A dual-stage early warning method and system for lithium battery thermal runaway based on sound signals

Abstract

The present invention relates to the technical field of battery thermal runaway warning, and provides a dual-level warning method and system for lithium battery thermal runaway based on sound signals, including: obtaining a battery sound signal sequence; identifying outliers in the battery sound signal sequence, and performing a first-level thermal runaway warning when an abnormal data point exists; extracting time-frequency domain features from the abnormal data point, and using an extreme gradient boosting tree algorithm optimized based on a sparrow optimization algorithm to identify whether there is thermal runaway expansion sound, thereby performing a second-level thermal runaway warning; the extreme gradient boosting tree algorithm optimized based on the sparrow optimization algorithm uses the sparrow optimization algorithm to adjust the number of iterations, learning rate, and decision tree depth of the extreme gradient boosting tree algorithm to the optimal parameters. A dual-level thermal runaway warning strategy is used to perform hierarchical identification of general anomalies or deep anomalies, effectively improving the recognition accuracy of early minor abnormal sound signals.

Description

Technical Field

[0001] The present invention relates to the technical field of battery thermal runaway warning technology, and in particular to a dual-stage warning method and system for lithium battery thermal runaway based on sound signals. Background Art

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

[0003] Lithium-ion batteries will experience performance degradation under conditions of electrical, thermal, and mechanical abuse, posing a risk of thermal runaway, which may eventually lead to safety accidents such as battery fires and explosions. Effectively detecting thermal runaway signals in the early stages of battery failure is an important means to avoid safety accidents. However, battery thermal runaway is extremely hidden in the early stages and highly sudden in the late stages, making it extremely difficult to accurately warn of it. Therefore, how to prevent problems before they occur and achieve advance warning of thermal runaway is a key technical problem that needs to be solved urgently.

[0004] At present, patents CN202410203187.9 and CN202410362650.4 use the temperature characteristics of lithium-ion batteries as early warning signals for battery thermal runaway. Patents CN202211310144.8 and CN202410285376.5 use the voltage characteristics of lithium-ion batteries as early warning signals for battery thermal runaway. Both of these identification methods can achieve the purpose of improving the safety of lithium-ion batteries.

[0005] However, temperature and voltage signals have obvious lags in battery thermal runaway, and are often not detected in the early stages of thermal runaway, resulting in delayed warnings, and early warnings of thermal runaway cannot be effectively and timely achieved. Patent CN111007461A builds a lithium battery thermal runaway positioning system based on acoustic signals based on the XGBoost algorithm, which achieves thermal runaway warnings by detecting the sound of safety valve discharge. However, since soft-pack batteries are not equipped with safety valves, the early sounds of thermal runaway are difficult to detect, which will cause the method to fail. And when the battery has a safety valve discharge sound, the battery is already in a dangerous state and the warning time is short. Therefore, how to capture weak sound signals in the early stages of thermal runaway remains to be solved. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a dual-level early warning method and system for lithium battery thermal runaway based on sound signals. A dual-level thermal runaway early warning strategy is adopted to perform hierarchical identification of general abnormalities or deep abnormalities. First, outlier detection is performed on the sound signal to perform preliminary screening of shallow faults. Then, a thermal runaway expansion sound identifier is constructed based on the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm. The sparrow optimization algorithm is used to adjust the optimal parameters to improve the recognition accuracy of early tiny abnormal sound signals; thereby forming a second-level thermal runaway early warning.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] In a first aspect, the present invention provides a dual-stage early warning method for thermal runaway of a lithium battery based on sound signals, comprising:

[0009] Get the battery sound signal sequence;

[0010] Identify outliers in the battery sound signal sequence and issue a first-level thermal runaway warning when abnormal data points are present;

[0011] The time-frequency domain features are extracted from the detected abnormal data points, and the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm is used to identify whether there is a thermal runaway expansion sound, so as to perform a secondary thermal runaway warning; the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm uses the sparrow optimization algorithm to adjust the optimal parameters of the number of iterations, learning rate and decision tree depth of the extreme gradient boosting tree algorithm.

[0012] As an optional implementation, an isolation forest algorithm is used to identify outliers, and the Mel spectrum is used as the identification feature. The average path length of each data point in the isolation tree is calculated, and the anomaly score is calculated based on the average path length. The abnormal data point is identified by comparing the anomaly score with the set anomaly threshold.

[0013] As an alternative implementation, the average path length is:

[0014] ;

[0015] The anomaly scores are:

[0016] ;

[0017] Among them, s is the sample, m is the Mel spectrum feature, is the sample expectation.

[0018] As an optional implementation, the extracted time domain features are root mean square energy, and the extracted frequency domain features are Mel-frequency cepstral coefficients, resonance peak energy, and spectrum centroid.

[0019] As an optional implementation method, the process of using the sparrow optimization algorithm to adjust the optimal parameters includes: setting the range of the number of iterations, learning rate and decision tree depth, each sparrow represents a parameter combination including the number of iterations, learning rate and decision tree depth, defining a fitness function, updating the population position through the sparrow's foraging and vigilance strategies, gradually approaching the optimal parameter combination until the maximum number of iterations is reached to obtain the optimal parameter combination.

[0020] As an optional implementation, the loss function of the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm is:

[0021] ;

[0022] in, is the loss function between the key dynamic feature value and the predicted value, is the regularization term; is the i-th key dynamic eigenvalue, for The corresponding predicted value; n is the number of abnormal data points;

[0023] The goal of the extreme gradient boosting tree algorithm based on the sparrow optimization algorithm is to minimize the expected value of the loss function by constructing a decision tree, Taylor-expanding and derivatizing the loss function, and finally obtaining the optimal solution of the objective function as follows:

[0024] ;

[0025] Among them, Le is the number of leaves in the decision tree, and are the first-order gradient function and the second-order gradient function respectively, is the penalty coefficient.

[0026] In a second aspect, the present invention provides a dual-stage early warning system for thermal runaway of a lithium battery based on sound signals, comprising:

[0027] An acquisition module is configured to acquire a battery sound signal sequence;

[0028] The first-level warning module is configured to identify outliers in the battery sound signal sequence and issue a first-level thermal runaway warning when there are abnormal data points;

[0029] The secondary warning module is configured to extract time-frequency domain features from the detected abnormal data points, and use an extreme gradient boosting tree algorithm optimized based on the sparrow optimization algorithm to identify whether there is a thermal runaway expansion sound, so as to perform a secondary warning of thermal runaway; the extreme gradient boosting tree algorithm optimized based on the sparrow optimization algorithm uses the sparrow optimization algorithm to adjust the optimal parameters of the number of iterations, learning rate and decision tree depth of the extreme gradient boosting tree algorithm.

[0030] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in the first aspect is completed.

[0031] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0032] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in the first aspect.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] Based on the extraction of multi-domain features of early-stage thermal runaway sound waves, combined with the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm, this invention proposes a dual-level thermal runaway warning method for lithium-ion soft-pack batteries based on multi-domain fusion of sound signals. Compared with the existing lithium battery thermal runaway warning method, this invention has the following advantages:

[0035] (1) The present invention does not require the establishment of a mathematical model of the battery, but only relies on directly measured sound signals. The sound signal characteristics have the advantages of early warning time and high sensitivity, and can provide early warning of thermal runaway of lithium-ion batteries when parameters such as current, voltage and temperature have not yet shown significant changes.

[0036] (2) The present invention adopts a two-level thermal runaway warning strategy to perform hierarchical identification of general abnormal or deep abnormal samples. First, the isolation forest algorithm is used to detect outliers in the battery sound signal sequence, so as to perform a first-level thermal runaway warning. The fault can be quickly processed through the threshold, and the shallow fault can be preliminarily screened. Then, in order to make full use of the dynamic characteristics of the sound of lithium-ion batteries during the charging and discharging process, the number of features is expanded, and multi-domain analysis is used for abnormal data points in the second-level thermal runaway anomaly detection. The time-frequency domain features are extracted as key dynamic features. Based on the key dynamic features, the extreme gradient boosting tree optimized by the sparrow optimization algorithm is used to realize the second-level thermal runaway warning, which effectively improves the recognition accuracy of early small abnormal sound signals.

[0037] (3) The present invention adopts an extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm for secondary thermal runaway warning. The number of iterations, learning rate and decision tree depth in the extreme gradient boosting tree algorithm are important parameters that affect the accuracy of the algorithm. Artificial adjustment of parameters generally cannot achieve the optimal parameter solution. Therefore, the sparrow optimization algorithm is used to adjust the optimal parameters, improve classification performance, speed up calculation efficiency, and be able to more quickly identify thermal runaway phenomena, thereby reserving sufficient time for taking safety measures.

[0038] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0040] Figure 1 A flow chart of a dual-stage early warning method for thermal runaway of a lithium battery based on multi-domain fusion of sound signals provided in Example 1 of the present invention;

[0041] Figure 2 A thermal runaway secondary warning flow chart provided in Example 1 of the present invention;

[0042] FIG3 (a) is a schematic diagram of outlier detection of sound samples in a 3.5C rate cycling thermal runaway experiment provided in Example 1 of the present invention;

[0043] FIG3( b ) is a schematic diagram of sound outlier detection in a 110% SOC cycle overcharge thermal runaway experiment provided in Example 1 of the present invention;

[0044] FIG3 (c) is a schematic diagram of outlier detection of sound samples in a 125% SOC cycle overcharge thermal runaway experiment provided in Example 1 of the present invention;

[0045] Figure 4 A schematic diagram of the generalization ability evaluation results based on the SSA-XGBoost algorithm provided in Example 1 of the present invention;

[0046] FIG5 (a) is a diagram of the classification results of the training set based on the SSA-XGBoost algorithm provided in Example 1 of the present invention;

[0047] FIG5( b ) is a diagram of the classification results of the test set based on the SSA-XGBoost algorithm provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0048] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0049] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0050] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "comprise" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0051] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0052] Example 1

[0053] This embodiment provides a dual-stage early warning method for thermal runaway of a lithium battery based on sound signals, comprising:

[0054] Get the battery sound signal sequence;

[0055] Identify outliers in the battery sound signal sequence and issue a first-level thermal runaway warning when abnormal data points are present;

[0056] The time-frequency domain features are extracted from the detected abnormal data points, and the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm is used to identify whether there is a thermal runaway expansion sound, so as to perform a secondary thermal runaway warning; the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm uses the sparrow optimization algorithm to adjust the optimal parameters of the number of iterations, learning rate and decision tree depth of the extreme gradient boosting tree algorithm.

[0057] The above-mentioned dual-level thermal runaway warning method is as follows: Figure 1 As shown, first, the isolation forest algorithm is used to detect outliers in the collected battery sound signal sequence. According to the set threshold, the sound points with abnormal scores are marked as abnormal data points, so as to perform a first-level thermal runaway warning. The fault can be quickly processed through the threshold, but only shallow faults can be preliminarily screened.

[0058] On the basis of abnormal data points detected by the first-level thermal runaway warning, the extreme gradient boosting tree algorithm optimized by the Sparrow Search Algorithm is used to perform second-level thermal runaway anomaly detection on samples of multi-dimensional key dynamic features. In the second-level thermal runaway anomaly detection, the abnormal data points are firstly analyzed in multiple domains. In the time domain, the root mean square energy (RMS) is selected as the key dynamic feature; in the frequency domain, the Mel-Frequency Cepstral Coefficients (MFCC), resonance peak energy and spectrum centroid are selected as the key dynamic features. Based on the above key dynamic features, the extreme gradient boosting tree (eXtreme Gradient Boosting, XGBoost) optimized by the Sparrow Search Algorithm (SSA) is used to distinguish the thermal runaway abnormal values ​​from the normal operating sound, thereby realizing the second-level thermal runaway warning.

[0059] The dual-stage thermal runaway warning method for lithium-ion soft-pack batteries proposed in the present invention does not require the establishment of a mathematical model of the battery, but only relies on directly measured sound sampling data. Based on multi-domain analysis, a sound recognition system is constructed by integrating multiple key dynamic features in the time domain and frequency domain, thereby improving the accuracy of system thermal runaway recognition.

[0060] The method of this embodiment is described in detail below.

[0061] Since the thermal runaway experiment of lithium-ion batteries has the characteristics of long duration and large amount of information, a fast-updating outlier monitor is required to realize real-time monitoring of the working status of lithium-ion batteries. In view of the high computational efficiency and strong robustness of the isolation forest algorithm, this embodiment establishes an outlier monitor based on the isolation forest for the sound data of the thermal runaway experiment of lithium-ion batteries.

[0062] Assume that there are s experimental data of thermal runaway experiment of lithium-ion battery. Firstly, multiple subsamples are randomly selected from the thermal runaway experimental data of lithium-ion battery, and an isolated tree is constructed for each subsample.

[0063] The expansion sound of lithium-ion soft-pack batteries during thermal runaway is weak, and there is no significant difference in the amplitude of the experimental data. Therefore, the Mel-spectrogram feature is selected as the recognition feature. The Mel-spectrogram feature is the spectrum representation of the audio signal on the Mel scale. The Mel scale is a frequency scale close to the human auditory perception. Therefore, the selection of the Mel-spectrogram feature can well reflect the differences in the human ear's perception of different frequencies.

[0064] The data is recursively split by randomly cutting the selected Mel-spectrogram features until the data points on the nodes are isolated. Since outliers are small in number and have large differences from normal values, they are more likely to be isolated in the isolation forest and have a shorter average path.

[0065] Therefore, the average path length of each data point in the isolation tree is calculated to identify abnormal data points. Average path length for:

[0066] (1)

[0067] Then, the average path length is used to calculate the anomaly score, which is:

[0068] (2)

[0069] Among them, s is the sample, m is the Mel spectrum feature, is the sample expectation.

[0070] In order to improve the stability of thermal runaway anomaly detection for lithium-ion batteries and reduce the recognition rate of sensitive anomalies, the anomaly threshold is set to 0.1% in this embodiment. Finally, based on the comparison between the anomaly score and the anomaly threshold, the abnormal data points where thermal runaway may occur are identified, thereby realizing the first-level anomaly detection of thermal runaway of lithium-ion batteries. When a first-level thermal runaway anomaly occurs, a fault warning should be issued to remind.

[0071] Since there are certain limitations in judging thermal runaway using only a single feature Mel spectrum, in order to fully utilize the dynamic characteristics of the sound during the charging and discharging process of lithium-ion batteries and expand the number of features, multi-domain analysis is used in the secondary anomaly detection of thermal runaway. In the time domain, the root mean square energy is selected as the key dynamic feature of the sound recognition system. In the frequency domain, the Mel frequency cepstral coefficient, resonance peak energy and spectrum centroid are selected as key dynamic features.

[0072] XGBoost is a gradient boosting decision tree algorithm that can automatically learn the complex relationships between key dynamic features and is suitable for processing high-dimensional data with nonlinear relationships. The Sparrow Optimization Algorithm can automatically search for the optimal hyperparameter combination of the XGBoost model, improve classification performance, and speed up its computational efficiency. Based on the above advantages, this embodiment selects the XGBoost algorithm (SSA-XGBoost) optimized by the Sparrow Optimization Algorithm for thermal runaway secondary anomaly detection.

[0073] When the XGBoost algorithm is used for classification, its core idea is to train the sound decision tree of each category based on an iterative method, gradually improve the model performance through the gradient boosting method, and thus achieve the best sound recognition effect.

[0074] First, the n abnormal data points obtained from the first-level anomaly detection of thermal runaway are defined as the data set for the second-level anomaly detection of thermal runaway. , represents the i-th key dynamic eigenvalue, express The corresponding predicted value is:

[0075] (3)

[0076] Among them, H is the number of sub-regression trees, is the corresponding learning function on the decision tree.

[0077] The corresponding objective function obtained through cumulative iteration is:

[0078] (4)

[0079] in, is the loss function between the real key dynamic feature value and the predicted value, is a regularization term to avoid overfitting.

[0080] The goal of the XGBoost algorithm is to minimize the expected value of the loss function by building a decision tree, so the Taylor expansion and derivation of equation (4) are performed, and the optimal solution of the objective function is finally obtained as:

[0081] (5)

[0082] Among them, Le is the number of leaves in the decision tree, and They are The first-order gradient function and the second-order gradient function, is the penalty coefficient.

[0083] In the XGBoost algorithm, the number of iterations, learning rate, and decision tree depth are important parameters that affect the accuracy of the algorithm. However, artificial adjustment of parameters generally cannot achieve the optimal parameter solution. Therefore, this embodiment uses the sparrow optimization algorithm to adjust the optimal parameters. First, initialize the sparrow optimization algorithm parameters, set the range of the number of iterations, learning rate, and decision tree depth, where each sparrow represents a parameter combination including the number of iterations, learning rate, and decision tree depth; define a fitness function to evaluate the performance of each sparrow; then, update the population position through the sparrow's foraging and vigilance strategies, and gradually approach the optimal parameter combination; repeat the process until the maximum number of iterations is reached, and finally output the optimal parameter combination.

[0084] Finally, the thermal runaway secondary anomaly detection flow chart based on the SSA-XGBoost algorithm is as follows: Figure 2As shown. The abnormal data points obtained by the thermal runaway first-level abnormality detection are used as input, and the SSA-XGBoost algorithm is used to identify whether the thermal runaway second-level abnormality detection value appears; if this value is identified, the power supply should be cut off immediately, ventilation and cooling should be carried out at the same time, the fault condition should be checked and the battery should be replaced.

[0085] In summary, this embodiment obtains a battery sound signal sequence and judges the collected sound signal through the above method; when the sound signal is diagnosed as a first-level thermal runaway abnormality, a fault warning is issued; when the sound signal is diagnosed as a second-level thermal runaway abnormality, the power supply should be quickly cut off and the battery should be checked, so as to achieve the purpose of thermal runaway diagnosis.

[0086] Use Cases:

[0087] Thermal runaway experiments were conducted on 3.9Ah lithium cobalt oxide soft-pack cells under different conditions in an accelerating calorimeter. The experimental conditions included 3.5C rate cycling, 110% battery state of charge (SOC) overcharge cycling, and 125% SOC overcharge cycling until thermal runaway occurred. During the experiment, the cell sound signals were sampled and recorded, and an outlier detection method based on isolation forest was used in the first-level anomaly detection of thermal runaway. The detection results under three experimental conditions were obtained, as shown in Figure 3 (a)-Figure 3 (c). Through the first-level anomaly detection of thermal runaway, sound samples that may contain battery thermal expansion sounds can be screened out. It can be seen that the scores corresponding to the detected abnormal points are significantly lower than other points.

[0088] After screening out suspicious abnormal sound samples, we conduct secondary anomaly detection for thermal runaway. In secondary anomaly detection, we perform root mean square solution and short-time Fourier transform on abnormal data through multi-domain analysis, and record the corresponding root mean square energy, Mel frequency cepstral coefficient, resonance peak energy and spectrum centroid. These four key dynamic features are used as input to the sound recognition system based on the SSA-XGBoost algorithm. 80% of the data set is used for training set and 20% for testing set, and the data is randomly divided.

[0089] Figure 4 The figure shows the generalization ability evaluation of the test set data in the sound evaluation system based on the SSA-XGBoost algorithm under different numbers of weak evaluators. Figure 4 It can be seen that the generalization ability of the system increases with the increase in the number of weak evaluators, and its generalization error continues to decrease. In particular, when more than 63 weak evaluators are used, the test set error is stable at 0.039, which shows that the model has good ensemble learning effect and generalization ability. Finally, the corresponding classification results of the training set and the test set are shown in Figure 5 (a)-Figure 5 (b).

[0090] In order to verify the accuracy of the experiment, the mean square error (MSE), accuracy and determination coefficient were used. As the evaluation index, the corresponding evaluation results are shown in Table 1.

[0091] Table 1 Evaluation results

[0092]

[0093] It can be seen that the sound recognition system based on SSA-XGBoost as a secondary anomaly detection of thermal runaway can effectively identify the tiny sound signals in the early stage of thermal runaway.

[0094] In summary, based on the acoustic wave information directly collected during the experiment, combined with multi-domain analysis, this paper uses the SSA-XGBoost algorithm to establish a dual-level thermal runaway warning strategy, which can realize the recognition of tiny sound signals in the early stage of thermal runaway of lithium-ion soft-pack batteries, improve the safety of lithium-ion soft-pack batteries during use, and has high application value.

[0095] Example 2

[0096] This embodiment provides a dual-stage early warning system for lithium battery thermal runaway based on sound signals, including:

[0097] An acquisition module is configured to acquire a battery sound signal sequence;

[0098] The first-level warning module is configured to identify outliers in the battery sound signal sequence and issue a first-level thermal runaway warning when there are abnormal data points;

[0099] The secondary warning module is configured to extract time-frequency domain features from the detected abnormal data points, and use an extreme gradient boosting tree algorithm optimized based on the sparrow optimization algorithm to identify whether there is a thermal runaway expansion sound, so as to perform a secondary warning of thermal runaway; the extreme gradient boosting tree algorithm optimized based on the sparrow optimization algorithm uses the sparrow optimization algorithm to adjust the optimal parameters of the number of iterations, learning rate and decision tree depth of the extreme gradient boosting tree algorithm.

[0100] As an optional implementation, an isolation forest algorithm is used to identify outliers, and the Mel spectrum is used as the identification feature. The average path length of each data point in the isolation tree is calculated, and the anomaly score is calculated based on the average path length. The abnormal data point is identified by comparing the anomaly score with the set anomaly threshold.

[0101] As an alternative implementation, the average path length is:

[0102] ;

[0103] The anomaly scores are:

[0104] ;

[0105] Among them, s is the sample, m is the Mel spectrum feature, is the sample expectation.

[0106] As an optional implementation, the extracted time domain features are root mean square energy, and the extracted frequency domain features are Mel-frequency cepstral coefficients, resonance peak energy, and spectrum centroid.

[0107] As an optional implementation method, the process of using the sparrow optimization algorithm to adjust the optimal parameters includes: setting the range of the number of iterations, learning rate and decision tree depth, each sparrow represents a parameter combination including the number of iterations, learning rate and decision tree depth, defining a fitness function, updating the population position through the sparrow's foraging and vigilance strategies, gradually approaching the optimal parameter combination until the maximum number of iterations is reached to obtain the optimal parameter combination.

[0108] As an optional implementation, the loss function of the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm is:

[0109] ;

[0110] in, is the loss function between the key dynamic feature value and the predicted value, is the regularization term; is the i-th key dynamic eigenvalue, for The corresponding predicted value; n is the number of abnormal data points;

[0111] The goal of the extreme gradient boosting tree algorithm based on the sparrow optimization algorithm is to minimize the expected value of the loss function by constructing a decision tree, Taylor-expanding and derivatizing the loss function, and finally obtaining the optimal solution of the objective function as follows:

[0112] ;

[0113] Among them, Le is the number of leaves in the decision tree, and are the first-order gradient function and the second-order gradient function respectively, is the penalty coefficient.

[0114] It should be noted that the above modules correspond to the steps described in Example 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer executable instructions.

[0115] In further embodiments, there is also provided:

[0116] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, it will not be described in detail here.

[0117] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0118] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0119] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in Example 1 is completed.

[0120] The method in Example 1 can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it is not described in detail here.

[0121] A computer program product includes a computer program, and when the computer program is executed by a processor, the method described in embodiment 1 is implemented.

[0122] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer executable instructions, such as instructions included in a program module, which are executed in a device on a real or virtual processor of the target to perform the process / method as described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functions of program modules can be combined or divided between program modules as needed. Machine executable instructions for program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.

[0123] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the computer or other programmable data processing device, causes the function / operation specified in the flow chart and / or block diagram to be implemented. The program code can be executed completely on a computer, partially on a computer, as an independent software package, partially on a computer and partially on a remote computer or completely on a remote computer or server.

[0124] In the context of the present invention, computer program codes or related data may be carried by any appropriate carrier to enable a device, apparatus or processor to perform the various processes and operations described above. Examples of carriers include signals, computer readable media, etc. Examples of signals may include electrical, optical, radio, acoustic or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0125] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0126] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A dual-stage early warning method for thermal runaway of lithium batteries based on sound signals, characterized in that: include: Get the battery sound signal sequence; Identify outliers in the battery sound signal sequence and issue a first-level thermal runaway warning when abnormal data points are present; The time-frequency domain features are extracted from the detected abnormal data points, and the extreme gradient boosting tree algorithm based on the sparrow optimization algorithm is used to identify whether there is thermal runaway expansion sound, so as to perform a secondary warning of thermal runaway; The extreme gradient boosting tree algorithm optimized based on the sparrow optimization algorithm is to use the sparrow optimization algorithm to adjust the optimal parameters of the number of iterations, learning rate and decision tree depth of the extreme gradient boosting tree algorithm; The isolation forest algorithm is used to identify outliers. The Mel spectrum is used as the identification feature. The average path length of each data point in the isolation tree is calculated. The anomaly score is calculated based on the average path length. The abnormal data points are identified by comparing the anomaly score with the set anomaly threshold. The extracted time domain features are root mean square energy, and the extracted frequency domain features are Mel-frequency cepstral coefficients, formant energy, and spectrum centroid.

2. A dual-stage early warning method for thermal runaway of a lithium battery based on sound signals as claimed in claim 1, characterized in that: The average path length is: ； The anomaly scores are: ； Among them, s is the sample, m is the Mel spectrum feature, is the sample expectation.

3. A dual-stage early warning method for thermal runaway of a lithium battery based on sound signals as claimed in claim 1, characterized in that: The process of adjusting the optimal parameters using the sparrow optimization algorithm includes: setting the range of the number of iterations, learning rate and decision tree depth, each sparrow represents a parameter combination including the number of iterations, learning rate and decision tree depth, defining the fitness function, updating the population position through the sparrow's foraging and vigilance strategies, and gradually approaching the optimal parameter combination until the maximum number of iterations is reached and the optimal parameter combination is obtained.

4. A dual-stage early warning method for thermal runaway of a lithium battery based on sound signals as claimed in claim 1, characterized in that: The loss function of the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm is: ； in, is the loss function between the key dynamic feature value and the predicted value, is the regularization term; is the i-th key dynamic eigenvalue, for The corresponding predicted value; n is the number of abnormal data points; The goal of the extreme gradient boosting tree algorithm based on the sparrow optimization algorithm is to minimize the expected value of the loss function by constructing a decision tree, Taylor-expanding and derivatizing the loss function, and finally obtaining the optimal solution of the objective function as follows: ； Among them, Le is the number of leaves in the decision tree, and are the first-order gradient function and the second-order gradient function respectively, is the penalty coefficient.

5. A dual-stage early warning system for thermal runaway of lithium batteries based on sound signals, characterized in that: include: An acquisition module is configured to acquire a battery sound signal sequence; The first-level warning module is configured to identify outliers in the battery sound signal sequence and issue a first-level thermal runaway warning when there are abnormal data points; The secondary warning module is configured to extract time-frequency domain features from the detected abnormal data points and use the extreme gradient boosting tree algorithm optimized by the sparrow optimization algorithm to identify whether there is thermal runaway expansion sound, so as to perform a secondary warning of thermal runaway; The extreme gradient boosting tree algorithm optimized based on the sparrow optimization algorithm is to use the sparrow optimization algorithm to adjust the optimal parameters of the number of iterations, learning rate and decision tree depth of the extreme gradient boosting tree algorithm; The isolation forest algorithm is used to identify outliers. The Mel spectrum is used as the identification feature. The average path length of each data point in the isolation tree is calculated. The anomaly score is calculated based on the average path length. The abnormal data points are identified by comparing the anomaly score with the set anomaly threshold. The extracted time domain features are root mean square energy, and the extracted frequency domain features are Mel-frequency cepstral coefficients, formant energy, and spectrum centroid.

6. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 4 is completed.

7. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 4.

8. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 4.

Images