Energy storage battery aging detection method and device based on cross-over difference algorithm

Abstract

The application provides an energy storage battery aging detection method and device based on a cross-over difference algorithm, and relates to the technical field of energy storage battery aging detection, and the specific steps comprise: dividing a sliding window into multiple cross-over windows, calculating extreme values, mean values and window residual sequences of feature parameters in each window, screening strong abnormal windows, constructing an abnormal window spread time trajectory, calculating a spread emergency coefficient, marking the difference correlation between feature parameters of each strong abnormal window based on a second preset rule, performing data processing on the strong correlation proportion and the spread emergency coefficient, generating an evaluation coefficient, and judging the aging degree of the energy storage battery. The multi-dimensional and multi-period difference correlation analysis lays a foundation for accurate screening of abnormal windows, significantly improves the accuracy of abnormal identification, and effectively improves the judgment ability of the abnormal emergency degree through dynamic analysis and scientific evaluation of the abnormal time sequence evolution and diffusion process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage battery aging detection, in particular to an energy storage battery aging detection method and device based on cross-over differential algorithm. BACKGROUND

[0002] As an important part of the new energy field, energy storage batteries are widely used in power storage, transportation and other fields. With the long-term use of energy storage batteries, their performance inevitably degrades and ages, leading to voltage and capacity decline, internal resistance increase, and thus affecting the safety and reliability of the energy storage system. Therefore, accurately and timely detecting the aging state of the energy storage battery is of great significance to ensure the stable operation of the energy storage system.

[0003] In the prior art, a kind of energy storage battery aging detection method based on cross-over differential algorithm with publication number CN116106775A, steps include: real-time acquisition of the electrical quantity data of energy storage battery, construct original data signal;The original data is processed according to cross-over differential algorithm, and a fault detection signal is constructed;The extreme value search is carried out to the obtained fault detection signal, and a fault judgment signal is constructed;Detect whether the fault judgment signal is greater than the set detection threshold value. If not, repeat step 1 to continue real-time acquisition of the electrical quantity data of energy storage battery, if yes, the energy storage battery is aging, and an alarm signal is sent. The method can quickly and reliably detect the aging failure of the energy storage battery, avoid the consistency problem of the energy storage battery caused by the aging of the energy storage battery, reduce the risk of safety caused by the aging of the energy storage battery, and realize the quick and reliable detection of the aging problem of single energy storage battery.

[0004] However, there are still the following deficiencies. As can be seen from the above statement, the existing method does not utilize the differential correlation between the characteristic parameters of the energy storage battery, resulting in lack of differentiation between strong correlation and weak correlation, difficulty in accurately identifying multiple parameter collaborative anomalies, affecting the screening effect of the abnormal window, making the abnormal judgment lack of accuracy, further, due to the inaccuracy of the abnormal window screening, the spread time trajectory of the abnormal window cannot be constructed, lacking dynamic tracing and analysis of the abnormal time evolution and diffusion process, unable to effectively evaluate the emergency degree and development trend of the abnormality, due to the lack of analysis of the time spread trajectory, limiting the comprehensive evaluation and timely warning ability of the aging risk of the energy storage battery.

[0005] The above information disclosed in the background section is only used to strengthen the understanding of the background of the present disclosure, therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] The application aims to provide a storage battery aging detection method and device based on a cross-over difference algorithm to solve the problems in the background art.

[0007] To achieve the above-mentioned purpose, the application provides the following technical solutions.

[0008] A storage battery aging detection method based on a cross-over difference algorithm, the specific steps comprising:

[0009] Collecting the characteristic parameters of the storage battery every hour within the past 24 hours to form time series data, and dividing the time series data into multiple cross-over windows using a sliding window method;

[0010] Obtaining the extreme value, mean value and standard deviation of the characteristic parameters in each window, and calculating the difference value of the characteristic parameters in each group of adjacent windows to obtain a residual sequence;

[0011] Comparing the extreme value, mean value and standard deviation of the characteristic parameters in each window with their normal interval range respectively, comparing the absolute value of the difference value of the characteristic parameters with a preset threshold value at the same time, screening out strong abnormal windows marked as data abnormality and difference abnormality at the same time, determining the initial abnormal window, tracing the spread window at the same time, constructing a spread time trajectory, and obtaining the spread emergency coefficient of each strong abnormal window based on the spread time trajectory and a first preset rule;

[0012] Based on a second preset rule, marking the difference correlation between the characteristic parameters in each strong abnormal window, including strong correlation and weak correlation, and counting the strong correlation proportion of the characteristic parameters in each strong abnormal window;

[0013] Data processing the strong correlation proportion and the spread emergency coefficient of the characteristic parameters in each strong abnormal window to generate an evaluation coefficient, comparing the evaluation coefficient with a preset threshold value, and judging the aging degree of the storage battery.

[0014] Further, a group of characteristic parameters is collected every hour, including voltage, capacity and internal resistance, a total of 24 groups of characteristic parameters, a group of characteristic parameters is represented in the form of a 3-row 1-column matrix, each row corresponds to a characteristic parameter, and the time series data is represented in the form of a 3-row 24-column matrix, each column represents a group of characteristic parameters, and the characteristic parameters in each column in the matrix are arranged in the order of collection time in sequence;

[0015] Using an overlapping window method with a sliding step of 3, the time series data is divided into 7 windows, and the time series data in each window is a 3-row 6-column matrix, respectively denoted as , is the time series data of the i-th window, is the index of the window, . ​

[0016] Furthermore, the residual sequence is obtained, and the specific logic is as follows:

[0017] For the seven windows obtained from the partitioning, the difference in feature parameters between the next and previous windows in adjacent windows is calculated to obtain the residual sequence represented in matrix form, based on the following formula:

[0018] ;

[0019] in, For the first The residual sequence of the nth window, which is the nth window. The window and the first The time series data difference of each window For the first The time series data for each window, and the residual sequence is also a 3x6 matrix.

[0020] Furthermore, strong anomaly windows that are simultaneously marked as data anomalies and difference anomalies are filtered out. The specific logic is as follows:

[0021] For each window, if the maximum, minimum, mean, and standard deviation of voltage, capacitance, and internal resistance in the time series data within the window all exceed the normal range, then the window is marked as an abnormal data window; otherwise, the window is marked as a normal data window.

[0022] If the absolute values ​​of the voltage, capacitance, and internal resistance differences in the residual sequence corresponding to the window are all greater than the preset threshold, then the window is marked as an abnormal differential window; otherwise, the window is marked as a normal differential window.

[0023] Based on the labeling results of each window, windows that are simultaneously labeled as data anomalies and difference anomalies are selected as strong anomaly windows.

[0024] The strong anomaly window with the smallest index is used as the initial anomaly window. At the same time, the propagation window is traced to construct the propagation time trajectory. The specific logic is as follows:

[0025] Sort all strong anomaly windows in ascending order of index value, select the smallest strong anomaly window as the initial anomaly window, and use the other strong anomaly windows as propagation windows, forming a propagation time trajectory from the initial anomaly window, in ascending order of index value, to the last propagation window.

[0026] The time intervals corresponding to each window are associated in the spread time trajectory, and the number of interval windows between adjacent strong anomaly windows is marked. The number of interval windows refers to the number of windows between two adjacent strong anomaly windows that are not marked as strong anomaly windows.

[0027] Furthermore, based on the propagation time trajectory and the first preset rule, the propagation urgency coefficient of each strong anomaly window is obtained. The specific logic is as follows:

[0028] Set the trajectory base coefficient, window progression coefficient, and interval correction coefficient to characterize the spread urgency coefficient of each strong anomaly window;

[0029] Among them, the trajectory baseline coefficient is used to reflect the overall risk level under the same spread time trajectory, and the trajectory baseline coefficient is set to be the same for all strong anomaly windows;

[0030] The window progression coefficient is used to characterize the degree of progression of each strong anomaly window in the propagation time trajectory. The assignment logic for the window progression coefficient of each strong anomaly window is as follows:

[0031] Based on the index of the strong anomaly window, its window progression coefficient is assigned a value. The smaller the index of the strong anomaly window, the smaller the window progression coefficient, and the larger the index of the strong anomaly window, the larger the window progression coefficient.

[0032] Among them, the interval correction coefficient reflects the influence of the number of interval windows between the current adjacent strong anomaly window and the previous strong anomaly window on the anomaly propagation speed.

[0033] The more interval windows there are, the slower the spread and the smaller the interval correction coefficient; conversely, the fewer the interval windows there are, the faster the spread and the larger the interval correction coefficient.

[0034] Based on the above three coefficients, the propagation urgency coefficient of each strong anomaly window is calculated using the following formula:

[0035] ;

[0036] in, For the first The urgency coefficient of the spread of a strong abnormal window. For the first The trajectory baseline coefficients of a strong anomaly window. For the first The window progression coefficient of a strongly anomalous window. For the first The interval correction coefficient for each strong anomaly window. The index for the strong exception window. , The number of strong exception windows;

[0037] In the formula, , , These are the weighting coefficients for the trajectory base coefficient, window progression coefficient, and interval correction coefficient, respectively. , , The specific value is determined based on the analytic hierarchy process (AHP).

[0038] Furthermore, based on the second preset rule, the differential correlation between the feature parameters of each strong anomaly window is labeled, and the specific logic is as follows:

[0039] For any strong anomaly window, if the voltage difference and capacity difference decrease continuously in its residual sequence, then the voltage-capacity relationship is marked as strong; otherwise, the voltage-capacity relationship is marked as weak.

[0040] In the residual sequence, if the voltage difference continuously decreases and the internal resistance difference continuously increases, then the voltage-internal resistance is marked as a strong correlation; otherwise, the voltage-internal resistance is marked as a weak correlation.

[0041] In the residual sequence, if the capacity difference continuously decreases and the internal resistance difference continuously increases, then the capacity-internal resistance is marked as a strong correlation; otherwise, the capacity-internal resistance is marked as a weak correlation.

[0042] Among them, continuous decrease or continuous increase refers to the characteristic parameter difference in the residual sequence showing a monotonically decreasing or increasing trend along the direction of increasing column number.

[0043] Furthermore, the strong correlation ratio and spread urgency coefficient of each strong anomaly window feature parameter are processed to generate evaluation coefficients, based on the following formula:

[0044] ;

[0045] in, The average evaluation coefficient for all strong anomaly windows. For the first The proportion of strong correlations among strong abnormal windows.

[0046] Furthermore, the evaluation coefficient is compared with a preset threshold to determine the degree of aging of the energy storage battery. The specific logic is as follows:

[0047] when At that time, the energy storage battery was determined to be slightly aged;

[0048] when At that time, the energy storage battery was determined to be moderately aged;

[0049] when At that time, the energy storage battery was determined to be severely aged;

[0050] in, To define the threshold for distinguishing between mild and moderate aging, The threshold for distinguishing between moderate and severe aging.

[0051] To achieve the above objectives, the present invention also provides the following technical solution:

[0052] The application discloses an energy storage battery aging detection device based on a cross-over differential algorithm.

[0053] A window division module is configured to collect feature parameters of the energy storage battery at each time point in the past 24 hours, form time series data, and divide the time series data into a plurality of cross-over windows by using a sliding window method.

[0054] A data processing module is configured to obtain extreme values, mean values and standard deviations of the feature parameters in each window, and calculate feature parameter difference values in each group of adjacent windows to obtain a residual sequence.

[0055] A trajectory construction module is configured to compare the extreme values, mean values and standard deviations of the feature parameters in each window with normal interval ranges, compare absolute values of the feature parameter difference values with a preset threshold value, screen out strong abnormal windows which are marked as data abnormality and differential abnormality at the same time, determine initial abnormal windows, trace spread windows, construct a spread time trajectory, and obtain a spread emergency coefficient of each strong abnormal window based on the spread time trajectory and a first preset rule.

[0056] A labeling module is configured to label differential correlation between the feature parameters in each strong abnormal window based on a second preset rule, including strong correlation and weak correlation, and count a strong correlation proportion of the feature parameters in each strong abnormal window.

[0057] A judgment module is configured to perform data processing on the strong correlation proportion and the spread emergency coefficient of the feature parameters in each strong abnormal window, generate an evaluation coefficient, compare the evaluation coefficient with a preset threshold value, and judge an aging degree of the energy storage battery.

[0058] Compared with the prior art, the application has the following beneficial effects:

[0059] The application collects feature parameters of the energy storage battery at each time point in the past 24 hours, forms time series data, divides the time series data into a plurality of cross-over windows by using a sliding window method, calculates extreme values, mean values and standard deviations of the feature parameters in the windows and a residual sequence, lays a foundation for accurate screening of abnormal windows through multi-dimensional and multi-period differential correlation analysis, and significantly improves the accuracy of abnormal identification.

[0060] Based on the accurately screened abnormal windows, initial abnormal windows and spread windows are further determined, a spread time trajectory of the abnormal windows is constructed, a time interval between adjacent strong abnormal windows is calculated, a spread emergency coefficient of each strong abnormal window is obtained based on the spread time trajectory and a preset rule, and the judgment ability for the emergency degree of the abnormality is effectively improved through dynamic analysis and scientific evaluation of abnormal time sequence evolution and diffusion processes.

[0061] Based on the second preset rule, the difference correlation between each strong abnormal window feature parameter is marked, the strong correlation proportion of each strong abnormal window feature parameter is counted, and the strong correlation proportion and the spread emergency coefficient are processed to generate an evaluation coefficient, which is used to judge the aging degree of the energy storage battery, can effectively evaluate the aging degree and development trend of the battery, and improve the evaluation ability of the energy storage battery aging by combining the analysis of the time sequence spread trajectory. BRIEF DESCRIPTION OF DRAWINGS

[0062] Figure 1 It is a whole method flowchart of the application;

[0063] Figure 2 It is a fitting diagram of the window progressive coefficient and the spread emergency coefficient of the application;

[0064] Figure 3 It is a fitting diagram of the interval correction coefficient and the spread emergency coefficient of the application;

[0065] Figure 4 It is a fitting diagram of the strong correlation proportion of the feature parameter and the evaluation coefficient of the application;

[0066] Figure 5 It is a fitting diagram of the spread emergency coefficient and the evaluation coefficient of the application;

[0067] Figure 6 It is a module composition block diagram of the application. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical scheme and advantages of the application more clear and obvious, the application will be further described in detail below combined with specific embodiments.

[0069] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the application should be understood as the usual meaning understood by those skilled in the art to which the application belongs. The "first", "second" and similar words used in the application do not represent any order, quantity or importance, but are only used to distinguish different components. "Include" or "contain" and similar words mean that the elements or objects before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connected" or "connected" and similar words are not limited to physical or mechanical connection, but can include electrical connection, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to represent relative positional relationship, when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0070] Embodiment:

[0071] Please refer to Figures 1-5The application provides a technical scheme:

[0072] A method for detecting aging of an energy storage battery based on a cross-over difference algorithm, and the specific steps include:

[0073] S1. Collecting characteristic parameters of the energy storage battery at each hour in the past 24 hours to form time series data, and dividing the time series data into a plurality of cross-over windows by using a sliding window method, and collecting a group of characteristic parameters every hour, which include voltage, capacity and internal resistance.

[0074] On the basis of the above embodiment, the collection method of the voltage, the capacity and the internal resistance is as follows:

[0075] An analog-to-digital converter (ADC) is connected to the positive and negative ports of the energy storage battery to collect the battery terminal voltage signal in real time, and the sampling frequency is set to once per hour, and 24 voltage data points are obtained in 24 hours.

[0076] The capacity is directly obtained by the battery management system of the battery.

[0077] The internal resistance of the battery reflects the internal electrochemical characteristics of the battery, and usually includes direct current resistance and alternating current resistance, and here the direct current resistance is calculated.

[0078] The pulse discharge method is used to apply a short-time pulse load at the battery terminal, measure the voltage change before and after the pulse, calculate the internal resistance, and the formula is as follows:

[0079] ;

[0080] Wherein, is the internal resistance of the energy storage battery, is the voltage before the pulse is applied, is the voltage when the pulse is applied, is the applied pulse current.

[0081] On the basis of the above embodiment, 24 groups of characteristic parameters are collected, and one group of characteristic parameters is represented in the form of a 3-row 1-column matrix:

[0082] ;

[0083] Wherein, is a group of characteristic parameters at the first time, are the voltage, the capacity and the internal resistance at the first time, respectively;

[0084] Each row corresponds to a characteristic parameter, and the time series data is represented in the form of a 3-row 24-column matrix, each column represents a group of characteristic parameters, and the characteristic parameters in each column of the matrix are arranged in the order of collection time;

[0085] Using an overlapping window method with a sliding step size of 3, the time series data is divided into 7 windows. The time series data in each window is a 3x6 matrix, denoted as follows: , For the first Time series data for each window, For the window index, .

[0086] S2. Obtain the extreme values, mean values, and standard deviations of the feature parameters within each window, and calculate the difference in feature parameters between each group of adjacent windows to obtain the residual sequence;

[0087] Based on the above embodiments, the residual sequence is obtained, and the specific logic is as follows:

[0088] For the seven windows obtained from the partitioning, the difference in feature parameters between the next and previous windows in adjacent windows is calculated to obtain the residual sequence represented in matrix form, based on the following formula:

[0089] ;

[0090] in, For the first The residual sequence of the nth window, which is the nth window. The window and the first The time series data difference of each window For the first The time series data for each window, and the residual sequence is also a 3x6 matrix.

[0091] S3. Compare the extreme values, mean values, and standard deviations of the feature parameters in each window with their normal ranges, and compare the absolute values ​​of the differences in the feature parameters with the preset thresholds. Select strong anomaly windows that are simultaneously marked as data anomalies and difference anomalies, determine the initial anomaly windows, trace the spread windows, construct the spread time trajectory, and obtain the spread urgency coefficient of each strong anomaly window based on the spread time trajectory and the first preset rule.

[0092] Based on the above embodiments, the normal range for the maximum, minimum, mean, and standard deviation of voltage, capacitance, and internal resistance is set as follows:

[0093] Based on a large amount of characteristic parameter data collected from healthy energy storage batteries during their normal operating cycles, statistical analysis was performed using the box plot method, specifically including:

[0094] The maximum and minimum values ​​are taken from the 95th and 5th percentiles, respectively, as the upper and lower limits of the normal range;

[0095] The mean is taken from the 50th percentile, representing the central trend of the normal range;

[0096] The standard deviation is calculated based on data in the interval from the 5th percentile to the 95th percentile, reflecting the fluctuation range of data in the normal interval.

[0097] The statistical data of the normal interval range is derived from a large number of new energy vehicle power batteries collected in actual operating environment, ensuring that the normal interval can accurately reflect the performance characteristics of new energy vehicle batteries.

[0098] On the basis of the above, the maximum value, minimum value, mean value, and standard deviation of each characteristic parameter (voltage, capacity, and internal resistance) have an independent normal interval range.

[0099] For voltage:

[0100] The maximum normal interval range is 3.9V-4.2V; the minimum normal interval range is 3.0V-3.3V; the mean normal interval range is 3.5V-3.8V; and the standard deviation normal interval range is 0-0.05V.

[0101] For capacity: the maximum normal interval range is 90Ah-100Ah; the minimum normal interval range is 20Ah-30Ah; the mean normal interval range is 60Ah-80Ah; and the standard deviation normal interval range is 0-5Ah.

[0102] For internal resistance: the maximum normal interval range is 8mΩ-12mΩ; the minimum normal interval range is 2mΩ-4mΩ; the mean normal interval range is 4mΩ-7mΩ; and the standard deviation normal interval range is 0-0.5mΩ.

[0103] On the basis of the above embodiment, strong abnormal windows that are simultaneously marked as data abnormal and differential abnormal are screened out, and the specific logic is as follows:

[0104] For each window, if the maximum value, minimum value, mean value, and standard deviation of voltage, capacity, and internal resistance in the time series data within the window all exceed the normal interval range, the window is marked as a data abnormal window, otherwise, the window is marked as a data normal window.

[0105] If the absolute value of the voltage, capacity, and internal resistance difference value in the residual sequence corresponding to the window is greater than a preset threshold, the window is marked as a differential abnormal window, otherwise, the window is marked as a differential normal window.

[0106] According to the marking result of each window, the window that is simultaneously marked as data abnormal and differential abnormal is screened out as a strong abnormal window.

[0107] The preset threshold is set as follows:

[0108] By collecting the window difference data (such as the change amount of voltage, capacity, and internal resistance) of a large number of healthy batteries in a normal working cycle;

[0109] The distribution of these differences is statistically normal or approximately normal;

[0110] The 95th percentile is taken as the preset threshold using the box plot method or quantile method, and the window is marked as a difference abnormal window if it exceeds the threshold.

[0111] On the basis of the above embodiments, the window marked as both data abnormal and difference abnormal is selected as a strong abnormal window according to the marking result of each window;

[0112] The strong abnormal window with the smallest index is taken as the initial abnormal window, and the spread window is traced back to construct a spread time track, and the specific logic is as follows:

[0113] All strong abnormal windows are sorted in ascending order of index value, and the smallest strong abnormal window is selected as the initial abnormal window, and the other strong abnormal windows are taken as the spread window, forming a spread time track from the initial abnormal window to the last spread window in ascending order of index value;

[0114] The time interval corresponding to each window is associated in the spread time track, and the track does not contain non-strong abnormal windows, but the number of interval windows between adjacent strong abnormal windows is marked, and the non-strong abnormal window refers to a window that is only data abnormal or only difference abnormal, or a normal window; wherein, the number of interval windows refers to the number of windows between two adjacent strong abnormal windows that are not marked as strong abnormal windows.

[0115] On the basis of the above, it should be noted that:

[0116] By constructing the spread time track, the process of abnormal state starting from the initial window and gradually spreading along the time axis can be clearly reflected, which helps to understand the evolution law and development trend of the abnormality, and effectively improves the judgment ability of the emergency degree of the abnormality, providing a reasonable basis for the determination of the subsequent recovery time.

[0117] On the basis of the above embodiments, Table 1 is set to express the change of the spread emergency coefficient of each strong abnormal window with the track base coefficient, window progression coefficient, and interval correction coefficient.

[0118] Table 1

[0119]

[0120] According to Table 1, the track base coefficient is a core index for measuring the stability and risk basis of the strong abnormal window track, and the values in the table are the average values of 24 sampling time points collected, which are adjusted according to the characteristics of the energy storage battery.

[0121] The window progression coefficient reflects the stage characteristics of the strong abnormal window advancing over time, and its value gradually increases from 0.15 to 0.75, and the corresponding spread emergency coefficient increases from 0.155 to 0.655 at the same time, so the window progression coefficient and the spread emergency coefficient are positively correlated.

[0122] The interval correction coefficient is a key parameter for adjusting the influence of the abnormal window interval, and its value gradually increases from 0.20 to 0.80, and the corresponding spread emergency coefficient increases from 0.155 to 0.655 at the same time, so the interval correction coefficient and the spread emergency coefficient are positively correlated.

[0123] According to Figure 2 It can be seen that the black square points of the spread emergency coefficient are also uniformly distributed along the fitting straight line, and as the window progression coefficient gradually increases, the spread emergency coefficient continues to linearly rise. The linear form of the fitting line clearly shows the linear driving effect of the window progression coefficient on the spread emergency coefficient, that is, the increase of the window progression coefficient will directly and linearly push up the spread emergency coefficient, and the positive correlation is stable and significant.

[0124] According to Figure 3 It can be seen that the black square points of the spread emergency coefficient are also highly consistent with the fitting straight line, and as the interval correction coefficient increases, the spread emergency coefficient shows a linear increasing trend. This shows that there is also a stable linear positive correlation between the interval correction coefficient and the spread emergency coefficient, and the change of the interval correction coefficient will linearly affect the size of the spread emergency coefficient.

[0125] On the basis of the above embodiment, based on the spread time trajectory and the first preset rule, the spread emergency coefficient of each strong abnormal window is obtained, and the specific logic is as follows:

[0126] Set the trajectory base coefficient, the window progression coefficient, and the interval correction coefficient to represent the spread emergency coefficient of each strong abnormal window.

[0127] The trajectory base coefficient is used to reflect the overall risk level under the same abnormal spread time trajectory.

[0128] Specifically, the abnormal spread time trajectory represents the development path of the abnormal state of the energy storage battery within a certain time range, and all the strong abnormal windows on the trajectory are in the same risk environment and evolution stage, so the trajectory base coefficients of these windows are set to be the same. The trajectory base coefficient reflects the overall risk environment intensity of the abnormal spread trajectory, and the greater the trajectory base coefficient, the higher the overall risk state of the energy storage battery system corresponding to the trajectory, and the greater the potential threat of abnormal spread.

[0129] The window progression coefficient is used to represent the progression degree of each strong abnormal window in the spread time trajectory, and the window progression coefficient of each strong abnormal window is assigned as follows:

[0130] Based on the index of the strong anomaly window, its window progression coefficient is assigned a value. The smaller the index value of the strong anomaly window, the smaller the window progression coefficient, and the larger the index value of the strong anomaly window, the larger the window progression coefficient.

[0131] The abnormal spread process is characterized by temporal continuity and accumulation. As time goes by, the abnormal state tends to gradually worsen or spread. The larger the window progression coefficient, the more significant and urgent the abnormal risk is, meaning that the strong abnormal window is in the later stage of the abnormal spread trajectory.

[0132] The interval correction coefficient reflects the influence of the number of interval windows between the current strong anomaly window and the previous strong anomaly window on the anomaly propagation speed. The number of interval windows refers to the number of windows between two adjacent strong anomaly windows that are not marked as strong anomaly windows.

[0133] The more interval windows there are, the slower the spread and the smaller the interval correction coefficient; conversely, the fewer the interval windows there are, the faster the spread and the larger the interval correction coefficient.

[0134] The spread urgency coefficient quantifies the urgency of an abnormal event and the speed at which the risk escalates. The interval correction coefficient serves as a quantitative indicator of the spread speed of an anomaly. The larger the value, the faster and more continuous the abnormal state develops over time, the more rapidly the risk accumulates, and the more significantly the urgency increases.

[0135] Based on the above three coefficients, the propagation urgency coefficient of each strong anomaly window is calculated using the following formula:

[0136] ;

[0137] in, For the first The spread urgency coefficient of a strong anomaly window is used to comprehensively evaluate the risk urgency of a strong anomaly window in the anomaly spread time trajectory by combining three index parameters: trajectory base coefficient, window progression coefficient, and interval correction coefficient. The larger the spread urgency coefficient, the higher the anomaly spread risk and the higher the urgency of the strong anomaly window.

[0138] By dynamically analyzing and scientifically evaluating the evolution and diffusion processes of anomalies, the ability to judge the urgency of anomalies has been effectively improved.

[0139] Based on this, it is necessary to further explain:

[0140] The abnormal propagation process of energy storage battery is a complex dynamic system with multiple factors, multiple stages and multiple rates. The trajectory-based coefficient reflects the overall risk situation of the system, the window progression coefficient embodies the time evolution depth of the abnormal state, and the interval correction coefficient quantifies the speed of abnormal propagation. Although the three factors interact in physical and chemical mechanisms, they can be regarded as independent risk contribution factors in a short time scale and local window. Based on the superposition principle in system dynamics, the influence of multiple independent risk factors can be regarded as linear superposition under certain conditions.

[0141] In addition, the trajectory-based coefficient, the window progression coefficient and the interval correction coefficient respectively depict the abnormal state from the macro risk environment, the time evolution and the propagation speed, and have strong independence in definition and physical meaning.

[0142] In summary, the function relationship between the propagation emergency coefficient and the trajectory-based coefficient, the window progression coefficient and the interval correction coefficient is quantified by the above function expression.

[0143] In the formula, is the trajectory-based coefficient of the mth strong abnormal window, is the window progression coefficient of the mth strong abnormal window, is the interval correction coefficient of the mth strong abnormal window, is the index of the strong abnormal window, , is the number of strong abnormal windows; In the formula, , ,

[0144] are the weight coefficients of the trajectory-based coefficient, the window progression coefficient and the interval correction coefficient respectively, , , The specific values of the trajectory-based coefficient, the window progression coefficient and the interval correction coefficient are determined according to the analytic hierarchy process, and the specific logic is as follows: Mark the trajectory-based coefficient, the window progression coefficient and the interval correction coefficient as 1, 2 and 3 respectively, determine the numerical value of the relative importance between each other by the nine-scale method, construct the judgment matrix, and mark the index of the trajectory-based coefficient as 1, the index of the window progression coefficient as 2, and the index of the interval correction coefficient as 3 to construct the judgment matrix

[0145]

[0146] ;

[0147] wherein, , both represent the index of the coefficient, and ,​​​​ a coefficient with index a coefficient with index the importance of the spread emergency coefficient, the specific value is determined by relevant experts using a 1-9 scoring method, a coefficient with index a coefficient with index the extreme importance of the spread emergency coefficient, a coefficient with index a coefficient with index the extreme unimportance of the spread emergency coefficient;

[0148] Divide each element value in the judgment matrix by the sum of its column to obtain a normalized judgment matrix, calculate the mean value of each row element value in the normalized judgment matrix, and take the mean value of the first row element value as the proportional coefficient of the trajectory base coefficient, take the mean value of the second row element value as the proportional coefficient of the window progression coefficient, and take the mean value of the third row element value as the proportional coefficient of the interval correction coefficient. Under the constraint condition that the sum of the scaled values is equal to 1, the three proportional coefficients are scaled equally, and the scaled values obtained are taken as the weights of the corresponding coefficients.

[0149] S4. Based on the second preset rule, the difference correlation between the feature parameters in each strong abnormal window is marked, including strong correlation and weak correlation, and the strong correlation proportion of the feature parameters in each strong abnormal window is counted;

[0150] Based on the above embodiment, based on the preset rule, the difference correlation between the feature parameters in each strong abnormal window is marked, and the specific logic is as follows:

[0151] In the residual sequence, if the voltage difference value continuously decreases and the capacity difference value continuously decreases, it indicates that there is a strong correlation or coupling relationship between the two in the battery aging process, and then the voltage-capacity is marked as strong correlation. When the voltage and capacity difference values have inconsistent trends, no obvious changes or no persistence (such as one side changes, the other side has no obvious changes or the change direction is opposite), it indicates that the correlation between the two parameters is weak, and the voltage-capacity is marked as weak correlation;

[0152] In the residual sequence, if the voltage difference value continuously decreases and the internal resistance difference value continuously increases, it indicates that there is a strong correlation or coupling relationship between the two in the battery aging process, and then the voltage-internal resistance is marked as strong correlation. When the voltage and internal resistance difference values have consistent trends, no obvious changes or no persistence (such as one side changes, the other side has no obvious changes or the change direction is the same), the voltage-internal resistance is marked as weak correlation;

[0153] If the capacity difference value continuously decreases and the internal resistance difference value continuously increases in the residual sequence, it indicates that there is a strong correlation or coupling relationship between the two in the battery aging process, and the capacity-internal resistance is marked as strong correlation. When the change trend of the capacity and internal resistance difference values is consistent, the change is not obvious or has no persistence (such as one side changes, and the other side has no obvious change or the change direction is the same), the capacity-internal resistance is marked as weak correlation.

[0154] In the above, it is necessary to point out that:

[0155] On the basis of the above, it is necessary to point out that:

[0156] By analyzing the correlation of voltage, capacity and internal resistance in the aging process, the strong correlation and weak correlation are effectively distinguished, and the internal mechanism of battery performance degradation can be more accurately revealed. For example, the strong correlation between voltage and capacity reflects the direct influence of capacity attenuation on voltage, and the strong correlation between voltage and internal resistance reflects the influence of internal resistance increase on battery output performance.

[0157] S5. Data processing is performed on the strong correlation proportion of each strong abnormal window characteristic parameter and the spread emergency coefficient to generate an evaluation coefficient. The evaluation coefficient is compared with a preset threshold value to determine the aging degree of the energy storage battery.

[0158] On the basis of the above embodiment, Table 2 is set to express the change of the evaluation coefficient of each strong abnormal window with the strong correlation proportion and the spread emergency coefficient of the characteristic parameter.

[0159] Table 2

[0160]

[0161] According to Table 2, the strong correlation proportion of the characteristic parameter increases from 0.5 to 0.74, the spread emergency coefficient increases from 0.51 to 0.75, and the evaluation coefficient synchronously increases from 0.26 to 0.56. The change trend of the three is the same, so the evaluation coefficient and the strong correlation proportion and the spread emergency coefficient of the characteristic parameter are positively correlated.

[0162] According to Figure 4 It can be seen that as the strong correlation proportion of the characteristic parameter gradually increases, the evaluation coefficient shows a stable upward trend as a whole, and the data points are closely distributed around the fitting line, indicating that the linear correlation between the two is strong. That is, the higher the strong correlation proportion of the characteristic parameter, the higher the evaluation coefficient.

[0163] According to Figure 5 It can be seen that when the spread emergency coefficient increases, the evaluation coefficient also shows a sustained upward trend, and the data points also fit the fitting line well, indicating that the positive correlation between the spread emergency coefficient and the evaluation coefficient is significant. The higher the spread emergency coefficient, the higher the evaluation coefficient.

[0164] Based on the above embodiments, the strong correlation ratio and spread urgency coefficient of each strong anomaly window feature parameter are processed to generate evaluation coefficients, based on the following formula:

[0165] ;

[0166] in, The average evaluation coefficient for all strong anomaly windows is used to evaluate the aging degree of energy storage batteries by combining two indicators: the strong correlation ratio of the characteristic parameters of strong anomaly windows and the propagation urgency coefficient. The larger the evaluation coefficient, the more severe the aging degree of the energy storage battery.

[0167] In the formula, For the first The proportion of strong correlations among strongly abnormal windows;

[0168] Based on the above, it should be noted that:

[0169] The strong correlation ratio of characteristic parameters within a strong anomaly window refers to the proportion of different characteristic parameters exhibiting strong correlation anomalies within that window. A larger ratio indicates that multiple key characteristic parameters are simultaneously abnormal and interconnected, reflecting a trend of coordinated deterioration of multiple performance indicators within the battery. Therefore, the strong correlation ratio is positively correlated with the aging degree of the energy storage battery; that is, the larger the strong correlation ratio, the more severe the battery aging.

[0170] The spread urgency coefficient indicates the speed and urgency of an abnormal state's propagation over time or space. A higher coefficient indicates a faster spread of the abnormal state within the battery system, a higher risk, and a more pronounced trend of battery health deterioration. Therefore, the spread urgency coefficient is positively correlated with the aging degree of the energy storage battery; that is, a higher spread urgency coefficient indicates more severe battery aging.

[0171] The strong correlation ratio and the spread urgency coefficient reflect the synergy of anomalous features and the urgency of the spread of anomalous states, respectively. Both are positively correlated indicators, and in practical applications, these two factors often influence and promote each other. Using a product form can reflect this synergistic amplification effect: the evaluation coefficient will increase significantly only when both indicators are large, accurately reflecting the severity of battery aging.

[0172] In summary, the above functional expression is used to express the functional relationship between the evaluation coefficient and the strong correlation ratio of the characteristic parameters and the spread urgency coefficient.

[0173] The evaluation coefficient can effectively assess the aging degree and development trend of batteries. Combined with the analysis of time-series propagation trajectory, it can improve the comprehensive assessment capability of energy storage battery aging.

[0174] On the basis of the above-mentioned embodiments, the evaluation coefficient is compared with a preset threshold value to judge the aging degree of the energy storage battery, and the specific logic is as follows:

[0175] When the evaluation coefficient is less than the threshold value, it is determined that the energy storage battery is slightly aged.

[0176] When the evaluation coefficient is greater than the threshold value, it is determined that the energy storage battery is moderately aged.

[0177] When the evaluation coefficient is greater than the threshold value, it is determined that the energy storage battery is moderately aged.

[0178] Wherein, the threshold value for dividing the slight aging and the moderate aging is The threshold value for dividing the moderate aging and the severe aging is

[0179] A large number of evaluation coefficient data samples of the energy storage battery in different aging stages (slight, moderate and severe) are collected, and the typical interval of the evaluation coefficient in different aging stages is determined through statistical analysis (such as distribution density estimation, cluster analysis, etc.), and the dividing point of the slight and moderate aging interval is selected as the threshold value , and the dividing point of the moderate and severe aging interval is selected as the threshold value .

[0180] Please refer to Figure 6 , the application also provides a technical solution:

[0181] An energy storage battery aging detection device based on a cross-over difference algorithm, the device is used to execute any one of the above-mentioned energy storage battery aging detection methods based on a cross-over difference algorithm, comprising:

[0182] A window division module is used to collect the feature parameters of the energy storage battery at each time within the past 24 hours to form time series data, and a sliding window method is used to divide the time series data into a plurality of cross-over windows.

[0183] A data processing module is used to obtain the extreme value, mean value and standard deviation of the feature parameters in each window, and calculate the feature parameter difference value in each group of adjacent windows to obtain a residual sequence.

[0184] A trajectory construction module is used to compare the extreme value, mean value and standard deviation of the feature parameters in each window with the normal interval range thereof, and compare the absolute value of the feature parameter difference value with a preset threshold value, filter out strong abnormal windows which are marked as data abnormality and difference abnormality at the same time, determine an initial abnormal window, trace a spreading window, construct a spreading time trajectory, and obtain a spreading emergency coefficient of each strong abnormal window based on the spreading time trajectory and a first preset rule.

[0185] ​​​​The labeling module is configured to label the difference correlation between the feature parameters in each strong abnormal window based on a second preset rule, including strong correlation and weak correlation, and to count the strong correlation proportion of the feature parameters in each strong abnormal window;

[0186] The judging module is configured to perform data processing on the strong correlation proportion of the feature parameters in each strong abnormal window and the spread emergency coefficient to generate an evaluation coefficient, compare the evaluation coefficient with a preset threshold, and judge the aging degree of the energy storage battery.

[0187] The above formulas are all dimensionless values calculated, the formulas are obtained by software simulation of a large number of collected data to obtain a formula of the most recent real situation, and the preset parameters in the formulas are set by a person skilled in the art according to the actual situation.

[0188] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized in the form of a computer program product wholly or partially. Those skilled in the art can realize that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be realized by electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software methods depends on the specific application and design constraints of the technical solutions.

[0189] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, which can be located in one place or distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0190] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for detecting aging of an energy storage battery based on a cross- over difference algorithm, characterized in that, The specific steps include: Collecting the characteristic parameters of the energy storage battery every hour in the past 24 hours to form time series data, and dividing the time series data into multiple cross-overlapping windows by using a sliding window method; Obtaining the extreme value, mean value and standard deviation of the characteristic parameters in each window, and calculating the difference value of the characteristic parameters in each group of adjacent windows to obtain a residual sequence; Comparing the extreme value, mean value and standard deviation of the characteristic parameters in each window with the normal interval range thereof respectively, comparing the absolute value of the difference value of the characteristic parameters with a preset threshold value, screening out strong abnormal windows marked as data abnormality and difference abnormality at the same time, determining initial abnormal windows, simultaneously tracing spreading windows, constructing a spreading time trajectory, and obtaining a spreading emergency coefficient of each strong abnormal window based on the spreading time trajectory and a first preset rule; Based on a second preset rule, marking the difference correlation between the characteristic parameters in each strong abnormal window, including strong correlation and weak correlation, and counting the strong correlation proportion of the characteristic parameters in each strong abnormal window; Data processing the strong correlation proportion and the spreading emergency coefficient of the characteristic parameters in each strong abnormal window to generate an evaluation coefficient, comparing the evaluation coefficient with a preset threshold value, and judging the aging degree of the energy storage battery; Based on the spreading time trajectory and the first preset rule, the spreading emergency coefficient of each strong abnormal window is obtained, and the specific logic is as follows: Set the trajectory base coefficient, window progression coefficient and interval correction coefficient to represent the spreading emergency coefficient of each strong abnormal window; The trajectory base coefficient is used to reflect the overall risk level under the same spreading time trajectory, and the trajectory base coefficients of all strong abnormal windows are set to be the same; The window progression coefficient is used to represent the progression degree of each strong abnormal window in the spreading time trajectory, and the window progression coefficient of each strong abnormal window is assigned as follows: Sort all strong abnormal windows according to the window index value from small to large, assign the window progression coefficient according to the serial number of the strong abnormal window, the smaller the serial number, the smaller the corresponding window progression coefficient, and the larger the serial number, the larger the corresponding window progression coefficient; The interval correction coefficient reflects the influence of the number of interval windows between the current adjacent strong abnormal window and the previous strong abnormal window on the abnormal spreading speed; The more the number of interval windows, the slower the spreading speed, and the smaller the interval correction coefficient; on the contrary, the fewer the number of interval windows, the faster the spreading speed, and the larger the interval correction coefficient; Based on the above three coefficients, the spreading emergency coefficient of each strong abnormal window is calculated, and the formula is as follows: wherein, is a spread emergency coefficient for the th strong anomaly window, is a trajectory base coefficient for the th strong anomaly window, is a window progression coefficient for the th strong anomaly window, is a gap correction coefficient for the th strong anomaly window, is an index of the strong anomaly window, , is a number of strong anomaly windows; In the formula, , , are weight coefficients of trajectory base coefficient, window progression coefficient, and interval correction coefficient, respectively, , , The specific values of are determined according to the analytic hierarchy process.

2. The energy storage battery aging detection method based on the cross-overlapping difference algorithm according to claim 1, characterized in that A group of characteristic parameters is collected every hour, including voltage, capacity and internal resistance, a total of 24 groups of characteristic parameters, a group of characteristic parameters is represented in the form of a 3-row 1-column matrix, each row corresponds to a characteristic parameter, and the time series data is represented in the form of a 3-row 24-column matrix, each column represents a group of characteristic parameters, and the characteristic parameters in each column of the matrix are arranged in the order of collection time; Using an overlapping window method with a sliding step size of 3, the time series data is divided into 7 windows. The time series data in each window is a 3x6 matrix, denoted as follows: , For the first Time series data for each window, For the window index, .

3. The cross-over superimposed differential algorithm based energy storage battery aging detection method according to claim 2, wherein, The residual sequence is obtained, and the specific logic is as follows: For the 7 windows obtained by division, the difference between the feature parameters of the previous window and the next window in the adjacent window is calculated to obtain a residual sequence represented in matrix form, and the formula is as follows: wherein, is the residual sequence of the th window, which is the time series data difference between the th window and the th window, is the time series data of the th window, and the residual sequence is also a 3-by-6 matrix.

4. The cross-over superimposed differential algorithm based energy storage battery aging detection method of claim 2, wherein, Strong abnormal windows marked as data abnormal and differential abnormal at the same time are screened out, and the specific logic is as follows: For each window, if the maximum value, minimum value, mean value and standard deviation of voltage, capacity and internal resistance in the time series data in the window are all out of the normal interval range, the window is marked as a data abnormal window, otherwise, the window is marked as a data normal window; If the absolute value of the voltage, capacity and internal resistance difference in the residual sequence corresponding to the window is greater than the preset threshold, the window is marked as a differential abnormal window, otherwise, the window is marked as a differential normal window; According to the marking result of each window, the window marked as data abnormal and differential abnormal at the same time is screened out as a strong abnormal window; The strong abnormal window with the smallest index is taken as the initial abnormal window, and the spread window is traced back to construct a spread time trajectory, and the specific logic is as follows: The strong abnormal window with the smallest serial number is selected as the initial abnormal window, and the other strong abnormal windows are taken as the spread windows to form a spread time trajectory from the initial abnormal window to the last spread window in ascending order of index value; In the spread time trajectory, the time intervals corresponding to each window are associated, and the number of interval windows between adjacent strong abnormal windows is marked, wherein the number of interval windows refers to the number of windows between two adjacent strong abnormal windows that are not marked as strong abnormal windows.

5. The energy storage battery aging detection method based on the cross-over differential algorithm according to claim 4, characterized in that, Based on the second preset rule, the differential correlation between the feature parameters of each strong abnormal window is marked, and the specific logic is as follows: For any strong abnormal window, in its residual sequence, if the voltage difference value continuously decreases and the capacity difference value continuously decreases, the voltage-capacity is marked as strong correlation, otherwise, the voltage-capacity is marked as weak correlation; In the residual sequence, if the voltage difference value continuously decreases and the internal resistance difference value continuously increases, the voltage-internal resistance is marked as strong correlation, otherwise, the voltage-internal resistance is marked as weak correlation; In the residual sequence, if the capacity difference value continuously decreases and the internal resistance difference value continuously increases, the capacity-internal resistance is marked as strong correlation, otherwise, the capacity-internal resistance is marked as weak correlation; Wherein, continuous decrease or continuous increase refers to that in the residual sequence, the feature parameter difference value presents a monotonous decreasing or increasing trend along the column number increasing direction.

6. The cross-over superimposed differential algorithm based energy storage battery aging detection method of claim 5, wherein, The strong correlation proportion of each strong abnormal window feature parameter and the spread emergency coefficient are processed to generate an evaluation coefficient, and the formula is as follows: wherein, is the average evaluation coefficient for all strong anomaly windows, is the strong correlation proportion of the th strong anomaly window.

7. The cross-over superimposed differential algorithm based energy storage battery aging detection method of claim 6, wherein, The evaluation coefficient is compared with the preset threshold to judge the aging degree of the energy storage battery, and the specific logic is as follows: When the energy storage battery is determined to be slightly aged; When the energy storage battery is determined to be moderately aged; When the energy storage battery is determined to be severely aged; wherein, is a threshold value for dividing mild aging from moderate aging, is a threshold value for dividing moderate aging from severe aging.

8. A device for detecting the aging of an energy storage battery based on a cross- overlapping difference algorithm, the device being used to perform the method for detecting the aging of an energy storage battery based on a cross-overlapping difference algorithm according to any one of claims 1-7, characterized in that, It includes: A window division module for collecting feature parameters of the energy storage battery at each time in the past 24 hours to form time series data, and dividing the time series data into multiple cross-over windows by using a sliding window method; A data processing module for obtaining the extreme value, mean value and standard deviation of the feature parameters in each window, and calculating the feature parameter difference in each group of adjacent windows to obtain a residual sequence; The trajectory construction module is configured to compare the extreme value, the mean value and the standard deviation of the feature parameters in each window with the normal interval range of the feature parameters respectively, compare the absolute value of the feature parameter difference with a preset threshold value, screen out strong abnormal windows which are marked as data abnormality and differential abnormality at the same time, determine initial abnormal windows, trace back spreading windows, construct a spreading time trajectory, and obtain a spreading emergency coefficient of each strong abnormal window based on the spreading time trajectory and a first preset rule; The labeling module is configured to label the differential correlation between the feature parameters in each strong abnormal window based on a second preset rule, including strong correlation and weak correlation, and count a strong correlation proportion of the feature parameters in each strong abnormal window; The judgment module is configured to perform data processing on the strong correlation proportion and the spreading emergency coefficient of the feature parameters in each strong abnormal window, generate an evaluation coefficient, compare the evaluation coefficient with a preset threshold value, and judge the aging degree of the energy storage battery.

Images