take the absolute value), and obtain the battery's rated capacity C. nominal Then the charge / discharge rate C rate,i = |I i | / C nominal In practical applications, the current may not be a constant value during this cycle; the effective value, maximum value, or average value of the current can be taken as I. i .

[0113] In this embodiment, after identifying a charge-discharge cycle from the state-of-charge sequence, the amplitude ΔDOD of that cycle is extracted. i Average State of Charge (SOC) mean,i Simultaneously, the charge / discharge rate C is calculated based on the current values ​​collected during this cycle. rate,i Using these three factors as input, and substituting them into a pre-calibrated multidimensional stress-damage function, the single-cycle damage amount d for this cycle is calculated. i The damage amount is then used for subsequent nonlinear accumulation to finally obtain the total damage and battery health status.

[0114] This embodiment constructs a multidimensional stress-damage function by simultaneously considering stress factors in three dimensions: cycle amplitude, average state of charge, and charge / discharge rate. This function can more comprehensively reflect the aging behavior of batteries under actual operating conditions. Especially under the complex operating conditions of alternating high state of charge float charging and shallow micro-cycles in data centers, the average state of charge affects the coupling degree between calendar aging and cycle aging, while the charge / discharge rate affects the stress changes of electrode materials and the lithium deposition rate. Incorporating these factors into the damage calculation can significantly improve the accuracy of predictions, avoid the systematic biases caused by one-dimensional models, and effectively improve the prediction accuracy problem of traditional single-dimensional damage functions under complex operating conditions.

[0115] This application also provides a battery health status prediction device. Please refer to [link to relevant documentation]. Figure 2 This is a schematic diagram of the battery health status prediction device provided in the embodiments of this application, as shown below. Figure 2 As shown, the battery health status prediction device 200 includes a data acquisition module 201, a cyclic extraction module 202, a damage calculation module 203, and an accumulation output module 204.

[0116] The data acquisition module 201 is used to acquire the battery's state of charge sequence in real time.

[0117] Specifically, the data acquisition module 201 can be connected to the state-of-charge estimation unit of the battery management system to receive the real-time output state-of-charge sequence. Optionally, the data acquisition module 201 can also directly collect the battery's voltage, current, and temperature data, and calculate the state-of-charge sequence using an internal estimation algorithm. The output of the data acquisition module 201 is connected to the input of the loop extraction module 202.

[0118] The cyclic extraction module 202 is used to extract cyclic parameters of at least one charge-discharge cycle from the state-of-charge sequence.

[0119] The output of the cyclic extraction module 202 is connected to the input of the damage calculation module 203.

[0120] The damage calculation module 203 is used to calculate the damage contribution of each charge-discharge cycle to battery aging based on each cycle parameter.

[0121] The output of the damage calculation module 203 is connected to the input of the cumulative output module 204.

[0122] The cumulative output module 204 is used to accumulate multiple damage contributions to obtain total damage, and to determine the health status of the battery based on the total damage.

[0123] Each module in this embodiment can be implemented by hardware circuits (such as digital signal control modules, field-programmable gate arrays, and application-specific integrated circuits), or by the control module executing software instructions stored in the memory. Through the coordinated operation of the above modules, the device can realize the battery health state prediction method described in any embodiment of this application, and has the beneficial effects of real-time performance, high accuracy, and scene adaptability.

[0124] In some embodiments, such as Figure 3 As shown, the cyclic extraction module 202 includes an extreme point extraction unit 2021 and a counting processing unit 2022.

[0125] The extreme point extraction unit 2021 is used to extract local maxima and local minima from the charged state sequence based on the dynamically adjusted peak and valley extraction threshold, forming a reversal point sequence.

[0126] The counting processing unit 2022 is used to perform online counting processing on the reversal point sequence, identify the charge and discharge cycles, and output the cycle parameters of each cycle (such as cycle amplitude and average state of charge).

[0127] As an optional embodiment, the counting processing unit 2022 can employ an online counting algorithm based on a stack structure. Specifically, the counting processing unit 2022 sequentially adds reversal points to the stack structure. Each time a new reversal point is added to the stack structure, it determines whether a charge / discharge cycle exists in the stack structure. When a charge / discharge cycle exists, the cycle parameters of the charge / discharge cycle are output, and the corresponding reversal point is removed from the stack structure.

[0128] As an optional embodiment, the counting processing unit 2022 uses the four-point method to determine whether there is a charge-discharge cycle in the stack structure: obtain four consecutive reversal points at the top of the stack, calculate the absolute value of the difference between two adjacent points, and when the three absolute values ​​of the difference satisfy the monotonically non-decreasing relationship, determine that the interval corresponding to the two middle reversal points forms a charge-discharge cycle.

[0129] In some embodiments, such as Figure 4 As shown, the extreme point extraction unit 2021 includes an adaptive threshold adjustment subunit 20210. The adaptive threshold adjustment subunit 20210 determines the operating mode by calculating the standard deviation of the state of charge within the sliding time window: when the standard deviation is less than a preset threshold, it is determined to be a float charging mode, and the peak-valley extraction threshold is set to the first threshold; when the standard deviation is greater than or equal to the preset threshold, it is determined to be a frequency modulation mode, and the peak-valley extraction threshold is set to the second threshold; wherein, the first threshold is greater than the second threshold.

[0130] In some embodiments, the cumulative output module 204 determines the corresponding nonlinear index based on the average state of charge for each charge-discharge cycle. For each cycle, the damage contribution of that cycle is subjected to a power operation (or a higher-level nonlinear transformation) with the nonlinear index as the power, and then the transformed damage contributions of each cycle are accumulated to obtain the total damage. When the average state of charge is higher than a preset reference value (e.g., 50%), the nonlinear index is less than 1, thereby amplifying the damage contribution under high charge state.

[0131] In some embodiments, the damage calculation module 203 internally stores a multidimensional stress damage function, whose independent variables include cycle amplitude, average state of charge, and charge / discharge rate. The damage calculation module 203 receives cycle parameters output by the cycle extraction module 202. If the cycle parameters do not include the charge / discharge rate, the damage calculation module 203 can also obtain the corresponding current value from the data acquisition module 201 or externally, and calculate the charge / discharge rate based on the ratio of the current value to the battery's rated capacity. Then, the cycle parameters are substituted into the pre-calibrated multidimensional stress damage function to calculate the single-cycle damage amount for each cycle.

[0132] This application also provides a battery management system; please refer to [link / reference]. Figure 5 This is a schematic diagram of the battery management system provided in an embodiment of this application. Figure 5As shown, the battery management system 300 includes a sensor assembly 301, a state of charge estimation module 302, and a control module 303.

[0133] Sensor assembly 301 is used to acquire voltage, current and temperature data of the battery.

[0134] Sensor assembly 301 may include a voltage sensor, a current sensor, and a temperature sensor. The voltage sensor is connected in parallel across the positive and negative terminals of the battery to detect the battery's terminal voltage in real time; the current sensor is connected in series in the battery's charging and discharging circuit to detect the battery's charging and discharging current in real time; the temperature sensor is attached to the battery surface or disposed inside the battery module to detect the battery's temperature in real time. Sensor assembly 301 outputs the collected voltage, current, and temperature data in the form of electrical signals.

[0135] The state of charge estimation module 302 is used to output a state of charge sequence in real time based on voltage, current and temperature data.

[0136] The state of charge (SOC) estimation module 302 is connected to the sensor assembly 301. The SOC estimation module 302 can be implemented using known estimation algorithms, such as the ampere-hour integration method, the extended Kalman filter method, the open-circuit voltage correction method, or a combination of these methods. The SOC estimation module 302 receives voltage, current, and temperature data collected by the sensor assembly 301, calculates the current SOC of the battery in real time at a predetermined sampling frequency (e.g., not less than 1 Hz), and outputs the SOC sequence in chronological order.

[0137] The control module 303 is connected to the state of charge estimation module 302 and is used to receive the state of charge sequence and output the battery health status. The control module 303 executes the battery health status prediction method as described in any of the foregoing embodiments of this application; the specific principles have been described in detail in the foregoing embodiments and will not be repeated here. The control module 303 receives the state of charge sequence output by the state of charge estimation module 302, and through online extraction of charge-discharge cycles, calculation of damage contribution, and accumulation of total damage, finally outputs the battery health status (e.g., remaining capacity percentage) and / or remaining service life.

[0138] The battery management system 300 provided in this embodiment integrates a sensor component 301, a state of charge estimation module 302, and a control module 303. It can monitor the battery's operating data in real time, predict the battery's health status online, and provide decision-making basis for the operation and maintenance management of data center energy storage systems.

[0139] This application also provides a battery module, please refer to the embodiments therein. Figure 6 This is a schematic diagram of the battery module provided in an embodiment of this application. Figure 6As shown, the battery module 400 includes one or more batteries 401, and a battery management system 300 connected to the one or more batteries 401.

[0140] Battery 401 is the basic energy storage unit of battery module 400. For data center energy storage applications, battery 401 can be a lithium iron phosphate battery, a ternary lithium battery, or other types of lithium-ion batteries. Battery module 400 may include multiple individual batteries 401, which can be connected in series, parallel, or a combination of series and parallel connections to meet voltage and capacity design requirements.

[0141] The sensor components 301 of the battery management system 300 are connected to each battery 401 or a series / parallel group of batteries 401 to collect voltage, current, and temperature data. The state-of-charge estimation module 302 and the control module 303 of the battery management system 300 process this data and output the health status of the battery module 400. The battery management system 300 may also include auxiliary functional modules such as equalization circuits and protection circuits to achieve equalization management and overcharge / over-discharge protection for individual battery cells.

[0142] The battery module 400 provided in this embodiment integrates the battery 401 and the battery management system 300 into a single unit, which can be directly applied to energy storage systems or used as a backup power module, making it easy to install, maintain and replace.

[0143] This application also provides an energy storage device; please refer to [link / reference]. Figure 7 This is a schematic diagram of the energy storage device provided in an embodiment of this application. Figure 7 As shown, the energy storage device 500 includes a battery array 501 and a battery management system 300 connected to the battery array 501.

[0144] The battery array 501 is composed of multiple battery modules 400 or batteries 401 connected in series and / or parallel to meet the voltage level and energy storage capacity requirements of data center energy storage systems. The positive and negative terminals of the battery array 501 are connected to the external power grid or load through a power conversion system to realize the storage and release of electrical energy.

[0145] The battery management system 300 is connected to the battery array 501. The battery management system 300 collects voltage, current, and temperature data of each battery module 400 or battery 401 in the battery array 501 through its sensor components 301, obtains the state of charge sequence through the state of charge estimation module 302, and then predicts the health status of the entire battery array 501 online through the control module 303. The battery management system 300 can also communicate with the upper-level controller (e.g., energy management system) of the energy storage device 500 to report health status information for the purpose of formulating charging and discharging strategies and operation and maintenance plans.

[0146] The energy storage device 500 can be applied to various scenarios. In one implementation, the energy storage device 500 is a stationary energy storage system, installed in power plants, substations, or data centers for grid frequency regulation, backup power, or peak shaving and valley filling. In another implementation, the energy storage device 500 is a mobile energy storage device, installed within a vehicle, to provide power or auxiliary power to the vehicle. When the energy storage device 500 is installed in a vehicle, the battery management system 300 sends health status information to the vehicle controller or dashboard via the vehicle's bus network to display the remaining battery life or indicate maintenance needs.

[0147] The energy storage device 500 provided in this embodiment adopts a battery management system 300 with real-time health status prediction function. The energy storage device 500 can monitor the health status of the energy storage battery in real time, detect abnormal battery cells in a timely manner, and avoid power outages or safety accidents caused by battery aging.

[0148] This application also provides an electronic device, which includes a memory and a processor.

[0149] The memory is used to store executable instructions. The memory may include volatile memory and / or non-volatile memory, such as random access memory, read-only memory, flash memory, hard disk, etc. The executable instructions stored in the memory include those for implementing the battery health state prediction method described in any embodiment of this application.

[0150] The processor can be a central processing unit, microcontroller, digital signal processor, field-programmable gate array, application-specific integrated circuit, or other hardware with data processing capabilities.

[0151] When the processor executes executable instructions in memory, it performs the following operations: Cyclic parameters for charge-discharge cycles are extracted based on the real-time acquired state-of-charge sequence of the battery. The damage contribution is calculated based on the parameters of each cycle. The cumulative damage contribution yields the total damage. The health status of the battery is determined based on the total damage.

[0152] The specific steps have been described in detail in the method embodiments and will not be repeated here.

[0153] The electronic device provided in this embodiment can be used as a standalone battery monitoring device, or it can be integrated into a battery management system, energy storage converter, or upper-level monitoring system to achieve real-time prediction of battery health status.

[0154] This application also provides a computer-readable storage medium storing executable instructions that, when executed by a processor, implement the battery health state prediction method described in any embodiment of this application.

[0155] Computer-readable storage media can be any tangible medium capable of storing program instructions, such as disks, optical discs, magneto-optical discs, solid-state drives, read-only memory, random access memory, flash memory, etc. This storage medium can be built into electronic devices or sold or distributed as a standalone product.

[0156] When the executable instructions in the storage medium are read and executed by the processor, the processor performs the following steps: Obtain the battery's state of charge sequence; Extract the cycle parameters of at least one charge-discharge cycle from the state-of-charge sequence; The contribution of each charge-discharge cycle to battery aging is calculated based on the parameters of each cycle. Accumulate multiple damage contributions to obtain the total damage; The health status of the battery is determined based on the total damage.

[0157] The specific steps have been described in detail in the method embodiments and will not be repeated here.

[0158] In the embodiments of this application, the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0159] In the embodiments of this application, the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0160] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0161] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. 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, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0162] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a targeted manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0163] 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.

[0164] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0165] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0166] Computer-readable media include both permanent and non-permanent, removable and non-removable media, which can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient media, such as modulated communication signals and carrier waves.

[0167] The embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for predicting battery health status, characterized in that, include: Based on the real-time acquired state of charge sequence of the battery, at least one charge-discharge cycle parameter is extracted from the state of charge sequence. Based on each cycle parameter, calculate the damage contribution of each charge-discharge cycle to battery aging. Accumulate multiple damage contributions to obtain the total damage; The health status of the battery is determined based on the total damage.

2. The method according to claim 1, characterized in that, The step of extracting cycle parameters for at least one charge-discharge cycle from the state-of-charge sequence includes: The peak and valley extraction thresholds are dynamically adjusted according to the battery's operating mode. Based on the adjusted peak-valley extraction threshold, a sequence of reversal points arranged in chronological order is obtained from the charged state sequence, wherein the reversal points are local maxima or local minima in the charged state sequence. The reversal point sequence is counted online to obtain the loop parameters.

3. The method according to claim 2, characterized in that, The online counting process for the inversion point sequence to obtain the loop parameters includes: The reversal points are added to the stack structure in sequence, and each time a new reversal point is added to the stack structure, it is determined whether there is a charge-discharge cycle in the stack structure. When a charge / discharge cycle exists, the cycle parameters of the charge / discharge cycle are output, and the corresponding inversion point is removed from the stack structure.

4. The method according to claim 3, characterized in that, The step of sequentially adding the reversal points to the stack structure, and determining whether a charging / discharging cycle exists in the stack structure each time a new reversal point is added to the stack structure, includes: Obtain four consecutive reversal points at the top of the stack structure, wherein the four reversal points include a first reversal point, a second reversal point, a third reversal point, and a fourth reversal point; Calculate the absolute value of the first difference between the first reversal point and the second reversal point, calculate the absolute value of the second difference between the second reversal point and the third reversal point, and calculate the absolute value of the third difference between the third reversal point and the fourth reversal point; When the absolute value of the first difference is not greater than the absolute value of the second difference, and the absolute value of the second difference is not greater than the absolute value of the third difference, the interval corresponding to the second reversal point and the third reversal point is determined to form the charge-discharge cycle.

5. The method according to claim 2 or 3, characterized in that, The step of dynamically adjusting the peak and valley extraction threshold according to the battery's operating mode includes: Calculate the standard deviation of the state of charge within the sliding time window; When the standard deviation is less than the preset threshold, it is determined to be in floating mode, and the peak-valley extraction threshold is set to the first threshold. When the standard deviation is greater than or equal to the preset threshold, it is determined to be a frequency modulation mode, and the peak-valley extraction threshold is set to the second threshold. Wherein, the first threshold is greater than the second threshold.

6. The method according to any one of claims 1 to 5, characterized in that, The cyclic parameters include cyclic amplitude and average state of charge; The accumulation of multiple damage contributions to obtain total damage includes: Based on the average state of charge of each charge-discharge cycle, determine the nonlinear index corresponding to the charge-discharge cycle; The damage contribution of each charge-discharge cycle is accumulated according to the corresponding nonlinear exponent to obtain the total damage.

7. The method according to claim 6, characterized in that, When the average state of charge deviates from the preset reference value, the nonlinear exponent deviates from 1.

8. The method according to any one of claims 1 to 7, characterized in that, The cycle parameters include cycle amplitude, average state of charge, and charge / discharge rate, wherein the charge / discharge rate is determined based on the ratio of the collected current value to the battery's rated capacity. The step of calculating the damage contribution of each charge-discharge cycle to battery aging based on each of the cycle parameters includes: Based on the pre-calibrated multidimensional stress damage function, the damage contribution of each charge-discharge cycle to battery aging is calculated, wherein the independent variable of the multidimensional stress damage function includes the cycle parameter.

9. A battery management system, characterized in that, include: Sensor components are used to collect voltage, current, and temperature data from energy storage batteries; The state of charge estimation module is used to output a state of charge sequence in real time based on the voltage, current and temperature data; A control module, connected to the state of charge estimation module, is used to execute the battery health state prediction method according to any one of claims 1 to 8.

10. A battery module, characterized in that, include: One or more batteries; The battery management system as described in claim 9, which is connected to the one or more batteries.

11. An energy storage device, characterized in that, include: Battery array; The battery management system as described in claim 9 is connected to the battery array.

12. An electronic device, characterized in that, include: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the battery health state prediction method according to any one of claims 1 to 8.

13. A computer-readable storage medium, characterized in that, It stores executable instructions for inducing the processor to execute the battery health state prediction method according to any one of claims 1 to 8.

Images