Mining lithium battery thermal runaway acoustic signal processing algorithm based on machine learning

Abstract

The embodiment of the invention provides a mining lithium battery thermal runaway acoustic signal processing algorithm based on machine learning. The mining lithium battery thermal runaway acoustic signal processing algorithm comprises the steps of synchronously obtaining a target acoustic signal and a battery environment parameter of a mining lithium battery in a thermal runaway process; preprocessing the target acoustic signal to obtain a preprocessed acoustic signal; performing spectral analysis, time domain analysis and wavelet transformation on the preprocessed acoustic signals, extracting thermal runaway related features, and correspondingly obtaining thermal runaway related key frequency features, vibration features and local frequency changes of the mining lithium battery; fusing the key frequency characteristic, the vibration characteristic and the local frequency change related to the thermal runaway and the battery environment parameters to obtain a thermal runaway fusion characteristic vector; and based on the thermal runaway fusion feature vector, performing thermal runaway monitoring on the mining lithium battery to obtain a thermal runaway monitoring result, and when an abnormal signal is monitored, automatically sending out early warning information and starting corresponding safety measures by the system.

Description

Technical Field

[0001] The present invention relates to the field of thermal runaway safety assessment of lithium-ion batteries, and in particular to, but not limited to, an acoustic signal processing algorithm for thermal runaway of mining lithium batteries based on machine learning. Background Art

[0002] With the widespread application of intelligent and automated technologies in the coal mining field, the use of lithium-ion batteries in underground mining equipment is growing. However, due to the poor thermal stability of lithium-ion battery materials, they are easily affected by the environment and thermal runaway occurs, such as pressure, temperature, humidity and other factors. The mine environment is characterized by high temperature, high humidity, high water vapor, dust pollution, poor ventilation and lack of heat dissipation performance, which poses challenges to the safe operation of mining lithium batteries.

[0003] Lithium-ion batteries have a complex multi-layer structure and poor thermal conductivity. When thermal runaway occurs, the temperature inside the lithium battery rises sharply and produces a large amount of gas, accompanied by expansion, valve opening, smoking, fire and other phenomena. Studies have attempted to monitor and warn of thermal runaway through changes in signals such as temperature, pressure, strain and gas inside the battery. However, in experiments, it was found that for common thermal runaway phenomena, the triggering of thermal runaway is often accompanied by the appearance of valve spraying sounds. In large-scale extreme thermal runaway, the explosion sound of the battery shell usually precedes the changes in temperature and voltage signals. Therefore, it is more accurate to judge the thermal runaway state of lithium batteries by sound signals. Summary of the invention

[0004] Based on the problems in the related technology, an embodiment of the present invention provides a machine learning-based acoustic signal processing algorithm for thermal runaway of mining lithium batteries, which makes up for the shortcomings of the existing technology in response speed and warning accuracy, and provides a more intelligent safety monitoring method for mining lithium battery systems.

[0005] The technical solution of the embodiment of the present invention is achieved as follows:

[0006] In a first aspect, an embodiment of the present invention provides an acoustic signal processing algorithm for thermal runaway of a mining lithium battery based on machine learning, the algorithm comprising:

[0007] Synchronously obtain the target acoustic signals and battery environmental parameters of mining lithium batteries during thermal runaway;

[0008] After preprocessing the target acoustic signal, a preprocessed acoustic signal is obtained;

[0009] The pre-processed acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively to extract features related to thermal runaway, and correspondingly obtain key frequency features, vibration features and local frequency changes related to thermal runaway of the mining lithium battery;

[0010] The key frequency characteristics, vibration characteristics and local frequency changes related to the thermal runaway and the battery environment parameters are fused, and a thermal runaway fusion feature vector is obtained using a feature-level fusion method;

[0011] Based on the thermal runaway fusion feature vector, a trained machine learning model is used to monitor the thermal runaway of the mining lithium battery to obtain a thermal runaway monitoring result. When an abnormal signal is detected, the system will automatically issue a warning message and initiate corresponding safety measures.

[0012] In a second aspect, an embodiment of the present invention provides a device for processing thermal runaway acoustic signals of a mining lithium battery based on machine learning, the device comprising:

[0013] An acquisition module is used to synchronously acquire the target acoustic signal and battery environmental parameters of mining lithium batteries during thermal runaway;

[0014] A preprocessing module, used for preprocessing the target acoustic signal to obtain a preprocessed acoustic signal;

[0015] An analysis module is used to perform spectrum analysis, time domain analysis and wavelet transform on the preprocessed acoustic signal, extract features related to thermal runaway, and obtain key frequency features, vibration features and local frequency changes related to thermal runaway of the mining lithium battery;

[0016] A fusion module, used to fuse the key frequency characteristics, vibration characteristics and local frequency changes related to the thermal runaway and the battery environment parameters, and obtain a thermal runaway fusion feature vector using a feature-level fusion method;

[0017] The monitoring module is used to perform thermal runaway monitoring on the mining lithium battery based on the thermal runaway fusion feature vector and adopt a trained machine learning model to obtain thermal runaway monitoring results. When an abnormal signal is detected, the system will automatically issue a warning message and initiate corresponding safety measures.

[0018] In some embodiments, the target acoustic signal includes instantaneous acoustic changes, abnormal acoustic signals and transmission acoustic signals; the acquisition module is also used to place a first sensor on the surface of the outer shell of the mining lithium battery; the first sensor is used to collect instantaneous acoustic changes when the gas leaks or the battery shell ruptures; a second sensor is placed at the heat dissipation part of the mining lithium battery; the second sensor is used to detect abnormal acoustic signals caused by poor heat dissipation; a third sensor is placed at the battery module connection of the mining lithium battery; the third sensor is used to collect acoustic signals transmitted between batteries, that is, the transmission acoustic signal.

[0019] In some embodiments, the preprocessing module is also used to filter the collected initial acoustic signal using a bandpass filter to obtain a filtered acoustic signal; use a Kalman filter to predict the historical acoustic signal and correct the filtered acoustic signal to extract the effective acoustic signal related to thermal runaway; enhance the effective acoustic signal to obtain an enhanced acoustic signal; segment the enhanced acoustic signal to obtain the preprocessed acoustic signal.

[0020] In some embodiments, the analysis module is also used to perform spectral analysis on the preprocessed acoustic signal using fast Fourier transform to obtain key frequency characteristics related to the thermal runaway; perform time domain analysis on the preprocessed acoustic signal, extract peak amplitude by traversing sampling points to identify instantaneous strong vibrations, calculate the root mean square value to quantify the overall vibration intensity and stability, and measure the signal duration to evaluate the significance of abnormal vibrations, thereby obtaining vibration characteristics related to the thermal runaway; perform time-frequency analysis on the preprocessed acoustic signal using wavelet transform to obtain local frequency changes related to the thermal runaway.

[0021] In some embodiments, the machine learning-based thermal runaway acoustic signal processing algorithm for mining lithium batteries is applied to a thermal runaway monitoring model, and the device further includes: a training module for inputting the preprocessed sample acoustic signal into the thermal runaway monitoring model for subsequent feature extraction and training; through the feature extraction layer of the thermal runaway monitoring model, the preprocessed sample acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively to extract features related to thermal runaway, and obtain sample key frequency features, sample vibration features and sample local frequency changes related to thermal runaway; through the feature fusion layer of the thermal runaway monitoring model, the preprocessed sample acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively to extract features related to thermal runaway, and obtain sample key frequency features, sample vibration features and sample local frequency changes related to thermal runaway; through the feature fusion layer of the thermal runaway monitoring model, the preprocessed sample acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively to extract features related to thermal runaway, and obtain sample key frequency features, sample vibration features and sample local frequency changes related to thermal runaway. The sample thermal runaway key frequency characteristics, the sample vibration characteristics and the sample local frequency change are feature fused, and a feature-level fusion method is used to obtain a sample thermal runaway fusion feature vector; through the information prediction layer of the thermal runaway monitoring model, based on the sample thermal runaway fusion feature vector, a trained machine learning model is used to perform thermal runaway monitoring on the sample thermal runaway monitoring task to obtain a sample thermal runaway monitoring result; the sample thermal runaway monitoring result is input into a preset loss model to obtain a loss result; based on the loss result, the model parameters in the thermal runaway monitoring model are corrected to obtain a trained thermal runaway monitoring model.

[0022] In some embodiments, the device also includes: an output module, which is also used to divide the working state of the mining lithium battery into multiple levels based on the thermal runaway monitoring result; based on each level and the preset warning mechanism, a corresponding alarm signal is output; the multiple levels include normal operation, low-level warning, intermediate warning and emergency warning; the normal operation means that the target acoustic signal and the thermal runaway monitoring result both indicate that the mining lithium battery is in a normal working state; the low-level warning means that there is a low-frequency vibration sound or a signal fluctuation of a short duration in the target acoustic signal, and no emergency treatment measures are triggered; the intermediate warning means that there is a strong abnormal signal in the target acoustic signal, and the thermal runaway monitoring result is to predict the possibility of thermal runaway of the mining lithium battery; the emergency warning means that there are typical thermal runaway features in the target acoustic signal, and the thermal runaway monitoring result is to determine that the mining lithium battery is on the edge of thermal runaway or is already in a thermal runaway state, and trigger emergency treatment measures.

[0023] In a third aspect, an embodiment of the present invention provides a mining lithium battery thermal runaway acoustic signal processing device based on machine learning, comprising: a memory for storing executable instructions; and a processor for implementing the above-mentioned mining lithium battery thermal runaway acoustic signal processing algorithm based on machine learning when executing the executable instructions stored in the memory.

[0024] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing executable instructions for causing a processor to execute the executable instructions to implement the above-mentioned machine learning-based thermal runaway acoustic signal processing algorithm for mining lithium batteries.

[0025] The machine learning-based thermal runaway acoustic signal processing algorithm for a mining lithium battery provided in an embodiment of the present invention first obtains a target acoustic signal and a battery environmental parameter of a mining lithium battery in a thermal runaway process; preprocesses the target acoustic signal to obtain a preprocessed acoustic signal; performs spectrum analysis, time domain analysis and wavelet transform on the preprocessed acoustic signal to obtain corresponding thermal runaway frequency characteristics, thermal runaway vibration characteristics and thermal runaway acoustic change characteristics of the mining lithium battery; fuses the thermal runaway frequency characteristics, thermal runaway vibration characteristics, thermal runaway acoustic change characteristics and battery environmental parameters to obtain a thermal runaway fusion feature vector; performs thermal runaway monitoring on the mining lithium battery based on the thermal runaway fusion feature vector to obtain a thermal runaway monitoring result. In this way, through acoustic signal monitoring, acoustic features can be captured in the early stage of thermal runaway, and the acoustic signals can be intelligently analyzed and classified with machine learning technology, which can effectively distinguish between normal and abnormal states and reduce false alarms and missed alarms. In addition, it can monitor and warn of thermal runaway of lithium-ion batteries earlier and more accurately in harsh mining environments, overcoming the shortcomings of existing monitoring methods in response speed and warning accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of an acoustic signal processing algorithm for thermal runaway of a mining lithium battery based on machine learning provided by an embodiment of the present invention;

[0027] Figure 2 It is a structural schematic diagram of a mining lithium battery thermal runaway acoustic signal processing system based on machine learning provided by an embodiment of the present invention;

[0028] Figure 3 It is a flow chart of another acoustic signal processing algorithm for thermal runaway of a mining lithium battery based on machine learning provided by an embodiment of the present invention;

[0029] Figure 4 A schematic diagram of the composition structure of a mining lithium battery thermal runaway acoustic signal processing device based on machine learning provided in an embodiment of the present invention;

[0030] Figure 5 A schematic diagram of the composition structure of a mining lithium battery thermal runaway acoustic signal processing device based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present invention. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.

[0032] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict. Unless otherwise defined, all technical and scientific terms used in the embodiments of the present invention have the same meaning as those commonly understood by those skilled in the art to which the embodiments of the present invention pertain. The terms used in the embodiments of the present invention are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0033] The following describes an exemplary application of the thermal runaway acoustic signal processing device for lithium batteries for mining based on machine learning in an embodiment of the present invention. The thermal runaway acoustic signal processing device for lithium batteries for mining based on machine learning provided in an embodiment of the present invention can be implemented as a terminal or a server. In one implementation, the thermal runaway acoustic signal processing device for lithium batteries for mining based on machine learning provided in an embodiment of the present invention can be implemented as various types of terminals such as laptops, tablet computers, desktop computers, and mobile devices; in another implementation, the thermal runaway acoustic signal processing device for lithium batteries for mining based on machine learning provided in an embodiment of the present invention can also be implemented as a server, wherein the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks (CDN, Content Delivery Network), and big data and artificial intelligence platforms. The terminal and the server can be directly or indirectly connected by wired or wireless communication, which is not limited in the embodiment of the present invention. Below, an exemplary application of the machine learning-based mining lithium battery thermal runaway acoustic signal processing device when implemented as a server will be described.

[0034] The embodiment of the present invention provides a mining lithium battery thermal runaway acoustic signal processing algorithm based on machine learning, see Figure 1 , Figure 1 This is a flow chart of an acoustic signal processing algorithm for thermal runaway of a lithium battery for mining based on machine learning provided by an embodiment of the present invention. Figure 1 The steps shown are explained.

[0035] Step S110, synchronously acquiring the target acoustic signal and battery environmental parameters of the mining lithium battery during the thermal runaway process.

[0036] In some embodiments, thermal runaway refers to a series of self-accelerating chemical reactions that occur in lithium batteries at high temperatures, ultimately leading to irreversible thermal runaway. These reactions generate a large amount of heat and gas, causing the battery temperature to rise sharply, and even causing combustion or explosion.

[0037] In some embodiments, the target acoustic signal refers to the acoustic signal data collected during the thermal runaway process of the mining lithium-ion battery, including vibration sound, gas leakage sound, explosion sound, etc. generated during the operation of the battery.

[0038] In some implementations, the battery environment parameters refer to parameters related to the battery environment, such as temperature, humidity, air pressure, and other environmental factors.

[0039] In the present invention, various types of sensor devices are installed near the battery module to ensure the stability and accuracy of signal acquisition.

[0040] Step S120: preprocess the target acoustic signal to obtain a preprocessed acoustic signal.

[0041] In some embodiments, preprocessing refers to performing bandpass filtering, noise suppression, signal enhancement, and signal segmentation on the target acoustic signal to remove background noise and retain acoustic features related to thermal runaway.

[0042] Here, a bandpass filter is used to filter the acoustic signal, retaining the frequency band signal from 300Hz to 10kHz and removing irrelevant low-frequency mechanical noise and high-frequency electromagnetic interference. The Kalman filter is used to predict historical data and correct real-time signals, effectively remove background noise in the mine environment, and extract effective signals related to thermal runaway. For weaker acoustic signals, gain adjustment is used for signal enhancement. The collected continuous acoustic signal is segmented into 0.1 second time windows, and each time window contains 1000 sample points, so that subsequent feature extraction and machine learning model training can be analyzed based on time series data.

[0043] Step S130, performing spectrum analysis, time domain analysis and wavelet transform on the preprocessed acoustic signal, extracting features related to thermal runaway, and correspondingly obtaining key frequency features, vibration features and local frequency changes related to thermal runaway of the mining lithium battery.

[0044] In some embodiments, spectrum analysis refers to converting time domain signals into frequency domain signals using fast Fourier transform (FFT), and extracting key frequency features related to thermal runaway in the frequency domain, such as main frequency, spectrum peak, frequency bandwidth, etc. Through these frequency features, acoustic changes in different stages of thermal runaway can be identified.

[0045] In some embodiments, time domain analysis refers to extracting time domain features of acoustic signals, including peak amplitude, root mean square value, duration, etc. of acoustic signals. Time domain features can effectively capture the intensity and duration of transient events (such as gas injection and explosion). The low-frequency vibration components in the acoustic signal are analyzed to identify the physical changes of battery materials during thermal runaway. These vibration features are of great significance for early detection of thermal runaway.

[0046] In some embodiments, wavelet transform refers to the ability to simultaneously analyze local features of a signal in time and frequency, capturing dynamic changes in frequency over time, that is, using wavelet transform to perform time-frequency analysis on complex non-stationary signals (such as popping sounds).

[0047] In the present invention, spectrum analysis, time domain analysis and wavelet transform are respectively performed on the preprocessed acoustic signal to obtain the thermal runaway frequency characteristics, thermal runaway vibration characteristics and thermal runaway acoustic change characteristics of the mining lithium battery.

[0048] Step S140, fusing the key frequency features, vibration features, local frequency changes and the battery environment parameters related to the thermal runaway, and using a feature-level fusion method to obtain a thermal runaway fusion feature vector.

[0049] In an embodiment of the present invention, the features extracted from spectrum analysis, time domain analysis and wavelet transform are fused to form a new thermal runaway fusion feature vector. The thermal runaway fusion feature vector includes the frequency peak in the frequency domain, the amplitude feature in the time domain, and the frequency change information in the time-frequency domain. A new feature set is generated by feature combination for subsequent machine learning model input. By multi-dimensionally fusing acoustic signals from different sources, the model's ability to identify thermal runaway in complex acoustic environments is enhanced.

[0050] Step S150, based on the thermal runaway fusion feature vector, the trained machine learning model is used to monitor the thermal runaway of the mining lithium battery to obtain the thermal runaway monitoring result. When an abnormal signal is detected, the system will automatically issue a warning message and initiate corresponding safety measures.

[0051] In the present invention, the trained machine learning model refers to the trained thermal runaway monitoring model.

[0052] In some embodiments, the trained thermal runaway monitoring model is applied to the acoustic signal real-time monitoring system of the mining lithium battery to monitor the operating status of the mining lithium battery at any time. Once an abnormal signal is detected, the acoustic signal real-time monitoring system will automatically issue a warning message and initiate safety measures through an audible and visual alarm or a remote control system. The early warning system supports multi-level warnings, issues warning instructions according to the severity of the acoustic signal, and links the mine automation safety system.

[0053] The machine learning-based thermal runaway acoustic signal processing algorithm for a mining lithium battery provided in an embodiment of the present invention first obtains a target acoustic signal and a battery environmental parameter of a mining lithium battery in a thermal runaway process; preprocesses the target acoustic signal to obtain a preprocessed acoustic signal; performs spectrum analysis, time domain analysis and wavelet transform on the preprocessed acoustic signal to obtain corresponding thermal runaway frequency characteristics, thermal runaway vibration characteristics and thermal runaway acoustic change characteristics of the mining lithium battery; fuses the thermal runaway frequency characteristics, thermal runaway vibration characteristics, thermal runaway acoustic change characteristics and battery environmental parameters to obtain a thermal runaway fusion feature vector; performs thermal runaway monitoring on the mining lithium battery based on the thermal runaway fusion feature vector to obtain a thermal runaway monitoring result. In this way, through acoustic signal monitoring, acoustic features can be captured in the early stage of thermal runaway, and the acoustic signals can be intelligently analyzed and classified with machine learning technology, which can effectively distinguish between normal and abnormal states and reduce false alarms and missed alarms. In addition, it can monitor and warn of thermal runaway of lithium-ion batteries earlier and more accurately in harsh mining environments, overcoming the shortcomings of existing monitoring methods in response speed and warning accuracy.

[0054] In some embodiments, the target acoustic signal includes an instantaneous acoustic change, an abnormal acoustic signal, and a transfer acoustic signal; the above step S110 can be implemented by the following steps S111 to S113:

[0055] Step S111, placing a first sensor on the outer shell surface of the mining lithium battery; the first sensor is used to collect instantaneous acoustic changes when the gas leaks or the battery shell ruptures.

[0056] Step S112, placing a second sensor at the heat dissipation location of the mining lithium battery; the second sensor is used to detect abnormal acoustic signals caused by poor heat dissipation.

[0057] Step S113, placing a third sensor at the connection point of the battery module of the mining lithium battery; the third sensor is used to collect the acoustic signal transmitted between batteries, that is, the transmitted acoustic signal.

[0058] In the present invention, a highly sensitive capacitive microphone sensor is used, and its frequency response range is 20Hz to 20kHz, which is suitable for capturing various types of acoustic signals generated during the thermal runaway of mining lithium batteries, and needs to have high temperature resistance and vibration resistance to adapt to the harsh environment of high temperature, high humidity, high pressure, etc. in the mine. Three sensors are arranged, and the sensors are fixed on the surface of the battery module using high temperature resistant industrial adhesives to ensure their stability under vibration and high temperature conditions, and connected to the data acquisition system through shielded cables to avoid electromagnetic interference. Sensor 1 is installed on the surface of the battery casing, close to the location where gas leakage or explosion may occur, to ensure that the instantaneous changes of these signals can be captured; sensor 2 is installed at the heat dissipation part of the battery to detect vibration and abnormal sound caused by poor heat dissipation; sensor 3 is installed at the connection of the battery module to capture the acoustic signals transmitted between batteries to determine the overall thermal runaway of the battery pack.

[0059] In the present invention, the method also includes synchronous collection of environmental parameters, using a thermocouple sensor to collect the temperature of the battery and its surrounding environment in real time, and a humidity sensor and an air pressure sensor to collect environmental humidity and air pressure changes respectively. The environmental parameters are updated once per second to ensure synchronization with the acoustic signal. Each collected data point is accompanied by an accurate timestamp to ensure that the environmental data and the acoustic signal can be accurately aligned in time, which is convenient for subsequent data analysis and feature fusion.

[0060] In some embodiments, the above step S120 may be implemented through steps S121 to S124:

[0061] Step S121: Use a bandpass filter to filter the collected initial acoustic signal to obtain a filtered acoustic signal.

[0062] Step S122: Use a Kalman filter to predict historical acoustic signals and to correct the filtered acoustic signals to extract effective acoustic signals related to thermal runaway.

[0063] Step S123: performing signal enhancement on the effective acoustic signal to obtain an enhanced acoustic signal.

[0064] Step S124: segment the enhanced acoustic signal to obtain the preprocessed acoustic signal.

[0065] In some embodiments, the above step S130 may be implemented by steps S131 to S133:

[0066] Step S131, performing spectrum analysis on the preprocessed acoustic signal by using fast Fourier transform to obtain key frequency features related to the thermal runaway.

[0067] Step S132, performing time domain analysis on the preprocessed acoustic signal, extracting the peak amplitude by traversing the sampling points to identify instantaneous strong vibration, calculating the root mean square value to quantify the overall vibration intensity and stability, and measuring the signal duration to evaluate the significance of the abnormal vibration, thereby obtaining the vibration characteristics related to the thermal runaway.

[0068] Step S133: performing time-frequency analysis on the preprocessed acoustic signal using wavelet transform to obtain local frequency changes related to the thermal runaway.

[0069] In some embodiments, a machine learning-based thermal runaway acoustic signal processing algorithm for mining lithium batteries is applied to a thermal runaway monitoring model, and the thermal runaway monitoring model is trained by the following steps: inputting the preprocessed sample acoustic signal into the thermal runaway monitoring model for subsequent feature extraction and training; performing spectrum analysis, time domain analysis and wavelet transform on the preprocessed sample acoustic signal through the feature extraction layer of the thermal runaway monitoring model, extracting features related to thermal runaway, and obtaining sample key frequency features, sample vibration features and sample local frequency changes related to thermal runaway; and performing feature fusion layer of the thermal runaway monitoring model. The key frequency characteristics of the sample thermal runaway, the sample vibration characteristics and the sample local frequency change are feature fused, and a feature-level fusion method is used to obtain a sample thermal runaway fusion feature vector; through the information prediction layer of the thermal runaway monitoring model, based on the sample thermal runaway fusion feature vector, a trained machine learning model is used to perform thermal runaway monitoring on the sample thermal runaway monitoring task to obtain a sample thermal runaway monitoring result; the sample thermal runaway monitoring result is input into a preset loss model to obtain a loss result; based on the loss result, the model parameters in the thermal runaway monitoring model are corrected to obtain a trained thermal runaway monitoring model.

[0070] In some embodiments, the algorithm also includes: based on the thermal runaway monitoring result, dividing the working state of the mining lithium battery into multiple levels; based on each level and the preset warning mechanism, outputting a corresponding alarm signal; the multiple levels include normal operation, low-level warning, intermediate warning and emergency warning; the normal operation means that both the target acoustic signal and the thermal runaway monitoring result indicate that the mining lithium battery is in a normal working state; the low-level warning means that there is a low-frequency vibration sound or a signal fluctuation of a short duration in the target acoustic signal, and no emergency treatment measures are triggered; the intermediate warning means that there is a strong abnormal signal in the target acoustic signal, and the thermal runaway monitoring result predicts that the mining lithium battery has the possibility of thermal runaway; the emergency warning means that there are typical thermal runaway features in the target acoustic signal, and the thermal runaway monitoring result determines that the mining lithium battery is on the edge of thermal runaway or is already in a thermal runaway state, and triggers emergency treatment measures.

[0071] The following describes an exemplary application of an embodiment of the present invention in a practical application scenario.

[0072] Figure 2 A schematic diagram of the structure of a mining lithium battery thermal runaway acoustic signal processing system based on machine learning provided in an embodiment of the present invention. The present invention includes a mining lithium-ion battery thermal runaway acoustic signal acquisition and preprocessing module, an acoustic signal feature extraction module, a machine learning modeling and training module, a real-time monitoring and early warning module, and a system adaptability and stability module. The preprocessed signal is passed to the feature extraction module, and the feature extraction provides input for the machine learning module, but the training of the machine learning model can be fed back to the feature extraction to optimize the data processing flow. The machine learning model is applied to the analysis of actual data through the real-time monitoring module, and the early warning information can also be used to optimize the model. The system adaptability and stability module can monitor and maintain system stability and provide feedback information to other modules.

[0073] Figure 3 A schematic flow chart of another algorithm for processing thermal runaway acoustic signals of a lithium battery for mining based on machine learning provided in an embodiment of the present invention, the method comprising:

[0074] Step 1: Collect acoustic signal data of mining lithium-ion batteries during thermal runaway, including vibration, gas leakage, and explosion sounds generated during battery operation. Data collection also includes parameters related to the battery environment, such as temperature, humidity, air pressure, and other environmental factors. The sensor equipment is installed near the battery module to ensure the stability and accuracy of signal acquisition.

[0075] Step 2: Preprocess the collected acoustic signals to remove background noise, retain the acoustic features related to thermal runaway, and use spectrum analysis, time domain analysis, wavelet transform and other methods to extract the features of the acoustic signals, including frequency, amplitude, duration, etc. According to different acoustic signal sources, perform data fusion to generate new feature combinations for input into subsequent machine learning models.

[0076] Step 3: Based on the extracted acoustic features, a thermal runaway identification model is established. The model is trained, verified, and tested using labeled experimental data. The convolutional neural network (CNN) algorithm is used to identify key signal features in the thermal runaway process. After the model training is completed, the model is optimized in combination with the results of experimental verification to ensure the accuracy and robustness of the model under actual working conditions.

[0077] Step 4: Apply the trained machine learning model to the real-time acoustic signal monitoring system of mining lithium batteries to monitor the battery operating status at any time. Once an abnormal signal is detected, the system will automatically issue an early warning message and initiate safety measures through sound and light alarms or remote control systems. The early warning system supports multi-level early warnings, issues early warning instructions according to the severity of the acoustic signal, and links the mine automation safety system.

[0078] Specifically, in step 1, a highly sensitive capacitive microphone sensor is used, and its frequency response range is 20Hz to 20kHz, which is suitable for capturing various types of acoustic signals generated during the thermal runaway of lithium batteries, and needs to have high temperature resistance and vibration resistance to adapt to the harsh environment of high temperature, high humidity, high pressure, etc. in the mine. Three sensors are arranged, and the sensors are fixed on the surface of the battery module using high temperature resistant industrial adhesives to ensure their stability under vibration and high temperature conditions, and connected to the data acquisition system through shielded cables to avoid electromagnetic interference. Sensor 1 is installed on the surface of the battery casing, close to the location where gas leakage or explosion may occur, to ensure that the instantaneous changes of these signals can be captured; sensor 2 is installed at the heat dissipation part of the battery to detect vibration and abnormal sound caused by poor heat dissipation; sensor 3 is installed at the connection of the battery module to capture the acoustic signals transmitted between batteries to determine the overall thermal runaway of the battery pack.

[0079] In step 2, acoustic signal preprocessing and feature extraction include:

[0080] (I) Signal acquisition device and parameter setting

[0081] (1) Data collection frequency setting:

[0082] According to the Nyquist sampling theorem, the sampling frequency f s Should not be lower than the highest frequency f of the signal max Twice, the specific formula is:

[0083] f s ≥2×f max ,

[0084] In this system, the highest frequency of the acoustic signal is set to 5kHz, so the sampling frequency is set to 10kHz to ensure that the fast-changing signals of thermal runaway are captured.

[0085] (2) Signal acquisition time window: Each acquisition cycle is 1 second, during which 10,000 sample points are collected to ensure that both instantaneous and continuous acoustic signals can be captured.

[0086] (3) Data transmission and storage: The collected signals are transmitted to the central data processing unit through a digital anti-interference protocol (such as RS485 or CAN bus). The data is stored in a buffer memory and transmitted in real time during the collection process to prevent signal loss.

[0087] (II) Signal preprocessing

[0088] (1) Bandpass filtering: Use a bandpass filter to filter the acoustic signal, retain the frequency band signal from 300Hz to 10kHz, and remove irrelevant low-frequency mechanical noise and high-frequency electromagnetic interference. The transfer function of the filter is defined as:

[0089]

[0090] (2) Noise suppression: Use the Kalman filter to predict historical data and correct real-time acoustic signals, effectively remove background noise in the mine environment, and extract effective signals related to thermal runaway. The recursive formula of the Kalman filter is:

[0091]

[0092] P k|k-1 =AP k-1|k-1 A T +Q,

[0093] in, is the predicted state, A is the state transfer matrix, B is the input control matrix, u k is the control input, P k|k-1 is the error covariance matrix, and Q is the process noise covariance.

[0094] (3) Signal enhancement: For weak acoustic signals, gain adjustment is used to enhance the signal. The signal enhancement formula is:

[0095] A out =G×A in

[0096] Among them, G is the gain coefficient. In this system, the gain is set to 2, that is, the signal amplitude is amplified by 2 times to ensure that weak signals can be effectively captured.

[0097] (4) Signal segmentation processing: The collected continuous acoustic signal is segmented into 0.1 second time windows, each of which contains 1,000 sample points, so that subsequent feature extraction and machine learning model training can be analyzed based on time series data.

[0098] (III) Synchronous collection of environmental parameters

[0099] Thermocouple sensors are used to collect the temperature of the battery and its surrounding environment in real time. Humidity sensors and air pressure sensors collect changes in environmental humidity and air pressure, respectively. Environmental parameters are updated once a second to ensure synchronization with acoustic signals. Each collected data point is accompanied by an accurate timestamp to ensure that environmental data and acoustic signals can be accurately aligned in time, facilitating subsequent data analysis and feature fusion.

[0100] (IV) Signal preprocessing results

[0101] After bandpass filtering and Kalman filtering, the acoustic signal that originally contained background noise is effectively purified. For the signal of gas leakage or explosion, its amplitude is amplified after gain processing to ensure that the characteristic signal is clearly visible. All pre-processed data are stored in the local data buffer and regularly transmitted to the central server for further analysis. Each set of acoustic signal data is aligned with the environmental parameter data to ensure that multi-dimensional factors can be comprehensively considered during data analysis.

[0102] (V) Feature extraction

[0103] (1) Spectral analysis: Use fast Fourier transform (FFT) to convert the time domain signal into the frequency domain signal, and extract the key frequency features related to thermal runaway in the frequency domain, such as main frequency, spectrum peak, frequency bandwidth, etc. Through these frequency features, the acoustic changes in different stages of thermal runaway can be identified. The Fourier transform formula is:

[0104]

[0105] Among them, X(f) is the frequency domain signal and x(t) is the time domain signal.

[0106] Through Fourier transform, the main frequency components in the signal can be identified and used to analyze the high-frequency popping sound or low-frequency vibration sound generated during thermal runaway.

[0107] (2) Time domain analysis: Extract the time domain features of the acoustic signal, including the peak amplitude, root mean square value, duration, etc. of the signal. The time domain features can effectively capture the intensity and duration of transient events (such as gas injection and explosion). Analyze the low-frequency vibration components in the acoustic signal to identify the physical changes of the battery material during the thermal runaway process. These vibration features are important for early detection of thermal runaway.

[0108] Peak amplitude formula: A peak = max(|x(t)|),

[0109] Root mean square (RMS) formula:

[0110] (3) Wavelet transform: Use wavelet transform to perform time-frequency analysis on complex non-stationary signals (such as popping sounds). Wavelet transform can analyze the local characteristics of the signal in both time and frequency, capturing the dynamic changes of frequency over time. The wavelet transform formula is:

[0111]

[0112] Among them, W ψ (a,b) is the transform coefficient of the signal under scale a and translation b, and ψ is the mother wavelet function. By adjusting a and b, the local frequency change of the signal can be analyzed.

[0113] Wavelet transform can analyze signals of different frequency bands with different resolutions, helping to identify changes in acoustic characteristics at different times during the thermal runaway process. For example, wavelet analysis can be used to identify low-frequency signals at the beginning of a gas leak and high-frequency signals during a burst.

[0114] (VI) Data Fusion

[0115] The features extracted from spectrum analysis, time domain analysis and wavelet transform are fused to form a new feature vector. The feature vector includes the frequency peak in the frequency domain, the amplitude feature in the time domain, and the frequency change information in the time-frequency domain. A new feature set is generated by feature combination for subsequent machine learning model input. By multi-dimensionally fusing acoustic signals from different sources, the model's ability to identify thermal runaway in complex acoustic environments is enhanced. The present invention adopts a feature-level fusion method, assuming x1, x2, ... x n For the feature vectors extracted from different sources, the fused feature vector is expressed as:

[0116] X fused =[x1,x1,x1,……,x n ],

[0117] Assume x1 is a 3D vector extracted from frequency domain analysis, such as [f peak ,f mean ,f bandwidth ], x2 is a 2D vector extracted from the time domain analysis, such as [A peak ,A duration ], then the result of feature-level fusion is:

[0118] X fused =[f peak ,f mean ,f bandwidth ,A peak ,A duration ].

[0119] In step 3, the establishment of the thermal runaway identification model includes:

[0120] 1. Data preparation and feature annotation

[0121] Dataset construction: The collected acoustic signals are extracted through feature extraction in step 2 to generate a complete feature set. Each feature set contains multidimensional features extracted from spectrum analysis, time domain analysis, wavelet transform and other methods. Each data sample consists of a feature vector X fused and the corresponding label y, where label y represents the state of the lithium battery (e.g., "normal operation" or "thermal runaway"). The dataset is then divided into training, validation, and test sets, with a common ratio of 80% for training, 10% for validation, and 10% for testing to ensure the generalization ability of the model.

[0122] Dataset balance: For samples of different categories (such as normal operation and thermal runaway signals), additional samples are generated through data augmentation techniques (such as adding noise or transforming the signal) to ensure the balance of the data set and avoid bias during model training.

[0123] (II) Convolutional Neural Network (CNN) Model Design

[0124] Input layer: The input layer of the model receives the multi-dimensional feature vector extracted from step 2. The feature vector includes a combination of acoustic features such as frequency, amplitude, duration, etc. The dimension of the input vector is equal to the feature dimension generated after feature-level fusion. For example, if the feature vector dimension is 100, the input layer contains 100 neurons.

[0125] Convolutional layer: The function of the convolutional layer is to automatically extract high-order features of the acoustic signal from the input data. Through the convolution operation, the model can identify local patterns in the acoustic signal, such as sudden sounds or abnormal vibrations in certain specific frequency bands. The convolutional layer formula is:

[0126]

[0127] in, is the output of the lth convolutional layer; is the convolution kernel; is the input of the previous layer, f is the activation function, using ReLU, f(x) = max(0,x).

[0128] Pooling layer: Use the pooling layer (usually the maximum pooling) to reduce the dimension of the convolutional layer output, prevent overfitting and reduce computational complexity. The pooling layer retains important features by performing the maximum value operation on the local area. The formula represents the maximum value selected in the 2×2 area. The maximum pooling layer formula is:

[0129]

[0130] Fully connected layer: The output of the convolutional layer and the pooling layer is flattened into a one-dimensional feature vector and input into the fully connected layer. The fully connected layer is used to further process the global features of the signal and generate classification predictions.

[0131] Output layer: The output layer uses the Softmax activation function to generate classification probabilities. The Softmax function converts the score of each classification into a probability distribution. The formula is as follows:

[0132]

[0133] Among them, z k is the score of the kth class in the output layer.

[0134] (III) Model training and optimization

[0135] Loss function: Use the cross entropy loss function to measure the error between the model prediction and the actual label. This loss function is suitable for classification tasks:

[0136]

[0137] Among them, y i is the true label, is the probability predicted by the model.

[0138] Optimizer: Use the Adam optimizer for gradient descent optimization. The Adam optimizer can dynamically adjust the learning rate and efficiently optimize large-scale data:

[0139]

[0140] Among them, θ t is the model parameter and η is the learning rate.

[0141] (IV) Model Validation and Cross-Validation

[0142] Use the validation set to evaluate the intermediate training results of the model to prevent the model from overfitting. After each epoch of model training, the validation set is used to evaluate its performance. K-fold cross validation is used to evaluate the generalization ability of the model. The dataset is divided into K subsets, and K-1 subsets are used for training each time. The remaining 1 subset is used for validation, and the cycle is repeated K times.

[0143] 5. Model evaluation and optimization

[0144] Performance indicators: Accuracy is used to evaluate the overall accuracy of model predictions:

[0145]

[0146] TP is true positive, TN is true negative, FP is false positive, and FN is false negative.

[0147] Accuracy:

[0148] Recall:

[0149] F1 score:

[0150] Model optimization: Optimize the model by adjusting the model's learning rate, convolution kernel size, number of convolution layers and other hyperparameters. You can use grid search or random search to automatically select the best hyperparameters. Use Dropout regularization technology during model training to randomly discard some neurons to prevent the model from overfitting. The Dropout ratio is usually set between 0.3 and 0.5.

[0151] In step 4, the purpose is to detect the operating status of the battery by real-time analysis of the collected acoustic signal data. Once an abnormal signal is detected, the early warning mechanism is triggered. The implementation of the real-time monitoring and early warning mechanism includes:

[0152] Architecture of real-time monitoring system:

[0153] Acoustic signal acquisition module: including vibration sound, gas leakage sound, explosion sound, etc. during battery operation. This module directly obtains real-time data from the sensor system designed in step 1.

[0154] Signal preprocessing module: Use bandpass filters, Kalman filters, etc. to denoise and enhance the acoustic signals collected in real time to ensure signal clarity and remove irrelevant background noise.

[0155] Feature extraction and data processing module: Use the spectrum analysis, time domain analysis and wavelet transform mentioned in step 2 to extract the features of the acoustic signal in real time, such as frequency peak, amplitude, duration, etc. These features will be passed as input to the trained CNN model for analysis.

[0156] Model prediction module: The preprocessed and feature-extracted signal data is input into the trained CNN model. The model makes predictions based on real-time data and outputs the current working status of the lithium battery.

[0157] 2. Design of early warning mechanism

[0158] Multi-level warning strategy: Based on the model prediction results, the system divides the working status of the lithium battery into multiple levels. The warning mechanism sends different alarm signals based on these levels. Typical warning levels are divided into the following categories:

[0159] Normal operation (no alarm): Both the acoustic signal and the model prediction result indicate that the battery is operating normally.

[0160] Low-level warning (Level 1 alarm): The system detects abnormal low-frequency vibration sounds or short-duration signal fluctuations, but does not immediately trigger emergency measures.

[0161] Intermediate warning (Level 2 alarm): There are strong abnormal signals in the acoustic signal characteristics (such as gas leakage sound or obvious vibration signals), and the model predicts that the battery may experience thermal runaway.

[0162] Emergency warning (Level 3 alarm): Typical thermal runaway characteristics (such as explosions and gas jets) appear in acoustic signals, and the system determines that the battery is on the verge of thermal runaway or has entered a thermal runaway state, triggering emergency handling measures.

[0163] When the system detects a Level 2 or Level 3 alarm, it will automatically trigger an audible and visual alarm to notify on-site staff to take appropriate measures. The system sends the warning signal to the mine central control system through a wireless communication module (such as LoRa, Wi-Fi or industrial Ethernet). The central control system can receive the alarm information and trigger linkage mechanisms, such as automatically cutting off the power supply and starting the cooling system.

[0164] Each time an early warning is triggered, the system automatically records the acoustic signal, prediction results and early warning level at that time. All data will be stored in the central data server as a basis for subsequent analysis and optimization of system performance.

[0165] 3. Real-time signal processing and model prediction

[0166] Data stream processing and model reasoning: The system uses a data stream processing architecture to process the continuous stream data of acoustic signals. The collected signal is first segmented using a sliding window method. The window length can be adjusted based on the experiment, usually 0.1 seconds. The signal data of each window is preprocessed (filtered, enhanced) and feature extracted to generate a feature vector.

[0167] The feature vector generated in each time window is input into the trained CNN model, and the model performs inference in real time and outputs the working status of the lithium battery.

[0168] The result of model inference is a probability distribution, for example:

[0169] P(normal)=0.95,P(thermal runaway)=0.05,

[0170] This prediction indicates that the lithium battery is working properly and the system will not trigger any alarms.

[0171] P(normal)=0.10,P(thermal runaway)=0.90

[0172] The system determines that the lithium battery is about to experience thermal runaway and immediate action must be taken.

[0173] The system determines whether the battery is in a thermal runaway state based on the output probability value and decides whether to issue a warning.

[0174] Result feedback and model adaptive adjustment: The system compares the model's prediction results with the actual lithium battery status. If the alarm issued by the system is confirmed, the system records the event and uses it for subsequent model adjustments. When the system detects false positives or false negatives multiple times, the system automatically adjusts the model parameters or retrains the model. The model can be fine-tuned by adding data sets to ensure the accuracy of the model under changing working conditions.

[0175] (IV) System stability and adaptability

[0176] The system regularly performs self-inspections on installed sensors, data acquisition devices, and wireless communication modules to ensure the normal operation of these devices. If a sensor failure or signal abnormality is detected, maintenance personnel will be automatically notified for inspection and replacement. The system's computing resources and processing capabilities are monitored in real time to ensure that the model reasoning speed and response time meet mine safety requirements. When the system load is too high, the system will improve performance by load balancing or optimizing the data processing pipeline. Anti-electromagnetic interference measures have been added to the system design to ensure that the acquisition and processing of acoustic signals are not interfered with in the complex electromagnetic environment of the mine. The system can automatically adjust the acquisition parameters and processing procedures based on factors such as temperature, humidity, and air pressure in the mine environment to ensure reliable operation in extreme environments.

[0177] Figure 4 is a schematic diagram of the composition structure of a mining lithium battery thermal runaway acoustic signal processing device based on machine learning provided in an embodiment of the present invention, such as Figure 4As shown, the thermal runaway acoustic signal processing device 400 for a mining lithium battery based on machine learning includes: an acquisition module 401, which is used to synchronously acquire a target acoustic signal and a battery environmental parameter of a mining lithium battery in a thermal runaway process; a preprocessing module 402, which is used to preprocess the target acoustic signal to obtain a preprocessed acoustic signal; an analysis module 403, which is used to perform spectrum analysis, time domain analysis and wavelet transform on the preprocessed acoustic signal, extract features related to thermal runaway, and obtain corresponding key frequency features, vibration features and local frequency changes related to thermal runaway of the mining lithium battery; a fusion module 404, which is used to fuse the key frequency features, vibration features and local frequency changes related to thermal runaway and the battery environmental parameters, and obtain a thermal runaway fusion feature vector using a feature-level fusion method; a monitoring module 405, which is used to perform thermal runaway monitoring on the mining lithium battery based on the thermal runaway fusion feature vector and a trained machine learning model to obtain a thermal runaway monitoring result. When an abnormal signal is detected, the system will automatically issue a warning message and initiate corresponding safety measures.

[0178] In some embodiments, the target acoustic signal includes instantaneous acoustic changes, abnormal acoustic signals and transmission acoustic signals; the acquisition module is also used to place a first sensor on the surface of the outer shell of the mining lithium battery; the first sensor is used to collect instantaneous acoustic changes when gas leaks or the battery shell ruptures; a second sensor is placed at the heat dissipation part of the mining lithium battery; the second sensor is used to detect abnormal acoustic signals caused by poor heat dissipation; a third sensor is placed at the battery module connection of the mining lithium battery; the third sensor is used to collect acoustic signals transmitted between batteries, that is, the transmission acoustic signal.

[0179] In some embodiments, the preprocessing module is also used to filter the collected initial acoustic signal using a bandpass filter to obtain a filtered acoustic signal; use a Kalman filter to predict the historical acoustic signal and correct the filtered acoustic signal to extract the effective acoustic signal related to thermal runaway; enhance the effective acoustic signal to obtain an enhanced acoustic signal; segment the enhanced acoustic signal to obtain the preprocessed acoustic signal.

[0180] In some embodiments, the analysis module is also used to perform spectral analysis on the preprocessed acoustic signal using fast Fourier transform to obtain key frequency characteristics related to the thermal runaway; perform time domain analysis on the preprocessed acoustic signal, extract peak amplitude by traversing sampling points to identify instantaneous strong vibrations, calculate the root mean square value to quantify the overall vibration intensity and stability, and measure the signal duration to evaluate the significance of abnormal vibrations, thereby obtaining vibration characteristics related to the thermal runaway; perform time-frequency analysis on the preprocessed acoustic signal using wavelet transform to obtain local frequency changes related to the thermal runaway.

[0181] In some embodiments, the machine learning-based thermal runaway acoustic signal processing algorithm for mining lithium batteries is applied to a thermal runaway monitoring model, and the device further includes: a training module for inputting the preprocessed sample acoustic signal into the thermal runaway monitoring model for subsequent feature extraction and training; through the feature extraction layer of the thermal runaway monitoring model, the preprocessed sample acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively to extract features related to thermal runaway, and obtain sample key frequency features, sample vibration features and sample local frequency changes related to thermal runaway; through the feature fusion layer of the thermal runaway monitoring model, the preprocessed sample acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively to extract features related to thermal runaway, and obtain sample key frequency features, sample vibration features and sample local frequency changes related to thermal runaway; through the feature fusion layer of the thermal runaway monitoring model, the preprocessed sample acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively to extract features related to thermal runaway, and obtain sample key frequency features, sample vibration features and sample local frequency changes related to thermal runaway. The sample thermal runaway key frequency characteristics, the sample vibration characteristics and the sample local frequency change are feature fused, and a feature-level fusion method is used to obtain a sample thermal runaway fusion feature vector; through the information prediction layer of the thermal runaway monitoring model, based on the sample thermal runaway fusion feature vector, a trained machine learning model is used to perform thermal runaway monitoring on the sample thermal runaway monitoring task to obtain a sample thermal runaway monitoring result; the sample thermal runaway monitoring result is input into a preset loss model to obtain a loss result; based on the loss result, the model parameters in the thermal runaway monitoring model are corrected to obtain a trained thermal runaway monitoring model.

[0182] In some embodiments, the device also includes: an output module, which is also used to divide the working state of the mining lithium battery into multiple levels based on the thermal runaway monitoring result; based on each level and the preset warning mechanism, a corresponding alarm signal is output; the multiple levels include normal operation, low-level warning, intermediate warning and emergency warning; the normal operation means that the target acoustic signal and the thermal runaway monitoring result both indicate that the mining lithium battery is in a normal working state; the low-level warning means that there is a low-frequency vibration sound or a signal fluctuation of a short duration in the target acoustic signal, and no emergency treatment measures are triggered; the intermediate warning means that there is a strong abnormal signal in the target acoustic signal, and the thermal runaway monitoring result is to predict the possibility of thermal runaway of the mining lithium battery; the emergency warning means that there are typical thermal runaway features in the target acoustic signal, and the thermal runaway monitoring result is to determine that the mining lithium battery is on the edge of thermal runaway or is already in a thermal runaway state, and trigger emergency treatment measures.

[0183] It should be noted that the description of the device of the embodiment of the present invention is similar to the description of the above method embodiment, and has similar beneficial effects as the method embodiment, so it is not repeated. For technical details not disclosed in the device embodiment, please refer to the description of the method embodiment of the present invention for understanding.

[0184] It should be noted that in the embodiment of the present invention, if the above-mentioned machine learning-based mining lithium battery thermal runaway acoustic signal processing algorithm is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions for a terminal to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read Only Memory), disk or optical disk and other media that can store program code. In this way, the embodiment of the present invention is not limited to any specific combination of hardware and software.

[0185] Correspondingly, an embodiment of the present invention provides a mining lithium battery thermal runaway acoustic signal processing device based on machine learning, Figure 5 is a schematic diagram of the composition structure of a mining lithium battery thermal runaway acoustic signal processing device based on machine learning provided in an embodiment of the present invention, such as Figure 5As shown, the mining lithium battery thermal runaway acoustic signal processing device 500 based on machine learning includes at least: a processor 501 and a computer-readable storage medium 502 configured to store executable instructions, wherein the processor 501 generally controls the overall operation of the mining lithium battery thermal runaway acoustic signal processing device 500 based on machine learning. The computer-readable storage medium 502 is configured to store instructions and applications executable by the processor 501, and can also cache data to be processed or processed by each module in the processor 501 and the mining lithium battery thermal runaway acoustic signal processing device 500 based on machine learning, which can be implemented through flash memory (FLASH) or random access memory (RAM, Random Access Memory).

[0186] An embodiment of the present invention provides a storage medium storing executable instructions, wherein the executable instructions are stored. When the executable instructions are executed by a processor, the processor will execute the algorithm provided by the embodiment of the present invention, for example, Figure 1 The algorithm shown.

[0187] In some embodiments, the storage medium can be a computer-readable storage medium, for example, a ferroelectric random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory, a magnetic surface memory, an optical disk, or a compact disk read-only memory (CD-ROM), etc.; it can also be various devices including one or any combination of the above memories.

[0188] In some embodiments, executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.

[0189] As an example, executable instructions may, but need not necessarily, correspond to a file in a file system, may be stored as part of a file storing other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or code portions). As an example, executable instructions may be deployed to be executed on one electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed at multiple locations and interconnected by a communication network.

[0190] The above description is only an embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement and improvement made within the spirit and scope of the present invention are included in the protection scope of the present invention.

[0191] It should be understood that "one embodiment" or "an embodiment" mentioned throughout the specification means that specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present invention. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in various embodiments of the present invention, the size of the serial number of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention. The serial numbers of the above-mentioned embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

[0192] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of further restrictions, an element defined by the statement "comprises one..." does not exclude the presence of other identical elements in the process, method, article or device including the element. In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed.

[0193] The above is only an embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A machine learning-based acoustic signal processing algorithm for thermal runaway of lithium batteries for mining, characterized in that: The algorithm includes: Synchronously obtain the target acoustic signals and battery environmental parameters of mining lithium batteries during thermal runaway; After preprocessing the target acoustic signal, a preprocessed acoustic signal is obtained; The pre-processed acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively to extract features related to thermal runaway, and correspondingly obtain key frequency features, vibration features and local frequency changes related to thermal runaway of the mining lithium battery; The key frequency characteristics, vibration characteristics and local frequency changes related to the thermal runaway and the battery environment parameters are fused, and a thermal runaway fusion feature vector is obtained using a feature-level fusion method; Based on the thermal runaway fusion feature vector, a trained machine learning model is used to monitor the thermal runaway of the mining lithium battery to obtain a thermal runaway monitoring result. When an abnormal signal is detected, the system will automatically issue a warning message and initiate corresponding safety measures.

2. The algorithm according to claim 1, characterized in that The target acoustic signal includes instantaneous acoustic changes, abnormal acoustic signals and transmission acoustic signals; the target acoustic signal and battery environmental parameters of the mining lithium battery in the thermal runaway process are obtained, including: A first sensor is placed on the surface of the outer shell of the mining lithium battery; the first sensor is used to collect instantaneous acoustic changes when gas leaks or the battery shell ruptures; A second sensor is placed at the heat dissipation part of the mining lithium battery; the second sensor is used to detect abnormal acoustic signals caused by poor heat dissipation; A third sensor is arranged at the connection of the battery modules of the mining lithium battery; the third sensor is used to collect the acoustic signal transmitted between the batteries, that is, the transmitted acoustic signal.

3. The algorithm according to claim 1, characterized in that After preprocessing the target acoustic signal, obtaining a preprocessed acoustic signal includes: A bandpass filter is used to filter the collected initial acoustic signal to obtain a filtered acoustic signal; A Kalman filter is used to predict historical acoustic signals and to correct the filtered acoustic signals to extract effective acoustic signals related to thermal runaway; Performing signal enhancement on the effective acoustic signal to obtain an enhanced acoustic signal; The enhanced acoustic signal is segmented to obtain the preprocessed acoustic signal.

4. The algorithm according to claim 1, characterized in that The pre-processed acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively to extract features related to thermal runaway, and corresponding key frequency features, vibration features and local frequency changes related to thermal runaway of the mining lithium battery are obtained, including: Performing spectrum analysis on the preprocessed acoustic signal using fast Fourier transform to obtain key frequency features related to the thermal runaway; Performing time domain analysis on the preprocessed acoustic signal, extracting peak amplitudes through traversing sampling points to identify instantaneous strong vibrations, calculating root mean square values ​​to quantify overall vibration intensity and stability, and measuring signal duration to evaluate the significance of abnormal vibrations, thereby obtaining vibration characteristics related to thermal runaway; The preprocessed acoustic signal is subjected to time-frequency analysis using wavelet transform to obtain the local frequency change related to the thermal runaway.

5. The algorithm according to claim 1, characterized in that The machine learning-based thermal runaway acoustic signal processing algorithm for mining lithium batteries is applied to a thermal runaway monitoring model, and the thermal runaway monitoring model is trained through the following steps: Inputting the preprocessed sample acoustic signal into the thermal runaway monitoring model for subsequent feature extraction and training; Through the feature extraction layer of the thermal runaway monitoring model, the preprocessed sample acoustic signal is subjected to spectrum analysis, time domain analysis and wavelet transform respectively, the features related to thermal runaway are extracted, and the key frequency features, sample vibration features and sample local frequency changes related to thermal runaway are obtained; Through the feature fusion layer of the thermal runaway monitoring model, the key frequency features of the sample thermal runaway, the sample vibration features and the sample local frequency changes are feature fused, and a sample thermal runaway fusion feature vector is obtained using a feature-level fusion method; Through the information prediction layer of the thermal runaway monitoring model, based on the sample thermal runaway fusion feature vector, a trained machine learning model is used to perform thermal runaway monitoring on the sample thermal runaway monitoring task to obtain a sample thermal runaway monitoring result; Inputting the sample thermal runaway monitoring result into a preset loss model to obtain a loss result; Based on the loss result, the model parameters in the thermal runaway monitoring model are corrected to obtain a trained thermal runaway monitoring model.

6. The algorithm according to claim 1, characterized in that The algorithm also includes: Based on the thermal runaway monitoring result, the working state of the mining lithium battery is divided into multiple levels; Based on each level and preset warning mechanism, output corresponding alarm signals; The multiple levels include normal operation, low-level warning, intermediate warning and emergency warning; the normal operation means that both the target acoustic signal and the thermal runaway monitoring result indicate that the mining lithium battery is in a normal working state; The low-level warning means that there is a low-frequency vibration sound or a signal fluctuation of short duration in the target acoustic signal, and no emergency treatment measures are triggered; The intermediate warning means that there is a strong abnormal signal in the target acoustic signal, and the thermal runaway monitoring result predicts the possibility of thermal runaway of the mining lithium battery; The emergency warning refers to the presence of typical thermal runaway characteristics in the target acoustic signal, and the thermal runaway monitoring result is to determine that the mining lithium battery is on the verge of thermal runaway or is already in a thermal runaway state, and trigger emergency treatment measures.

7. A mining lithium battery thermal runaway acoustic signal processing device based on machine learning, characterized in that: The device comprises: An acquisition module is used to synchronously acquire the target acoustic signal and battery environmental parameters of mining lithium batteries during thermal runaway; A preprocessing module, used for preprocessing the target acoustic signal to obtain a preprocessed acoustic signal; An analysis module is used to perform spectrum analysis, time domain analysis and wavelet transform on the preprocessed acoustic signal, extract features related to thermal runaway, and obtain key frequency features, vibration features and local frequency changes related to thermal runaway of the mining lithium battery; A fusion module, used to fuse the key frequency characteristics, vibration characteristics and local frequency changes related to the thermal runaway and the battery environment parameters, and obtain a thermal runaway fusion feature vector using a feature-level fusion method; The monitoring module is used to perform thermal runaway monitoring on the mining lithium battery based on the thermal runaway fusion feature vector and adopt a trained machine learning model to obtain thermal runaway monitoring results. When an abnormal signal is detected, the system will automatically issue a warning message and initiate corresponding safety measures.

8. A mining lithium battery thermal runaway acoustic signal processing device based on machine learning, characterized in that: include: A memory for storing executable instructions; A processor, for implementing the acoustic signal processing algorithm for thermal runaway of a mining lithium battery based on machine learning as described in any one of claims 1 to 6 when executing the executable instructions stored in the memory.

9. A computer-readable storage medium storing executable instructions for causing a processor to execute the executable instructions to implement the machine learning-based thermal runaway acoustic signal processing algorithm for mining lithium batteries as described in any one of claims 1 to 6.

Images