A lithium ion battery charging safety monitoring method based on magnetic field intensity sensing

Abstract

The application discloses a lithium ion battery charging safety monitoring method based on magnetic field intensity sensing, considers the coupling of electric-thermal-magnetic multi-physical fields, establishes a battery pack magnetic field sensing model; based on the current-magnetic field intensity data set of each single battery simulated by the battery pack magnetic field sensing model, a BP neural network is established and trained; the magnetic field intensity data of the battery pack charging process are monitored online, as the input of the trained BP neural network, the local current of each single battery is output, and whether the charging process is abnormal is judged by the local current of each single battery. The application can effectively monitor the charging process of the lithium ion battery pack, and exclude the hidden dangers of the local faults in the charging process caused by the inconsistency of single batteries, the lack of local parameters in the overall monitoring method of the battery pack.

Description

Technical Field

[0001] This invention relates to the field of new energy vehicle charging technology, specifically to a method for monitoring the charging safety of lithium-ion batteries based on magnetic field strength sensing. Background Technology

[0002] With the rapid growth in the number of new energy vehicles, safety issues are becoming increasingly prominent, making ensuring charging safety a crucial challenge that cannot be ignored. According to relevant research statistics, most new energy vehicle accidents occur during the charging process. Several factors increase the likelihood of charging accidents. First, battery overheating may occur during charging; due to the chemical reactions that release heat during charging, if heat cannot be effectively dissipated, the battery temperature will rise, exceeding safe limits and potentially causing an accident. Second, battery aging is also a potential risk factor; as usage time increases, the performance of the battery pack gradually declines, its capacity decreases, and its internal resistance increases, thus affecting the stability during charging and discharging. Aging batteries are more susceptible to overcharging, over-discharging, and other charging control errors, increasing the risk of accidents. In addition, inconsistent aging rates among individual cells within the battery pack can also lead to accidents; performance differences between individual cells can cause voltage imbalances during charging, potentially leading to overcharging or over-discharging of certain cells and thus causing an accident.

[0003] Safety monitoring during the charging process is crucial for ensuring the safety of new energy vehicle charging. Real-time monitoring of the charging process allows for the timely detection of anomalies, enabling appropriate measures to prevent accidents. Traditional charging safety protection mechanisms primarily rely on the electric vehicle's Battery Management System (BMS). The BMS monitors voltage, temperature, and charge / discharge status, and manages these parameters to ensure safe operation. However, existing BMS technology struggles to effectively assess the inconsistencies between individual cells within the battery pack. Therefore, relying solely on traditional BMS for battery charging safety monitoring carries the risk of failing to accurately detect and prevent overcharging and over-discharging of individual cells.

[0004] In recent years, many new methods and studies on safety monitoring during lithium-ion battery charging have emerged. YAMANAKA et al. were among the first to propose monitoring charging by monitoring temperature changes. However, in actual charging, changes in ambient temperature affect the battery's heat dissipation, thus altering the relationship between temperature changes and the state of charge. Therefore, different temperature conditions may lead to inaccurate monitoring of the state of charge, reducing the universality of this temperature change monitoring method. JOHNSON et al. were among the first to propose a parallel circuit for overcharge protection with a parallel voltage regulator. When the battery is fully charged, the parallel circuit can cut off the main circuit current. However, considering that power batteries age to varying degrees with increasing mileage, the voltage of a fully charged battery is lower than the factory-set voltage. In this case, if the battery is actually fully charged but has not reached the preset cutoff voltage of the protection circuit, the circuit will not intervene in the charging process, and overcharging risk will still occur. Therefore, protection circuits alone cannot ensure that the battery is free from overcharging throughout its entire lifespan. ZHAO et al., based on real-vehicle data of new energy vehicle power batteries, proposed an improved fault diagnosis strategy for BMS using machine learning algorithms and a 3σ multi-level screening strategy (3σ-MSS). In summary, existing research and technologies for achieving safe monitoring during battery charging are not yet perfect. They suffer from limitations such as difficulty in monitoring local battery packs, insufficient accuracy and reliability, and high computational complexity. In particular, most methods indirectly infer the state of the battery pack by using overall parameters such as battery pack voltage and temperature, without taking into account the differences between individual cells within the battery pack. This makes it impossible to provide specific information about individual cells and achieve accurate monitoring of local characteristics of the battery pack.

[0005] Magnetic field strength sensing offers a new approach for safe monitoring of lithium-ion battery charging processes. In recent years, related research and applications have gradually increased. Existing research assesses the internal performance state of lithium batteries by measuring changes in the external induced magnetic field, providing a non-invasive method for battery monitoring and diagnosis. However, it has limitations, such as research objects being limited to single lithium batteries and simulation models not considering the effects of temperature. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a lithium-ion battery charging safety monitoring method based on magnetic field strength sensing. This method achieves accurate monitoring of the entire charging process of the power battery pack through magnetic field sensing and a BP neural network.

[0007] The present invention achieves the above-mentioned technical objectives through the following technical means.

[0008] A method for monitoring the charging safety of lithium-ion batteries based on magnetic field strength sensing:

[0009] Considering the coupling of electro-thermal-magnetic multiphysics fields, a magnetic field sensing model for the battery pack is established.

[0010] Based on the simulation data of current-magnetic field intensity of each individual battery cell using the magnetic field sensing model of the battery pack, a BP neural network is established and trained.

[0011] The magnetic field strength data of the battery pack during the online charging process is used as the input of the trained BP neural network to output the local current of each individual cell. The local current of each individual cell is used to determine whether the charging process is abnormal.

[0012] The coupling of the electro-thermal-magnetic multiphysics fields is as follows: An electrochemical model of a lithium-ion battery is established based on the lithium-ion battery interface, and the distribution of current density in the entire battery electrochemical model is solved; a fluid heat transfer interface is introduced, and the heat conduction equation of the electrochemical-thermal multiphysics interface is used to calculate the internal temperature change of the battery during charging; a magnetic field interface is introduced, and the spatial magnetic field is calculated using the internal current distribution and temperature of the lithium battery as initial conditions, achieving unidirectional coupling of the electric and thermal fields to the magnetic field; then, the internal current distribution and temperature distribution of the lithium battery are recalculated based on the spatial magnetic field, achieving reverse coupling of the magnetic field to the electric and thermal fields; the recalculated current distribution and temperature are then used as conditions to calculate the magnetic field, and this process is repeated until the difference between the average values ​​of two adjacent spatial magnetic field data is less than a set value.

[0013] Furthermore, the charging process is judged by the local current of each individual battery cell. Specifically, the local current of each individual battery cell is compared with the mean value of the current normal distribution model. If the difference exceeds the set threshold, an abnormal alarm will be triggered immediately.

[0014] Furthermore, by referring to the charging current safety parameters, it can be determined whether the abnormality may damage the battery pack. If the abnormality exceeds the safety range, corresponding measures are taken.

[0015] Furthermore, an air domain is added to the external magnetic field sensing model of the battery pack.

[0016] Furthermore, when establishing the battery pack magnetic field sensing model, the battery is divided into a grid, including 37,076 grid vertices, 67,832 prisms, 8,896 triangles, 10,578 quadrilaterals, and 3,356 edge cells.

[0017] Furthermore, the hidden layer of the BP neural network contains 18 neurons.

[0018] Furthermore, the battery pack magnetic field sensing model is established in COMSOL software, and the boundary conditions required for the model are set.

[0019] Furthermore, the set value is 0.1uT.

[0020] The beneficial effects of this invention are as follows: By deriving the influence of temperature on the spatial magnetic field during the charging process of lithium-ion batteries, considering the changes in the spatial magnetic field caused by the heat generated during battery charging, the coupling relationship between the electro-magnetic-thermal fields is studied. The spatial magnetic field excited during the charging process is calculated using the air domain of an external device, and a battery pack magnetic field sensing model is established based on the finite element analysis method, deepening the research on the spatial magnetic field generated by lithium-ion batteries. Based on the simulation dataset of current-magnetic field strength of each individual battery cell from the battery pack magnetic field sensing model, a BP neural network is established and trained. The magnetic field strength data of the battery pack during the charging process is monitored online, serving as the input to the trained BP neural network, outputting the local current of each individual battery cell, and judging whether the charging process is abnormal based on the local current of each individual battery cell. This invention, based on magnetic field strength sensing and a BP neural network, can effectively monitor the charging process of lithium-ion battery packs, perceive the current magnitude of each individual battery cell in the battery pack online, identify charging anomalies, and locate abnormal local current cells. This eliminates the hidden dangers of local faults during the charging process caused by inconsistencies in individual batteries and the lack of local parameters in the overall battery pack monitoring method, thus helping to improve the safety of the charging process of vehicle power batteries. Attached Figure Description

[0021] Figure 1 This is a schematic diagram illustrating the principle of generating the battery-induced magnetic field according to the present invention.

[0022] Figure 2 This is the grid division diagram of the single cell described in this invention;

[0023] Figure 3 This is a magnetic field sensing diagram of a single battery cell described in this invention;

[0024] Figure 4 This is a comparative experimental diagram of the charging voltage curves of the battery described in this invention;

[0025] Figure 5 This is a diagram showing the location of the magnetic field measurement described in this invention.

[0026] Figure 6(a) shows the measurement effect of the laboratory exposed environment described in this invention;

[0027] Figure 6(b) shows the measurement effect of the magnetic field shielding device described in this invention;

[0028] Figure 7 This is a comparison chart of the experimental and simulation results described in this invention;

[0029] Figure 8 This is the mesh partitioning diagram of the 2s2p battery pack described in this invention;

[0030] Figure 9 This is a flowchart of the lithium-ion battery charging safety monitoring system based on magnetic field strength sensing as described in this invention.

[0031] Figure 10This is a diagram showing the training results of the neural network described in this invention;

[0032] Figure 11 This is a graph illustrating the neural network training process described in this invention.

[0033] Figure 12 This is a diagram of the normal distribution model of the single-cell current described in this invention. Detailed Implementation

[0034] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the scope of protection of the present invention is not limited thereto.

[0035] This invention discloses a lithium-ion battery charging safety monitoring method based on magnetic field strength sensing. First, by studying the coupling relationship between the electro-magnetic-thermal fields, a spatial magnetic field excited during the charging process is calculated using an external air domain, establishing a battery pack magnetic field sensing model. This model is simulated using COMSOL finite element software, and a finite number of unknowns determined after mesh independence verification are used to approximate an infinite number of unknowns to achieve accurate analysis of the real system. Then, the simulation results of the battery pack magnetic field sensing model are compared and verified with experimental data to ensure the model's accuracy and effectiveness. Next, based on the simulated current-magnetic field strength dataset of each individual battery cell, a BP neural network is established and trained, and its effectiveness in using magnetic field strength sensing to infer the current of individual batteries is evaluated. The magnetic field strength data during the battery pack charging process is monitored online and used as input to the trained BP neural network, outputting the local current of each individual battery cell. The local current of each individual battery cell is used to determine whether the charging process is abnormal.

[0036] A method for monitoring the charging safety of lithium-ion batteries based on magnetic field strength sensing, specifically including the following steps:

[0037] Step (1), Establishment of the magnetic field sensing model for lithium-ion single cell

[0038] The magnetic field generated during the charging process of a lithium-ion battery is formed by the superposition of the electromagnetic magnetic field excited by the current inside the battery and the magnetic field caused by the magnetization of the magnetic materials inside the battery.

[0039] During the electrochemical reaction in a lithium-ion battery, current flows through the positive and negative electrodes and current collectors. This internal current generates an induced magnetic field in space, such as... Figure 1 As shown.

[0040] Lithium-ion batteries may contain magnetic materials such as iron and nickel. These materials may undergo magnetization during battery charging, meaning their molecular magnetic moments rearrange to respond to an external magnetic field. This magnetization generates a magnetic field around the magnetic material, creating a magnetic field distribution in space.

[0041] Existing lithium-ion battery magnetic field sensing models only consider electromagnetic coupling and do not account for the influence of temperature on magnetic field sensing, thus failing to fully and accurately describe changes in the internal magnetic field of the battery. The charging process of lithium-ion batteries generates a significant amount of heat, causing substantial changes in the internal temperature of the battery, which in turn affects the magnetic field by influencing parameters such as the conductivity of the battery materials.

[0042] For part B of the magnetic field generated by the internal current of the battery internal The equation for the electric field generated is as follows:

[0043]

[0044] Where: J represents the nonlinear current density distribution, which can be expressed as the product of the electrolyte conductivity σ and the electric field strength E; the electric field strength E can be expressed as the negative of the potential gradient, and under simplifying assumptions, the gradient calculation can be simplified to a partial derivative calculation; and according to the Arrhenius equation, the electrolyte conductivity σ is temperature-dependent, therefore, the magnetic field part B generated by the current inside the battery... internal It is temperature-dependent. The derivation of the equation for the electro-magnetic field is as follows:

[0045] J=σ×E (2)

[0046]

[0047]

[0048]

[0049]

[0050] Where σ0 is the conductivity at the reference temperature, Ea is the activation energy, k is the Boltzmann constant, T represents the temperature, μ0 is the magnetic permeability in vacuum (constant), and V represents the internal volume of the battery. It represents electric potential, and x, y, and z represent the three directions in a rectangular coordinate system.

[0051] Equation (6) shows that temperature affects the magnetic field generated by the current.

[0052] Temperature changes also affect the magnetization of materials inside the battery. In lithium-ion batteries, the magnetization characteristics of the positive and negative electrode materials are closely related to temperature. As the temperature increases during the charging process of a lithium-ion battery, the magnetism of the materials changes, leading to a corresponding change in the distribution of the magnetic field. Specifically, the magnetization characteristics of magnetic materials are derived by considering Curie's law, which describes the relationship between the magnetization characteristics of magnetic materials and temperature. According to Curie's law, the magnetization intensity of a magnetic material is inversely proportional to temperature, as shown in the following equation:

[0053]

[0054] Where M represents magnetization and C is a constant. Equation (7) shows that magnetization decreases with increasing temperature. Meanwhile, the magnetic field strength of the magnetic material generating a spatial magnetic field and the magnetization have the following relationship:

[0055] B_material=μ0×(H+M) (8)

[0056] Where B represents the magnetic field strength, H represents the applied magnetic field strength, and M represents the magnetization intensity.

[0057] Combining formulas (7) and (8) above, the effect of temperature on the magnetic field can be derived. Substituting M from Curie's law into the formula for magnetic field strength, we get:

[0058]

[0059] Therefore, temperature changes can not only affect the electromagnetism inside the battery, but also indirectly affect the distribution of the magnetic field by changing the magnetization of the materials.

[0060] Based on the above derivation, it can be seen that when calculating the spatial magnetic field excited during the charging process of a lithium-ion battery, in addition to considering the electromagnetic magnetic field excited by the current inside the battery and the magnetization of magnetic materials, the influence of temperature changes during battery charging on the magnetic field sensing model must also be considered. This requires more complex electro-magnetic-thermal multiphysics coupling to obtain more accurate results.

[0061] This invention uses COMSOL Multiphysics 6.1 software to establish a magnetic field sensing model of a lithium-ion single-cell battery. Utilizing the software's lithium-ion battery interface, magnetic field interface, and fluid heat transfer interface, the boundary conditions required for the magnetic field sensing model of the lithium-ion single-cell battery can be characterized. The dimensions and materials of the single-cell battery are shown in Tables 1 and 2.

[0062] Table 1 Geometric Dimensions of Individual Cells

[0063]

[0064] Table 2 Battery Material Settings

[0065]

[0066] In COMSOL Multiphysics 6.1, the lack of built-in multiphysics interfaces for lithium-ion batteries and fluid heat transfer and magnetic fields prevents the coupling of electro-thermal-magnetic multiphysics fields. Therefore, this invention proposes a coupling strategy that considers the influence of temperature on magnetic field perception to achieve coupling of electro-thermal-magnetic multiphysics fields. First, an electrochemical model of a lithium-ion battery is established based on the lithium-ion battery interface, and the distribution of current density in the entire battery electrochemical model is solved. Next, a fluid heat transfer interface is introduced, and the heat conduction equation of the built-in electrochemical-thermal multiphysics interface is used to calculate the internal temperature change of the battery during charging. Subsequently, a magnetic field interface is introduced, using the current distribution and temperature information inside the lithium battery as initial conditions, and the spatial magnetic field is calculated by solving Maxwell's equations, achieving unidirectional coupling of the electric and thermal fields to the magnetic field. Then, based on the spatial magnetic field data, the current distribution and temperature distribution inside the lithium battery are recalculated, achieving reverse coupling of the magnetic field to the electric and thermal fields. The magnetic field is then recalculated using the recalculated current distribution and temperature information as conditions. This process is repeated iteratively, continuously updating the current density, battery internal temperature, and spatial magnetic field distribution until the difference between the average values ​​of the two sets of spatial magnetic field data is less than 0.1 μT, ultimately achieving electro-thermal-magnetic multiphysics coupling. Considering the significant workload of iterative calculations, this invention utilizes Matlab, writing code to call COMSOL software to improve simulation efficiency. The simulation model not only considers the influence of current density but also incorporates temperature changes during lithium-ion charging, making the calculation results more closely resemble actual conditions.

[0067] When establishing a magnetic field sensing model for a single lithium-ion battery, it is necessary to perform grid division of the single battery, such as... Figure 2 As shown, since the simplified geometry of a lithium-ion battery can be approximated as a thin sheet-like stack along the z-axis, and considering the requirements of reducing the number of meshes, optimizing the solution process, and reducing the difficulty of the solution, the model adopts a sweeping method to construct the mesh in the thickness direction.

[0068] Mesh independence is an important concept in numerical simulations for evaluating the impact of mesh refinement on the solution. In this invention, the mesh independence of a lithium-ion magnetic field sensing model built using COMSOL software was verified to ensure the high credibility and reliability of the results. To verify mesh independence, computational analysis was performed on progressively refined meshes generated on the same geometry using appropriate algorithms and parameter settings, ensuring accurate capture of geometric features. By comparing simulation results under different meshes, this invention evaluated the impact of mesh size on the accuracy and precision of the solution, as shown in Table 3. It was found that when the mesh size reached 34282, further mesh refinement had little impact on the calculated current density.

[0069] Table 3 Comparison of simulation results before and after mesh refinement.

[0070]

[0071] When the final individual cell mesh is divided into 19,964 vertices, 34,282 prisms, 4,496 triangles, 6,069 quadrilaterals, and 1,468 edge elements, the average element quality reaches 0.7937. This indicates that the mesh generation and solution method used can provide reliable results, and the computation time is within 10 minutes.

[0072] Considering the characteristics of magnetic field research, an air domain will be added to the magnetic field sensing model of a lithium-ion single-cell battery. Introducing an air domain allows for a more comprehensive investigation of the magnetic field characteristics induced by lithium-ion battery charging in space, which is significant for understanding the formation mechanism and influencing factors of the magnetic field during charging. A schematic diagram of the single-cell battery magnetic field sensing model after adding the air domain is shown below. Figure 3 As shown.

[0073] Step (2), Establishment of the magnetic field sensing model of the 2s2p battery pack

[0074] While sensing the magnetic field of a single battery cell is significant for exploring the principles of magnetic field generation in lithium-ion batteries, extending the charging safety monitoring method proposed in this invention to the battery pack level is more practically meaningful. The 2s2p connection is fundamental to the complex connection of individual cells within a battery pack. The method proposed based on magnetic field strength sensing for detecting the entire charging process is typical and can be extended to explore magnetic field distribution, coupling effects, and optimization strategies under more complex connection methods.

[0075] Analogous to the magnetic field sensing model of a single lithium-ion battery cell, when the mesh setting of the magnetic field sensing model of a 2s2p battery pack includes 37076 mesh vertices, 67832 prisms, 8896 triangles, 10578 quadrilaterals, and 3356 edge cells, such as... Figure 8 As shown, the average unit mass reaches 0.8036, and the calculation time is within 18 minutes.

[0076] Step (3), Experimental verification of the battery pack magnetic field sensing model

[0077] Using laboratory charge-discharge equipment and modeling simulation, partial voltage curves for 1C constant current charging were obtained simultaneously, such as... Figure 4 As shown in the figure. Comparative analysis reveals that the trends of the simulation curve and the experimental curve are similar, but the voltage of the simulation curve is slightly higher than that of the experimental data. This may be due to the simplification made in the battery pack magnetic field sensing model, resulting in the actual battery's internal resistance being slightly larger than that of the model battery. Based on the above comparison results, the parameter settings in the battery pack magnetic field sensing model are reasonable, and the battery pack magnetic field sensing model can be used to accurately simulate the charging and discharging process of the actual battery.

[0078] To verify the authenticity and accuracy of the established battery pack magnetic field sensing model in simulating battery magnetic fields, we selected... Figure 5 The locations of magnetic field measurements in the battery pack are shown, and the experimental and simulation results are compared.

[0079] Laboratory magnetic field measurements were performed using a triaxial fluxgate sensor, model HWT3100-485. The sensor measurement results were presented as magnetic flux density membrane |B| and the components of magnetic flux density B in the x, y, and z directions. The parameters are shown in Table 4.

[0080] Table 4 Sensor Parameters

[0081] Viter Intelligence Serial port 485 800uT~+800uT 13nT / LSB 9600 (default)

[0082] When measuring magnetic fields in the laboratory, due to the characteristics of the magnetic field and the relatively weak electromagnetic field, a magnetic field shielding device is needed to eliminate the influence of geomagnetic fluctuations and electromagnetic interference from surrounding electrical equipment on the battery magnetic field measurement. To verify the shielding effect of this device, the fluxgate sensor was placed in an exposed laboratory environment and in a magnetically shielded device for long-term continuous measurements. Figure 6(a) shows the measurement results in the exposed laboratory environment. Affected by the geomagnetic field and the complex environment of various electrical appliances in the laboratory, the three-axis components of the magnetic field strength all exhibit irregular and large fluctuations. Figure 6(b) shows the measurement results in the magnetically shielded device. Compared with the laboratory environment, the magnetic field is significantly more stable, and the background noise is reduced to less than 1 uT, indicating a significant shielding effect.

[0083] By comparison Figure 7 The measurement results show that, under the same charging conditions, the simulated three-axis components of the magnetic field strength are close to the experimental results. Although there are some differences in the values ​​along the z-axis, the overall trend is the same, and the values ​​remain on the same order of magnitude. Therefore, it can be concluded that the model is effective and reliable in simulating and studying the actual spatial magnetic field distribution of batteries.

[0084] Step (4), Establishment of the BP neural network

[0085] Based on the simulation dataset of individual cell current-magnetic field strength using a battery pack magnetic field sensing model covering the entire charging process including charging anomalies, a backpropagation (BP) neural network was established and trained. The effectiveness of the BP neural network in inferring individual cell current using magnetic field strength sensing was then evaluated. The principle and process are as follows: Figure 9 As shown.

[0086] This invention uses simulated data to train a backpropagation (BP) neural network, covering a large number of local anomalies and changes during charging, enabling the neural network to better understand and capture the features and patterns during the battery pack charging process. Therefore, the amount of training data is considered sufficiently rich, contributing to improved training effectiveness and realism of the neural network. Furthermore, a hidden layer structure with 18 neurons is chosen. The hidden layer is a key component in the neural network, providing non-linear feature extraction and representation capabilities. By adjusting the number of neurons in the hidden layer, the complexity and learning ability of the network can be controlled. By rationally designing the number of neurons in the hidden layer, the accuracy and generalization ability of the model can be improved while balancing complexity and learning ability. The training results are as follows... Figure 10 As shown.

[0087] Analysis of the training results shows that the R-values ​​all exceeded 0.99, indicating that the neural network model can explain most of the variability of the dependent variable in the training, validation, and test sets, demonstrating high predictive accuracy. Furthermore, this also indicates that the model achieved similar predictive performance on different datasets, exhibiting good generalization ability. This means that the model can also provide accurate predictions for unseen data, possessing a certain degree of universality and reliability.

[0088] Further analysis of the BP neural network training process, such as Figure 11 As shown, the curves for the training set, validation set, and test set highly overlap and converge, leading to the following conclusion:

[0089] (1) The model has good fitting ability: During the training process, the model can fit the training data well and show similar performance on the unseen validation set and test set. This means that the neural network model can effectively learn the features and patterns in the training data.

[0090] (2) The model is neither overfitted nor underfitted: Overfitting means that the model is too complex and overfits the training set data, resulting in poor performance on unknown data; underfitting means that the model is too simple and fails to fit an accurate pattern on the training set; the consistency of the three curves shows that the model has appropriate complexity, enabling it to achieve highly accurate prediction results on various datasets.

[0091] (3) The model has strong generalization ability: Generalization ability refers to the model's ability to predict unknown data, that is, the model's performance on unseen data; highly overlapping and convergent curves show that the model can generalize well to new data samples and has high reliability and adaptability.

[0092] Therefore, the training process curves further demonstrate the effectiveness and reliability of the BP neural network in solving the problem of local current monitoring throughout the battery pack charging process.

[0093] Step (5): Determine the data processing scheme

[0094] A magnetic field sensor is used to monitor the magnetic field strength data during the battery pack charging process online, and this data is used as input to a trained BP neural network to output the local current of each individual battery cell. Based on the local current of each individual battery cell, a data processing scheme is further designed to achieve safe monitoring and alarm functions during the charging process. First, by collecting sufficient amounts of individual battery cell current data under the same charging condition covering abnormal situations, a normal distribution model of the current of each individual battery cell is established. Figure 12 As shown.

[0095] Secondly, to monitor for anomalies, the local current of each individual battery cell is compared with the normal current distribution model. Based on the 3σ principle of normal distribution, a threshold value three times the standard deviation is set. The local current of each individual battery cell is compared with the mean of the normal current distribution model; if the difference exceeds this threshold, an anomaly alarm is immediately triggered. Simultaneously, by referring to safe charging current parameters, it can be determined whether the anomaly may damage the battery pack. If the anomaly exceeds the safe range, rapid measures are taken, such as stopping charging or reducing the charging rate, to ensure the safety of the battery pack.

[0096] In summary, to ensure the safety of lithium-ion battery charging, this invention proposes a lithium-ion battery charging safety monitoring method based on magnetic field strength sensing. This method involves establishing a battery pack magnetic field sensing model, considering the influence of temperature during modeling, and coupling electro-thermal-magnetic multi-physics fields. The model simulates the current-magnetic field strength of each individual battery cell, training a backpropagation (BP) neural network. The magnetic field strength data during the battery pack charging process is monitored online and used as input to the trained BP neural network, outputting the local current of each individual battery cell. The local current of each individual battery cell is then used to determine whether the charging process is abnormal. Experimental results show that the BP neural network, verified through simulation and experiment, can accurately predict the local current of each battery cell, with a correlation coefficient as high as 0.99063 with the actual current. This effectively monitors the charging process of the lithium-ion battery pack, sensing the current magnitude of each individual battery cell online, thereby identifying charging anomalies and locating abnormal local current cells. This eliminates the hidden dangers of local faults during charging caused by inconsistencies in individual battery cells and the lack of local parameters in overall battery pack monitoring methods, thus contributing to improving the safety of vehicle-mounted power battery charging.

[0097] The embodiments described above are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for monitoring the charging safety of lithium-ion batteries based on magnetic field strength sensing, characterized in that: Considering the coupling of electro-thermal-magnetic multiphysics fields, a magnetic field sensing model for the battery pack is established. Based on the simulation data of current-magnetic field intensity of each individual battery cell using the magnetic field sensing model of the battery pack, a BP neural network is established and trained. The magnetic field strength data of the battery pack during the online charging process is used as the input of the trained BP neural network to output the local current of each individual cell. The local current of each individual cell is used to determine whether the charging process is abnormal. The coupling of the electro-thermal-magnetic multiphysics fields is as follows: An electrochemical model of a lithium-ion battery is established based on the lithium-ion battery interface, and the distribution of current density in the entire battery electrochemical model is solved; a fluid heat transfer interface is introduced, and the heat conduction equation of the electrochemical-thermal multiphysics interface is used to calculate the internal temperature change of the battery during the charging process; a magnetic field interface is introduced, and the spatial magnetic field is calculated using the internal current distribution and temperature of the lithium battery as initial conditions, realizing the unidirectional coupling of the electric and thermal fields to the magnetic field; then, the internal current distribution and temperature distribution of the lithium battery are recalculated based on the spatial magnetic field, realizing the reverse coupling of the magnetic field to the electric and thermal fields. Then, the recalculated current distribution and temperature are used as conditions to calculate the magnetic field. This process is repeated until the difference between the average values ​​of two adjacent spatial magnetic field data is less than the set value.

2. The lithium-ion battery charging safety monitoring method according to claim 1, characterized in that, The charging process is judged by the local current of each individual battery cell. Specifically, the local current of each individual battery cell is compared with the mean value of the current normal distribution model. If the difference exceeds the set threshold, an abnormal alarm will be triggered immediately.

3. The lithium-ion battery charging safety monitoring method according to claim 2, characterized in that, By referring to the charging current safety parameters, it can be determined whether the abnormality may damage the battery pack. If the abnormality exceeds the safety range, corresponding measures should be taken.

4. The lithium-ion battery charging safety monitoring method according to claim 1, characterized in that, An air domain is added to the external magnetic field sensing model of the battery pack.

5. The lithium-ion battery charging safety monitoring method according to claim 1, characterized in that, When establishing the magnetic field sensing model of the battery pack, the battery is divided into a grid, including 37,076 grid vertices, 67,832 prisms, 8,896 triangles, 10,578 quadrilaterals and 3,356 edge cells.

6. The lithium-ion battery charging safety monitoring method according to claim 1, characterized in that, The hidden layer of the BP neural network contains 18 neurons.

7. The lithium-ion battery charging safety monitoring method according to claim 1, characterized in that, The battery pack magnetic field sensing model was built in COMSOL software, and the required boundary conditions were set for the model.

8. The lithium-ion battery charging safety monitoring method according to claim 1, characterized in that, The set value is 0.1uT.

Images