Battery liquid leakage detection method and detection system

Abstract

The invention discloses a battery liquid leakage detection method and detection system, and belongs to the technical field of batteries. The battery liquid leakage detection method comprises the following steps: step 1, acquiring impedance information of a battery in real time, generating an ELS curve of the battery, and extracting leakage information of the battery based on the ELS curve; 2, the leakage information is preprocessed, and leakage characteristics in different frequency states are decomposed from the leakage information; 3, pre-training a prediction model; and 4, inputting the leakage characteristics into the prediction model to generate the leakage probability of the battery. In the technical scheme provided by the invention, when the leakage of the battery is detected, the impedance information of the battery is mainly collected. The impedance information of the battery can be directly acquired by measuring the voltage and the current, and the acquisition mode of the impedance information is simple and the acquisition precision is high. And whether the battery leaks liquid or not can be judged without obvious liquid leakage of the battery, so that the safety of the battery in use can be improved.

Description

Technical Field

[0001] The present application relates to the field of battery technology, and in particular to a battery leakage detection method and detection system. Background Art

[0002] Lithium batteries have high voltage, high energy density, and long charge and discharge life. They are an excellent energy storage body. They are widely used in many fields. Due to the high energy density of lithium batteries, safety has always been an important issue in the development of lithium batteries. There is a large amount of electrolyte in lithium batteries. When the electrolyte leaks, it is easy to cause a fire. After a single lithium battery catches fire, it will quickly ignite other lithium batteries, causing the lithium battery stack to explode.

[0003] In order to improve the safety performance of lithium batteries, corresponding sensors are installed when using lithium batteries to monitor whether there is leakage of lithium battery electrolyte. However, this monitoring method is less timely. The use of liquid sensors to determine whether there is leakage requires a large amount of electrolyte to overflow to the effective monitoring position before the alarm can be triggered. Therefore, the early warning capability is quite limited. When the electrolyte leakage is detected, the battery has already leaked significantly and has the conditions for fire.

[0004] In addition, gas detectors can also be used to detect products produced after electrolyte leakage, such as carbon dioxide, carbon monoxide, hydrogen, etc., to reflect whether the electrolyte has leaked. Unlike using liquid sensors to detect whether electrolyte leakage has occurred, using gas detectors does not require the electrolyte to overflow to an effective detection position, but still requires the electrolyte leakage to reach a certain level.

[0005] In summary, the above two methods of electrolyte leakage detection are both limited by the leakage amount of the electrolyte, and cannot issue a timely warning when a small leakage of the electrolyte occurs. Summary of the invention

[0006] The content of this application is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this application is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0007] As a first aspect of the present application, in order to solve the technical problems mentioned in the above background technology section, some embodiments of the present application provide a battery leakage detection method, comprising the following steps:

[0008] Step 1: Obtain the impedance information of the battery in real time, generate the ELS curve of the battery, and extract the leakage information of the battery based on the ELS curve;

[0009] Step 2: Preprocess the leakage information to decompose the leakage features under different frequency states;

[0010] Step 3: Pre-train the prediction model;

[0011] Step 4: Input the leakage characteristics into the prediction model to generate the leakage probability of the battery.

[0012] In the technical solution provided in the present application, when detecting whether the battery is leaking, the impedance information of the battery is collected. The impedance information of the battery can be directly obtained by measuring the voltage and current, and the method of obtaining the impedance information is relatively simple and the acquisition accuracy is high. After the battery leaks, the impedance information will change due to the loss of electrolyte. Compared with the normally aged battery, the impedance information characteristics of the battery with slow loss of electrolyte are different. Based on this, it is judged whether the battery has a leakage phenomenon. In this way, it is possible to judge whether the battery is leaking without requiring obvious leakage of the battery, thereby improving the safety performance of the battery.

[0013] Step 1 includes the following steps:

[0014] Step 11: Monitor the SOC value of the battery in real time;

[0015] Step 12: When the preset SOC measurement node is reached, an excitation current signal i(t) is sent to the battery, i=Isin(wt+Q i ), where I is the preset current value, t is the time, w represents the angular frequency, Q i is the current parameter;

[0016] Measure the voltage across the battery to get the voltage response signal u(t); u(t) = Usin(wt+Q u ), Q u is the voltage parameter;

[0017] Calculate the impedance diagram based on i(t) and u(t);

[0018] Z=Z re +jZ im ;

[0019] Among them, Z re is the real part of impedance, Z im is the imaginary part of impedance, j is the imaginary unit;

[0020] Step 13: Take Z re is the horizontal axis, Z im Generate an impedance graph of the battery for the vertical axis, and use the impedance graphs at different SOC values ​​as leakage information.

[0021] When using impedance characteristics to determine whether a battery is leaking, it is necessary to ensure the accuracy of impedance measurement. The accuracy of impedance measurement is related to many factors. In order to ensure the detection effect, it is necessary to filter out low-precision data. For this purpose, this application provides the following technical solutions:

[0022] In step 13, the impedance diagram at each SOC value is tested for validity. If the validity of the impedance diagram is low, the impedance diagram at the SOC value is remeasured.

[0023] This application will perform validity checks on the impedance graph, so some inaccurate impedance information can be removed and the detection accuracy can be increased.

[0024] Further, step 13 includes the following steps:

[0025] Step 131: Obtain the impedance diagram to be tested and decompose the impedance diagram into Z re (w) and Z im (w), Z im (w) is the imaginary part of the impedance at angular frequency w, Z re (w) is the real part of the impedance at the angular frequency;

[0026] Step 132: Based on Z im (w) Calculate the theoretical Z re ˋ(w);

[0027]

[0028] Based on Z re (w) Calculate the theoretical Z im `(w);

[0029]

[0030] Step 133: Based on Theory Z im `(w) and Z im (w) Error and Z re (w) and Z re `(w) The error screening impedance diagram detection results are obtained to obtain the impedance diagram at different frequencies.

[0031] In the technical solution provided in the present application, the intrinsic connection between the imaginary part and the real part of the impedance, that is, the causal response between the impedance spectrum and the current signal, and the specific relationship between the imaginary part and the real part of the impedance spectrum are utilized to verify whether the measured impedance is accurate, thereby increasing the accuracy of impedance information collection.

[0032] Impedance information mainly reflects the capacitance characteristics of the battery, that is, the change of impedance at different frequencies, and reflects the impedance characteristics of the battery under high frequency conditions. In practice, it cannot reflect the changing state of the electrochemical reaction inside the battery, so it cannot reflect the characteristics of the electrochemical changes of the battery after leakage.

[0033] The impedance information of the battery will change during the aging process. Although the degree of change is inconsistent with the impedance information when the battery electrolyte leaks, there is a high similarity. The simple impedance information cannot accurately reflect the battery's dynamic process, especially when the impedance information of different frequencies is superimposed together. When analyzing the impedance information, the battery's dynamic process will be lost.

[0034] Step 2 includes the following steps:

[0035] Step 21: Obtain leakage information of the battery, and for each impedance graph in the leakage information, extract the original features of the impedance graph;

[0036] The original features include the starting resistance R1, the arc vertex resistance R2, the inflection point resistance R3, and the low-frequency fitting straight line f(re);

[0037] Among them, the starting resistance R1 is the point where the imaginary part of the impedance diagram starts to rise, the arc vertex resistance R2 is the point where the imaginary part of the impedance diagram starts to decrease, the inflection point resistance R3 is the inflection point where the imaginary part of the impedance diagram starts to rise for the second time, and the low-frequency fitting straight line f(re) is the fitting straight line of the subsequent part of the inflection point resistance in the impedance diagram.

[0038] Step 22: For the impedance diagrams of different frequencies at each SOC value, extract the frequency feature Z(w);

[0039] τ = lgj;

[0040] Where w is the angular frequency, t is the relaxation time, j is the imaginary unit, R1 is the starting resistance, g(τ) is the relaxation time distribution function, and lg is the logarithm to the base 10;

[0041] Step 23: Extract peak points V1, V2, V3, and V4 from Z(w);

[0042] Step 24: Take V1, V2, V3, V4, the starting resistance R1, the arc vertex resistance R2, the inflection point resistance R3 and the low-frequency fitting straight line f(re) as leakage features.

[0043] In the technical solution provided by the present application, in order to allow the leakage characteristics to accurately reflect the electrochemical characteristics inside the battery, the impedance information in the frequency domain is converted into the relaxation time distribution in the time domain. This conversion process is based on the assumption that the electrochemical model is a series connection of an ohmic resistor and a polarization impedance, and the polarization impedance is expressed as a continuous series RC parallel circuit. A single ideal polarization process is generally expressed by a parallel resistor and capacitor, corresponding to (relaxation time) τ = RC. In practice, the capacitor C is generally represented by a constant phase angle element, which is represented by a relaxation time distribution function g(τ) centered on the characteristic time constant τ in the relaxation time distribution diagram. This modeling analysis can effectively separate the overlapping processes of the battery dynamics process in the frequency spectrum through time domain conversion and the solution of the relaxation time distribution function, thereby more accurately reflecting the characteristics of the battery dynamics process.

[0044] During the use of batteries, multiple batteries are generally combined together as a battery pack, so in practice, the prediction accuracy of the battery needs to be guaranteed. To this end, in practice, in order to ensure safety, it is inclined to increase the prediction threshold to avoid omissions. When the neural network model predicts whether the battery is leaking, the internal feature transmission process is basically not visible, so when it is used, it will lead to the inability to trust the model's detection results too accurately. To this end, the application provides the following technical solutions:

[0045] Furthermore, step 3 includes the following steps:

[0046] Step 31: Obtain a set of training samples P in advance, each training sample P includes a leakage feature X and a label Y, and the label y includes leakage and non-leakage; construct the original question;

[0047] y i (∈x i +b)>1, or y i (∈x i +b)<-1; i=1, 2…n, n is an integer, n>1, i represents the sample index, ∈ is the normal vector of the hyperplane, b is the intercept of the hyperplane, x i is the leakage feature of the i-th training sample, y i is the set of corresponding category labels;

[0048] Step 32: Introduce Lagrangian function:

[0049] i and j represent the index of the training samples, α, α i , α j They represent Lagrange multipliers, L represents Lagrange function, and K represents kernel function;

[0050] Step 33: Solve the dual problem and find the optimal Lagrange multiplier αi ;

[0051]

[0052] max a W(α) represents the maximization of the objective function W(α) with respect to the Lagrange multiplier α i The value of , W(α) is the objective function;

[0053] The following constraints are met:

[0054]

[0055] 0≤α i ≤C, i=1……n,

[0056] K(x i , x j ) is the kernel function, which is used to calculate the inner product between samples, and C represents the penalty function;

[0057]

[0058] u i is a slack variable, which is used to indicate the extent to which each sample is allowed to violate the interval constraint;

[0059] Step 34: Continuously adjust the Lagrange multiplier α i Until the verification accuracy reaches the maximum value.

[0060] In the technical solution provided by this application, the SVM model is used to design the prediction model. Compared with the neural network model, the decision boundary of the SVM model is easier to understand and explain, so it is easier to adjust the key factors when the SVM model makes decisions in practice, and the decision method of the SVM model can be adjusted in the required direction to try to avoid missed detection problems. In addition, compared with other prediction models, the decision boundary of the SVM model is less affected by abnormal points and noise points, and the model is more stable.

[0061] When training the SVM model, the Lagrange multiplier needs to be adjusted continuously. Each adjustment of the Lagrange multiplier requires multiple trainings. Therefore, when training the SVM model, the training is very difficult. In addition, because the number of features introduced in this solution is too large, the training is difficult. Therefore, in this solution, the following technical solutions are adopted:

[0062] Further, step 34 includes the following steps:

[0063] Step 341: taking all value ranges of Lagrange multipliers as the search space d;

[0064] Set the particle swarm size N;

[0065] Initialize particle position G h ; h represents the index of the particle;

[0066] Set the maximum number of iterations T;

[0067] At the tth iteration, the position vector of the particle is

[0068] Step 342: Continue iterating, and for each iteration:

[0069] Calculate the convergence factor Coefficient vector Swing coefficient vector

[0070]

[0071] For each particle G h , calculate the fitness value f(G h ), sort all the current particles according to the fitness value to obtain the optimal particle L1, the suboptimal particle L2 and the third particle L3;

[0072] Select the position of the current optimal solution as the target position, and calculate the distance vector between each particle and the target position.

[0073] Update particle positions:

[0074]

[0075] in is the coefficient vector, t is the current iteration number, Indicates the current target position, Indicates the current particle position, Represents the distance between the particle and the target position;

[0076] Step 343: After updating several times:

[0077] For each particle, calculate the distance vector between it and the current optimal particle L1, the second-best particle L2, and the third particle L3 respectively

[0078] Update the temporary position of the particle according to the distance coefficient;

[0079]

[0080] Update the temporary position of the particle based on the distance and coefficient vector

[0081]

[0082] in, is the distance between the optimal particle L1, the suboptimal particle L2, the third particle L3 and other particles, are the positions of other particles, They represent the current positions of the optimal particle L1, the suboptimal particle L2, and the third particle L3, respectively. and is the corresponding coefficient vector, Indicates the current particle position;

[0083]

[0084] Step 343: When the maximum number of iterations T is reached, the iteration ends. When the obtained objective function value is lower than a preset value, the iteration ends.

[0085] In this scheme, based on the particle swarm algorithm, the positions of the optimal particle L1, the suboptimal particle L2 and the third particle L3 are introduced as the target positions. When optimizing in the space d of the entire problem dimension, the remaining examples are guided to continuously approach the target position. Therefore, in the entire optimization process, we can start from a global perspective and avoid the balance between the local and the global during optimization, thereby increasing the accuracy and convergence ability of the model.

[0086] When using particle swarm to find the best solution, there is a certain luck factor in iterating to the best solution. is determined by, and the coefficient vector And through the convergence factor Determine. The convergence factor changes linearly during convergence. Therefore, the entire optimization process is actually a linear process, which will lead to low convergence ability of the model during the optimization process. To this end, this application provides the following technical solutions:

[0087] Furthermore,

[0088] In this solution, the convergence factor is modified from linear decrease to nonlinear decrease (exponential decrease), which can increase the diversity of the convergence factor change. Therefore, in the early stage of iteration, the convergence speed of the convergence factor is slow, and it is used to find the global optimal solution. As the iteration proceeds, the convergence factor decreases faster, and the local optimal solution can be better found. For this reason, in this solution, by adjusting the convergence factor, both the local optimal solution and the global optimal solution have sufficient performance.

[0089] As a second aspect of the present application, the present application provides a battery leakage detection system comprising:

[0090] An information collection device, used to collect impedance information of each battery;

[0091] The prediction device predicts the leakage probability of the battery according to the aforementioned battery leakage detection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] The drawings constituting a part of this application are used to provide a further understanding of this application, so that other features, purposes and advantages of this application become more obvious. The illustrative embodiment drawings and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application.

[0093] In addition, throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that the components and elements are not necessarily drawn to scale.

[0094] In the attached picture:

[0095] Figure 1 The figure is a flow chart of a battery leakage detection method.

[0096] Figure 2 The impedance spectrum.

[0097] Figure 3 is the peak position diagram of Z(w).

[0098] Figure 4 for Change curves during linear decrease and exponential decrease. DETAILED DESCRIPTION

[0099] Embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not intended to limit the scope of protection of the present application.

[0100] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0101] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0102] Reference Figure 1 , a battery leakage detection method, comprising the following steps:

[0103] Step 1: Obtain the impedance information of the battery in real time, generate the ELS curve of the battery, and extract the leakage information of the battery based on the ELS curve.

[0104] Step 11: Monitor the SOC value of the battery in real time;

[0105] The SOC value of the battery can be calculated by detecting the battery inflow in the battery. The detection of SOC is a prior art and will not be discussed here.

[0106] Step 12: When the preset SOC measurement node is reached, an excitation current signal i(t) is sent to the battery, i=Isin(wt+Q i ), where I is the preset current value, t is the time, w represents the angular frequency, Q i is the current parameter;

[0107] Measure the voltage across the battery to get the voltage response signal u(t); u(t) = Usin(wt+Q u ), Q u is the voltage parameter;

[0108] Calculate impedance Z based on i(t) and u(t);

[0109] Z=Z re +jZ im ;

[0110] Among them, Z re is the real part of impedance, Z im is the imaginary part of impedance and j is the imaginary unit.

[0111] Step 13: Take Z re is the horizontal axis, Z im Generate the impedance diagram of the battery for the vertical axis, and use the impedance diagrams at different SOC values ​​as leakage information. Figure 2 shown.

[0112] In this solution, the measurement of impedance is related to the frequency of the current. Specifically, when measuring the impedance graph, it is not to obtain a single impedance information at a certain SOC value, but multiple impedance information at multiple SOC values.

[0113] First, the battery is pre-set to different SOC values ​​during a charge or discharge process, such as 10%, 30%, 50%, 60%, and 90%. When the battery reaches these SOC values, impedance information is acquired. Each time the impedance information is acquired, the excitation current signal i(t) generates multiple angular frequencies, so the impedance information finally obtained includes the SOC at the frequencies of different excitation current signals i(t) at multiple SOC values.

[0114] Step 13 includes the following steps:

[0115] Step 131: Obtain the impedance diagram to be tested and decompose the impedance diagram into Z re (w) and Z im (w), Z im (w) is the imaginary part of the impedance at angular frequency w, Z re (w) is the real part of the impedance at the angular frequency;

[0116] Step 132: Based on Z im (w) Calculate the theoretical Z re `(w);

[0117]

[0118] Based on Z re (w) Calculate the theoretical Z im `(w);

[0119]

[0120] Step 133: Based on Theory Z im `(w) and Z im (w) Error and Z re (w) and Z re `(w) The error screening impedance diagram detection results are obtained to obtain the impedance diagram at different frequencies.

[0121] Step 133 mainly involves screening the impedance diagram, mainly using the specific relationship between the real part and the imaginary part in the impedance diagram. If the relationship is met, it means that the impedance calculation is accurate, otherwise it is inaccurate.

[0122] Step 2: Preprocess the leakage information to decompose the leakage features under different frequency states.

[0123] There are relatively many curves for leaked information. Each time the battery is charged, multiple SOC value nodes are recorded. At the same time, multiple current frequencies are used to measure impedance for each node, so there are multiple information curves. Such data is very complicated. For this reason, this application provides the following solutions:

[0124] Step 2 includes the following steps:

[0125] Step 21: Obtain leakage information of the battery, and for each impedance graph in the leakage information, extract the original features of the impedance graph;

[0126] The original features include the starting resistance R1, the arc vertex resistance R2, the inflection point resistance R3, and the low-frequency fitting straight line f(re);

[0127] Among them, the starting resistance R1 is the point where the imaginary part of the impedance diagram starts to rise, the arc vertex resistance R2 is the point where the imaginary part of the impedance diagram starts to decrease, the inflection point resistance R3 is the inflection point where the imaginary part of the impedance diagram starts to rise for the second time, and the low-frequency fitting straight line f(re) is the fitting straight line of the subsequent part of the inflection point resistance in the impedance diagram; f(re) is a fitting straight line, and its fixed item information has been determined by the inflection point, and its slope is mainly recorded during analysis.

[0128] Step 22: For the impedance diagrams of different frequencies at each SOC value, extract the frequency feature Z(w);

[0129] τ = lgj;

[0130] Where w is the angular frequency, t is the relaxation time, j is the imaginary unit, R1 is the starting resistance, g(τ) is the relaxation time distribution function, and lg is the logarithm to the base 10;

[0131] Step 23: Extract peak points V1, V2, V3, and V4 from Z(w).

[0132] like Figure 3 As shown, there are 4 specific peak points in Z(w), and the positions and peak values ​​of the peak points are different in different Z(w), but each Z(w) has 4 peak points.

[0133] Step 24: Take V1, V2, V3, V4, the starting resistance R1, the arc vertex resistance R2, the inflection point resistance R3 and the low-frequency fitting straight line f(re) as leakage features.

[0134] In this scheme, the features in the impedance information are extracted, and the impedance information is frequency-converted, the impedance change characteristics at different frequencies are compared, and the changes of the impedance information at different levels are further extracted.

[0135] Step 3: Pre-train the prediction model.

[0136] When predicting whether there is leakage, this solution uses the SVM model for prediction. Specifically:

[0137] Step 31: Obtain a set of training samples P in advance, each training sample P includes a leakage feature X and a label Y, and the label y includes leakage and non-leakage; construct the original question;

[0138] y i (∈x i +b)>1, or y i (∈x i +b)<-1; i=1, 2…n, n is an integer, n>1, i represents the sample index, ∈ is the normal vector of the hyperplane, b is the intercept of the hyperplane, x iis the leakage feature of the i-th training sample, y i is the set of corresponding category labels;

[0139] Step 32: Introduce Lagrangian function:

[0140] i and j represent the index of the training samples, α, α i , α j They represent Lagrange multipliers, L represents Lagrange function, and K represents kernel function;

[0141] Step 33: Solve the dual problem and find the optimal Lagrange multiplier α i ;

[0142]

[0143] max a W(α) represents the maximization of the objective function W(α) with respect to the Lagrange multiplier α i The value of , W(α) is the objective function;

[0144] The following constraints are met:

[0145]

[0146] 0≤α i ≤C, i=1……n,

[0147] K(x i , x j ) is the kernel function, which is used to calculate the inner product between samples, and C represents the penalty function;

[0148]

[0149] u i is a slack variable, which is used to indicate the extent to which each sample is allowed to violate the interval constraint;

[0150] Step 34: Continuously adjust the Lagrange multiplier α i Until the verification accuracy reaches the maximum value.

[0151] Step 34 includes the following steps:

[0152] Step 341: taking all value ranges of Lagrange multipliers as the search space d;

[0153] Set the particle swarm size N;

[0154] Initialize particle position G h ; h represents the index of the particle;

[0155] Set the maximum number of iterations T;

[0156] At the tth iteration, the position vector of the particle is

[0157] Step 342: Continue iterating, and for each iteration:

[0158] Calculate the convergence factor Coefficient vector Swing coefficient vector

[0159]

[0160] For each particle G h , calculate the fitness value f(G h ), sort all the current particles according to the fitness value to obtain the optimal particle L1, the suboptimal particle L2 and the third particle L3;

[0161] Select the position of the current optimal solution as the target position, and calculate the distance vector between each particle and the target position.

[0162] Update particle positions:

[0163]

[0164] in is the coefficient vector, t is the current iteration number, Indicates the current target position, Indicates the current particle position, Represents the distance between the particle and the target position;

[0165] Step 343: After updating several times:

[0166] For each particle, calculate the distance vector between it and the current optimal particle L1, the second-best particle L2, and the third particle L3 respectively

[0167] Update the temporary position of the particle according to the distance coefficient;

[0168]

[0169] Update the temporary position of the particle based on the distance and coefficient vector

[0170]

[0171] in, is the distance between the optimal particle L1, the suboptimal particle L2, the third particle L3 and other particles, are the positions of other particles, They represent the current positions of the optimal particle L1, the suboptimal particle L2, and the third particle L3, respectively. and is the corresponding coefficient vector, Indicates the current particle position;

[0172]

[0173] Step 343: When the maximum number of iterations T is reached, the iteration ends. When the obtained objective function value is lower than a preset value, the iteration ends.

[0174] In some embodiments,

[0175] Specifically, in the aforementioned factual manner, It is a linear change during iteration, and in this scheme, As the iteration progresses, the number of words will change. When the number of words changes, in the initial iteration, The rate of change is small, which makes it easy to optimize from a global perspective to avoid falling into a local optimal solution. In subsequent iterations, The greater the rate of change of , the better the local consideration can be, which increases the convergence rate. Figure 4 shown.

[0176] Step 4: Input the leakage characteristics into the prediction model to generate the leakage probability of the battery.

[0177] In step 3, the optimization process of the prediction model is provided, and in step 2, the preprocessing method of the leakage features is provided. Therefore, in practice, it is only necessary to collect the leakage features and then input the leakage features into the prediction model to generate the leakage probability of the battery.

[0178] Embodiment 2: A battery leakage detection system includes an information collection device and a prediction device. The information collection device is used to collect impedance information of each battery. The prediction device predicts the battery leakage probability based on the above-mentioned battery leakage detection method.

[0179] The information collection device is mainly used to obtain the impedance information of the battery. Generally speaking, in order to ensure that the impedance information can be obtained, a current generator and a voltage detector are required, the current generator is used to generate an alternating current signal, and the voltage detector is used to detect the voltage at both ends of the battery. The prediction device has a built-in prediction model, which uses the method in Example 1 to perform electrolyte leakage detection.

[0180] The above description is only some preferred embodiments of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present application is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with (but not limited to) the technical features with similar functions disclosed in the embodiments of the present application to form a technical solution.

Claims

1. A battery leakage detection method, characterized in that: include: Step 1: Obtain the impedance information of the battery in real time, generate the ELS curve of the battery, and extract the leakage information of the battery based on the ELS curve; Step 2: Preprocess the leakage information to decompose the leakage features under different frequency states; Step 3: Pre-train the prediction model; Step 4: Input the leakage characteristics into the prediction model to generate the leakage probability of the battery.

2. The battery leakage detection method according to claim 1, characterized in that: Step 1 includes the following steps: Step 11: Monitor the SOC value of the battery in real time; Step 12: When the preset SOC measurement node is reached, an excitation current signal i(t) is sent to the battery, i=Isin(wt+Q i ), where I is the preset current value, t is the time, w represents the angular frequency, Q i is the current parameter; Measure the voltage across the battery to get the voltage response signal u(t); u(t) = Usin(wt+Q u ), Q u is the voltage parameter; Calculate impedance Z based on i(t) and u(t); Z=Z re +jZ im ; Among them, Z re is the real part of impedance, Z im is the imaginary part of impedance, j is the imaginary unit; Step 13: Take Z re is the horizontal axis, Z im Generate an impedance graph of the battery for the vertical axis, and use the impedance graphs at different SOC values ​​as leakage information.

3. The battery leakage detection method according to claim 2, characterized in that: In step 13, the impedance diagram at each SOC value is tested for validity. If the validity of the impedance diagram is low, the impedance diagram at the SOC value is remeasured.

4. The battery leakage detection method according to claim 3, characterized in that: Step 13 includes the following steps: Step 131: Obtain the impedance diagram to be tested and decompose the impedance diagram into Z re (w) and Z im (w), Z im (w) is the imaginary part of the impedance at angular frequency w, Z re (w) is the real part of the impedance at the angular frequency; Step 132: Based on Z im (w) Calculate the theoretical Z re `(w); Based on Z re (w) Calculate the theoretical Z im `(w); Step 133: Based on Theory Z im `(w) and Z im (w) Error and Z re (w) and Z re `(w) The error screening impedance diagram detection results are obtained to obtain the impedance diagram at different frequencies.

5. The battery leakage detection method according to claim 3, characterized in that: Step 2 includes the following steps: Step 21: Obtain leakage information of the battery, and for each impedance graph in the leakage information, extract the original features of the impedance graph; The original features include the starting resistance R1, the arc vertex resistance R2, the inflection point resistance R3, and the low-frequency fitting straight line f(re); Among them, the starting resistance R1 is the point where the imaginary part of the impedance diagram starts to rise, the arc vertex resistance R2 is the point where the imaginary part of the impedance diagram starts to decrease, the inflection point resistance R3 is the inflection point where the imaginary part of the impedance diagram starts to rise for the second time, and the low-frequency fitting straight line f(re) is the fitting straight line of the subsequent part of the inflection point resistance in the impedance diagram; Step 22: For the impedance diagrams of different frequencies at each SOC value, extract the frequency feature Z(w); τ=lgj; Where w is the angular frequency, τ is the relaxation time, j is the imaginary unit, R1 is the starting resistance, g(τ) is the relaxation time distribution function, and lg is the logarithm to the base 10; Step 23: Extract peak points V1, V2, V3, and V4 from Z(w); Step 24: Take V1, V2, V3, V4, the starting resistance R1, the arc vertex resistance R2, the inflection point resistance R3 and the low-frequency fitting straight line f(re) as leakage features.

6. The battery leakage detection method according to claim 2, characterized in that: Step 3 includes the following steps: Step 31: Obtain a set of training samples P in advance, each training sample P includes a leakage feature X and a label Y, and the label y includes leakage and non-leakage; construct the original question; y i (∈x i +b)>1, or y i (∈x i +b)<-1; i=1, 2…n, n is an integer, n>1, i represents the sample index, ∈ is the normal vector of the hyperplane, b is the intercept of the hyperplane, x i is the leakage feature of the i-th training sample, y i is the set of corresponding category labels; Step 32: Introduce Lagrangian function: i and j represent the index of the training samples, α, α i , α j They represent Lagrange multipliers, L represents Lagrange function, and K represents kernel function; Step 33: Solve the dual problem and find the optimal Lagrange multiplier α i ; max a W(α) represents the maximization of the objective function W(α) with respect to the Lagrange multiplier α i The value of , W(α) is the objective function; The following constraints are met: 0≤α i ≤C,i=1……n, K(x i , x j ) is the kernel function, which is used to calculate the inner product between samples, and C represents the penalty function; u i is a slack variable, which is used to indicate the extent to which each sample is allowed to violate the interval constraint; Step 34: Continuously adjust the Lagrange multiplier α i Until the verification accuracy reaches the maximum value.

7. The battery leakage detection method according to claim 6, characterized in that: Step 34 includes the following steps: Step 341: taking all value ranges of Lagrange multipliers as the search space d; Set the particle swarm size N; Initialize particle position G h ; h represents the index of the particle; Set the maximum number of iterations T; At the tth iteration, the position vector of the particle is Step 342: Continue iterating, and for each iteration: Calculate the convergence factor Coefficient vector Swing coefficient vector For each particle G h , calculate the fitness value f(G h ), sort all the current particles according to the fitness value to obtain the optimal particle L1, the suboptimal particle L2 and the third particle L3; Select the position of the current optimal solution as the target position, and calculate the distance vector between each particle and the target position. Update particle positions: in is the coefficient vector, t is the current iteration number, Indicates the current target position, Indicates the current particle position, Represents the distance between the particle and the target position; Step 343: After updating several times: For each particle, calculate the distance vector between it and the current optimal particle L1, the second-best particle L2, and the third particle L3 respectively. Update the temporary position of the particle according to the distance coefficient; Update the temporary position of the particle based on the distance and coefficient vector in, is the distance between the optimal particle L1, the suboptimal particle L2, the third particle L3 and other particles, are the positions of other particles, They represent the current positions of the optimal particle L1, the suboptimal particle L2, and the third particle L3, respectively. and is the corresponding coefficient vector, Indicates the current particle position; Step 343: When the maximum number of iterations T is reached, the iteration ends. When the obtained objective function value is lower than a preset value, the iteration ends.

8. The battery leakage detection method according to claim 1, characterized in that:

9. A battery leakage detection system, characterized in that: include: An information collection device, used to collect impedance information of each battery; A prediction device generates a battery leakage probability according to the battery leakage detection method according to any one of claims 1 to 8.

Images