Method and device for comprehensive and accurate quantitative diagnosis of micro-overcharge faults of lithium ion batteries

Abstract

The present application belongs to the field of lithium ion batteries, in order to solve the problem that the current ion battery micro-overcharge fault diagnosis method is difficult to be applied online, a lithium ion battery micro-overcharge fault comprehensive accurate quantitative diagnosis method and device are provided. The lithium ion battery micro-overcharge fault comprehensive accurate quantitative diagnosis method comprises discharging, standing, charging and standing operations on the lithium ion battery in turn, and determining the low-frequency excitation parameters; obtaining the voltage response sequence of the lithium ion battery after the low-frequency excitation parameters are applied, and extracting the local incremental features and global morphological features of the voltage response sequence based on the domain knowledge and the minimum random convolution kernel transformation respectively; based on the local incremental features and the global morphological features, combining the pre-trained deep learning network, obtaining the micro-overcharge prediction value, and then according to the micro-overcharge degree range, obtaining the quantitative micro-overcharge degree. It is easy to be applied online, and can accurately quantify the micro-overcharge degree.

Description

Technical Field

[0001] This invention belongs to the field of lithium-ion batteries, and particularly relates to a method and device for comprehensive and accurate quantitative diagnosis of micro-overcharge faults in lithium-ion batteries. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Battery abuse is a significant cause of thermal runaway accidents in lithium batteries. Micro-overcharging is a typical and extremely insidious form of battery abuse, where a fully charged battery is incorrectly charged again with a further amount of charge. This can be induced by defects in the battery management system (BMS) charging strategy, sensor malfunctions, and battery inconsistencies. Batteries that have undergone micro-overcharging exhibit varying degrees of current collector damage, SEI film decomposition, and anode lithium plating at the microscopic level, leading to a decrease in usable capacity, abnormal lifespan degradation, and the potential for thermal runaway. Effective diagnosis of micro-overcharging is crucial. However, these microscopic phenomena are difficult to reflect through macroscopic parameters, making micro-overcharging faults highly concealed and extremely challenging to diagnose.

[0004] Currently, methods for diagnosing micro-overcharge faults in lithium-ion batteries fall into two categories: non-in-situ diagnostic methods and in-situ diagnostic methods. Non-in-situ diagnostic methods can employ scanning electron microscopy, transmission electron microscopy, X-ray diffraction, etc., but the experimental equipment is expensive and requires damaging the battery pack structure, making online application difficult. In-situ diagnostic methods are mainly divided into model-based methods and data-driven methods. For model-based methods, existing technology provides a method for diagnosing micro-overcharge using electrochemical impedance spectroscopy (EIS) combined with a model. However, EIS has extremely high requirements for sampling accuracy and frequency, especially the sampling frequency, which is currently ten to tens of times higher than that of battery management systems (BMS), making online application difficult. For data-driven methods, existing technology provides a method for detecting micro-overcharge in lithium-ion battery modules based on pressure characteristics. However, integrating an accurate and reliable stress sensor into the BMS still faces cost, structural, and technical bottlenecks.

[0005] In summary, current methods for diagnosing micro-overcharge faults in lithium-ion batteries are not suitable for the low-frequency sampling scenarios of actual BMS, require additional sensors, have complex parameter identification processes, and may even require damage to the battery pack structure, making them difficult to apply online. Summary of the Invention

[0006] To address the aforementioned technical issues, this invention provides a comprehensive and accurate quantitative diagnostic method and device for micro-overcharge faults in lithium-ion batteries. This method can fully excite the battery in a short time and is more suitable for low-frequency sampling scenarios in actual BMS. The diagnostic process uses only voltage signals, requiring no additional sensors or complex parameter identification processes, making it easy to apply online. It not only has the advantages of high detection rate, high detection accuracy, and low false alarm rate, but also accurately quantifies the degree of micro-overcharge.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The first aspect of the present invention provides a comprehensive and accurate quantitative diagnostic method for micro-overcharge faults in lithium-ion batteries.

[0009] In one or more embodiments, a comprehensive and accurate quantitative diagnosis method for micro-overcharge faults in lithium-ion batteries is provided, including:

[0010] The lithium-ion battery was subjected to discharge, rest, charge and rest operations in sequence. The zero-state response of the lithium-ion battery during discharge, the zero-input response of the discharge, the zero-state response of the charge and the zero-input response of the charge were obtained accordingly, and the low-frequency excitation parameters were determined.

[0011] The voltage response sequence of a lithium-ion battery after applying low-frequency excitation with defined parameters is obtained. Local incremental features and global morphological features of the voltage response sequence are extracted based on domain knowledge and minimal random convolution kernel transformation, respectively.

[0012] Based on local incremental features and global morphological features, combined with a pre-trained deep learning network, a micro-overfill prediction value is obtained. Then, based on the range of micro-overfill degree, a quantified micro-overfill degree is obtained. The deep learning network is used to characterize the relationship between the local incremental features and global morphological features and the optimal label of the micro-overfill degree.

[0013] As one implementation method, the process of selecting the optimal label for the degree of micro-overcharge is as follows:

[0014] By simulating different degrees of micro-overcharge conditions, all observable characteristics of lithium-ion batteries are obtained and a potential label matrix of micro-overcharge degree is formed. The label with the minimum loss is calculated by the loss function and taken as the best label of micro-overcharge degree.

[0015] As one implementation method, the monotonic constraint loss is calculated based on the battery degradation law of the potential label matrix.

[0016] As one implementation method, a loss function is obtained in the process of selecting the optimal label for the degree of micro-overcharge by accumulating the overcharge level and overcharge cycle loss.

[0017] As one implementation method, in order to keep the battery charge constant before and after excitation during the determination of low-frequency excitation parameters, the following needs to be met:

[0018] ;

[0019] in, Q and Q r These are the charge / discharge capacity and the rated capacity, respectively. Q Less than or equal to Qmax % of the state of charge; I n This represents the discharge amplitude of the battery when it is at rest. t ab The duration of continuous discharge. I p This refers to the charging amplitude when the battery is idle. t cd This refers to the continuous charging time.

[0020] As one implementation method, a voltage constraint is introduced to determine the settling time during the process of determining the low-frequency excitation parameters:

[0021] ;

[0022] in, V R It is a mapping from time to response voltage. t The settling time is defined as time t, with the start of the settling period as time 0. - and 0 + This represents the instant before time t and the instant after time 0; V s The steady-state voltage after a set resting time; objective function f ( t The goal is to cover more voltage changes in a shorter time. V f It is the voltage coverage factor; the constraint function g(t) equals the voltage change covered during the resting period divided by the voltage change over a long resting period, and g(t) is greater than or equal to V f .

[0023] As one implementation method, the voltage increment of each node is selected as a local incremental feature.

[0024] As one implementation method, in the deep learning model process, the global morphological features are compressed through the first fully connected layer to enhance the effective data density; the SOC is injected into the feature stream through a conditional network; the hidden layers of global morphological features and local incremental features are fused through a cross-attention mechanism to obtain an attention score; the score is converted into a probability distribution through the Softmax function to form an attention weight matrix; the value vector is weighted and summed using the attention weight matrix, and then added to the query vector, and the hidden layer information of the local incremental features is weighted and aggregated into the hidden layer of the global morphological features to achieve dynamic feature selection; the consistency of the feature distribution is enhanced through the LayerNorm function, and the fused features are output; the fused features are passed through a fully connected layer to output the predicted value.

[0025] A second aspect of the present invention provides a comprehensive and accurate quantitative diagnostic device for micro-overcharge faults in lithium-ion batteries.

[0026] In one or more embodiments, a comprehensive and accurate quantitative diagnostic device for micro-overcharge faults in lithium-ion batteries includes:

[0027] The low-frequency excitation parameter determination module is used to sequentially perform discharge, rest, charge and rest operations on the lithium-ion battery, and obtain the discharge zero-state response, discharge zero-input response, charging zero-state response and charging zero-input response of the lithium-ion battery, thereby determining the low-frequency excitation parameters.

[0028] The feature extraction module is used to obtain the voltage response sequence of a lithium-ion battery after applying low-frequency excitation with defined parameters. It extracts local incremental features and global morphological features of the voltage response sequence based on domain knowledge and minimal random convolution kernel transformation, respectively.

[0029] The micro-overfill prediction module is used to obtain micro-overfill prediction values ​​based on local incremental features and global morphological features, combined with a pre-trained deep learning network. Then, based on the range of micro-overfill levels, the quantified micro-overfill degree is obtained. The deep learning network is used to characterize the relationship between the local incremental features and global morphological features and the optimal label of the micro-overfill degree.

[0030] A third aspect of the present invention provides an electronic device.

[0031] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the comprehensive and accurate quantitative diagnosis method for micro-overcharge faults of lithium-ion batteries as described above.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] The present invention provides a comprehensive and accurate quantitative diagnosis method for micro-overcharge faults in lithium-ion batteries. The method involves sequentially discharging, resting, charging, and resting the lithium-ion battery to obtain the zero-state response, zero-input response, zero-state response, and zero-input response of the lithium-ion battery. This allows for the determination of low-frequency excitation parameters. The low-frequency excitation method is easy to implement, can fully excite the battery in a short time, and is more suitable for low-frequency sampling scenarios in actual BMS.

[0034] The diagnostic process of this invention uses only voltage signals, requiring no additional sensors or complex parameter identification processes, thus possessing high practical value. This invention fully extracts multi-scale feature datasets of voltage response sequences and accurately maps micro-overcharge tags based on an intelligent deep feature fusion network. It not only has the advantages of high detection rate, high detection accuracy, and low false alarm rate, but also accurately quantifies the degree of micro-overcharging. Attached Figure Description

[0035] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0036] Figure 1 This is a flowchart illustrating the comprehensive and accurate quantitative diagnosis method for micro-overcharge faults in lithium-ion batteries according to an embodiment of the present invention.

[0037] Figure 2 This is a schematic diagram of the structure of the lithium-ion battery micro-overcharge fault comprehensive and accurate quantitative diagnosis device according to an embodiment of the present invention;

[0038] Figure 3 This is a schematic diagram of an electronic device according to an embodiment of the present invention;

[0039] Figure 4 This is a schematic diagram illustrating the principle of comprehensive and accurate quantitative diagnosis of micro-overcharge faults in lithium-ion batteries according to an embodiment of the present invention.

[0040] Figure 5 This is the basic waveform and second-order equivalent circuit model of the embodiments of the present invention;

[0041] Figure 6 This refers to the charging capacity of each battery in different charging cycles according to an embodiment of the present invention.

[0042] Figure 7 This is a tag matrix representing the charge cycle-overcharge degree according to an embodiment of the present invention;

[0043] Figure 8 These are the prediction results of the network model in this embodiment of the invention on the test set;

[0044] Figure 9 This is the confusion matrix of the diagnostic results in this embodiment of the invention. Detailed Implementation

[0045] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0046] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0047] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0048] Figure 1 This is a flowchart illustrating a comprehensive and accurate quantitative diagnosis method for micro-overcharge faults in lithium-ion batteries, as described in an embodiment of the present invention. (Combined with...) Figure 1 and Figure 4 The comprehensive and accurate quantitative diagnosis method for micro-overcharge faults in lithium-ion batteries shown in this embodiment may include the following steps S101 to S103.

[0049] The specific implementation process of steps S101 to S103 is as follows:

[0050] Step S101: Perform discharge, rest, charge and rest operations on the lithium-ion battery in sequence, and obtain the discharge zero-state response, discharge zero-input response, charging zero-state response and charging zero-input response of the lithium-ion battery, so as to determine the low-frequency excitation parameters.

[0051] By applying deterministic excitation to the battery and measuring its response, the battery characteristics can be stably reflected, thus effectively solving the problem of poor applicability caused by overfitting in data-driven methods. This invention proposes a low-frequency excitation method, the basic waveform of which is as follows: Figure 5 As shown. Figure 5 As shown, there are nine nodes in the voltage response within one excitation current cycle. At time a, the current changes from 0 to... I n The voltage response jumps from node 1 to node 2, and then slowly decreases from node 2 to node 3 under the influence of the discharge current; at time b, the current changes from... I n When the voltage changes to 0, the voltage response jumps from node 3 to node 4, and then slowly increases from node 4 to node 5 under polarization; at time c, the current changes from 0 to... I p The voltage response jumps from node 5 to node 6, and then slowly increases from node 6 to node 7 under the influence of the charging current; at time d, the current changes from... I p When the voltage drops to 0, the voltage response jumps from node 7 to node 8, and then slowly decreases from node 8 to node 9 under polarization. Based on the second-order equivalent circuit model and Kirchhoff's laws, the following can be obtained progressively: Figure 5 The terminal voltages of nodes 1 to 9 in the medium voltage response VFor a low-frequency excitation cycle, firstly, while the battery is at rest, an amplitude of [value missing] is applied. I n The discharge lasts for a period of time. t ab This is done to obtain the battery's zero-state discharge response. Then, the battery is allowed to rest. t bc To obtain the zero-input discharge response. Then, an amplitude of... I p Duration is t cd Charge the battery to obtain a state-of-charge response. Finally, allow the battery to rest. t de This allows for the acquisition of the battery's zero-input charging response. In this way, low-frequency excitation can fully excite the battery, resulting in a voltage response sequence that includes the complete discharge and charge responses. The information contained within can support micro-overcharge diagnostics.

[0052] To ensure that the battery charge remains constant before and after excitation, the following must be satisfied:

[0053] (1);

[0054] in, Q and Q r These are the charge / discharge capacity and the rated capacity, respectively. To avoid voltage interference caused by changes in SOC, Q It should be less than or equal to Q max % of SOC. A voltage constraint is introduced to determine the resting time:

[0055] (2);

[0056] in, V R It is a mapping from time to response voltage. t The settling time is defined as time t, with the start of the settling period as time 0. - and 0 + This represents the instant before time t and the instant after time 0. V s The steady-state voltage after a set resting time. Objective function. f ( t The goal is to cover more voltage changes in a shorter time. Simultaneously, to ensure sufficient voltage response information is obtained, a voltage coverage factor is introduced. V f The constraint function g(t) equals the voltage change during the resting period divided by the voltage change over a long resting period; g(t) should be greater than or equal to... V f.

[0057] Step S102: Obtain the voltage response sequence of the lithium-ion battery after applying low-frequency excitation with determined parameters, and extract the local incremental features and global morphological features of the voltage response sequence based on domain knowledge and minimal random convolution kernel transformation, respectively.

[0058] Current excitation induces a voltage response. Local incremental features and global morphological features of the voltage response sequence are extracted based on domain knowledge and minimal random convolution kernel transformation, respectively.

[0059] like Figure 5 As shown, based on the second-order equivalent circuit model and Kirchhoff's laws, the following can be obtained progressively. Figure 5 Terminal voltages of nodes 1-9 in the medium voltage response V :

[0060] (3);

[0061] in, V OCV Open circuit voltage, V C1 and V C2 R1 and R2 are the voltages across polarization capacitors C1 and C2, respectively. R0 is the equivalent ohmic internal resistance, and R1 and R2 are the equivalent polarization internal resistances. V 2 to V 9. By performing incremental calculations sequentially, we can obtain:

[0062] (4);

[0063] The voltage subscript indicates the direction of the incremental calculation. This is the parameter vector of the RC network. f is an implicit function of the variable to the voltage increment, and its subscript is the same as the subscript of the voltage increment. Clearly, I n , I p , t ab Settling time t bc , t cd , t de When fixed, V 21 , V 43 , V 65 , V 87 It can directly reflect the internal resistance of the ohm. V 32 , V54 , V 76 , V 98 It also has a direct mapping relationship with x. Therefore, the voltage increment is not only easy to obtain, but also contains information about the parameters in the equivalent circuit, eliminating the need for a complex parameter identification process. Thus, the voltage increment at each node is selected as the local increment feature F. LI ,Right now:

[0064] (5);

[0065] Minimal random convolution kernel transform possesses characteristics such as random fixed-length convolution kernels, multi-scale dilation, no need for backpropagation, and positive proportional features, enabling it to extract multi-scale features of time series quickly and with high accuracy.

[0066] When the sampling rate is f s At that time, a voltage response sequence can be written as S =[ s 1, s 2, s 3,… s m ].in, s Let m be a sampling point, and m = ⌊ t ae / f s ⌋. To emphasize morphological characteristics, S Max-Min normalization is performed to obtain S L Furthermore, by n A dataset consisting of voltage response sequences M Can be written as M =[ S L1 ; S L2 ;…; S Ln The process of feature extraction using minimal random convolution kernel transformation is as follows:

[0067] (6);

[0068] Firstly M and feature length The Fit function takes the input as input and outputs the converter's parameters, including the expansion coefficient vector. D , eigenvectors of each expansion coefficient N D and bias vector B . D and N DGenerated based on logarithmic intervals, where N D The sum of all elements should equal . B The length is The Fit function runs only once, and then freezes the obtained transformer parameters.

[0069] Then, with M T , D , N D , B Using this as input, the global morphological characteristics of the voltage response are obtained. F GM . M T It is a dataset of voltage response sequences from which features are to be extracted. F GM Number of rows and M T Same, number of columns For a single voltage response sequence, its global morphological characteristics can be represented as follows:

[0070] (7);

[0071] in, The subscript represents an element in a single voltage response sequence. k The index of the element, with a value from 1 to... .

[0072] Step S103: Based on local incremental features and global morphological features, and combined with a pre-trained deep learning network, obtain micro-overfill prediction values, and then obtain the quantified micro-overfill degree according to the range of micro-overfill degree; wherein, the deep learning network is used to characterize the relationship between the local incremental features and global morphological features and the optimal label of micro-overfill degree.

[0073] Specifically, the process of selecting the optimal label for slight overfill is as follows:

[0074] By simulating different degrees of micro-overcharge conditions, all observable characteristics of lithium-ion batteries are obtained and a potential label matrix of micro-overcharge degree is formed. The label with the minimum loss is calculated by the loss function and taken as the best label of micro-overcharge degree.

[0075] Microscopic phenomena caused by micro-overcharging are difficult to reflect using macroscopic parameters. Therefore, in in-situ diagnostics, using macroscopic parameters to select micro-overcharging labels is particularly important. Overcharge cycles can be performed at different overcharge levels (i.e., different charging cut-off voltages) to simulate varying degrees of micro-overcharging conditions. Then, observable characteristics of individual cells are obtained through offline testing as potential labels for the degree of micro-overcharging. Specifically, after each overcharge operation, capacity testing is used to obtain characteristics such as charge / discharge capacity and constant voltage / constant current charging capacity as potential labels. For any observable characteristic with the potential to become the best label, the set of observable characteristics for all degrees of micro-overcharging is determined. C It can be written as:

[0076] (8);

[0077] in m and n These represent the overcharge level and overcharge cycle, respectively; a value of 0 for either indicates no fault. Clearly, C for m +1 line n A matrix with +1 column. For C Subtract the corresponding first column element from all elements to obtain a new increment matrix. C This is done to eliminate initial value offsets during fault-free conditions and to emphasize how observable characteristics change with overcharge cycles. Then, for... C I Obtain the label matrix by performing Max-Min normalization. L .

[0078] To select proxy tags, the tag matrix was analyzed based on battery degradation patterns. L Calculate the monotonic constraint loss:

[0079] (9);

[0080] in Loss L and Loss C These are overcharge level and overcharge cycle loss, respectively. L Label matrix L The elements in the array are indexed to indicate their row and column. `Func` is the function name, which is mapped to a Max function (maximum value function) and a Min function (minimum value function) when the overall labels are negatively correlated with the fault severity. When `Func` is mapped to a Min function, for the same overcharge level or overcharge cycle, the label should exhibit monotonically non-decreasing property with respect to the overcharge cycle or overcharge level. Loss values ​​will be accumulated if the constraint is not met. The analysis process is similar when `Func` is mapped to a Max function. Finally, Loss L and LossC The total loss is obtained by adding the losses in these two directions. Loss T .

[0081] For the latent label matrix formed by all observable properties, perform... Loss T Calculate and select the label with the minimum loss as the best label.

[0082] Extract F LI and F GM After identifying the features and determining the optimal labels with slight overfilling, a deep learning network is constructed to fit the non-linear relationship between features and labels. First, a fully connected layer is used to apply the F... GM Features are compressed to enhance effective data density:

[0083] (10);

[0084] in h GM , W 1. b 1 represents the hidden layer, weight matrix, and offset vector, respectively. Then, the SOC is injected into the feature stream through a conditional network; the process can be represented as:

[0085] (11);

[0086] in h SOC and h LI These are hidden layers based on SOC conditions and local incremental features, respectively. This represents the element-wise product. F SOC It is the battery's State of Charge (SOC) feature. Then, it is fused through a cross-attention mechanism. h GM and h LI , can be represented as:

[0087] (12);

[0088] in, Q , K , V These are the query, key, and value vectors, respectively. d k for Q , K , V Dimensions. Through Q and K dot product calculation h GM right h LIThe similarity at each position is the attention score. Then, the scores are converted into a probability distribution using the Softmax function to form the attention weight matrix. A Next, using A right V Perform a weighted sum, then with Q Adding means that h LI Information is weighted and aggregated h GM In this process, dynamic feature selection is implemented. Furthermore, the consistency of feature distribution is enhanced through the LayerNorm function, outputting the fused feature F. A .

[0089] Finally, F A The predicted value is output after the fully connected layer. :

[0090] (13);

[0091] In the micro-overcharge diagnosis process, firstly, a low-frequency excitation is applied to the battery to obtain a voltage response sequence. Then, the extracted multi-scale features of the voltage response are input into a network to obtain a micro-overcharge prediction value. The predicted value can be normalized to a range of 0-1 or 0-100 to intuitively quantify the degree of micro-overcharge. Alternatively, a threshold can be manually selected. T h To identify whether there is a history of micro-overcharging.

[0092] The experiment used ternary lithium batteries with rated capacity, charge / discharge cutoff voltage, and maximum charge / discharge rate of 1.5Ah, 4.2 / 2.5V, and 4C / 20C, respectively.

[0093] Q max The preferred setting is 1, take I n , I p All are 3A, duration t ab , t cd Both are 20 seconds. This is calculated at this point. Q It equals 1.1%, satisfying the charge constraint of formula (1). Voltage coverage factor V f The preferred setting is 0.6. The SOC range was obtained through charge / discharge resting experiments. V R and V s When the resting time t bc , tde When the value is within 10 seconds, in formula (2) f ( t It reaches its maximum value and then monotonically decays over time. However, it does not satisfy the condition in formula (2) within 10 seconds. g ( t Constraints. When the resting time... t bc , t de When the value is within 40 seconds, g ( t The value is approximately 0.68, which satisfies formula (2). In summary, the LFE parameters are determined as follows: t ab , t cd Both are 20 seconds. t bc , t de Both last 40 seconds, and the charging / discharging current is 3A.

[0094] Batteries numbered B1 to B6 were subjected to ten charging cycles at cutoff voltages of 4.2V, 4.2V, 4.3V, 4.4V, 4.6V, and 4.8V respectively. The charging capacity (charged capacity to rated capacity) of each battery in different charging cycles was recorded as follows: Figure 3 As shown, the charging capacity of B1 and B2 remains stable at 100%. The charging capacities of B3, B4, and B5 are also relatively stable, fluctuating slightly around 106.6%, 112.1%, and 120.6%, respectively. The charging capacity of B6 fluctuates around 127% for the first six charging cycles, then steadily increases to 153.8% after the seventh cycle.

[0095] Charging capacity (the ratio of charged capacity to rated capacity) such as Figure 6 As shown, the charging capacity of B1 and B2 remains stable at 100%. The charging capacities of B3, B4, and B5 are also relatively stable, fluctuating slightly around 106.6%, 112.1%, and 120.6%, respectively. The charging capacity of B6 fluctuates around 127% for the first six charging cycles, then steadily increases to 153.8% after the seventh cycle.

[0096] After each charging cycle, standard tests were performed to obtain potential labels. Then, within the 10% to 90% SOC range, 20 low-frequency excitations were injected every 10% SOC to obtain voltage response sequences at full SOC. Through these steps, 10680 voltage response sequences were obtained as a dataset.

[0097] Through capacity testing after each charging cycle, several observable characteristics were obtained as potential tags: charging capacity, discharging capacity, constant voltage capacity, and constant voltage capacity ratio (capacity during constant voltage phase divided by charging capacity). The bidirectional monotonic loss of these potential tags was calculated using formulas (8) and (9), and the results are shown in Table 1. Therefore, constant voltage capacity was selected as the optimal tag for micro-overcharging, and its overcharge cycle-overcharge degree tag matrix is ​​shown in Table 1. L like Figure 7 As shown.

[0098] Table 1. Two-way monotonic loss for different observable characteristics;

[0099]

[0100] It can be seen that the maximum label value for a healthy battery is 0.183, while the minimum label value for a slightly overcharged battery is 0.213. Therefore, the diagnostic threshold can be selected within this range. T h Furthermore, the tag matrix L The range is expanded from 0 to 1 to 0 to 100% to visually reflect the degree of micro-overcharge.

[0101] The dataset was divided into training and test sets in a 7:3 ratio. F was obtained. GM Time settings f n The value is 1680. The prediction results of the network model on the test set are as follows: Figure 8 As shown, the predicted value matches the actual value very well. The calculated root mean square error (RMSE) between the two is as low as 2.18%, proving the network's accurate quantification capability for micro-overcharging.

[0102] Furthermore, by setting a diagnostic threshold, the performance of the proposed method in diagnosing the presence or absence of micro-overcharge history can be analyzed from a binary classification perspective. T h When set to 19.3%, the confusion matrix of the diagnostic results is as follows: Figure 9 As shown, positive (P) and negative (N) represent the presence and absence of micro-overcharge faults, respectively.

[0103] The true positive rate, representing the detection rate of micro-overcharging, is calculated using TP / (TP+FN). The precision rate, reflecting the detection accuracy of micro-overcharging, is calculated using TP / (TP+FP). Simultaneously, the false positive rate, representing the false alarm rate of micro-overcharging, can be calculated using FP / (FP+TN). The calculated detection rate, detection accuracy, and false alarm rate are 98.17%, 98.56%, and 2.441%, respectively, demonstrating the superior performance of the proposed method in the diagnosis of micro-overcharging.

[0104] The method of this invention can not only be adapted to the existing BMS framework in terms of sampling requirements and sensor requirements, but also achieve micro-overcharge identification with high detection rate, high detection accuracy and low false alarm rate. Furthermore, it can quantify the degree of micro-overcharge with low RMSE, providing a comprehensive, accurate and quantitative digital diagnosis method for micro-overcharge faults.

[0105] This invention provides a comprehensive and accurate quantitative diagnosis method for micro-overcharge faults in lithium-ion batteries. First, low-frequency excitation parameters are determined based on a co-constraint of charge and voltage, thereby obtaining the battery's voltage response sequence under a fixed low-frequency excitation. Second, a micro-overcharge degree label is determined based on a bidirectional monotonic loss function. Then, on one hand, a minimal random convolutional kernel transform is used to efficiently extract global morphological features of the voltage sequence; on the other hand, physically meaningful local incremental features of the voltage sequence are extracted based on domain knowledge, providing multi-scale data support for diagnosis. Finally, a deep learning network based on a cross-attention mechanism is constructed to fuse multi-scale features, ultimately achieving intelligent digital identification and degree quantification of micro-overcharge.

[0106] like Figure 2 As shown, the comprehensive and accurate quantitative diagnostic device for micro-overcharge faults in lithium-ion batteries provided in this embodiment of the invention can be implemented in software. The comprehensive and accurate quantitative diagnostic device for micro-overcharge faults in lithium-ion batteries includes the following software modules: low-frequency excitation parameter determination module 201, feature extraction module 202, and micro-overcharge prediction module 203.

[0107] The following is an introduction to the functions of each software module in the comprehensive and accurate quantitative diagnostic device for micro-overcharge faults in lithium-ion batteries:

[0108] The low-frequency excitation parameter determination module 201 is used to sequentially perform discharge, rest, charge and rest operations on the lithium-ion battery, and obtain the discharge zero-state response, discharge zero-input response, charging zero-state response and charging zero-input response of the lithium-ion battery, thereby determining the low-frequency excitation parameters.

[0109] The feature extraction module 202 is used to obtain the voltage response sequence of a lithium-ion battery after applying low-frequency excitation with determined parameters, and extracts local incremental features and global morphological features of the voltage response sequence based on domain knowledge and minimal random convolution kernel transformation, respectively.

[0110] The micro-overfill prediction module 203 is used to obtain micro-overfill prediction values ​​based on local incremental features and global morphological features, combined with a pre-trained deep learning network, and then obtain the quantified micro-overfill degree according to the range of micro-overfill degree; wherein, the deep learning network is used to characterize the relationship between the local incremental features and global morphological features and the optimal label of micro-overfill degree.

[0111] The structure of the electronic device according to an embodiment of the present invention will be described in detail below. Figure 3This is a schematic diagram of the composition structure of an electronic device provided in an embodiment of the present invention. It can be understood that... Figure 3 The diagram shows only an exemplary structure of the electronic device, not the entire structure. Some or all of the structures shown may be implemented as needed.

[0112] The electronic device provided in this embodiment of the invention includes: at least one processor 301, a memory 302, a user interface 303, and at least one network interface 304. The various components in the comprehensive and accurate quantitative diagnostic device for micro-overcharge faults in lithium-ion batteries are coupled together via a bus system 305. It can be understood that the bus system 305 is used to realize the connection and communication between these components. In addition to a data bus, the bus system 305 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 3 The general designated all buses as Bus System 305.

[0113] The user interface 303 may include a monitor, keyboard, mouse, trackball, click wheel, buttons, touchpad, or touch screen.

[0114] It is understood that memory 302 can be volatile memory or non-volatile memory, or both. In this embodiment of the invention, memory 302 is capable of storing data to support the operation of the terminal. Examples of this data include any computer programs used to operate on the terminal, such as operating systems and applications. The operating system includes various system programs, such as framework layers, core library layers, driver layers, etc., used to implement various basic services and handle hardware-based tasks. Applications can include various applications.

[0115] In some embodiments, the lithium-ion battery micro-overcharge fault comprehensive and accurate quantitative diagnosis device 300 provided in this embodiment of the invention can be implemented using a combination of hardware and software. As an example, the lithium-ion battery micro-overcharge fault comprehensive and accurate quantitative diagnosis device 300 provided in this embodiment of the invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the lithium-ion battery micro-overcharge fault comprehensive and accurate quantitative diagnosis method provided in this embodiment of the invention. For example, the processor in the form of a hardware decoding processor can employ one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0116] As an example, processor 301 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., wherein the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0117] As an example of the hardware implementation of the lithium-ion battery micro-overcharge fault comprehensive and accurate quantitative diagnosis device 300 provided in this embodiment of the invention, the device provided in this embodiment of the invention can be directly executed by a processor 301 in the form of a hardware decoding processor. For example, it can be executed by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components to implement the lithium-ion battery micro-overcharge fault comprehensive and accurate quantitative diagnosis method provided in this embodiment of the invention.

[0118] The memory 302 in this embodiment of the invention is used to store various types of data to support the operation of the comprehensive and accurate quantitative diagnostic device for micro-overcharge faults in lithium-ion batteries, or to store data for execution. Figure 1 The program code for the method shown. Examples of this data include: any executable instructions for operation on a comprehensive and accurate quantitative diagnostic device for micro-overcharge faults in lithium-ion batteries, such as executable instructions that can be included in the executable instructions to implement the comprehensive and accurate quantitative diagnostic method for micro-overcharge faults in lithium-ion batteries according to embodiments of the present invention.

[0119] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including functions for executing... Figure 1 The program code for the method shown. In such an embodiment, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by the central processing unit, it performs the various functions defined in the apparatus of this application.

[0120] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0121] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A comprehensive and accurate quantitative diagnostic method for micro-overcharge faults in lithium-ion batteries, characterized in that, include: The lithium-ion battery was subjected to discharge, rest, charge and rest operations in sequence. The zero-state response of the lithium-ion battery during discharge, the zero-input response of the discharge, the zero-state response of the charge and the zero-input response of the charge were obtained accordingly, and the low-frequency excitation parameters were determined. The voltage response sequence of a lithium-ion battery after applying low-frequency excitation with defined parameters is obtained. Local incremental features and global morphological features of the voltage response sequence are extracted based on domain knowledge and minimal random convolution kernel transformation, respectively. Based on local incremental features and global morphological features, combined with a pre-trained deep learning network, a micro-overfill prediction value is obtained, and then the quantified micro-overfill degree is obtained according to the range of micro-overfill degree; wherein, the deep learning network is used to characterize the relationship between the local incremental features and global morphological features and the optimal label of micro-overfill degree. In determining the low-frequency excitation parameters, to ensure that the battery charge remains unchanged before and after excitation, the following must be satisfied: ; in, Q and Q r These are the charge / discharge capacity and the rated capacity, respectively. Q Less than or equal to Q max % of the state of charge; I n This represents the discharge amplitude of the battery when it is at rest. t ab The duration of continuous discharge. I p This refers to the charging amplitude when the battery is idle. t cd For continuous charging time; In determining the low-frequency excitation parameters, a voltage constraint is introduced to determine the settling time: ； in, V R It is a mapping from time to response voltage; t The settling time is defined as time t, with the start of the settling period as time 0. - and 0 + This represents the instant before time t and the instant after time 0; V s The steady-state voltage after a set resting time; objective function f ( t The goal is to cover more voltage changes in a shorter time. V f It is the voltage coverage factor; the constraint function g(t) equals the voltage change covered during the resting period divided by the voltage change over a long resting period, and g(t) is greater than or equal to V f .

2. The method for comprehensive and accurate quantitative diagnosis of micro-overcharge faults in lithium-ion batteries as described in claim 1, characterized in that, The process of selecting the optimal label for slight overcharge is as follows: By simulating different degrees of micro-overcharge conditions, all observable characteristics of lithium-ion batteries are obtained and a potential label matrix of micro-overcharge degree is formed. The label with the minimum loss is calculated by the loss function and taken as the best label of micro-overcharge degree.

3. The method for comprehensive and accurate quantitative diagnosis of micro-overcharge faults in lithium-ion batteries as described in claim 2, characterized in that, The monotonic constraint loss is calculated based on the battery degradation law for the potential label matrix.

4. The method for comprehensive and accurate quantitative diagnosis of micro-overcharge faults in lithium-ion batteries as described in claim 2, characterized in that, The loss function is obtained by accumulating the overcharge level and overcharge cycle loss in the process of selecting the optimal label for the degree of micro-overcharge.

5. The method for comprehensive and accurate quantitative diagnosis of micro-overcharge faults in lithium-ion batteries as described in claim 1, characterized in that, The voltage increment at each node is selected as a local incremental feature.

6. The method for comprehensive and accurate quantitative diagnosis of micro-overcharge faults in lithium-ion batteries as described in claim 1, characterized in that, In the deep learning model, the global morphological features are compressed through the first fully connected layer to enhance the effective data density; the SOC is injected into the feature stream through a conditional network; the hidden layers of global morphological features and local incremental features are fused through a cross-attention mechanism to obtain an attention score; the score is converted into a probability distribution through the Softmax function to form an attention weight matrix; the value vector is weighted and summed using the attention weight matrix, and then added to the query vector to apply weights to the hidden layer information of local incremental features and aggregate it into the hidden layer of global morphological features, thus achieving dynamic feature selection; the LayerNorm function is used to enhance the consistency of feature distribution and output the fused features; the fused features are then passed through a fully connected layer to output the predicted value.

7. A comprehensive and accurate quantitative diagnostic device for micro-overcharge faults in lithium-ion batteries, characterized in that, It is implemented using the steps of the comprehensive and accurate quantitative diagnosis method for micro-overcharge faults in lithium-ion batteries as described in any one of claims 1-6; including: The low-frequency excitation parameter determination module is used to sequentially perform discharge, rest, charge and rest operations on the lithium-ion battery, and obtain the discharge zero-state response, discharge zero-input response, charging zero-state response and charging zero-input response of the lithium-ion battery, thereby determining the low-frequency excitation parameters. The feature extraction module is used to obtain the voltage response sequence of a lithium-ion battery after applying low-frequency excitation with defined parameters. It extracts local incremental features and global morphological features of the voltage response sequence based on domain knowledge and minimal random convolution kernel transformation, respectively. The micro-overfill prediction module is used to obtain micro-overfill prediction values ​​based on local incremental features and global morphological features, combined with a pre-trained deep learning network. Then, based on the range of micro-overfill levels, the quantified micro-overfill degree is obtained. The deep learning network is used to characterize the relationship between the local incremental features and global morphological features and the optimal label of the micro-overfill degree.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the comprehensive and accurate quantitative diagnosis method for micro-overcharge faults in lithium-ion batteries as described in any one of claims 1-6.

Images