Battery thermal runaway early warning method, device and equipment and storage medium

Abstract

The invention belongs to the technical field of battery safety management, and discloses a battery thermal runaway early warning method, device and equipment and a storage medium. The method comprises the following steps: acquiring an acoustic signal in a prefabricated cabin energy storage system, and extracting acoustic signal features according to a weighted dynamic Mel-frequency cepstral coefficient to obtain acoustic features; when the acoustic signal classification network model identifies that the acoustic feature is a thermal runaway acoustic signal, positioning the difference value of the acoustic signal to obtain the position of a target single battery; detecting the position of the target single battery to obtain gas concentration; when the gas concentration exceeds a preset concentration threshold value, early warning of thermal runaway of the battery is carried out; according to the method, the target position of the fault battery monomer is quickly positioned and the gas concentration is directionally monitored for early warning by combining the acoustic signal characteristics extracted by the weighted dynamic Mel-frequency cepstral coefficient with the acoustic signal classification network and the difference positioning algorithm, and the battery thermal runaway false alarm or missing alarm risk is remarkably reduced based on an acoustic-gas cooperative triggering mechanism; accurate early warning of the thermal runaway of the battery is realized.

Description

Technical Field

[0001] This invention relates to the field of battery safety management technology, and in particular to a battery thermal runaway early warning method, device, equipment, and storage medium. Background Technology

[0002] In recent years, electrochemical energy storage accidents have occurred frequently, posing significant safety risks to large-scale energy storage applications. Thermal runaway and its propagation in individual battery cells, modules, and systems are key causes of these accidents. Currently, the thermal runaway detection capabilities for large-scale lithium iron phosphate battery energy storage systems are insufficient, failing to adequately analyze data related to early-stage thermal runaway. This results in inaccurate thermal runaway detection, making it impossible to accurately detect batteries at risk of thermal runaway. Therefore, it is necessary to strengthen the detection of thermal runaway states in electrochemical energy storage systems, improve the risk monitoring and early warning capabilities for safe operation, refine corresponding safety control measures, and enhance the safety and reliability of electrochemical energy storage systems in engineering, commercial, and large-scale applications, thereby supporting the construction of new power systems. Summary of the Invention

[0003] In view of this, it is necessary to provide a method, device, equipment and storage medium for early warning of battery thermal runaway, so as to solve the technical problem that it is currently impossible to accurately detect battery thermal runaway.

[0004] To address the above problems, this invention provides a battery thermal runaway early warning method, comprising: Acoustic signals are acquired inside the prefabricated cabin energy storage system, and acoustic features are obtained by extracting features from the acoustic signals based on the weighted dynamic Mel-Cepstral coefficients. The prefabricated cabin energy storage system includes multiple individual battery locations. When the acoustic feature is identified as a thermal runaway acoustic signal based on a well-trained acoustic signal classification network model, the acoustic signal is used for differential localization to obtain the location of the target single cell. The gas concentration was obtained by detecting the position of the target single cell based on the ring-down spectrum of an open optical fiber ring cavity; When the gas concentration exceeds a preset concentration threshold, a battery thermal runaway warning is issued.

[0005] In one possible implementation, acquiring the acoustic signal inside the prefabricated energy storage system and extracting acoustic features from the acoustic signal based on the weighted dynamic Mel-frequency cepstral coefficients includes: Acquire acoustic signals from inside the prefabricated energy storage system, pre-emphasize and frame the acoustic signals to obtain multiple frame signals; Window function processing and Fourier transform are performed on each frame signal to obtain the signal spectrum of each frame signal; The Mel frequency is obtained by frequency conversion of the signal spectrum using a Mel scale filter. The dynamic features of the Mel frequency are extracted, and the acoustic features are obtained by weighted calculation based on the dynamic features.

[0006] In one possible implementation, the step of extracting the dynamic features of the Mel frequency and performing a weighted calculation based on the dynamic features to obtain acoustic features includes: The Mel frequency is subjected to logarithmic and discrete cosine transform processing to obtain static Mel cepstral coefficients; Extract the first-order and second-order dynamic features of the static Mel-frequency cepstral coefficients; The acoustic features are obtained by weighted fusion of the static Mel-frequency cepstral coefficients, the first-order dynamic features, and the second-order dynamic features.

[0007] In one possible implementation, when the acoustic feature is identified as a thermal runaway acoustic signal based on a well-trained acoustic signal classification network model, the differential localization of the acoustic signal to obtain the location of the target single cell includes: When the acoustic feature is identified as a thermal runaway acoustic signal based on a well-trained acoustic signal classification network model, the acoustic signal acquisition time points of the acoustic signal are obtained at different sound acquisition devices. The time delay of the acoustic signal is obtained based on the time point of the acoustic signal acquisition. The location of the target single battery cell is obtained by calculating the difference between the original positions of each sound acquisition device and the time delay.

[0008] In one possible implementation, the step of calculating the target single-cell battery location based on the difference between the original positions of each sound acquisition device and the distance difference between the sound signal and each sound acquisition device includes: Based on the spherical interpolation algorithm, the reference coordinates of the target single cell are obtained from the original positions of each sound acquisition device and the time delay. A set of objective equations is constructed based on the reference coordinates, and a reference solution is obtained by solving the set of objective equations using the least squares method. The three-dimensional spatial coordinates corresponding to the reference solution are used as the location of the target single cell.

[0009] In one possible implementation, before performing differential localization on the acoustic signal to obtain the location of the target single cell when the acoustic feature is identified as a thermal runaway acoustic signal according to a well-trained acoustic signal classification network model, the method further includes: Acquire different types of training sound signals, extract features from the training sound signals, and obtain training features; Obtain a preset convolutional neural network, wherein the learning rate of the preset convolutional neural network is a preset learning rate M, and the weights of the weighted dynamic MFCC are preset weights P; The preset convolutional neural network is trained using the training features to obtain a fully trained acoustic signal classification network model. The acoustic features are input into the fully trained acoustic signal classification network model to obtain acoustic signal categories, which include thermal runaway acoustic signals.

[0010] In one possible implementation, the gas concentration includes C2H4 concentration, C2H6 concentration, CO concentration, and H2 concentration; The step of issuing a battery thermal runaway warning when the gas concentration exceeds a preset concentration threshold includes: If the CO concentration is greater than or equal to a preset CO concentration or the H2 concentration is greater than or equal to a preset H2 concentration threshold, an early warning of battery thermal runaway will be issued. If the C2H4 concentration is greater than or equal to a preset C2H4 concentration threshold or the C2H6 concentration is greater than or equal to a preset C2H6 concentration threshold, a battery thermal runaway warning will be issued.

[0011] Furthermore, to achieve the above objectives, the present invention also proposes a battery thermal runaway early warning device, the battery thermal runaway early warning device comprising: The acquisition module is used to acquire acoustic signals inside the prefabricated cabin energy storage system, and to extract acoustic features from the acoustic signals based on the weighted dynamic Mel-frequency cepstral coefficients. The prefabricated cabin energy storage system includes multiple individual battery locations. The localization module is used to perform differential localization on the acoustic signal to obtain the location of the target single cell when the acoustic feature is identified as a thermal runaway acoustic signal according to the well-trained acoustic signal classification network model. The concentration detection module is used to detect the position of the target single cell based on the ring-down spectrum of the open optical fiber ring cavity to obtain the gas concentration; The early warning module is used to provide early warning of battery thermal runaway when the gas concentration exceeds a preset concentration threshold.

[0012] Furthermore, to achieve the above objectives, the present invention also proposes an electronic device comprising: a memory, a processor, a display, and a battery thermal runaway warning program stored in the memory and executable on the processor, the battery thermal runaway warning program being configured to implement the steps of the battery thermal runaway warning method as described above.

[0013] In addition, to achieve the above objectives, the present invention also proposes a storage medium storing a battery thermal runaway early warning program, wherein when the battery thermal runaway early warning program is executed by a processor, it implements the steps of the battery thermal runaway early warning method described above.

[0014] The beneficial effects of the above implementation method are as follows: by combining the acoustic signal features extracted by weighted dynamic Mel-Cepstral coefficients with an acoustic signal classification network and a differential localization algorithm, the faulty battery cell can be quickly located while identifying the characteristic acoustic signals of thermal runaway. Open optical path fiber ring cavity ring decay spectroscopy technology is used to perform directional gas monitoring at the target location. Based on a preset concentration threshold, an early warning is triggered. The acoustic-gas coordinated triggering mechanism significantly reduces the risk of false alarms or missed alarms by a single sensor, solves the problem of insufficient accuracy in traditional thermal runaway detection, and realizes accurate early warning of battery thermal runaway. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the first embodiment of the battery thermal runaway early warning method of the present invention; Figure 2 This is a schematic diagram illustrating the process of constructing a recognition classifier to classify and recognize acoustic signals in the first embodiment of the battery thermal runaway early warning method of the present invention. Figure 3 This is a complete schematic diagram of the battery thermal runaway early warning process in the first embodiment of the battery thermal runaway early warning method of the present invention; Figure 4 This is a flowchart illustrating the second embodiment of the battery thermal runaway early warning method of the present invention; Figure 5 This is a schematic diagram of the microphone arrangement and sound source location in the second embodiment of the battery thermal runaway early warning method of the present invention; Figure 6 This is a schematic diagram of sound source localization using spherical interpolation in the second embodiment of the battery thermal runaway early warning method of the present invention; Figure 7 This is a structural block diagram of the first embodiment of the battery thermal runaway early warning device of the present invention.

[0017] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention; Detailed Implementation The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] In the description of the embodiments of this application, unless otherwise stated, "a plurality of" means two or more.

[0019] In this embodiment of the invention, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, apparatus, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such process, method, product or device.

[0020] The naming or numbering of steps in the embodiments of the present invention does not mean that the steps in the method flow must be executed in the time / logical order indicated by the naming or numbering. The execution order of the named or numbered process steps can be changed according to the technical purpose to be achieved, as long as the same or similar technical effect can be achieved.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] The executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a battery thermal runaway early warning device. The following description uses a battery thermal runaway early warning device as an example to illustrate this embodiment and the subsequent embodiments.

[0023] This invention provides a battery thermal runaway early warning method, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the battery thermal runaway early warning method of the present invention.

[0024] In this embodiment, the battery thermal runaway early warning method includes steps S10~S40: Step S10: Acquire the acoustic signal inside the prefabricated cabin energy storage system, and extract the acoustic features from the acoustic signal based on the weighted dynamic Mel-frequency cepstral coefficients. The prefabricated cabin energy storage system includes multiple individual battery locations.

[0025] It is understandable that a prefabricated energy storage system can be understood as a battery compartment with a large number of battery modules, which may include battery modules, battery management system, temperature control system and fire protection system, etc.

[0026] It should be noted that the sound signal can be acquired by pre-installed microphones. The prefabricated energy storage system has multiple sets of microphones installed inside to collect sound signals, and each set can include four microphones set in different locations.

[0027] It should be noted that the weighted dynamic Mel-Cepstral coefficients are an optimization of the dynamic Mel-Cepstral coefficients. In thermal runaway acoustic detection, not all frequency components or all cepstral coefficients are equally important. Thermal runaway may generate acoustic fingerprints in specific frequency bands, requiring different importance weights to be assigned to different Mel-Cepstral filter channels or different cepstral coefficients. This weight is not arbitrary; it is usually obtained through training and learning with a large amount of data. In this embodiment, it can be set based on experience.

[0028] It should be understood that the acoustic features obtained by extracting features from the acoustic signal based on the weighted dynamic Mel-Cepstral coefficients are fused with first-order and second-order difference coefficients describing the dynamic changes of sound, and key features of the target acoustic event (such as battery thermal runaway) are highlighted by applying optimized weights to different frequencies or coefficients.

[0029] In one feasible implementation, step S10 may include steps A11 to A14: Step A11: Acquire the acoustic signal inside the prefabricated cabin energy storage system, pre-emphasize and frame the acoustic signal to obtain multiple frame signals.

[0030] It should be noted that pre-emphasis can be understood as enhancing the high-frequency part of the attenuated sound signal through a high-pass filter, making the high-frequency characteristics more prominent.

[0031] It should be understood that the sound source itself (such as a ruptured battery), the propagation medium, and the microphone all have a strong attenuation effect on high-frequency signals, resulting in very weak high-frequency components and a low signal-to-noise ratio in the sound signal. Pre-emphasis can compensate for this attenuation, making the high-frequency features more prominent. Furthermore, many key acoustic events (such as minute cracks in the battery casing or the initial ejection of gas) contain rich high-frequency information, and enhancing the high-frequency components helps subsequent algorithms capture these features more effectively.

[0032] It should be noted that framing can be understood as slicing a continuous long-duration sound signal into multiple short-time frames, i.e., multiple frame signals.

[0033] It should be understood that an acoustic signal is a long-term continuous signal, and the characteristics of the acoustic signal (such as the spectrum) change with time, which is called a non-stationary signal. However, within a sufficiently short time period (e.g., 10-40 milliseconds), the characteristics of sound can be considered relatively stable (quasi-stationary). To use the same method to analyze acoustic signals, it is necessary to ensure that each frame of the signal being analyzed is stable.

[0034] Furthermore, each frame of signal needs to be transformed from the time domain to the frequency domain, i.e., a Fourier transform is performed. The theoretical basis of the Fourier transform requires that the signal be stationary.

[0035] Step A12: Perform window function processing and Fourier transform on each frame signal to obtain the signal spectrum of each frame signal.

[0036] Understandably, applying a window function to each frame signal can be done by multiplying each frame signal with a window function, which can effectively reduce the discontinuities at both ends of each frame signal and make each frame signal more complete.

[0037] It should be understood that the Fourier transform in this embodiment refers to the Fast Fourier Transform (FFT). The Fast Fourier Transform can analyze which frequency components are contained in each frame signal, convert the time domain signal into a frequency domain representation, and obtain the frequency distribution of the frame signal.

[0038] It should be noted that the spectrum can be understood as a report of sound components, with a horizontal axis and a vertical axis. The horizontal axis is the frequency (unit: Hz), from 0 to the sampling rate / 2; the vertical axis is the amplitude or energy (usually converted to decibel dB scale, which is more in line with human ear perception).

[0039] Among them, a specific value is determined based on the horizontal and vertical axes, which represents the energy / decibels of a specific frequency range.

[0040] Step A13: Perform frequency conversion on the signal spectrum according to the Mel scale filter to obtain the Mel frequency.

[0041] It should be noted that the Mel-scale filter can be represented by the following formula:

[0042] in, f For frequency, f mei The frequency is Mel.

[0043] Step A14: Extract the dynamic features of the Mel frequency, and perform weighted calculations based on the dynamic features to obtain the acoustic features.

[0044] It should be noted that in acoustic detection of thermal runaway, not all frequency components or all cepstral coefficients are equally important. Thermal runaway may produce acoustic fingerprints in specific frequency bands (for example, crackling or hissing sounds at certain frequencies may be more reliable signs).

[0045] Furthermore, different cepstral coefficients carry different levels of information importance. Generally, lower-order coefficients contain more information about the spectral shape, while higher-order coefficients contain more detailed information and may also contain more noise.

[0046] It should be understood that weighting can be interpreted as assigning different importance weights to different Mel filter channels or different cepstral coefficients. These weights are obtained through training and learning with a large amount of data and can be set according to actual needs.

[0047] It should be noted that the step of extracting the dynamic features of the Mel frequency and performing weighted calculations based on the dynamic features to obtain acoustic features includes: performing logarithmic and discrete cosine transform processing on the Mel frequency to obtain static Mel cepstral coefficients; extracting the first-order and second-order dynamic features of the static Mel cepstral coefficients; and performing weighted fusion of the static Mel cepstral coefficients, the first-order dynamic features, and the second-order dynamic features to obtain acoustic features.

[0048] The process of taking the logarithm and performing discrete cosine transform on the Mel frequency to obtain the static Mel cepstral coefficients can be understood as the human ear's perception of sound intensity being logarithmic (i.e., when the volume doubles, the loudness we perceive is not linear). Taking the logarithm can compress the dynamic range, making the characteristics more consistent with human hearing.

[0049] Understandably, the Discrete Cosine Transform (DCT) can concentrate energy on the first few coefficients, and the resulting coefficients are independent of each other, which is very beneficial for subsequent classifiers to learn and classify.

[0050] Among them, static Mel-frequency cepstral coefficients (static MFCC) only describe the characteristics of a single frame, but sound changes over time, so it is necessary to further extract the dynamic change information of the sound signal, namely the first-order dynamic features and the second-order dynamic features.

[0051] It should be noted that the first-order dynamic feature is calculated by taking the first-order difference (i.e., derivative) between several adjacent static MFCCs, which represents the rate at which the MFCC coefficients change over time; the second-order dynamic feature can be understood as taking the difference again from the first-order dynamic feature, which represents the acceleration of the change of the first-order difference coefficients, describes the trend of the change rate itself, and makes the dynamic feature richer and more refined.

[0052] The weighted fusion of the static Mel-Cepstral coefficients, the first-order dynamic features, and the second-order dynamic features can be understood as performing weighted calculations based on pre-set weights for the static Mel-Cepstral coefficients, the first-order dynamic features, and the second-order dynamic features to obtain the acoustic features.

[0053] It is worth noting that a series of weighted dynamic Mel-Cepstral coefficient filter banks simulate the non-uniform perception characteristics of sound by the human ear, and combined with the characteristics of critical bandwidth, a filter bank generally uses 22-28 filters, and the most commonly used filter type is the triangular filter.

[0054] In practice, each group of collected acoustic signals is processed by noise reduction and effective segment cutting. Finally, multiple 2-second acoustic signals are extracted from each sample for feature extraction. These signals are then preprocessed, Fourier transformed, logarithmic energy calculated, and DCT operated on sequentially. The preprocessing includes pre-emphasis, framing, and windowing of the acoustic signals. The main purpose is to improve the signal-to-noise ratio of the acoustic signals, smooth the spike signals between adjacent frames, and perform periodic processing on the acoustic signals.

[0055] In this embodiment, high-frequency attenuation is compensated by pre-emphasis and framing, and non-stationary signals are transformed into analyzable short-time stationary segments. Further windowing and Fourier transform are applied to convert the sound from the time domain to the frequency domain, resulting in a spectrum that reveals the frequency distribution of sound energy. Using a Mel-scale filter bank, the linear spectrum is mapped onto a non-linear scale that conforms to the characteristics of human hearing, highlighting the resolution of the low-frequency region that the human ear is sensitive to, enhancing the discriminative power of features. Dynamic features (first-order and second-order differences) reflecting sound changes are extracted and weighted and fused, which not only describes the static characteristics of sound but also captures its dynamic behavior. Through weighting, the features most relevant to thermal runaway are highlighted, effectively improving the anti-interference ability and classification accuracy of the final acoustic features in real noisy environments.

[0056] The above are merely feasible implementations of step S10 provided in this embodiment. This embodiment does not specifically limit the specific implementation of step S10.

[0057] Step S20: When the acoustic feature is identified as a thermal runaway acoustic signal based on the well-trained acoustic signal classification network model, the acoustic signal is differentially located to obtain the location of the target single cell.

[0058] Understandably, the fully trained acoustic signal classification network model is obtained by training the AlexNet neural network with sample data, which is the collection of different types of signal characteristics generated within the prefabricated cabin energy storage system.

[0059] It should be understood that acoustic features can include various types of acoustic signals, such as interference signals, current signals, and so on.

[0060] It should be noted that differential positioning of the acoustic signal can be performed by using the Time Difference of Arrival (TDOA) positioning solution method to perform spherical interpolation for thermal runaway single cell positioning.

[0061] Understandably, by using TDOA technology, the three-dimensional coordinates of the acoustic signal characteristics are calculated by calculating the time difference between the arrival of the acoustic signal at sensors located at different positions inside the cabin, with an accuracy down to the extreme precision of a single battery cell.

[0062] In one possible implementation, steps A201 to A204 may be included before step S20: Step A201: Obtain different types of training sound signals, extract features from the training sound signals, and obtain training features.

[0063] It should be noted that by simulating the thermal runaway valve process of the battery in the energy storage compartment under various working conditions, a large amount of acoustic signal data was collected. After extracting features from the data, a database of different feature matrices was constructed, and a multidimensional feature parameter dataset was built.

[0064] It should be noted that the multidimensional feature parameter dataset extracts MFCC acoustic feature parameters, i.e., training features, to simulate the human ear's perception mechanism of sound signals. Both types of parameters can accurately characterize its key acoustic properties, providing a reliable basis for subsequent recognition tasks.

[0065] It should be emphasized that the key steps in building multiple recognition classifiers to classify and recognize sound signals are to extract features from the sound signals in the dataset and optimize the parameters of the recognition model, and to train the features into a convolutional neural network using different pattern recognition methods.

[0066] It should be noted that different types of training sound signals can be acquired by obtaining sound data of the safety valves of the batteries used in the energy storage compartment during thermal runaway. This data includes, but is not limited to, the location of the valve battery in different positions of the battery pack, and the distance of the valve position from the sound sampling point. Subsequently, the system analyzes and preprocesses the sampled sound through a built-in program, sampling as much data as possible to improve the accuracy of subsequent recognition.

[0067] In practical implementation, the process of constructing a recognition classifier to classify and recognize sound signals is as follows: Figure 2 As shown in the figure, the dataset was established by first simulating an experiment of thermal runaway caused by overheating. Acoustic signals in the early stage of thermal runaway were collected during the experiment. The collected acoustic signals in the early stage of thermal runaway were preprocessed by pre-emphasis, framing, windowing and noise fusion. After preprocessing, feature extraction was performed, and the extracted features were used as the training dataset. A recognition classifier was built based on the training dataset. The parameters of the recognition classifier were optimized. The results of each parameter optimization were compared with the evaluation index to complete the parameter training of the recognition classifier and obtain the parameter features, classifier model and optimal classification parameters.

[0068] Step A202: Obtain a preset convolutional neural network. The learning rate of the preset convolutional neural network is a preset learning rate M, and the weights of the weighted dynamic MFCC are preset weights P.

[0069] It should be noted that the default convolutional neural network is a deep convolutional neural network (AlexNet network), which introduces a linear rectified function (also known as the ReLU activation function). The default learning rate of the CNN network is set to 0.01. The default weights in the weighted dynamic MFCC can be set to 1 / 3 and 1 / 6. The structure of the convolutional neural network can be divided into an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.

[0070] It should be emphasized that at this point, the network has the strongest ability to identify positive and negative examples in the dataset. It can effectively avoid misidentifying other interfering acoustic signals as early acoustic signals of thermal runaway of lithium batteries, and will not misjudge thermal runaway signals as interference signals and thus miss them. At this point, the early warning capability of the recognition system is the strongest, and this parameter combination is the best recognition parameter combination for CNN networks.

[0071] Step A203: Train the preset convolutional neural network using the training features to obtain a fully trained acoustic signal classification network model.

[0072] It should be noted that each training feature is pre-labeled as either a thermal runaway acoustic signal or another interference signal.

[0073] Specifically, it can be understood as repeatedly training an initialized pre-defined convolutional neural network using a large amount of labeled, high-quality training features, i.e., sound feature data. Through an iterative cycle of "forward prediction - loss calculation - backpropagation - parameter update", a neural network that can accurately distinguish between thermal runaway sounds and normal environmental noise is finally obtained.

[0074] Step A204: Input the acoustic features into the fully trained acoustic signal classification network model to obtain acoustic signal categories, including thermal runaway acoustic signals.

[0075] Understandably, the acoustic characteristics here refer to the acoustic characteristics of the sound signals collected by the microphone in actual situations after processing.

[0076] In this embodiment, training data of multiple types of acoustic signals covering normal operating conditions and thermal runaway states are extracted, and a high-quality feature dataset with strong representation and strong resistance to interference is constructed by using weighted dynamic MFCC feature extraction. A convolutional neural network with a preset learning rate M and preset weights P is used to ensure the stability and convergence efficiency of the training process, while the weighting mechanism highlights the acoustic features most relevant to thermal runaway. The preset network is trained using the training features to obtain a fully trained classification model, which can automatically learn the complex mapping relationship between acoustic features and thermal runaway states. The real-time extracted acoustic features are input into the trained model, and the output includes classification results containing thermal runaway acoustic signals, realizing millisecond-level identification of thermal anomaly acoustic events. This provides accurate and reliable trigger signals for subsequent localization and gas verification, significantly reducing the risk of false alarms and false negatives.

[0077] The above are merely feasible implementation methods prior to step S20 provided in this embodiment. This embodiment does not specifically limit the specific implementation methods prior to step S20.

[0078] Step S30: Detect the position of the target single cell based on the ring-down spectrum of the open optical fiber ring cavity to obtain the gas concentration.

[0079] It should be noted that open-path fiber ring cavity ring decay spectroscopy (also known as open-path FLRDS) is an extremely sensitive trace gas detection technology. It performs final gas verification on the location of the individual thermal runaway cell and combines sound and gas to provide early warning of thermal runaway, greatly improving the accuracy and reliability of the warning.

[0080] It should be emphasized that the open optical path FLRDS technology is an improvement on the "cavity ring-down gas sensing technology". The main components include a laser, a function generator, a single-mode fiber, a coupler, an isolator, a collimator, a gas cavity with a built-in collimator (simulating open optical path gas detection), a photodetector, and an oscilloscope, etc., to detect gases related to thermal runaway in lithium iron phosphate battery energy storage chambers.

[0081] In practice, the concentration of gas released during battery thermal runaway is quantified based on the open optical path FLRDS; an open optical path design method based on fiber optic collimator (C-lens collimator) is used, and the open optical path parameters of the C-lens collimator are optimized for use.

[0082] Step S40: When the gas concentration exceeds a preset concentration threshold, a battery thermal runaway warning is issued.

[0083] It should be noted that the gases related to thermal runaway in the lithium iron phosphate battery energy storage chamber are C2H4, C2H6, CO, and H2.

[0084] In practice, an alarm is triggered when CO ≥ 100 ppm, and an early signal is triggered when H2 ≥ 900 ppm. Power is cut off and ventilation is activated when either CO or H2 reaches a certain level. An alarm is triggered when either C2H4 ≥ 2 ppm or C2H6 ≥ 5 ppm.

[0085] It should be understood that gas concentration detection can be achieved by first sampling the gas, and during the real-time gas acquisition phase, the sampling methods are respectively the top diffused open optical path and the bottom evacuation micro-flow sampling.

[0086] In one feasible implementation, step S40 may include steps A41-A42: Step A41: If the CO concentration is greater than or equal to a preset CO concentration or the H2 concentration is greater than or equal to a preset H2 concentration threshold, then an early warning of battery thermal runaway is issued.

[0087] It should be noted that the preset CO concentration can be 100 ppm.

[0088] It should be emphasized that the battery thermal runaway alarm will trigger a first-level linkage to produce corresponding actions, which are immediate power cut-off and forced ventilation.

[0089] It should be noted that the preset H2 concentration threshold can be 900 ppm. When the H2 concentration is greater than or equal to 900 ppm, it can be determined that it is an early thermal runaway.

[0090] It should be emphasized that the early warning of battery thermal runaway will also perform a first-level linkage action of immediate power cut-off and forced ventilation. The first-level linkage will be executed if either CO or H2 is triggered. The two do not inhibit each other, but rather verify each other. A difference of more than 20% will trigger a fault code.

[0091] Step A43: If the C2H4 concentration is greater than or equal to a preset C2H4 concentration threshold or the C2H6 concentration is greater than or equal to a preset C2H6 concentration threshold, a battery thermal runaway warning is issued.

[0092] It should be noted that the preset C2H4 concentration threshold can be set to 2 ppm, and the preset C2H6 concentration threshold can be set to 5 ppm.

[0093] To further clarify, battery thermal runaway early warning can be either an audible or visual warning. The audible and visual warning is a level two warning, which corresponds to a level two linkage action, namely, power reduction operation and notification of maintenance personnel.

[0094] It should be emphasized that triggering either C2H4 or C2H6 will trigger a Level 2 warning; triggering either CO or H2 will trigger a Level 1 linkage. The two do not inhibit each other, but rather verify each other. A difference of more than 20% will trigger a fault code.

[0095] It should be noted that gas values ​​can be displayed in real time both outside the energy storage compartment and on the mobile phones and computers of maintenance personnel. Once a certain concentration is reached, an alarm will be issued. It features real-time monitoring, accurate judgment, and timely alarm, which can significantly improve the safety and reliability of the power plant.

[0096] In this embodiment, the early warning is divided into two levels based on gas characteristics, enabling refined management of the thermal runaway process. At the same time, multiple characteristic gases are used as the basis for judgment, forming internal cross-validation, which avoids missed reports due to the failure of a certain gas sensor or the unknown source of a single gas, and provides more accurate early warning of thermal runaway.

[0097] The above are merely feasible implementations of step S40 provided in this embodiment. This embodiment does not specifically limit the specific implementation of step S40.

[0098] It is worth noting that this embodiment comprehensively judges the safety valve rupture caused by gas production during thermal runaway of a lithium iron phosphate battery. It locates the leak while simultaneously capturing sound and detecting gas concentration to obtain a gas leak warning level. This invention can be applied to the operation and maintenance of energy storage stations to detect and warn of early thermal runaway. This method facilitates battery status monitoring and significantly improves the safety and reliability of the power station.

[0099] In practical implementation, a complete battery thermal runaway early warning process can be referenced. Figure 3 , Figure 3 The system identifies and locates the thermal runaway battery cells inside the prefabricated energy storage system based on acoustic signals. Then, it quantifies the gas concentration released during thermal runaway based on the open optical path FLRDS method. Finally, it locates the battery based on the acoustic characteristics of the battery during thermal runaway and provides a battery fault thermal runaway early warning result based on the quantitative results of the gas detection device.

[0100] This embodiment provides a battery thermal runaway early warning method. By combining acoustic signal features extracted from weighted dynamic Mel-Cepstral coefficients with an acoustic signal classification network and a differential localization algorithm, the method can quickly locate faulty battery cells while identifying characteristic acoustic signals of thermal runaway. Open-path fiber optic ring cavity ring-down spectroscopy is used to perform directional gas monitoring at the target location. An early warning is triggered based on a preset concentration threshold. The acoustic-gas coordinated triggering mechanism significantly reduces the risk of false alarms or missed alarms from a single sensor, solves the problem of insufficient accuracy in traditional thermal runaway detection, and achieves accurate early warning of battery thermal runaway.

[0101] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 4 Step S20, the battery thermal runaway early warning method further includes steps S21~S23: Step S21: When the acoustic feature is identified as a thermal runaway acoustic signal based on the fully trained acoustic signal classification network model, the acoustic signal acquisition time points of the acoustic signal are obtained at different sound acquisition devices.

[0102] It should be noted that the classification of acoustic features by a fully trained acoustic signal classification network model is used as a prerequisite to avoid redundant localization or unnecessary localization calculations.

[0103] It should be noted that after identifying the acoustic signal as thermal runaway, further location needs to be determined based on the acoustic signal, the gas concentration at the location where the acoustic signal was generated needs to be determined, and the battery thermal runaway needs to be assessed from two directions.

[0104] It should be understood that the sound acquisition device refers to a microphone array installed in a prefabricated cabin according to a specific layout. The spatial coordinates of each microphone are precisely measured in advance.

[0105] It should be noted that the sound signal acquisition time point refers to the time point at which the same acoustic feature is acquired by various sound acquisition devices.

[0106] Step S22: Obtain the time delay of the sound signal based on the sound signal acquisition time point.

[0107] It should be noted that when a sound signal is emitted from the target sound source location, the time required for the sound wave to travel to microphones at different distances and angles is different. Therefore, the sound signals received by each microphone will have a phase difference. The TDOA-based positioning algorithm uses the phase difference to calculate the time delay value of the sound signal at the sound source location to travel to each microphone.

[0108] Furthermore, the time delay is multiplied by the speed of sound to obtain the distance difference between each microphone, and the equation based on the distance difference is solved.

[0109] The calculation involves determining the time delay of the target sound signal propagating to each microphone location, where the microphone arrangement and sound source location are as follows: Figure 5 As shown, m 1. m 2. m 3. m 4 are four-element microphones arranged along the long and short sides of the battery storage compartment. When the sound source is at position A, s 1. s 2. s 3. s 4 represents the distance from the sound source to the microphone, with the microphone as the reference point. m For reference, the time it takes for the sound source to reach the microphone is 1. t 1. The time of arrival at other microphones is recorded as follows: t 2. t 3. t4. Sound source arrives m 2. m The time difference of 3 is denoted as τ 21 , τ 21 = τ 2- τ 1, and so on, where τ 21 , τ 31 , τ 41 That is, the time delay value.

[0110] Furthermore, the time delay value is substituted into the algorithm to determine the positional relationship between the sound source location and the time delay values ​​of each microphone.

[0111] Step S23: Calculate the difference between the original positions of each sound acquisition device and the time delay to obtain the position of the target single battery cell.

[0112] In practice, a microphone is set up. m 1. Coordinate is ( x i , y i , z i ), known time delay value τ 21 , τ 31 , τ 41 If the speed of sound is 340 m / s, then the coordinates of the sound source location can be obtained by solving the following equation:

[0113] but , i =1, 2, 3, 4, d i The time delay value can be obtained by solving for it. At this time, the unknowns are the three coordinate values ​​of the sound source, three independent equations, and three unknowns to be solved. The sound source coordinates can be obtained from this system of equations. Note that the unit of time delay should be consistent when calculating.

[0114] Furthermore, since this system of equations is nonlinear and difficult to solve precisely, spherical interpolation is chosen as the most suitable sound source localization algorithm for lithium battery energy storage compartments. A schematic diagram of spherical interpolation for sound source localization can be found in [reference needed]. Figure 6 , Figure 6 middle m 0 is the origin of the coordinate system. r s It is the spatial vector of the sound source.s and m 0 spatial distance, r i It's a microphone. m i and reference microphone m Spatial distance between 0, d i0 Indicates two microphones m i , m 1. Spatial difference in distance from the sound source location T The transpose of the matrix can be expressed as follows: .

[0115] At the same time, based on the assumptions , Substitute It can then be transformed into:

[0116] The results were:

[0117] Since the sound signals were collected on-site during the experiment, errors were unavoidable. d i0 The accuracy is limited, and the above formula will generally not equal 0, thus introducing an error. ,make:

[0118] Set N Each microphone has a relative to the reference microphone. N Given a delay value, we can obtain... N The equations can be rewritten in matrix form as follows:

[0119] in,

[0120] like r s If it already exists, then

[0121] at this time s The mean square error is the smallest, among which Therefore, s Substitution We can obtain:

[0122] Polynomial expansion yields:

[0123] in:

[0124] but The two roots are:

[0125] Will Substituting the two roots, we can obtain the coordinates of the sound source. S SI :

[0126] In one feasible implementation, step S23 may include steps A231 to A233: Step A231: Based on the spherical interpolation algorithm, the reference coordinates of the target single cell are obtained according to the original position of each sound acquisition device and the time delay.

[0127] It should be understood that the sound signals collected by multiple microphones arrive at different microphones at different times, resulting in a time delay. Given the speed of sound, the time difference is converted into a distance difference by multiplying the distance difference by the speed of sound.

[0128] It should be noted that in three-dimensional space, the distance difference between one microphone and another is a constant. d The TDOA localization algorithm, which uses phase difference to calculate the time delay of the sound signal from the sound source to each microphone, calculates the distance difference between the microphones by multiplying it by the speed of sound. An equation is then set up based on this distance difference and solved. However, this system of equations is nonlinear and difficult to solve precisely. To solve this system, spherical interpolation is used. Based on the existing coordinates and their distance relationships, an error equation system is determined as the objective function. This objective function is then solved using the least squares method, and the final sound source location is determined based on the least squares solution.

[0129] Step A232: Construct a set of target equations based on the reference coordinates, and solve the set of target equations using the least squares method to obtain a reference solution.

[0130] It should be noted that the objective function can be solved using the least squares method to obtain the location of the sound source.

[0131] It should be noted that there are multiple microphone pairs in this embodiment, which can list multiple equations to form an overdetermined system of equations. The overdetermined system of equations usually does not have a unique solution because there are small errors in the measurement values ​​(time difference / distance difference). By using the least squares method, an optimal solution is found that minimizes the "sum of squares of errors" between this solution and all equations, which is the reference solution.

[0132] Step A233: Use the three-dimensional spatial coordinates corresponding to the reference solution as the target single cell position.

[0133] It should be noted that the reference solution is a three-dimensional spatial coordinate. By comparing this three-dimensional spatial coordinate with the battery map of the prefabricated compartment, it is possible to accurately locate which specific battery cell, module, or cluster has experienced thermal runaway.

[0134] In this embodiment, the nonlinear equation processed by spherical difference calculation is used to overcome measurement errors by using the least squares method, and finally outputs the most likely and accurate sound source location coordinates in three-dimensional space, providing an accurate location reference for further gas concentration detection.

[0135] The above are merely feasible implementations of step S23 provided in this embodiment. This embodiment does not specifically limit the specific implementation of step S23.

[0136] It is important to emphasize that during the real-time sound monitoring phase, a multi-level architecture is used to achieve closed-loop control of "sound-location-response," ensuring that the response is completed within the millisecond time window when the spray valve opens. Four microphones are placed in a group at the four corners of the energy storage compartment (about 1.5m above the ground) to collect fan and air conditioner noise as reference signals for active noise reduction. The dataset includes various types of scene noise (footsteps, electrical noise, electric drill noise, etc.). Once the target sound is detected, an alarm is immediately issued, and spatial positioning begins. Maintenance personnel receive text, image, and voice broadcasts on their mobile phones or computers in real time.

[0137] This embodiment provides a battery thermal runaway early warning method. Through a collaborative mechanism of acoustic recognition triggering and high-precision time-delay positioning, the positioning calculation is triggered after the acoustic classification model confirms that there is a high risk of thermal runaway, which effectively reduces the amount of computation. At the same time, by utilizing the time difference of sound waves arriving at sensor arrays located at different positions in the cabin, the specific physical coordinates of the individual battery cell that has experienced thermal runaway can be accurately located. Gas concentration detection is further performed on the positioning coordinates. Through dual verification of sound and gas concentration, rapid and accurate identification of thermal runaway in the prefabricated cabin is achieved.

[0138] To better implement the battery thermal runaway early warning method in the embodiments of the present invention, based on the battery thermal runaway early warning method, correspondingly, as follows: Figure 7 As shown, this embodiment of the invention also provides a battery thermal runaway early warning device, the battery thermal runaway early warning device 700 comprising: Acquisition module 701 is used to acquire acoustic signals inside the prefabricated cabin energy storage system and extract acoustic features from the acoustic signals based on weighted dynamic Mel-Cepstral coefficients. The positioning module 702 is used to perform differential positioning on the acoustic signal to obtain the location of the target single cell when the acoustic feature is identified as a thermal runaway acoustic signal according to the well-trained acoustic signal classification network model. The concentration detection module 703 is used to detect the position of the target single cell based on the ring-down spectrum of the open optical fiber ring cavity to obtain the gas concentration; The early warning module 704 is used to provide an early warning of battery thermal runaway when the gas concentration exceeds a preset concentration threshold.

[0139] The battery thermal runaway early warning device 700 provided in the above embodiments can realize the technical solutions described in the above battery thermal runaway early warning method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above battery thermal runaway early warning method embodiments, and will not be repeated here.

[0140] like Figure 8 As shown, the present invention also provides an electronic device 800. The electronic device 800 includes a processor 801, a memory 802, and a display 803. Figure 8 Only some components of the electronic device 800 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0141] In some embodiments, memory 802 may be an internal storage unit of electronic device 800, such as a hard disk or memory of electronic device 800. In other embodiments, memory 802 may also be an external storage device of electronic device 800, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 800.

[0142] Furthermore, the memory 802 may include both internal storage units of the electronic device 800 and external storage devices. The memory 802 is used to store application software and various types of data installed on the electronic device 800.

[0143] In some embodiments, processor 801 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 802 or process data, such as the battery thermal runaway early warning method of the present invention.

[0144] In some embodiments, display 803 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 803 is used to display information from electronic device 800 and to display a visual user interface. Components 801-803 of electronic device 800 communicate with each other via a system bus.

[0145] In some embodiments of the present invention, when the processor 801 executes the battery thermal runaway early warning program in the memory 802, the following steps can be implemented: acquiring the acoustic signal inside the prefabricated cabin energy storage system; extracting acoustic features from the acoustic signal based on the weighted dynamic Mel-Cepstral coefficients; when the acoustic feature is identified as a thermal runaway acoustic signal based on a well-trained acoustic signal classification network model, performing differential localization on the acoustic signal to obtain the location of the target single battery cell; detecting the location of the target single battery cell based on the ring-down spectrum of an open optical fiber ring cavity to obtain the gas concentration; and issuing a battery thermal runaway early warning when the gas concentration exceeds a preset concentration threshold.

[0146] It should be understood that when the processor 801 executes the battery thermal runaway warning program in the memory 802, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0147] Furthermore, the embodiments of the present invention do not specifically limit the type of electronic device 800 mentioned. Electronic device 800 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, electronic device 800 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0148] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the battery thermal runaway early warning method provided by the above methods. The method includes: acquiring acoustic signals inside the prefabricated cabin energy storage system; extracting acoustic features from the acoustic signals based on weighted dynamic Mel-Cepstral coefficients to obtain acoustic features; when the acoustic features are identified as thermal runaway acoustic signals according to a well-trained acoustic signal classification network model, performing differential localization on the acoustic signals to obtain the location of the target single cell; detecting the location of the target single cell based on the ring-down spectrum of an open optical fiber ring cavity to obtain the gas concentration; and issuing a battery thermal runaway early warning when the gas concentration exceeds a preset concentration threshold.

[0149] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0150] The battery thermal runaway early warning method provided by the present invention has been described in detail above. Specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of ​​the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for early warning of battery thermal runaway, characterized in that, include: Acoustic signals are acquired inside the prefabricated cabin energy storage system, and acoustic features are obtained by extracting features from the acoustic signals based on the weighted dynamic Mel-frequency cepstral coefficients. The prefabricated cabin energy storage system includes multiple individual battery locations. When the acoustic feature is identified as a thermal runaway acoustic signal based on a well-trained acoustic signal classification network model, the acoustic signal is used for differential localization to obtain the location of the target single cell. The gas concentration was obtained by detecting the position of the target single cell based on the ring-down spectrum of an open optical fiber ring cavity; When the gas concentration exceeds a preset concentration threshold, a battery thermal runaway warning is issued.

2. The battery thermal runaway early warning method as described in claim 1, characterized in that, The process of acquiring acoustic signals inside the prefabricated energy storage system and extracting acoustic features from the acoustic signals based on weighted dynamic Mel-frequency cepstral coefficients includes: Acquire acoustic signals from inside the prefabricated energy storage system, pre-emphasize and frame the acoustic signals to obtain multiple frame signals; Window function processing and Fourier transform are performed on each frame signal to obtain the signal spectrum of each frame signal; The Mel frequency is obtained by frequency conversion of the signal spectrum using a Mel scale filter. The dynamic features of the Mel frequency are extracted, and the acoustic features are obtained by weighted calculation based on the dynamic features.

3. The battery thermal runaway early warning method as described in claim 2, characterized in that, The step of extracting the dynamic features of the Mel frequency and performing weighted calculations based on the dynamic features to obtain acoustic features includes: The Mel frequency is subjected to logarithmic and discrete cosine transform processing to obtain static Mel cepstral coefficients; Extract the first-order and second-order dynamic features of the static Mel-frequency cepstral coefficients; The acoustic features are obtained by weighted fusion of the static Mel-frequency cepstral coefficients, the first-order dynamic features, and the second-order dynamic features.

4. The battery thermal runaway early warning method as described in claim 1, characterized in that, When identifying the acoustic feature as a thermal runaway acoustic signal based on a well-trained acoustic signal classification network model, the method of differential localization of the acoustic signal to obtain the location of the target single cell includes: When the acoustic feature is identified as a thermal runaway acoustic signal based on a well-trained acoustic signal classification network model, the acoustic signal acquisition time points of the acoustic signal are obtained at different sound acquisition devices. The time delay of the acoustic signal is obtained based on the time point of the acoustic signal acquisition. The location of the target single battery cell is obtained by calculating the difference between the original positions of each sound acquisition device and the time delay.

5. The battery thermal runaway early warning method as described in claim 4, characterized in that, The method of calculating the difference between the original positions of each sound acquisition device and the time delay to obtain the position of the target single battery cell includes: Based on the spherical interpolation algorithm, the reference coordinates of the target single cell are obtained from the original positions of each sound acquisition device and the time delay. A set of objective equations is constructed based on the reference coordinates, and a reference solution is obtained by solving the set of objective equations using the least squares method. The three-dimensional spatial coordinates corresponding to the reference solution are used as the location of the target single cell.

6. The battery thermal runaway early warning method as described in claim 4, characterized in that, Before performing differential localization on the acoustic signal to obtain the location of the target single cell when identifying the acoustic feature as a thermal runaway acoustic signal based on a well-trained acoustic signal classification network model, the method further includes: Acquire different types of training sound signals, extract features from the training sound signals, and obtain training features; Obtain a preset convolutional neural network, wherein the learning rate of the preset convolutional neural network is a preset learning rate M, and the weights of the weighted dynamic MFCC are preset weights P; The preset convolutional neural network is trained using the training features to obtain a fully trained acoustic signal classification network model. The acoustic features are input into the fully trained acoustic signal classification network model to obtain acoustic signal categories, which include thermal runaway acoustic signals.

7. The battery thermal runaway early warning method as described in claim 1, characterized in that, The gas concentrations include C2H4 concentration, C2H6 concentration, CO concentration, and H2 concentration; The step of issuing a battery thermal runaway warning when the gas concentration exceeds a preset concentration threshold includes: If the CO concentration is greater than or equal to a preset CO concentration or the H2 concentration is greater than or equal to a preset H2 concentration threshold, an early warning of battery thermal runaway will be issued. If the C2H4 concentration is greater than or equal to a preset C2H4 concentration threshold or the C2H6 concentration is greater than or equal to a preset C2H6 concentration threshold, a battery thermal runaway warning will be issued.

8. A battery thermal runaway early warning device, characterized in that, The battery thermal runaway early warning device includes: The acquisition module is used to acquire acoustic signals inside the prefabricated cabin energy storage system, and to extract acoustic features from the acoustic signals based on the weighted dynamic Mel-Cepstral coefficients. The prefabricated cabin energy storage system includes multiple individual battery locations. The localization module is used to perform differential localization on the acoustic signal to obtain the location of the target single cell when the acoustic feature is identified as a thermal runaway acoustic signal according to the well-trained acoustic signal classification network model. The concentration detection module is used to detect the position of the target single cell based on the ring-down spectrum of the open optical fiber ring cavity to obtain the gas concentration; The early warning module is used to provide early warning of battery thermal runaway when the gas concentration exceeds a preset concentration threshold.

9. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps of the battery thermal runaway early warning method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the battery thermal runaway early warning method as described in any one of claims 1 to 7.

Images