A battery self-discharge abnormality diagnosis method

Abstract

The present application relates to the technical field of lithium ion battery, specifically to a battery self-discharge abnormal diagnosis method, comprising the following steps: for a healthy lithium ion battery, in a preset battery state of charge point set, the battery is placed in a temperature box, scanning at a rate of 0.1 ℃ / min in a temperature range of 5 ℃, synchronously collecting open-circuit voltage sequence and temperature sequence, and using linear regression to calculate the slope of the voltage sequence to the temperature sequence, the present application realizes accurate identification and positioning of self-discharge abnormality by constructing a dual-diagnosis mechanism of battery thermodynamic fingerprint and spatial heat field distribution, establishes a benchmark entropy change spectrum line of healthy battery, and provides a high-sensitivity thermodynamic reference system for abnormal diagnosis, the entropy change coefficient directly reflects the lattice structure and electrochemical reaction of electrode material, and the spectrum line form can capture subtle thermodynamic state changes caused by early faults such as trace metal lithium dissolution or electrolyte decomposition.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery technology, and in particular to a method for diagnosing abnormal battery self-discharge. Background Technology

[0002] The field of lithium-ion battery technology refers to rechargeable battery systems that use lithium ions as charge carriers. This field encompasses the core components of the battery, including the negative electrode, which is typically graphite; the positive electrode, which is a transition metal oxide containing lithium (such as lithium iron phosphate or lithium nickel cobalt manganese oxide); the non-aqueous electrolyte for transporting lithium ions; and the separator for separating the positive and negative electrodes.

[0003] Current technologies primarily rely on monitoring external battery characteristics such as open-circuit voltage drop rate, capacity decay, or increased internal resistance. These methods typically only passively detect anomalies when self-discharge has progressed to a significant level, leading to a marked deterioration in overall battery performance. This results in a significant diagnostic lag, missing the optimal window for early intervention. For instance, when an internal micro-short circuit has just formed, the resulting leakage current is extremely small, having a negligible impact on the overall battery voltage and capacity in a short period, making it difficult to detect using traditional monitoring methods. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method for diagnosing abnormal battery self-discharge.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for diagnosing abnormal battery self-discharge, comprising the following steps:

[0006] For healthy lithium-ion batteries, in a preset set of battery state of charge points, the batteries are placed in a temperature chamber and the temperature range of 5℃ is scanned at a rate of 0.1℃ / min. The open circuit voltage sequence and temperature sequence are collected simultaneously. The slope of the voltage sequence against the temperature sequence is calculated by linear regression to obtain the entropy change coefficient value of each battery state of charge point. Then, all entropy change coefficient values ​​are connected by spline interpolation to establish a reference entropy change spectrum.

[0007] Based on the aforementioned baseline entropy change spectral line, all peaks and valleys in the spectral line are located using the first derivative zero-finding method. The peak and valley positions are determined as characteristic battery state of charge points. The entropy change coefficient of the lithium-ion battery under test, which is expected to undergo lithium plating, is repeatedly measured at the characteristic battery state of charge points. At the same time, the battery surface is continuously photographed with an infrared thermal imager at a frequency of 25 frames / second, and the temperature reading of each pixel at each time point is recorded to obtain the entropy change data and the sequence thermal image.

[0008] Preferably, the method further includes:

[0009] Based on the reference entropy change spectrum and the entropy change data to be measured and the sequence thermal image, the entropy change data to be measured and the reference entropy change spectrum are subtracted at the same battery state of charge point to obtain the entropy change deviation value set. At the same time, the time series temperature value of each pixel coordinate in the sequence thermal image is averaged to obtain the average temperature map. For each pixel in the average temperature map, the average temperature of the four neighboring pixels is calculated to obtain the entropy change deviation feature and the heat flow gradient map.

[0010] Based on the entropy change deviation characteristics and the heat flow gradient map, the Euclidean norm of the entropy change deviation numerical set is calculated, and the pixel with the largest absolute value in the heat flow gradient map is identified as a hot spot caused by diaphragm damage. The hot spot intensity value is extracted, and the Euclidean norm and the hot spot intensity value are weighted and summed to construct an evaluation index. The evaluation index is compared with a preset diagnostic threshold to obtain a self-discharge abnormality diagnosis conclusion.

[0011] Preferably, the step of obtaining the reference entropy variation spectral line specifically includes:

[0012] For healthy lithium-ion batteries, in a preset set of battery state of charge points, the battery is placed in a temperature chamber and the temperature range of 5°C is scanned at a rate of 0.1°C / min. Discrete open-circuit voltage values ​​and battery temperature values ​​are recorded synchronously at a sampling frequency of 1Hz to form paired time series data pairs and obtain temperature and pressure time series data.

[0013] Based on the temperature and pressure time series data, for each battery state of charge point, the corresponding open-circuit voltage sequence is used as the dependent variable and the temperature sequence is used as the independent variable. A straight line is fitted by the least squares method, and the slope of the straight line is used as the entropy change coefficient value of the battery state of charge point. The entropy change coefficient values ​​of all battery state of charge points are summarized to generate a discrete entropy change point set.

[0014] Based on the discrete entropy change point set, piecewise fitting is performed between adjacent entropy change coefficient values, while ensuring the continuity of the first and second derivatives at the connection points, connecting the discrete points into a smooth curve, and establishing a benchmark entropy change spectrum.

[0015] Preferably, the steps for acquiring the entropy change data to be tested and the sequence thermal image are as follows:

[0016] Based on the aforementioned reference entropy-variable spectral line, the gradient value of each point on the spectral line is calculated using the central difference method to form a first derivative sequence. Then, the first derivative sequence is traversed to locate all battery state of charge points where the gradient value changes from positive to negative and from negative to positive, thus obtaining a set of characteristic state of charge points.

[0017] Based on the set of characteristic state of charge points, the lithium-ion battery under test is placed in a temperature chamber, and a temperature scan is performed at each characteristic state of charge point. The open-circuit voltage and temperature data are recorded. Then, the least squares method is applied to the data at each point to calculate the slope of voltage with respect to temperature, and the discrete entropy change value under test is generated.

[0018] Based on the discrete entropy change value to be measured, the surface of the battery is continuously photographed at a frequency of 25 frames / second using an infrared thermal imager during the measurement process. The temperature reading of each pixel at each time point is stored in a three-dimensional matrix. The discrete entropy change value to be measured is paired with the three-dimensional matrix to obtain the entropy change data to be measured and the sequence thermal image.

[0019] Preferably, the steps for obtaining the entropy change deviation feature and the heat flux gradient map are as follows:

[0020] Based on the reference entropy change spectrum line and the entropy change data to be measured and the entropy change data to be measured in the sequence thermal image, the set of characteristic charge state points is traversed. At each point, the value of the entropy change data to be measured is subtracted from the value of the corresponding point on the reference entropy change spectrum line, and all the differences are stored in a new array to obtain the set of entropy change deviation values.

[0021] Based on the entropy change data to be measured and the sequence thermal images in the sequence thermal images, the temperature value of each pixel coordinate in the image is extracted at all time points, the average of the temperature values ​​is calculated, and the average is assigned to the pixel point with the same coordinate in the new image to generate an average temperature map.

[0022] Preferably, the step of obtaining the entropy change deviation feature and the heat flux gradient map further includes:

[0023] Based on the average temperature map and the entropy change deviation value set, each pixel of the average temperature map is traversed, the average temperature of the four adjacent pixels (top, bottom, left, and right) is calculated, and the temperature value of the center pixel is subtracted from the average temperature value to obtain the difference. All the differences are used to form a new image to obtain the entropy change deviation feature and heat flow gradient map.

[0024] Preferably, the steps for obtaining the self-discharge abnormality diagnosis conclusion are as follows:

[0025] Based on the entropy deviation characteristics and heat flow gradient map, the entropy deviation values ​​are extracted, each value in the set of entropy deviation values ​​is squared, and then all the squared values ​​are added together to get the sum. Finally, the square root of the sum is calculated to obtain the entropy deviation norm.

[0026] Based on the entropy change deviation feature and the heat flow gradient map, each pixel in the heat flow gradient map is traversed to find the pixel corresponding to the maximum value, and the maximum absolute value is taken as the hot spot intensity value.

[0027] Preferably, the step of obtaining the self-discharge abnormality diagnostic conclusion further includes:

[0028] Based on the entropy change deviation norm and the hotspot intensity value, the entropy change deviation norm is multiplied by a weighting coefficient, and the hotspot intensity value is multiplied by a weighting coefficient to construct an evaluation index. The evaluation index is then compared with a preset diagnostic threshold to obtain a self-discharge abnormality diagnosis conclusion.

[0029] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0030] This invention achieves precise identification and localization of self-discharge anomalies by constructing a dual diagnostic mechanism based on battery thermodynamic fingerprints and spatial thermal field distribution. A baseline entropy change spectrum for healthy batteries is established, providing a highly sensitive thermodynamic reference for anomaly diagnosis. The entropy change coefficient, as a direct reflection of the electrode material's lattice structure and electrochemical reactions, can capture subtle thermodynamic state changes caused by early faults such as trace lithium dissolution or electrolyte decomposition. This diagnostic method based on intrinsic electrochemical parameters reveals potential risks earlier than traditional macroscopic characterization relying solely on voltage or capacity decay. Furthermore, by locating spectral peaks and valleys as feature detection points, diagnostic efficiency is improved, avoiding lengthy testing across the entire SOC range. Combining entropy change spectrum deviation analysis with spatiotemporal processing of high-resolution sequential thermal images, environmental noise is effectively suppressed through time averaging, and weak hotspot signals generated by localized self-discharge are amplified by calculating spatial gradients, successfully sharpening the diffuse temperature field into a clear localization basis. Ultimately, the comprehensive evaluation index, which integrates entropy change deviation norm and hotspot intensity value, quantifies the severity of anomalies, providing an objective and reliable numerical basis for diagnostic decisions and improving the accuracy, sensitivity, and location precision of early fault diagnosis in lithium-ion batteries. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the steps of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0033] Please see Figure 1 This invention provides a technical solution, a method for diagnosing abnormal battery self-discharge, comprising the following steps:

[0034] For healthy lithium-ion batteries, in a preset set of battery state of charge points, the batteries are placed in a temperature chamber and the temperature range of 5℃ is scanned at a rate of 0.1℃ / min. The open circuit voltage sequence and temperature sequence are collected simultaneously. The slope of the voltage sequence against the temperature sequence is calculated by linear regression to obtain the entropy change coefficient value of each battery state of charge point. Then, all entropy change coefficient values ​​are connected by spline interpolation to establish a reference entropy change spectrum.

[0035] Based on the baseline entropy change spectrum, all peaks and valleys in the spectrum are located using the first derivative zero-finding method. The peak and valley positions are determined as the characteristic battery state of charge points. The entropy change coefficient of the lithium-ion battery under test, which is expected to undergo lithium plating, is repeatedly measured at the characteristic battery state of charge points. At the same time, the surface of the battery is continuously photographed with an infrared thermal imager at a frequency of 25 frames / second, and the temperature reading of each pixel at each time point is recorded to obtain the entropy change data and the sequence thermal image.

[0036] Based on the benchmark entropy change spectrum and the entropy change data to be measured and the sequence thermal image, the difference between the entropy change data to be measured and the benchmark entropy change spectrum at the same battery state of charge point is calculated to obtain the set of entropy change deviation values. At the same time, the time series temperature value of each pixel coordinate in the sequence thermal image is averaged to obtain the average temperature map. For each pixel in the average temperature map, the average temperature of the four neighboring pixels is calculated to obtain the entropy change deviation characteristics and heat flow gradient map.

[0037] Based on the entropy change deviation characteristics and the heat flow gradient map, the Euclidean norm of the set of entropy change deviation values ​​is calculated, and the pixel with the largest absolute value in the heat flow gradient map is identified as the hot spot caused by diaphragm damage. The hot spot intensity value is extracted, and the Euclidean norm and the hot spot intensity value are weighted and summed to construct an evaluation index. The evaluation index is compared with the preset diagnostic threshold to obtain the diagnosis conclusion of self-discharge abnormality.

[0038] The specific steps for obtaining the baseline entropy variation spectral line are as follows:

[0039] For healthy lithium-ion batteries, in a preset set of battery state of charge points, the battery is placed in a temperature chamber and the temperature range of 5°C is scanned at a rate of 0.1°C / min. Discrete open-circuit voltage values ​​and battery temperature values ​​are recorded synchronously at a sampling frequency of 1Hz to form paired time series data pairs and obtain temperature and pressure time series data.

[0040] Based on temperature and pressure time series data, for each battery state of charge point, the corresponding open-circuit voltage sequence is used as the dependent variable and the temperature sequence is used as the independent variable. A straight line is fitted by the least squares method, and the slope of the straight line is used as the entropy change coefficient of the battery state of charge point. The entropy change coefficients of all battery state of charge points are summarized to generate a discrete entropy change point set.

[0041] Based on the discrete entropy change point set, piecewise fitting is performed between adjacent entropy change coefficient values, while ensuring the continuity of the first and second derivatives at the connection points, connecting the discrete points into a smooth curve to establish a benchmark entropy change spectrum.

[0042] Specifically, a healthy lithium-ion battery is defined as one battery from a selected batch that has not been cycled or abused, has a capacity retention rate of no less than 99% after three standard charge-discharge cycles, and has a DC internal resistance deviation within 3% of the nominal value. First, a preset set of battery state-of-charge (SOC) points covering the main operating range of the battery is established. This set starts at 10% SOC and increases in 5% increments up to 90% SOC, specifically {10%, 15%, 20%, ...,}. 90%} Next, the healthy lithium-ion battery was securely placed in a high-precision temperature chamber. A voltage acquisition terminal was connected using a four-wire method, and a calibrated T-type thermocouple was tightly attached to the geometric center of the battery's maximum surface area. Subsequently, for each state of charge point in the aforementioned set, for example, 50%, the battery was precisely charged and discharged to that state using coulomb integration combined with the open-circuit voltage correspondence. Then, it was left to stand for at least 4 hours until the change in its open-circuit voltage was less than 1 millivolt over 30 consecutive minutes, confirming that the battery had reached electrochemical equilibrium. The temperature control program of the temperature chamber was then activated, performing a temperature scan of 5°C around the center temperature point of 25°C at a strictly constant rate of 0.1°C / min, i.e., linearly increasing the temperature from 22.5°C to 27.5°C. During the entire 50-minute scan, a data acquisition system synchronized with the voltmeter and thermometer continuously recorded the battery's open-circuit voltage and surface temperature at a sampling frequency of 1Hz, recording the data at each time point. Each corresponds to an open-circuit voltage reading. and a temperature reading This forms a group containing 3000 The time series of data pairs is obtained by repeatedly performing this operation at each point in the preset set of battery state of charge points, and finally collecting the time series data pairs under all state of charge points to obtain temperature and pressure time series data.

[0043] Based on temperature and pressure time-series data, for each battery state of charge point, there are independent time-series data pairs, for example, 3000 data pairs corresponding to the 50% state of charge point. The set of data points, where From 1 to 3000, the open-circuit voltage sequence As the dependent variable, the temperature series As the independent variable, a linear relationship model is established. Linear regression fitting is performed using the least squares method, the specific calculation process of which involves minimizing the sum of squared residuals. In order to make Minimum slope The following formula is used for calculation: in, This represents the total number of samples at this state of charge point, which is 3000. and The first Temperature and open-circuit voltage values ​​at each sampling time. It is the sum of the products of temperature and voltage. It is the sum of all temperature values. It is the sum of all voltage values. It is the sum of the squares of all temperature values, from the 3000 collected data. Substitute the data into the formula to calculate the slope value. The slope value This is defined as the entropy change coefficient value at the 50% state of charge point. This calculation process is applied sequentially to all preset battery state of charge points, such as 10%, 15%, up to 90%, thereby calculating a corresponding entropy change coefficient value for each state of charge point. Finally, all state of charge points and their corresponding entropy change coefficient values ​​are summarized in the form of data pairs, such as {(10%, 0.3mV / K), (15%, 0.35mV / K), ...}, generating a discrete entropy change point set.

[0044] Based on the discrete entropy change point set, which is a set of multiple (charge state, entropy change coefficient) data pairs, a cubic spline interpolation method is used to fit a piecewise function between these discrete points. For any two adjacent data points in the discrete entropy change point set... and ,in Represents the state of charge. Representing the entropy change coefficients, construct a cubic polynomial. To describe arrive To determine the four coefficients for the curve within the interval, the following series of constraints must be met: First, the function must pass through all data points, that is, for all intervals... ,satisfy as well as Secondly, at each internal data point At this point, the first derivatives of the two polynomials on the left and right sides must be equal, that is... At the same time, the second derivative must also be continuous, that is... These conditions collectively ensure a smooth transition at the connection points. To solve the entire system of equations, two additional boundary conditions are needed. Here, natural spline boundary conditions are used, which force the second derivative of the curve to be zero at the two endpoints (10% and 90% charge state). and By solving the system of linear equations formed by all the above conditions, the coefficients of all piecewise cubic polynomials can be uniquely determined. Combining all these piecewise cubic polynomials together forms a smooth curve that is continuous throughout the entire domain of charged states (10% to 90%) and whose first and second derivatives are also continuous, thus establishing a reference entropy variation spectrum.

[0045] The specific steps for acquiring the entropy change data to be measured and the sequential thermal images are as follows:

[0046] Based on the reference entropy variation spectral line, the gradient value of each point on the spectral line is calculated using the central difference method to form a first derivative sequence. Then, the first derivative sequence is traversed to locate all battery state of charge points where the gradient value changes from positive to negative and from negative to positive, thus obtaining a set of characteristic state of charge points.

[0047] Based on the characteristic state of charge point set, the lithium-ion battery under test is placed in a temperature chamber, and temperature scanning is performed at each characteristic state of charge point. The open circuit voltage and temperature data are recorded. Then, the least squares method is applied to the data at each point to calculate the slope of voltage with respect to temperature, and the discrete entropy change value under test is generated.

[0048] Based on the discrete entropy change value to be measured, the surface of the battery is continuously photographed at a frequency of 25 frames / second using an infrared thermal imager during the measurement process. The temperature reading of each pixel at each time point is stored in a three-dimensional matrix. The discrete entropy change value to be measured is paired with the three-dimensional matrix to obtain the entropy change data to be measured and the sequence thermal image.

[0049] Specifically, based on the reference entropy-variable spectral line, that is, the continuous function obtained by cubic spline interpolation. ,in To determine the state of charge, initially use a step size of 0.1% state of charge. ,Right now Discretize the continuous function to generate a high-density point set. Subsequently, to calculate the gradient value at each sampling point on the spectrum, the central difference method was used. Find the first derivative for each sampling point. Its gradient value Through formula Perform approximate calculations, where and These are functions in The values ​​at adjacent sampling points are used to perform this calculation on all internal sampling points, forming a sequence of first derivatives corresponding to the high-density point set. This sequence of first derivatives is then traversed to locate all local extrema. Specifically, this is achieved by checking the signs of two adjacent gradient values; when a pair of adjacent gradient values ​​is found... and The product of is negative, that is At that time, it was determined that in and There exists an extreme point if and Then locate a peak point, if and Then, a valley point is located. To avoid pseudo-extreme points caused by computational noise, a gradient threshold is introduced. mV / K / SOC, only when and Only when the sign change is confirmed as a valid extreme point, the state of charge values ​​corresponding to all such peaks and valleys are collected. For example, if extreme values ​​are identified at 23.4%, 45.8%, 67.2%, and 88.1% state of charge, these values ​​are summarized to obtain a set of characteristic state of charge points.

[0050] Based on a set of characteristic state of charge (SOC) points, such as those containing {23.4%, 45.8%, 67.2%, 88.1%}, the lithium-ion battery under test, which is expected to undergo lithium plating, is installed in the same temperature chamber and measuring equipment environment as when testing healthy batteries. The SOC of the battery is sequentially adjusted to each point in the characteristic SOC point set. For the 45.8% point, a high-precision charge-discharge machine is used to precisely adjust the battery SOC to 45.8% using coulometric measurement. The battery is then left to stand for at least 4 hours until the voltage fluctuation is less than 1 mV within 30 minutes. Next, a temperature scanning procedure identical to that for healthy batteries is performed: scanning a 5°C temperature range in the temperature chamber at a rate of 0.1°C / min, while simultaneously acquiring open-circuit voltage and battery temperature data at a frequency of 1 Hz. This yields a set of temperature-pressure time-series data specific to this characteristic point. Then, the least squares method is immediately applied to this newly acquired data, with the calculation formula exactly the same as before. in and Using time-series data obtained from measurements at the current feature point of the battery under test, this formula is used to calculate the entropy change coefficient of the battery under test at a state of 45.8% charge. This complete process of adjusting the state of charge, settling, temperature scanning, and slope calculation is repeated once at each point (23.4%, 67.2%, 88.1%) in the set of characteristic state of charge points. Finally, all characteristic state of charge points and their corresponding measured entropy change coefficient values ​​are summarized to generate the discrete entropy change value to be measured.

[0051] During the aforementioned temperature scan of the lithium-ion battery under test at each characteristic state of charge (SOC) point to generate the discrete entropy change value, an infrared thermal imager is simultaneously activated. The imager's lens, through the infrared window of the temperature chamber, is aimed directly at and completely covers the largest plane of the battery under test. Its emissivity parameter is calibrated according to the battery surface coating material, for example, set to 0.95. The focal length and temperature measurement range are also pre-adjusted. During the 50-minute temperature scan at each SOC point, the infrared thermal imager continuously captures the surface temperature distribution map of the battery at a fixed frequency of 25 frames per second. Each frame is a two-dimensional array, for example, 640x480 pixels. Each element in the array represents the real-time temperature reading of the corresponding pixel. For a single 50-minute measurement, a total of [number missing] frames are collected. The 75,000 frames of thermal images are stacked sequentially over time to form a three-dimensional matrix with dimensions (640, 480, 75,000). This matrix completely records the temperature change of each pixel on the battery surface over time during the measurement of a single characteristic state of charge point. This three-dimensional matrix is ​​the sequential thermal image of that point. This acquisition process is repeated at all characteristic state of charge points along with the entropy change measurement. After the measurement is completed, the discrete entropy change value to be measured generated in the previous step is paired with the sequential thermal image acquired synchronously during the measurement. That is, each characteristic state of charge point is associated with an entropy change value and a corresponding three-dimensional thermal image matrix, thus obtaining the entropy change data to be measured and the sequential thermal image.

[0052] The specific steps for obtaining the entropy change deviation characteristics and heat flux gradient map are as follows:

[0053] Based on the baseline entropy change spectrum and the entropy change data to be measured in the sequence thermal image, the set of characteristic charge state points is traversed. At each point, the value of the entropy change data to be measured is subtracted from the value of the corresponding point on the baseline entropy change spectrum, and all the differences are stored in a new array to obtain the set of entropy change deviation values.

[0054] Based on the entropy change data to be measured and the sequential thermal images in the sequence thermal images, the system traverses each pixel coordinate in the image, extracts the temperature value of the pixel coordinate at all time points, calculates the average of the temperature values, and assigns the average to the pixel point with the same coordinate in the new image to generate an average temperature map.

[0055] Based on the average temperature map and the set of entropy change deviation values, we traverse each pixel of the average temperature map, calculate the average temperature of the four adjacent pixels (top, bottom, left, and right), and then subtract the average temperature from the temperature value of the center pixel to obtain the difference. We then construct a new image from all the differences to obtain the entropy change deviation features and the heat flow gradient map.

[0056] Specifically, based on the baseline entropy-variable spectral line, this is a continuous function obtained through cubic spline interpolation. The system consists of the entropy change data to be measured and the entropy change data to be measured in the sequence thermal images. This is a discrete set containing multiple data pairs (characteristic state of charge, entropy change coefficient value). First, a traversal program is initiated. This program uses the set of characteristic state of charge points, such as {23.4%, 45.8%, 67.2%, 88.1%}, as an index, and processes each point sequentially. For the first point, 23.4%, the program first searches for and extracts the corresponding entropy change coefficient value in the entropy change data to be measured, denoted as... Meanwhile, the program substitutes the state of charge value of 0.234 into the reference entropy variation spectral function. In the calculation, the reference entropy change coefficient value at that point is obtained. Then, the difference between the two values ​​is calculated. This difference represents the entropy change deviation at the 23.4% state of charge point. The program then automatically moves to the next feature point at 45.8% and repeats the exact same operation, searching for the value to be measured. Calculate the baseline value And calculate the difference. This process continues until all points in the characteristic charge state point set have been processed. The program will then calculate all the differences during this process. The values ​​are stored sequentially in a newly created one-dimensional floating-point array to obtain the set of entropy deviation values.

[0057] Based on the entropy change data to be measured and the sequential thermal images in the sequence thermal images, which are multiple three-dimensional matrices acquired and stored during the measurement at each characteristic state of charge point (e.g., {23.4%, 45.8%, 67.2%, 88.1%}), these separately acquired sequential thermal images are first logically stitched together in the time dimension. For example, if 75,000 frames of images are acquired at each point, the four points will form a total of [data missing]. The overall dataset of frame thermal images has a logical structure of a very large three-dimensional matrix with dimensions (width, height, total number of frames), for example (640, 480, 300000). Next, a two-dimensional floating-point matrix with the same size as a single frame thermal image (e.g., 640x480) and all elements set to zero is initialized to store the final average temperature map. Then, a nested loop is started, with the outer loop iterating through all row coordinates of the image. (From 0 to 479), the inner loop iterates through all column coordinates. (From 0 to 639), in each inner loop, for the current pixel coordinates The program will extract all temperature readings of the pixel at all 300,000 time points from the ultra-large 3D matrix, forming a one-dimensional time series vector containing 300,000 temperature values. Then, the arithmetic mean of all temperature values ​​in the vector is calculated, i.e. After the calculation is completed, the average temperature value obtained will be... Assign coordinates to the newly created two-dimensional matrix. The pixels, after the double loop is completed, each pixel value in the newly created two-dimensional matrix represents the time-averaged temperature during the entire multi-point measurement process, generating an average temperature map.

[0058] Based on the average temperature map, this is a two-dimensional matrix, denoted as... Its dimensions are And the set of values ​​whose entropy changes deviate from the set of values, which is a one-dimensional array, denoted as First, initialize a new two-dimensional matrix with the same size as the average temperature map and a floating-point data type. This is used to store the calculated heat flux gradient. Then, each pixel in the average temperature map is traversed through a nested double loop. ,in Row coordinates (from 0 to ), Column coordinates (from 0 to Within the loop, for the current center pixel... The program will access it. ,Down ,Left and right The temperature values ​​of four adjacent pixels are calculated, and the arithmetic mean of these four values ​​is calculated. For pixels on image boundaries, such as pixels in the first row, if there are no upper neighbors, then when calculating the average, only the existing neighbors (left, right, and bottom) are considered. That is, the denominator is 3 instead of 4. After calculating the neighborhood average temperature, the temperature value of the center pixel is used. Subtract the average temperature of the neighborhood to obtain the difference. This difference represents the degree to which the temperature at the center point is salient to its surroundings, i.e., an approximation of the heat flux gradient. Store in a new matrix The corresponding position After traversing all pixels, the matrix This is the heat flux gradient map. Finally, the calculated heat flux gradient map matrix is... The set of values ​​that deviate from the entropy change obtained in the previous step The data is packaged into a composite data structure to obtain entropy change deviation characteristics and heat flow gradient maps.

[0059] The specific steps for obtaining a diagnosis of self-discharge abnormality are as follows:

[0060] Based on the entropy change deviation characteristics and heat flow gradient map, the entropy change deviation values ​​are extracted, each value in the set of entropy change deviation values ​​is squared, and then all the squared values ​​are added together to get the sum. Finally, the square root of the sum is calculated to obtain the entropy change deviation norm.

[0061] Based on the entropy change deviation characteristics and the heat flow gradient map, we traverse each pixel in the heat flow gradient map, find the pixel corresponding to the maximum value, and take the maximum absolute value as the hot spot intensity value.

[0062] Based on the entropy change deviation norm and hotspot intensity value, the entropy change deviation norm is multiplied by a weighting coefficient, and the hotspot intensity value is multiplied by a weighting coefficient to construct an evaluation index. The evaluation index is then compared with a preset diagnostic threshold to obtain a self-discharge abnormality diagnosis conclusion.

[0063] Specifically, based on the entropy change deviation feature and the heat flux gradient map, which is a composite data structure containing the heat flux gradient map and the entropy change deviation numerical set, the entropy change deviation numerical set is first extracted from this data structure. This set is a one-dimensional array whose elements are the entropy change deviation values ​​calculated at each characteristic state of charge point, for example... ,in This represents the total number of characteristic charge state points. Then, each value in this array is squared to obtain a new array of squared values. Then, calculate the sum of all elements in this new array to obtain a scalar value, namely the sum of squares. Finally, the square root operation is performed on this sum, and the formula is as follows: in, This is the desired deviation of the entropy change from the norm. The number of characteristic state-of-charge points. For the first The entropy deviation of a characteristic charged state point is given by the norm, which is numerically equal to the Euclidean length of the multidimensional vector representing the entropy deviation, thus yielding the entropy deviation norm.

[0064] Based on entropy change deviation features and heat flow gradient maps, this data structure includes a heat flow gradient map, which is a two-dimensional matrix corresponding to the pixels on the battery surface. Each element in the matrix This represents the heat flux gradient at that location, and its value can be positive or negative. First, a floating-point variable is initialized to store the largest absolute value found so far, with an initial value of 0. Then, two integer variables are initialized to record the corresponding pixel coordinates. Subsequently, the program traverses the heat flux gradient map matrix using nested loops. Each pixel in In each loop, the gradient value of the current pixel is read. And calculate its absolute value. Then, this calculated absolute value is compared with the currently stored maximum absolute value. If the absolute value of the current pixel is greater than the stored maximum absolute value, the maximum absolute value variable is updated to the current value, and the coordinate variable is updated simultaneously. This process continues until all pixels in the matrix have been checked. After the loop ends, the final value stored in the maximum absolute value variable is the intensity of the most significant location of the heat flow gradient on the entire battery surface, and this maximum absolute value is used as the hot spot intensity value.

[0065] Based on the entropy change deviation norm and the hotspot intensity value, the entropy change deviation norm is first multiplied by its corresponding weighting coefficient. And multiply the hotspot intensity value by its corresponding weighting coefficient. Then, the two products are added together to construct the evaluation index, the calculation formula of which is: ,in As evaluation indicators, For entropy change deviates from norm, This represents the intensity value of the hotspot and its weighting coefficient. and The settings are based on statistical analysis of experimental data from a large number of battery samples (e.g., 100 healthy batteries and 50 batteries known to have varying degrees of self-discharge faults). The aim is to balance the dimensions of the two indicators and reflect their contribution to the diagnostic results. For example, if the analysis finds that the entropy change deviates from the norm in the range of 0 to 0.1 mV / K, and the hotspot intensity in the range of 0 to 2.0 °C, to ensure that their contributions are comparable, a setting can be made. , This ensures that the typical values ​​of both evaluation indicators are within the range of 0 to 1. Subsequently, the calculated evaluation indicators... The results were compared to a preset diagnostic threshold, which was an evaluation metric calculated based on the aforementioned 100 healthy battery samples. The diagnostic threshold is determined by the statistical distribution of the 100 evaluation indicators. Specifically, the method involves calculating the mean and standard deviation of these indicators. For example, if the mean is 0.4 and the standard deviation is 0.1, the diagnostic threshold can be set as the mean plus three times the standard deviation. This setting ensures that the probability of a healthy battery being falsely identified as abnormal is less than 0.15%. Finally, the evaluation metrics of the battery under test are... Compared to this diagnostic threshold of 0.7, if If so, the battery is determined to have an abnormal self-discharge. If so, it is judged as normal, and the diagnosis of abnormal self-discharge is obtained.

[0066] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for diagnosing abnormal battery self-discharge, characterized in that, Includes the following steps: For healthy lithium-ion batteries, in a preset set of battery state of charge points, the batteries are placed in a temperature chamber and the temperature range of 5℃ is scanned at a rate of 0.1℃ / min. The open circuit voltage sequence and temperature sequence are collected simultaneously. The slope of the voltage sequence against the temperature sequence is calculated by linear regression to obtain the entropy change coefficient value of each battery state of charge point. Then, all entropy change coefficient values ​​are connected by spline interpolation to establish a reference entropy change spectrum. Based on the aforementioned baseline entropy change spectral line, all peaks and valleys in the spectral line are located using the first derivative zero-finding method. The peak and valley positions are determined as characteristic battery state of charge points. The entropy change coefficient of the lithium-ion battery under test, which is expected to undergo lithium plating, is repeatedly measured at the characteristic battery state of charge points. At the same time, the battery surface is continuously photographed with an infrared thermal imager at a frequency of 25 frames / second, and the temperature reading of each pixel at each time point is recorded to obtain the entropy change data and the sequence thermal image.

2. The battery self-discharge abnormality diagnosis method according to claim 1, characterized in that, The method further includes: Based on the reference entropy change spectrum and the entropy change data to be measured and the sequence thermal image, the entropy change data to be measured and the reference entropy change spectrum are subtracted at the same battery state of charge point to obtain the entropy change deviation value set. At the same time, the time series temperature value of each pixel coordinate in the sequence thermal image is averaged to obtain the average temperature map. For each pixel in the average temperature map, the average temperature of the four neighboring pixels is calculated to obtain the entropy change deviation feature and the heat flow gradient map. Based on the entropy change deviation characteristics and the heat flow gradient map, the Euclidean norm of the entropy change deviation numerical set is calculated, and the pixel with the largest absolute value in the heat flow gradient map is identified as a hot spot caused by diaphragm damage. The hot spot intensity value is extracted, and the Euclidean norm and the hot spot intensity value are weighted and summed to construct an evaluation index. The evaluation index is compared with a preset diagnostic threshold to obtain a self-discharge abnormality diagnosis conclusion.

3. The battery self-discharge abnormality diagnosis method according to claim 1, characterized in that, The specific steps for obtaining the reference entropy variation spectral line are as follows: For healthy lithium-ion batteries, in a preset set of battery state of charge points, the battery is placed in a temperature chamber and the temperature range of 5°C is scanned at a rate of 0.1°C / min. Discrete open-circuit voltage values ​​and battery temperature values ​​are recorded synchronously at a sampling frequency of 1Hz to form paired time series data pairs and obtain temperature and pressure time series data. Based on the temperature and pressure time series data, for each battery state of charge point, the corresponding open-circuit voltage sequence is used as the dependent variable and the temperature sequence is used as the independent variable. A straight line is fitted by the least squares method, and the slope of the straight line is used as the entropy change coefficient value of the battery state of charge point. The entropy change coefficient values ​​of all battery state of charge points are summarized to generate a discrete entropy change point set. Based on the discrete entropy change point set, piecewise fitting is performed between adjacent entropy change coefficient values, while ensuring the continuity of the first and second derivatives at the connection points, connecting the discrete points into a smooth curve, and establishing a benchmark entropy change spectrum.

4. The battery self-discharge abnormality diagnosis method according to claim 1, characterized in that, The specific steps for acquiring the entropy change data to be measured and the sequence thermal images are as follows: Based on the aforementioned reference entropy-variable spectral line, the gradient value of each point on the spectral line is calculated using the central difference method to form a first derivative sequence. Then, the first derivative sequence is traversed to locate all battery state of charge points where the gradient value changes from positive to negative and from negative to positive, thus obtaining a set of characteristic state of charge points. Based on the set of characteristic state of charge points, the lithium-ion battery under test is placed in a temperature chamber, and a temperature scan is performed at each characteristic state of charge point. The open-circuit voltage and temperature data are recorded. Then, the least squares method is applied to the data at each point to calculate the slope of voltage with respect to temperature, and the discrete entropy change value under test is generated. Based on the discrete entropy change value to be measured, the surface of the battery is continuously photographed at a frequency of 25 frames / second using an infrared thermal imager during the measurement process. The temperature reading of each pixel at each time point is stored in a three-dimensional matrix. The discrete entropy change value to be measured is paired with the three-dimensional matrix to obtain the entropy change data to be measured and the sequence thermal image.

5. The battery self-discharge abnormality diagnosis method according to claim 2, characterized in that, The specific steps for obtaining the entropy change deviation feature and the heat flux gradient map are as follows: Based on the reference entropy change spectrum line and the entropy change data to be measured and the entropy change data to be measured in the sequence thermal image, the set of characteristic charge state points is traversed. At each point, the value of the entropy change data to be measured is subtracted from the value of the corresponding point on the reference entropy change spectrum line, and all the differences are stored in a new array to obtain the set of entropy change deviation values. Based on the entropy change data to be measured and the sequence thermal images in the sequence thermal images, the temperature value of each pixel coordinate in the image is extracted at all time points, the average of the temperature values ​​is calculated, and the average is assigned to the pixel point with the same coordinate in the new image to generate an average temperature map.

6. The battery self-discharge abnormality diagnosis method according to claim 5, characterized in that, The steps for obtaining the entropy change deviation feature and the heat flux gradient map also include: Based on the average temperature map and the entropy change deviation value set, each pixel of the average temperature map is traversed, the average temperature of the four adjacent pixels (top, bottom, left, and right) is calculated, and the temperature value of the center pixel is subtracted from the average temperature value to obtain the difference. All the differences are used to form a new image to obtain the entropy change deviation feature and heat flow gradient map.

7. The battery self-discharge abnormality diagnosis method according to claim 2, characterized in that, The specific steps for obtaining the self-discharge abnormality diagnosis conclusion are as follows: Based on the entropy deviation characteristics and heat flow gradient map, the entropy deviation values ​​are extracted, each value in the set of entropy deviation values ​​is squared, and then all the squared values ​​are added together to get the sum. Finally, the square root of the sum is calculated to obtain the entropy deviation norm. Based on the entropy change deviation feature and the heat flow gradient map, each pixel in the heat flow gradient map is traversed to find the pixel corresponding to the maximum value, and the maximum absolute value is taken as the hot spot intensity value.

8. The battery self-discharge abnormality diagnosis method according to claim 7, characterized in that, The steps for obtaining the self-discharge abnormality diagnostic conclusion also include: Based on the entropy change deviation norm and the hotspot intensity value, the entropy change deviation norm is multiplied by a weighting coefficient, and the hotspot intensity value is multiplied by a weighting coefficient to construct an evaluation index. The evaluation index is then compared with a preset diagnostic threshold to obtain a self-discharge abnormality diagnosis conclusion.

Images