Lithium ion single battery monitoring system and method based on weak magnetic detection technology

Abstract

The application provides a lithium ion single battery monitoring system and method based on a weak magnetic detection technology, comprising: a multi-channel magnetic field measurement module, which is used for measuring magnetic signal data of a lithium ion single battery through a distributed magnetic flux gate sensor array; a magnetic signal acquisition module, which is used for acquiring the magnetic signal data generated by the lithium ion single battery measured by the multi-channel magnetic field measurement module; and a signal processing module, which is used for converting the preprocessed magnetic signal data into a curve graph and completing classification evaluation of a health state of the lithium ion single battery in combination with a battery health state evaluation model. The technical scheme of the application has an accurate classification and identification rate of 87.5% for lithium ion pilot or small test batteries in different health states and before and after different working conditions, has application space in the fields of electric vehicles and the like, and has important significance for guaranteeing safe operation of equipment.

Description

Technical Field

[0001] This invention belongs to the field of battery health status monitoring technology, specifically relating to a lithium-ion single-cell battery monitoring system and method based on weak magnetic field detection technology. Background Technology

[0002] Currently, the main magnetic correlation monitoring methods for single-cell lithium-ion batteries include magnetic resonance (MR) technology and magnetic field measurement technology. MR technology includes nuclear magnetic resonance (NMR), electron paramagnetic resonance (EPR), and magnetic resonance imaging (MRI). NMR is a spectroscopic analysis method based on the interaction between atomic nuclear magnetic moments and an applied magnetic field, capable of quantitatively characterizing the changes in chemical substances during lithium-ion battery cycling. EPR is a spectroscopic analysis method based on the interaction between unpaired electrons and an applied magnetic field, capable of detecting free radicals, transition metal ions, and other paramagnetic substances in materials, thus applicable to monitoring and researching lithium-ion battery materials, reaction mechanisms, and failure mechanisms. MRI is a non-destructive imaging method based on the principle of NMR, providing high-resolution spatial information for visually monitoring the internal microstructure and dynamic processes of lithium-ion batteries. Magnetic field measurement technology, which has emerged in recent years and is applied to lithium-ion batteries, works by using magnetic sensors to measure the external magnetic field of the battery and analyzing changes in the external magnetic field to assess the battery's performance status.

[0003] In recent years, magnetic correlation monitoring (MRM) technology has emerged, providing an innovative solution for lithium-ion battery state monitoring. Compared to traditional detection methods, this technology has significant advantages: firstly, it can achieve multi-dimensional real-time monitoring using low-cost equipment; secondly, relying on the visualization of magnetic field distribution, it can intuitively reflect the internal current distribution characteristics of the battery; and thirdly, the spatial arrangement of the magnetic sensor array can be flexibly configured according to the battery system characteristics and monitoring requirements. However, current research still faces some challenges:

[0004] Current technologies for dynamic monitoring of battery charge-discharge cycles lack statistically significant repeatable experimental data and have not conducted repeated full charge-discharge cycle studies on batteries. Furthermore, in-depth analysis of individual magnetic induction intensity curves is lacking. While single-channel magnetic induction intensity curves, as direct representations of electrochemical processes, exhibit strong correlations between their dynamic evolution and the coupling effects of multiple physical fields within the battery, their related time-frequency domain characteristics have not been fully analyzed. Current state assessment based on magnetic signals has not yet established a diagnostic model that integrates electromagnetic time-domain characteristics with battery health status. There is a lack of dedicated extraction algorithms for the time-domain characteristics of magnetic induction curves, as well as feature classification models based on machine learning algorithms. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a lithium-ion single-cell battery monitoring system and method based on weak magnetic detection technology. This system performs in-situ non-destructive weak magnetic monitoring of the complete charge-discharge cycle of a single lithium-ion battery and classifies and evaluates the health status of the lithium-ion single-cell battery based on the magnetic induction data curve.

[0006] A lithium-ion single-cell battery monitoring system based on weak magnetic field detection technology includes:

[0007] A multi-channel magnetic field measurement module is used to measure the magnetic signal data of lithium-ion single cells through a distributed fluxgate sensor array;

[0008] The magnetic signal acquisition module is used to acquire magnetic signal data generated by lithium-ion single-cell batteries as measured by the multi-channel magnetic field measurement module.

[0009] The signal processing module is used to convert the preprocessed magnetic signal data into a curve and combine it with the battery health status assessment model to complete the classification and assessment of the health status of lithium-ion single cells.

[0010] Preferably, the multi-channel magnetic field measurement module uses a distributed fluxgate sensor array whose detection sensitive direction is perpendicular to the cylindrical end face, and the actual measured magnetic signal data are the external induced magnetic field signals at different positions on the surface of the lithium-ion single battery cell at the same time during the charge and discharge cycle.

[0011] Preferably, the multi-channel magnetic field measurement module includes:

[0012] The influencing factor acquisition unit is used to obtain the influencing factors of the external induced magnetic field of a lithium-ion single cell based on Maxwell's equations, magnetic permeability, and the electrochemical reaction principle of a lithium-ion single cell; wherein, the influencing factors include the internal current of the lithium-ion single cell and the magnetic susceptibility of the lithium-ion single cell material, and the magnetic susceptibility is an interference signal;

[0013] The magnetic field change acquisition unit is used to obtain the change of the magnetic field on the surface of the lithium-ion single cell based on the internal current of the lithium-ion single cell.

[0014] The monitoring location acquisition unit is used to scan the external induced magnetic field of the lithium-ion single cell during the charge and discharge cycle using a distributed fluxgate sensor array, and obtain the location where the change of the magnetic field on the surface of the lithium-ion single cell meets the preset conditions and is far away from the interference signal as the monitoring location.

[0015] Preferably, the signal processing module includes:

[0016] The signal preprocessing unit is used to select 100 magnetic signal data within one second, remove the highest and lowest values, and calculate the average value of the remaining magnetic signal data as the magnetic field monitoring data value for this second.

[0017] The curve generation unit is used to generate a magnetic field change data curve based on the magnetic field monitoring data value and the corresponding time data.

[0018] The feature extraction unit is used to extract the piecewise time-domain magnetic feature values ​​of the magnetic field change data curve;

[0019] The model building unit is used to construct a battery health status assessment model by combining the segmented time-domain magnetic feature values ​​with the k-NN algorithm.

[0020] Preferably, the model building unit includes:

[0021] A dataset construction subunit is used to divide the segmented time-domain magnetic feature values ​​into a training dataset and a test dataset; wherein, the training dataset consists of segmented time-domain magnetic feature values ​​that label the health status categories of lithium-ion single-cell batteries; and the test dataset consists of segmented time-domain magnetic feature values ​​to be evaluated.

[0022] The k-value determination sub-unit is used to determine the k-value for the k-NN algorithm through cross-validation.

[0023] The Euclidean distance calculation subunit is used to calculate the Euclidean distance between the piecewise temporal magnetic eigenvalues ​​in the test dataset and the training dataset using the k-NN algorithm with a determined k value.

[0024] The neighbor sample acquisition subunit is used to obtain neighbor sample data points in the training dataset that meet the similarity threshold with the test dataset based on the Euclidean distance.

[0025] The health status assessment subunit is used to obtain the health status assessment result of the lithium-ion single cell battery in the test dataset based on the health status category of the neighboring sample data points.

[0026] Preferably, the improvement to the k-NN algorithm includes introducing an automatic shuffling layer and a data normalization transpose layer, and adding a confusion matrix to evaluate the health status assessment results of lithium-ion single-cell batteries.

[0027] This invention also provides a method for monitoring lithium-ion single-cell batteries based on weak magnetic field detection technology, and the system described above includes:

[0028] Magnetic signal data of lithium-ion single cells are measured using a distributed fluxgate sensor array.

[0029] The magnetic signal data generated by a single lithium-ion battery cell was acquired by a multi-channel magnetic field measurement module.

[0030] The preprocessed magnetic signal data is converted into a curve, and the health status of lithium-ion single cells is classified and evaluated by combining the battery health status assessment model.

[0031] Preferably, the method for obtaining the monitoring position of the distributed fluxgate sensor array includes:

[0032] Based on Maxwell's equations, magnetic permeability, and the electrochemical reaction principle of lithium-ion cells, the influencing factors of the external induced magnetic field of lithium-ion cells are obtained; among them, the influencing factors include the internal current of the lithium-ion cell and the magnetic susceptibility of the lithium-ion cell material, wherein the magnetic susceptibility is an interference signal.

[0033] Based on the internal current of the lithium-ion single cell, the change in the surface magnetic field of the lithium-ion single cell is obtained.

[0034] A distributed fluxgate sensor array is used to scan the external induced magnetic field of a lithium-ion cell during the charge-discharge cycle. The location where the change in the magnetic field on the surface of the lithium-ion cell meets the preset conditions and is far away from the interference signal is used as the monitoring location.

[0035] Compared with existing technologies, the beneficial effects of this invention are as follows: By designing a specially packaged magnetic detection sensor device, this invention can achieve weak magnetic non-destructive monitoring of single lithium-ion batteries with different shapes, sizes, and requirements during charge-discharge cycles. Based on the magnetic field monitoring data, a rough judgment can be made on the health status of lithium-ion batteries. Combining magnetic feature value extraction methods with the k-NN algorithm model, accurate classification and evaluation of the health status of lithium-ion batteries can be achieved. As a supplement to existing monitoring technologies, it can achieve non-destructive in-situ monitoring of single lithium-ion batteries during complete charge-discharge cycles. Combined with feature value extraction methods and the k-NN algorithm model, the accuracy rate of classifying and identifying lithium-ion batteries under different health states and before and after different operating conditions reaches 87.5%. This invention has application potential in fields such as electric vehicles and is of great significance for ensuring the safe operation of equipment. Attached Figure Description

[0036] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a schematic diagram of the lithium-ion single-cell battery monitoring system based on weak magnetic field detection technology according to an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram of the components of a lithium-ion single-cell battery monitoring system based on weak magnetic field detection technology according to an embodiment of the present invention;

[0039] Figure 3This is a schematic diagram of the probe in an embodiment of the present invention;

[0040] Figure 4 This is a schematic diagram of the packaging according to an embodiment of the present invention;

[0041] Figure 5 This is a typical magnetic anomaly diagram according to an embodiment of the present invention;

[0042] Figure 6 This is a schematic diagram of the scanning direction in an embodiment of the present invention;

[0043] Figure 7 This is a monitoring diagram according to an embodiment of the present invention;

[0044] Figure 8 This is a schematic diagram of magnetic monitoring data curves according to an embodiment of the present invention;

[0045] Figure 9 The following are the magnetic monitoring data results of battery No. 16 in Embodiment 1 of the present invention; wherein, (a) is the pilot-scale weak magnetic monitoring result diagram, and (b) is the small-scale weak magnetic monitoring result diagram;

[0046] Figure 10 The figures show the electrochemical monitoring data results of the pilot and small-scale batteries in Example 16 of this invention; where (a) is the electrochemical monitoring results during the charging stage and (b) is the electrochemical monitoring results during the discharging stage.

[0047] Figure 11 The graph shows the performance evaluation results of the training dataset in this embodiment of the invention; where (a) is a line graph of the prediction results and (b) is the confusion matrix.

[0048] Figure 12 The graph shows the performance evaluation results of the test dataset in this embodiment of the invention; where (a) is a line graph of the prediction results and (b) is the confusion matrix. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] Example 1

[0052] like Figure 1 , Figure 2As shown, the lithium-ion single-cell battery monitoring system based on weak magnetic field detection technology includes: a multi-channel magnetic field measurement module, a magnetic signal acquisition module, and a signal processing module (host computer); the host computer and the magnetic signal acquisition module are connected via Ethernet.

[0053] The multi-channel magnetic field measurement module is used to measure the magnetic signal data of lithium-ion single cells through a distributed fluxgate sensor array.

[0054] A further implementation involves a distributed fluxgate sensor array used in the multi-channel magnetic field measurement module. The detection sensitivity direction of this array is perpendicular to the cylindrical end face. The actual measured magnetic signal data represents the external induced magnetic field signals at different locations on the surface of a single lithium-ion battery cell at the same time during charge-discharge cycles. This allows for the simultaneous acquisition of magnetic field signals from multiple sensors. Depending on the size and shape of the battery being measured and the monitoring requirements, fluxgate sensors with different numbers of channels can be selected. The structure of the multi-channel fluxgate sensor is as follows: Figure 3 As shown.

[0055] Based on the location and structure of the actual weld seam being tested, a certain length of cable is reserved for connecting the sensor array and the magnetic signal acquisition board module.

[0056] This system also includes a probe fixture for encapsulating and securing the sensor, which is encapsulated within the fixture. The encapsulation is made of a nylon material with very weak magnetic and electrical conductivity to ensure no impact on the monitored magnetic signal. The fixture structure is as follows: Figure 4 As shown. Screws are used to secure the sensor to the outside of the fixture housing to prevent the probe's packaging position or magnetism and conductivity from affecting the monitoring results.

[0057] A further embodiment of the invention includes a multi-channel magnetic field measurement module comprising:

[0058] The influencing factor acquisition unit is used to obtain the influencing factors of the external induced magnetic field of a lithium-ion cell based on Maxwell's equations, magnetic permeability, and the electrochemical reaction principle of lithium-ion cells. These influencing factors include the internal current of the lithium-ion cell and the magnetic susceptibility of the lithium-ion cell material, with magnetic susceptibility serving as an interference signal. Specifically, the expression for the current law in Maxwell's equations is as follows:

[0059]

[0060] In the formula, the left side ∮ l H·dl is the line integral of the magnetic field along the closed path l, representing the circulation of the magnetic field along the closed path, i.e., the induced magnetic field generated by the current inside the battery in space. Here, H represents the magnetic field strength, l represents the closed loop (current path), and dl represents the displacement vector. The first term on the right-hand side, ∫∫... sJ·ds represents the conduction current, J represents the current density, s is an arbitrary surface enclosed by l, and ds represents the area infinitesimal vector; the second term on the right-hand side of the equation... This represents the displacement current that varies with time. ds represents the reciprocal of the electric displacement field D with respect to time t, i.e., the density of the displacement current, and ds also represents the area infinitesimal vector.

[0061] Positive electrode electrochemical reaction formula:

[0062]

[0063] Negative electrode electrochemical reaction formula:

[0064]

[0065] Overall reaction formula:

[0066]

[0067] The relationship between the induced magnetic field and the current in a circuit: ∮ l H·dl=∫∫ s J·ds

[0068] Relationship between changes in magnetic field and changes in magnetic permeability:

[0069] B=μ0(1+χ m )H

[0070] μ0 — Vacuum permeability.

[0071] χ m —Magnetic susceptibility represents the ability of a material to be magnetized by an external magnetic field.

[0072] H represents the magnetic field strength.

[0073] B—represents magnetic flux density.

[0074] The magnetic field change acquisition unit is used to obtain the change of the magnetic field on the surface of the lithium-ion single cell based on the internal current of the lithium-ion single cell.

[0075] The monitoring location acquisition unit is used to scan the external induced magnetic field of a lithium-ion cell during the charge-discharge cycle using a distributed fluxgate sensor array. The monitoring location is defined as the position where the magnetic field change on the battery surface meets preset conditions and is far from interference signals. The preset conditions are the position where the magnetic field change on the battery surface is most significant and stable.

[0076] like Figure 6The following monitoring points were ultimately selected: Monitoring points should avoid the battery edges (battery edges are in contact with air and may have more complex designs and physical structures; for example, edges may contain connection points, contact points, and casing, which can lead to irregular or restricted current flow. Therefore, the current distribution at the edges often differs from that in the battery center, resulting in more unstable changes in the edge magnetic field); avoid battery electrodes (changes in the magnetic susceptibility of the electrode material during charging and discharging can cause significant changes in the battery's induced magnetic field, interfering with the monitoring results); and cover the battery as much as possible (to avoid invalid monitoring results due to problems with a single probe). The placement method can be designed according to the shape and size of the battery being monitored and the monitoring requirements. For example, based on the size and shape of the battery being monitored, a three-channel probe is selected. After the probe is inserted into the casing and fixed with screws, the casing is placed parallel to the front surface of the battery, tightly fitted to the center of the battery, and then fixed in place. Figure 7 As shown.

[0077] The weak magnetic field monitoring technology used in this invention is based on magnetic field signals, i.e., the physical quantity of magnetic induction intensity (unit: Tesla, T; in actual measurement, it is generally nT, 1 nT = 10⁻⁶). -9 T) measurement is a non-destructive monitoring technology that does not require an excitation field. When the intrinsic magnetic properties of the monitored object (such as magnetic susceptibility and coercivity) change, when internal defects exist, or when parameters within the object that cause changes in the magnetic field (such as voltage, current, and temperature) change, magnetic anomaly signals are generated. Figure 5As shown. Magnetic anomaly is defined as an upward or downward bulge in the magnetic induction intensity signal curve measured by the sensor. During the electrochemical reaction process of lithium-ion batteries, significant current distributions exist in the positive and negative electrode materials and current collectors. Based on Maxwell's equations, the movement of charges and the time-varying characteristics of current within the battery will induce a magnetic field in its surrounding space. Given the low time-varying gradient of the current distribution within lithium-ion batteries (lithium ion migration in the electrolyte depends on diffusion, and its diffusion coefficient is very small, making it difficult for the lithium-ion current distribution to change rapidly during charging and discharging, resulting in a slow change in current over time), the time-varying gradient of the current distribution within lithium-ion batteries is relatively low. Here, "low" is a relative concept, without an absolutely fixed threshold. In specific research and applications, a relatively reasonable range will be determined through experimental measurement and data analysis based on different battery systems, operating conditions, and specific performance indicators of interest, to judge whether the gradient is at a "low" level. For example... For some lithium-ion batteries, when the rate of change of current density is less than a certain value per unit time and unit space (e.g., less than 1 mA per square centimeter per second), it may be considered that the time-varying gradient of current distribution is low. However, this value will vary depending on the specific battery. Furthermore, the circuit exhibits significant large-scale structural features. The large-scale structural features of the internal circuit of a lithium-ion battery mainly include the following aspects: electrode structure, current collector, internal battery connection structure, and overall battery packaging structure. It can be reasonably inferred that its induced magnetic field mainly originates from the internal current. During the charging and discharging of a lithium-ion battery, the change in the internal current directly affects the change in the external induced magnetic field. The change in the internal current is the source of the change in the external magnetic field, and the change in the internal current represents the change in various parameters of the battery. Therefore, by placing a high-precision distributed magnetic sensor array on the surface of a lithium-ion cell, the external induced magnetic field during the complete charge-discharge cycle of the battery can be continuously monitored. By analyzing the changes and patterns in the monitored magnetic field data, the internal changes of the lithium-ion battery during the charge-discharge cycle can be obtained, thereby assessing the health status of the lithium-ion battery.

[0078] The magnetic signal acquisition module is used to acquire magnetic signal data generated by lithium-ion single cells obtained from the multi-channel magnetic field measurement module.

[0079] The signal processing module is used to convert the preprocessed magnetic signal data into a curve and combine it with the battery health status assessment model to complete the classification and assessment of the health status of lithium-ion single cells.

[0080] A further embodiment of the embodiment includes a signal processing module comprising:

[0081] The signal preprocessing unit is used to select 100 magnetic signal data within one second, remove the highest and lowest values, and calculate the average value of the remaining magnetic signal data as the magnetic field monitoring data value for this second.

[0082] The curve generation unit is used to generate magnetic field change data curves based on magnetic field monitoring data values ​​and corresponding time data.

[0083] The feature extraction unit is used to extract the segmented time-domain magnetic feature values ​​of the magnetic field change data curve. Specifically, the monitored magnetic field data is read, and the first column of data is read and converted into X-axis time values ​​in seconds (s); the data in columns 2-N are read and converted into Y-axis magnetic field magnitude values ​​in nanoteslas (nT). The data is segmented by inputting the X-axis values ​​of the segment points, and the segmented time-domain feature values ​​are extracted by the programmed program, including variance, standard deviation, root mean square, peak-to-peak value, and kurtosis.

[0084] Specifically, the first step is to read and preprocess the raw magnetic field data, including merging data worksheets, deleting NaN values ​​from each channel through looping to ensure data continuity, and downsampling the processed data by 100 times to reduce the amount of data.

[0085] Then, for each channel's data, the user manually specifies four dividing points to divide the data into three intervals, and calculates the time-domain statistical characteristics (variance, standard deviation, mean, peak-to-peak value) and waveform characteristics (kurtosis, root mean square, kurtosis factor) for each interval. Five magnetic field characteristic values ​​are extracted using the following formula to characterize the battery's charging and discharging state information:

[0086] F1 = Var(x) = E[(x-μ)] 2 ]

[0087]

[0088] F5=ptp=max(data(interval))-min(data(interval))

[0089] Where x represents the sample value, μ represents the sample mean, N represents the total number of samples, data(interval) represents the data set within the specified interval, F1 represents the degree of data variation, F2 represents the average degree of data deviation from the mean, F3 represents the magnitude of the data "amplitude," often used to describe the energy or fluctuation amplitude of a signal, F4 is used to measure the sharpness of the data distribution or the thickness of its tails, and F5 is used to represent the total amplitude of data or signal fluctuations, often used to describe the maximum range of signal variation. These five feature values ​​are used to construct a feature vector F = (F1, F2, F3, F4, F5), and finally, the time-domain waveforms of the three channels are plotted to visually demonstrate the data characteristics.

[0090] The model building unit is used to construct a battery health status assessment model by combining piecewise time-domain magnetic eigenvalues ​​with the k-NN algorithm.

[0091] Specifically, the raw magnetic field monitoring data stored in the host computer is converted into a curve. This conversion requires data preprocessing: 100 magnetic field data points are selected every second, the highest and lowest values ​​are removed, and the average of the remaining magnetic field data is calculated as the magnetic field monitoring data value for that second. The time data of the raw magnetic field monitoring data stored in the host computer is the actual time, in the format of hours:minutes:seconds. This time data is converted into second data starting from 0 using code. Next, the curve of the monitored magnetic field changes is analyzed. The curve can be divided into five stages, such as... Figure 8 As shown: The first stage is a gradual and stable increase or slight decrease in magnetic field intensity, corresponding to the constant current charging stage of the battery charging and discharging process; the second stage is a sudden increase in magnetic field intensity followed by a gradual stabilization, corresponding to the constant voltage charging stage of the battery charging and discharging process; the third stage is a sharp increase in magnetic field intensity, corresponding to the pause stage between charging and discharging in the battery charging and discharging process; the fourth stage is the cessation of the sharp increase in magnetic field intensity, followed by a decrease and a horizontal trend, or a direct horizontal trend, corresponding to the constant current discharging stage of the battery charging and discharging process; the fifth stage is the end of the horizontal trend in magnetic field intensity, followed by a sharp decrease, corresponding to the pause stage after discharging in the battery charging and discharging process. After the five stages are completed, the battery completes one charge-discharge cycle. Based on the above, the changing patterns of the induced magnetic field intensity monitoring data are summarized; then, feature extraction is performed on the monitored magnetic field change data curves. Based on the characteristics and curve patterns of the magnetic field monitoring data: the magnetic field monitoring data is collected at a fixed frequency, so the time-domain feature values ​​of the data curve are extracted; the charging and discharging of the battery have different stages, so the data curve is segmented for time-domain magnetic feature value extraction; finally, the extracted segmented time-domain magnetic feature values ​​are input into the constructed dataset, and the health status of the monitored lithium-ion battery is classified and evaluated by combining the k-NN algorithm model.

[0092] A further implementation method is that the model building unit includes:

[0093] The dataset construction subunit is used to divide the segmented time-domain magnetic feature values ​​into training and testing datasets. The training dataset consists of segmented time-domain magnetic feature values ​​that label the health status categories of lithium-ion single-cell batteries, while the testing dataset consists of segmented time-domain magnetic feature values ​​to be evaluated. Specifically, the battery health status categories are divided and the sample order is randomly shuffled to avoid the data order affecting the fairness of the model. Then, training and testing sets are extracted according to the battery health status categories to ensure that the proportion of each category in the training and testing datasets is consistent.

[0094] The k-value determination sub-unit is used to determine the k-value for the k-NN algorithm through cross-validation.

[0095] The Euclidean distance calculation subunit is used to calculate the Euclidean distance between the piecewise temporal magnetic eigenvalues ​​in the test dataset and the training dataset using the k-NN algorithm with a determined k value.

[0096] The nearest sample acquisition subunit is used to obtain neighboring sample data points in the training dataset that meet the similarity threshold with the test dataset based on Euclidean distance.

[0097] The health status assessment subunit is used to obtain the health status assessment results of lithium-ion single-cell batteries in the test dataset based on the health status categories of neighboring sample data points.

[0098] Specifically, the dataset is normalized (k-NN is based on distance metrics, and differences in feature scales can lead to biases in distance calculation); then the k-NN model is trained, with the nearest neighbor set to 1, meaning the nearest neighbor is selected for classification. The built-in k-NN classifier automatically calculates the distance between samples, and the category is determined by "majority voting," thus enabling predictions for both the training and test datasets.

[0099] According to the formula:

[0100] The accuracy of the training and test datasets is calculated and displayed through a prediction comparison chart. A confusion matrix (rows: true class, columns: predicted class, diagonal elements: number of correct classifications, off-diagonal elements: number of misclassifications) is then used to show the misclassification rates between classes. The evaluation criterion is to compare the test data point with the training data point closest to it within the selected "k" value range. The type of the training data point is then used as the evaluation result for the test data point, and the evaluation format is a rank category (good / poor).

[0101] A further implementation involves improving the k-NN algorithm by introducing an automatic shuffling layer and a data normalization transpose layer, and adding a confusion matrix to evaluate the health status assessment results of lithium-ion single-cell batteries.

[0102] Specifically, the core of the automatic shuffling function is to use the `randperm` function, based on a pseudo-random number generator and the Fisher-Yates shuffling algorithm, to generate a random permutation sequence containing all sample indices. The rows of the dataset are then rearranged according to this sequence to eliminate biases such as class concentration, temporal dependencies, or experimental batches that may arise from the original data order. This process first obtains the total number of samples in the dataset, then uses the algorithm to generate random indices, mapping them row by row to randomize the sample order while maintaining the correspondence between features and labels. For imbalanced data, stratified sampling can be combined after shuffling to ensure a balanced proportion of each state class in the training and test datasets. For time-series battery cycling data, block shuffling can be used to preserve local dependencies rather than completely shuffling the sample order. For multi-table joins, the same random index sequence must be used for synchronous shuffling to avoid sample and feature misalignment. This operation has low computational complexity and can be reproducible by setting a random seed, effectively improving the fairness and reliability of model training and evaluation.

[0103] Normalization uses the following formula: Scaling the data to the [0,1] range ensures that different features have the same scale, thereby avoiding the excessive influence of certain features on the model due to their large numerical range.

[0104] Data transposition is necessary because the k-NN classifier requires the input feature matrix to be in sample × feature format, and the normalized data format to be in feature × sample format. Without transposition, the classification logic would be completely wrong. Transposition is to avoid sample dimension confusion, to allow the normalization function to correctly handle the dimension of each feature, and to ensure that all sample values ​​of the same feature are scaled uniformly.

[0105] The evaluation criterion is to compare the test data point with the training data point that is closest to the selected "k" value range. The type of the training data point is the evaluation result of the test data point, and the evaluation form is a grade category (good / poor).

[0106] Example 2

[0107] This invention also provides a method for monitoring lithium-ion single-cell batteries based on weak magnetic field detection technology, using the system of Example 1, comprising:

[0108] Magnetic signal data of lithium-ion single cells are measured using a distributed fluxgate sensor array.

[0109] The magnetic signal data generated by a single lithium-ion battery cell was acquired by a multi-channel magnetic field measurement module.

[0110] The preprocessed magnetic signal data is converted into a curve, and combined with the battery health status assessment model, the health status of lithium-ion single cells is classified and assessed.

[0111] A further implementation method for obtaining the monitored position of a distributed fluxgate sensor array includes:

[0112] Based on Maxwell's equations, magnetic permeability, and the electrochemical reaction principle of lithium-ion cells, the influencing factors of the external induced magnetic field of lithium-ion cells are obtained. Among them, the influencing factors include the internal current of the lithium-ion cell and the magnetic susceptibility of the lithium-ion cell material, with the magnetic susceptibility being an interference signal.

[0113] Based on the internal current of a lithium-ion cell, the change in the surface magnetic field of the lithium-ion cell is obtained.

[0114] A distributed fluxgate sensor array is used to scan the external induced magnetic field of a lithium-ion cell during the charge-discharge cycle. The location where the change in the magnetic field on the surface of the lithium-ion cell meets the preset conditions and is far away from interference signals is used as the monitoring location.

[0115] This embodiment also provides specific operating steps for the system of Embodiment 1:

[0116] (1) First, based on the battery parameters, the battery charge and discharge cycle operation steps and specific parameters need to be designed in advance.

[0117] (2) Determine the position of the magnetic sensor probe on the battery based on the shape and size of the battery and the monitoring requirements.

[0118] (3) After connecting and fixing the battery to be monitored to the magnetic sensor, place it at the monitoring position, then open the host computer software for weak magnetic monitoring, check the connection status, and if there is any abnormality, reconnect the battery to the instrument and check the battery status.

[0119] (4) The fourth step is to add the designed parameters and operating steps to the host computer software and set the data saving format, that is, to complete the setting of the battery charging and discharging process parameters.

[0120] (5) Fifth step, select the monitoring battery channel and set the name of the step to be executed. After confirmation, enter the experimental information of the monitored battery (including the monitoring battery information, monitoring date and conditions); then click the multi-channel magnetic field measurement system software to receive and save. The magnetic measurement system starts to collect data synchronously and stores the data in the host computer.

[0121] (6) After the battery charge-discharge cycle monitoring ends, export the magnetic field monitoring data and electrochemical monitoring data stored in the host computer, organize them into new data documents, and use a self-written plotting program to plot the data into curves and compare and analyze them.

[0122] Example 3

[0123] This embodiment monitored and verified five groups of lithium-ion batteries. Since the differences between the magnetic field monitoring data curves of each group were extremely subtle, to present the monitoring results more concisely and effectively, the magnetic field monitoring data curve of battery number 16 was selected as the final monitoring and verification result graph (in this embodiment, a 3-channel fluxgate sensor was selected for monitoring based on the size and shape of the tested battery and the monitoring requirements). The magnetic monitoring data result graph of battery number 16 is shown below. Figure 9 (a) and Figure 9 As shown in (b), the electrochemical monitoring data results of battery No. 16 are as follows: Figure 10 (a) and Figure 10 As shown in (b).

[0124] The results of the electrochemical monitoring data verification were compared with the results of the monitoring method proposed in this invention, and the monitoring results of this invention are more intuitive and obvious.

[0125] The magnetic monitoring data curves for the small-scale and pilot-scale tests show a significant difference within the red circle area. The curve for the small-scale battery exhibits sharp fluctuations within the red circle, while the curve for the pilot-scale battery is smooth and stable within the red circle. Compared to the magnetic monitoring data curves, the electrochemical monitoring data curves, although showing minor differences in data size or duration, do not exhibit the significant fluctuations seen in the magnetic monitoring results.

[0126] Based on the monitored data of induced magnetic field intensity changes during the charge-discharge cycle of lithium-ion batteries, a time-domain magnetic signal feature extraction method was used to extract magnetic signal feature values ​​(in this embodiment, five feature values ​​were extracted from each of the three curve segments, for a total of fifteen feature values). An input magnetic signal feature value dataset was constructed, with category "1" representing lithium-ion batteries in pilot-scale conditions and category "2" representing lithium-ion batteries in small-scale conditions. Then, the k-NN algorithm was used to classify and identify the dataset, with k set to "2". 70% of the data was used as training data in the training dataset (TrainData), and the remaining 30% was used as test data in the prediction dataset (TestData) to train the model. Finally, simulation tests were performed on the training and test datasets, and performance evaluation results (including prediction accuracy and confusion matrix) were output. Figure 11 This is a graph showing the performance evaluation results of the training dataset in an embodiment of the present invention; wherein, Figure 11 (a) is a line graph of the prediction results. Figure 11 (b) is the confusion matrix; Figure 12 This is a graph showing the performance evaluation results of the test dataset in an embodiment of the present invention; wherein, Figure 12 (a) is a line graph of the prediction results. Figure 12 (b) is the confusion matrix.

[0127] The verification results for single lithium-ion batteries under different health conditions in the same environment are as follows:

[0128] The data consisted of 49 sets of battery data in the same environment but with different health states, and 56 sets of data in the small-scale trial (with poor health states), for a total of 105 sets of data. Classification results showed that the model's prediction accuracy was high, with a prediction accuracy of 94.5% on the training dataset. Considering precision, recall, and F1 score, the model demonstrated strong ability to correctly identify and misclassify classes "1" and "2," exhibiting good overall performance. On the test dataset, the model achieved a prediction accuracy of 87.5%. Considering precision, recall, and F1 score, the model showed some misclassification for class "1" and some missed classifications for class "2." While the overall performance was lower than on the training set, the difference was small, indicating that the model possesses good generalization ability and performs relatively stably in predicting new data.

[0129] In summary, the technical solution of this invention has high application value in the engineering field. The weak magnetic field monitoring system and method proposed in this invention, as a supplement to existing monitoring technologies, can achieve non-destructive in-situ monitoring of a single lithium-ion battery during a complete charge-discharge cycle. Combined with feature value extraction methods and the k-NN algorithm model, the health status of the battery is classified and evaluated. The accuracy rate for classifying and identifying lithium-ion batteries under different health states and before and after different operating conditions reaches 87.5%. Ultimately, the battery is classified as a pilot-scale battery or a small-scale battery, thus determining the health status of the lithium-ion battery. This invention has potential applications in fields such as electric vehicles and is of great significance for ensuring the safe operation of equipment.

[0130] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A lithium-ion single-cell battery monitoring system based on weak magnetic field detection technology, characterized in that, include: A multi-channel magnetic field measurement module is used to measure the magnetic signal data of lithium-ion single cells through a distributed fluxgate sensor array; The magnetic signal acquisition module is used to acquire magnetic signal data generated by lithium-ion single-cell batteries as measured by the multi-channel magnetic field measurement module. The signal processing module is used to convert the preprocessed magnetic signal data into a curve and combine it with the battery health status assessment model to complete the classification and assessment of the health status of lithium-ion single cells. The multi-channel magnetic field measurement module uses a distributed fluxgate sensor array whose detection sensitive direction is perpendicular to the cylindrical end face. The actual measured magnetic signal data are the external induced magnetic field signals at different positions on the surface of the lithium-ion single battery at the same time during the charge and discharge cycle. The multi-channel magnetic field measurement module includes: The influencing factor acquisition unit is used to obtain the influencing factors of the external induced magnetic field of a lithium-ion single cell based on Maxwell's equations, magnetic permeability, and the electrochemical reaction principle of a lithium-ion single cell; wherein, the influencing factors include the internal current of the lithium-ion single cell and the magnetic susceptibility of the lithium-ion single cell material, and the magnetic susceptibility is an interference signal; The magnetic field change acquisition unit is used to obtain the change of the magnetic field on the surface of the lithium-ion single cell based on the internal current of the lithium-ion single cell. The monitoring location acquisition unit is used to scan the external induced magnetic field of the lithium-ion single cell during the charge and discharge cycle using a distributed fluxgate sensor array, and obtain the location where the change of the magnetic field on the surface of the lithium-ion single cell meets the preset conditions and is far away from the interference signal as the monitoring location. The signal processing module includes: The signal preprocessing unit is used to select 100 magnetic signal data within one second, remove the highest and lowest values, and calculate the average value of the remaining magnetic signal data as the magnetic field monitoring data value for this second. The curve generation unit is used to generate a magnetic field change data curve based on the magnetic field monitoring data value and the corresponding time data. The feature extraction unit is used to extract the piecewise time-domain magnetic feature values ​​of the magnetic field change data curve; The model building unit is used to construct a battery health status assessment model by combining the segmented time-domain magnetic feature values ​​with the k-NN algorithm. The model building unit includes: A dataset construction subunit is used to divide the segmented time-domain magnetic feature values ​​into a training dataset and a test dataset; wherein, the training dataset consists of segmented time-domain magnetic feature values ​​that label the health status categories of lithium-ion single-cell batteries; and the test dataset consists of segmented time-domain magnetic feature values ​​to be evaluated. The k-value determination sub-unit is used to determine the k-value for the k-NN algorithm through cross-validation. The Euclidean distance calculation subunit is used to calculate the Euclidean distance between the piecewise temporal magnetic eigenvalues ​​in the test dataset and the training dataset using the k-NN algorithm with a determined k value. The neighbor sample acquisition subunit is used to obtain neighbor sample data points in the training dataset that meet the similarity threshold with the test dataset based on the Euclidean distance. The health status assessment subunit is used to obtain the health status assessment result of the lithium-ion single cell battery in the test dataset based on the health status category of the neighboring sample data points.

2. The system according to claim 1, characterized in that, The improvements to the k-NN algorithm include the introduction of an automatic shuffling layer and a data normalization transpose layer, and the addition of a confusion matrix to evaluate the health status assessment results of lithium-ion single-cell batteries.

3. A method for monitoring lithium-ion single-cell batteries based on weak magnetic field detection technology, using the system described in any one of claims 1-2, characterized in that, include: Magnetic signal data of lithium-ion single cells are measured using a distributed fluxgate sensor array. The magnetic signal data generated by a single lithium-ion battery cell was acquired by a multi-channel magnetic field measurement module. The preprocessed magnetic signal data is converted into a curve, and the health status of lithium-ion single cells is classified and evaluated by combining the battery health status assessment model.

4. The method according to claim 3, characterized in that, The method for obtaining the monitored position of the distributed fluxgate sensor array includes: Based on Maxwell's equations, magnetic permeability, and the electrochemical reaction principle of lithium-ion cells, the influencing factors of the external induced magnetic field of lithium-ion cells are obtained; among them, the influencing factors include the internal current of the lithium-ion cell and the magnetic susceptibility of the lithium-ion cell material, wherein the magnetic susceptibility is an interference signal. Based on the internal current of the lithium-ion single cell, the change in the surface magnetic field of the lithium-ion single cell is obtained. A distributed fluxgate sensor array is used to scan the external induced magnetic field of a lithium-ion cell during the charge-discharge cycle. The location where the change in the magnetic field on the surface of the lithium-ion cell meets the preset conditions and is far away from the interference signal is used as the monitoring location.

Images