Lithium ion battery thermal runaway early warning method

Abstract

The invention discloses a lithium ion battery thermal runaway early warning method, and belongs to the technical field of battery safety monitoring, and the method comprises the steps: synchronously obtaining a low-frequency sound wave signal, a temperature signal, a voltage signal and a stress-strain signal of a lithium ion battery; noise reduction processing is carried out on the low-frequency sound wave signals, time-frequency feature extraction is carried out on the low-frequency sound wave signals after noise reduction processing, and feature frequency band energy corresponding to a thermal runaway early event is obtained; a thermal runaway sensitivity coefficient is calculated based on the characteristic frequency band energy, and when the thermal runaway sensitivity coefficient is larger than a first coefficient threshold value, thermal runaway early warning is triggered; and on the basis of the thermal runaway sensitivity coefficient, the temperature signal, the voltage signal and the stress-strain signal, constructing a feature vector, and inputting the feature vector into a trained long-short-term memory network and attention mechanism hybrid model for classification to obtain a thermal runaway early warning level. The method can solve the problems of installation difficulty, detection lag and high false alarm rate in the prior art.

Description

Technical Field

[0001] The present invention relates to a lithium-ion battery thermal runaway early warning method, belonging to the technical field of battery safety monitoring. Background Art

[0002] In response to the high risk of lithium-ion batteries, researchers have developed a variety of technologies to monitor and warn of thermal runaway accidents in lithium-ion batteries to ensure their safe use. However, existing lithium-ion battery warning technologies have several limitations.

[0003] Traditional temperature and voltage sensors, as well as acoustic warning technologies for safety valve opening, trigger alarms only during the intense phase of thermal runaway (temperatures > 150°C), failing to meet early warning requirements. Gas detection methods are susceptible to air flow, and the concentrations of characteristic gases like CO and H2 are below the detection limit in the early stages of thermal runaway. Traditional temperature and new stress-strain warning systems require intrusion into battery cells and modules at the detection end, which is destructive to the battery structure and increases sensor tolerance requirements, making it difficult to retrofit and update lithium-ion batteries already in production. Existing acoustic emission technology primarily monitors high-frequency mechanical waves greater than 20kHz and is unable to capture critical low-frequency physical and chemical reaction signals such as electrolyte decomposition.

[0004] These limitations result in the inability of existing early warning technologies to provide timely and reliable warnings, posing hidden dangers to the large-scale application of lithium-ion batteries. Summary of the Invention

[0005] The purpose of the present invention is to provide a lithium-ion battery thermal runaway early warning method, which solves the problems of difficult installation, delayed detection and high false alarm rate in the prior art by extracting low-frequency acoustic wave features and fusing multimodal data.

[0006] In order to achieve the above object, the present invention provides the following technical solutions: In a first aspect, the present invention provides a lithium-ion battery thermal runaway early warning method, comprising: Synchronously acquire low-frequency acoustic wave signals, temperature signals, voltage signals, and stress-strain signals of lithium-ion batteries; Perform noise reduction on the low-frequency acoustic wave signal and extract the time-frequency features of the low-frequency acoustic wave signal after noise reduction to obtain the characteristic frequency band energy corresponding to the early stage of thermal runaway events; Calculate the thermal runaway sensitivity coefficient based on the energy of the characteristic frequency band. When the thermal runaway sensitivity coefficient is greater than the first coefficient threshold, a thermal runaway warning is triggered. Based on the thermal runaway sensitivity coefficient, temperature signal, voltage signal and stress-strain signal, a feature vector is constructed, and the feature vector is input into the trained long-short-term memory network and attention mechanism hybrid model for classification to obtain the thermal runaway warning level.

[0007] In combination with the first aspect, further, synchronously acquiring the low-frequency acoustic wave signal, temperature signal, voltage signal, and stress-strain signal of the lithium-ion battery includes: Through the non-invasive arrangement of acoustic sensing arrays, temperature sensing modules, voltage sensing modules and stress and strain sensing modules, the low-frequency acoustic wave signals, temperature signals, voltage signals and stress and strain signals of lithium-ion batteries are synchronously collected at a preset sampling rate.

[0008] In combination with the first aspect, further, the acoustic sensing array includes MEMS sensors arranged on two diagonal sides of the battery pack or battery module; The temperature sensing module includes a thermocouple and an infrared thermometer arranged on the surface of the single battery; The voltage sensing module includes a voltage probe connected to the output terminal of the battery module and a differential amplifier circuit with optical coupling isolation; The stress and strain sensing module includes a stress sensor arranged on the surface of the battery module and a single-bridge strain sensor arranged on the surface of the single battery.

[0009] In combination with the first aspect, further, performing noise reduction processing on the low-frequency sound wave signal includes: Constructing a transfer function based on the spatial arrangement of the acoustic sensing array; Based on the transfer function, a normalized least mean square adaptive filter is used to cancel the noise of the low-frequency sound wave signal. Among them, the noise transfer function for: ; in, represents the ambient noise spectrum collected by the reference sensor, Represents the target mixed signal spectrum collected by the main sensor.

[0010] In combination with the first aspect, further, time-frequency feature extraction is performed on the low-frequency sound wave signal after noise reduction to obtain the characteristic frequency band energy corresponding to the early thermal runaway event, including: Perform Morlet wavelet transform on the low-frequency acoustic signal after noise reduction to extract the energy of the characteristic frequency band corresponding to the early stage of thermal runaway events; Among them, the characteristic frequency band energy corresponding to the early thermal runaway events includes the electrolyte boiling event energy, the diaphragm rupture event energy and the shell deformation event energy.

[0011] Combined with the first aspect, further, the calculation formula of the thermal runaway sensitivity coefficient is: ; in, represents the thermal runaway sensitivity coefficient, represents the energy of the electrolyte boiling event, represents the energy of the diaphragm rupture event, represents the shell deformation event energy, Indicates the total sound energy in the low frequency band.

[0012] Combining the first aspect, the hybrid model of long short-term memory network and attention mechanism includes: An input module for receiving a feature vector; The temporal feature extraction module is used to capture the temporal dependency of feature vectors through two layers of long short-term memory network units to obtain the hidden state sequence; The attention weighting module is used to calculate the weight distribution of historical hidden states based on the hidden state sequence and generate a weighted feature vector; Feature fusion module, used to reduce the dimension and perform nonlinear transformation on the weighted feature vector through the fully connected layer; The output module is used to generate a thermal runaway warning probability distribution through an activation function based on the weighted feature vector after dimensionality reduction and nonlinear transformation, and output a thermal runaway warning level based on the thermal runaway warning probability distribution.

[0013] Combined with the first aspect, further, the hybrid model of long short-term memory network and attention mechanism is trained using multimodal time series data of thermal runaway of lithium-ion batteries through weighted cross entropy loss function and Adam optimizer.

[0014] Combined with the first aspect, further, the thermal runaway warning levels include: Observation level, triggered when a single modality data is abnormal; Action level: triggered when at least two modal data are abnormal and the thermal runaway sensitivity coefficient is greater than the second coefficient threshold.

[0015] In a second aspect, the present invention provides a lithium-ion battery thermal runaway warning system, comprising: The sensing layer is used to synchronously obtain the low-frequency acoustic wave signal, temperature signal, voltage signal and stress and strain signal of the lithium-ion battery; The processing layer is used to perform noise reduction on the low-frequency sound wave signal and extract the time-frequency features of the low-frequency sound wave signal after noise reduction to obtain the characteristic frequency band energy corresponding to the early stage of thermal runaway events; The decision layer is used to calculate the thermal runaway sensitivity coefficient based on the energy of the characteristic frequency band. When the thermal runaway sensitivity coefficient is greater than the first coefficient threshold, a thermal runaway warning is triggered; and it is used to construct a feature vector based on the thermal runaway sensitivity coefficient, temperature signal, voltage signal and stress-strain signal, and input the feature vector into the trained long-short-term memory network and attention mechanism hybrid model for classification to obtain the thermal runaway warning level.

[0016] In a third aspect, the present invention provides a computer device, comprising: Storage medium for storing computer programs; A processor is used to execute a computer program to implement the lithium-ion battery thermal runaway early warning method according to any one of the first aspects.

[0017] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the lithium-ion battery thermal runaway warning method according to any one of the first aspects.

[0018] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the lithium-ion battery thermal runaway warning method according to any one of the first aspects.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The lithium-ion battery thermal runaway warning method provided by the present invention synchronously collects low-frequency acoustic wave signals and multi-physical quantity data, constructs a thermal runaway sensitivity coefficient through dynamic noise suppression and time-frequency feature extraction, and combines multi-dimensional data coupling analysis. It uses a hybrid model of long-short-term memory network and attention mechanism to achieve multi-source spatiotemporal feature fusion, outputs a thermal runaway warning level, can identify early characteristic signals, and trigger a warning 5-7 minutes before thermal runaway. The signal-to-noise ratio is improved by more than 40dB, and the response delay is less than 200ms. It can significantly improve the safety of lithium-ion battery application scenarios and reduce the risk of fire accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is an architecture diagram of a lithium-ion battery thermal runaway warning system provided by an embodiment of the present invention; Figure 2 This is a flow chart of a lithium-ion battery thermal runaway early warning method provided by an embodiment of the present invention; Figure 3 This is a diagram showing the evolution of low-frequency acoustic wave signals during a thermal runaway process provided by an embodiment of the present invention; Figure 4 This is a spectrum diagram of a low-frequency acoustic wave signal during a thermal runaway process provided by an embodiment of the present invention; Figure 5 This is a multimodal data fusion early warning decision logic diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The technical solution of the present application will be further described in detail below in conjunction with specific implementation methods.

[0022] The embodiments of the present application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application and are not to be construed as limiting the present application. The embodiments of the present application and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0023] The present invention provides a method for early warning of thermal runaway of a lithium-ion battery, comprising: Synchronously acquire low-frequency acoustic wave signals, temperature signals, voltage signals, and stress-strain signals of lithium-ion batteries; Perform noise reduction on the low-frequency acoustic wave signal and extract the time-frequency features of the low-frequency acoustic wave signal after noise reduction to obtain the characteristic frequency band energy corresponding to the early stage of thermal runaway events; Calculate the thermal runaway sensitivity coefficient based on the energy of the characteristic frequency band. When the thermal runaway sensitivity coefficient is greater than the first coefficient threshold, a thermal runaway warning is triggered. Based on the thermal runaway sensitivity coefficient, temperature signal, voltage signal and stress-strain signal, a feature vector is constructed, and the feature vector is input into the trained long-short-term memory network and attention mechanism hybrid model for classification to obtain the thermal runaway warning level.

[0024] In the early stages of thermal runaway, such as when SEI film decomposition and lithium deposition occur at the negative electrode, characteristic low-frequency acoustic waves ranging from 0.1Hz to 100Hz are generated. However, their signal intensity is only 1 / 1000th of the ambient noise. Furthermore, in the early stages of thermal runaway in lithium-ion batteries, the low-frequency acoustic signal accounts for the majority of the energy in the entire frequency band. Furthermore, the acoustic signal has excellent penetration, allowing detection outside the battery pack. Using this signal can prevent damage to the original battery module structure.

[0025] The lithium-ion battery thermal runaway warning method provided in the embodiment of the present application can effectively identify early abnormal low-frequency acoustic wave signals during the use of lithium-ion batteries, reduce the probability of lithium-ion battery thermal runaway caused by backward warning technology, such as the probability of an accident in which a whole cabin fire is induced by battery thermal runaway in an energy storage power station, and greatly improve the fire safety of lithium-ion battery application scenarios; the current mainstream battery module thermal runaway risk warning technology has a high application cost, and mainly uses a temperature signal, acoustic signal, electrical signal and combustible gas concentration coupling method to alarm on the eve of or when the disaster is triggered. The warning timing is relatively delayed, the disaster control is difficult, the modification cost is high, and it is easy to cause serious disaster losses. Therefore, the recognition technology of lithium-ion battery thermal runaway low-frequency acoustic wave signals can effectively identify the early low-frequency acoustic wave signals of thermal runaway disasters, which has significant benefits for the smooth operation of lithium-ion batteries.

[0026] In one possible embodiment, the first coefficient threshold is 0.15.

[0027] In this embodiment, when the thermal runaway sensitivity coefficient is greater than the first coefficient threshold value of 0.15 for three consecutive times, a thermal runaway warning is triggered.

[0028] In one possible embodiment, synchronously acquiring low-frequency acoustic wave signals, temperature signals, voltage signals, and stress and strain signals of a lithium-ion battery specifically includes: synchronously acquiring low-frequency acoustic wave signals, temperature signals, voltage signals, and stress and strain signals of the lithium-ion battery at a preset sampling rate through a non-invasively arranged acoustic sensing array, a temperature sensing module, a voltage sensing module, and a stress and strain sensing module, and using time-division multiplexing technology to reduce multi-channel crosstalk, thereby acquiring multiple types of signals, mainly low-frequency acoustic wave signals, during the operation of the lithium-ion battery.

[0029] Non-invasive includes physical non-invasive and signal non-invasive. Physical non-invasive means that the probe only contacts the external electrodes of the battery, and signal non-invasive means that no additional current is injected during the acquisition process, which does not affect the chemical state of the battery.

[0030] In this embodiment, the acoustic sensing array includes MEMS sensors arranged diagonally across the battery pack or module. For example, when used within an energy storage compartment, the sensors are arranged diagonally within the compartment to form an array. This facilitates monitoring the acoustic characteristics of all battery modules within the compartment and lays the foundation for subsequent updates to the positioning algorithm. The temperature sensing module includes thermocouples and infrared thermometers arranged on the surface of the individual cells. Six thermocouples are installed on the surface of each battery within the module to monitor all-around temperature changes in the battery, and in conjunction with the infrared thermometer, fully identify the battery temperature evolution pattern. The voltage sensing module includes a voltage probe connected to the output of the battery module and a differential amplifier circuit with optocoupler isolation. The voltage probe is connected non-invasively, and the differential amplifier circuit is integrated after the voltage probe. The stress and strain sensing module includes a stress sensor arranged on the surface of the battery module and a single-bridge strain sensor arranged on the surface of the individual cells to monitor stress and strain signals that occur earlier than temperature changes.

[0031] Specifically, the 0.1Hz~100Hz low-frequency acoustic wave signal, temperature signal, voltage signal and stress-strain signal of the lithium-ion battery are synchronously collected at a sampling rate of 100Hz.

[0032] The 0.1Hz~100Hz low-frequency acoustic wave signal covers the acoustic emission spectrum range of key physical and chemical reactions in the early stage of thermal runaway: ①0.1Hz~5Hz corresponds to the early thermal runaway event: SEI film decomposition; ②5Hz~10Hz corresponds to the early thermal runaway event: electrolyte boiling; ③20Hz~30Hz corresponding to the early thermal runaway event is: diaphragm rupture; ④The early thermal runaway event corresponding to 50Hz~80Hz is: shell deformation.

[0033] The lithium-ion battery thermal runaway warning method provided in this embodiment can focus on low-frequency acoustic wave signals, construct an acoustic fingerprint map including three types of events: electrolyte boiling, diaphragm rupture, and shell deformation, and effectively construct a low-frequency acoustic wave disaster map.

[0034] In a possible embodiment, performing noise reduction processing on the low-frequency sound wave signal specifically includes the following steps: Step 1: Construct a transfer function based on the spatial arrangement of the acoustic sensor array; In this embodiment, the noise transfer function for: ; in, represents the ambient noise spectrum collected by the reference sensor, Represents the target mixed signal spectrum collected by the main sensor.

[0035] : The noise spectrum collected by the reference sensor is placed in a non-battery area (such as the module bracket). It only contains environmental noise (fan / mechanical vibration) and provides a pure noise sample for building a noise baseline.

[0036] : The mixed signal spectrum collected by the main sensor, arranged on the battery surface, contains the target sound wave signal + environmental noise, and is the signal source to be processed, from which the effective sound wave needs to be separated.

[0037] : Noise transfer function, which characterizes the propagation characteristics of noise from the reference point to the main point, is in complex form (including amplitude / phase), establishes a noise space mapping relationship, and guides the filter to generate a noise cancellation signal.

[0038] Reference sensors are placed diagonally across the battery pack to maximize ambient noise capture and minimize interference with battery signals. The main sensor, attached to the battery surface, directly monitors low-frequency sound waves from 0.1Hz to 100Hz. The transfer function addresses the randomness of time-domain noise and uses a normalized least mean square adaptive filter to adjust coefficients in real time to accommodate noise fluctuations (such as fan startup and shutdown).

[0039] Step 2: Based on the transfer function, a normalized least mean square adaptive filter is used to cancel the noise of the low-frequency sound wave signal.

[0040] In this embodiment, the step size parameter of the normalized least mean square adaptive filter is satisfy ,in, is the autocorrelation matrix of the input signal.

[0041] The normalized least mean square adaptive filter is used to cancel the noise of low-frequency sound wave signals, which can improve the signal-to-noise ratio by more than or equal to 40dB. Figure 2 As shown in the figure, under the background noise of the energy storage cabin (fan / inverter interference), the filter makes the lower limit of effective signal detection reach 0.01Pa.

[0042] In one possible embodiment, time-frequency feature extraction is performed on the low-frequency sound wave signal after noise reduction processing to obtain characteristic frequency band energy corresponding to the early thermal runaway event, which specifically includes: performing Morlet wavelet transform on the low-frequency sound wave signal after noise reduction processing to extract characteristic frequency band energy corresponding to the early thermal runaway event.

[0043] In this embodiment, the characteristic frequency band energy corresponding to the early thermal runaway event includes the electrolyte boiling event energy, the diaphragm rupture event energy, and the shell deformation event energy.

[0044] The calculation formula of thermal runaway sensitivity coefficient is: ; in, represents the thermal runaway sensitivity coefficient, represents the energy of the electrolyte boiling event, represents the energy of the diaphragm rupture event, represents the shell deformation event energy, Indicates the total sound energy in the low frequency band.

[0045] Specifically, the energy of the electrolyte boiling event corresponds to the sound energy in the frequency band of 5Hz~10Hz, the energy of the diaphragm rupture event corresponds to the sound energy in the frequency band of 20Hz~30Hz, the energy of the shell deformation event corresponds to the sound energy in the frequency band of 50Hz~80Hz, and the total sound energy in the low-frequency band is the total sound energy in the low-frequency band of 0.1Hz~100Hz.

[0046] Extracting the energy of the characteristic frequency band corresponding to the early stage of thermal runaway events and calculating the thermal runaway sensitivity coefficient specifically includes the following steps: Step a: Directed extraction of key event energy; Only three characteristic frequency bands strongly associated with thermal runaway were selected (not all frequency bands were randomly selected). Electrolyte boiling (5Hz-10Hz), diaphragm rupture (20Hz-30Hz), and housing deformation (50Hz-80Hz) are the earliest detectable events of thermal runaway.

[0047] Step b: Calculate the total energy benchmark; Normalization: Eliminate the influence of environmental noise fluctuations (such as The sudden increase could be an external vibration); Function of the denominator: Converts absolute energy values into relative proportions to improve cross-scenario applicability.

[0048] Step c: Construct sensitivity coefficients.

[0049] Molecular focus on disaster events: Quantify core risks; Ratio meaning: When When it rises, it indicates that the proportion of disaster event energy has increased (even if the total energy remains unchanged).

[0050] Example: Normal state: ; Early stage of thermal runaway: (The first coefficient threshold is reached).

[0051] In this embodiment, the low-frequency acoustic wave signal during the thermal runaway process evolves as follows: Figure 3 As shown, the spectrum of the low-frequency acoustic wave signal during thermal runaway is as follows: Figure 4 shown.

[0052] In one possible embodiment, the hybrid model of the long short-term memory network and the attention mechanism specifically includes: An input module for receiving a feature vector; The temporal feature extraction module is used to capture the temporal dependency of feature vectors through two layers of long short-term memory network units to obtain the hidden state sequence; The attention weighting module is used to calculate the weight distribution of historical hidden states based on the hidden state sequence and generate a weighted feature vector; Feature fusion module, used to reduce the dimension and perform nonlinear transformation on the weighted feature vector through the fully connected layer; The output module is used to generate a thermal runaway warning probability distribution through an activation function based on the weighted feature vector after dimensionality reduction and nonlinear transformation, and output a thermal runaway warning level based on the thermal runaway warning probability distribution.

[0053] In this embodiment, the long short-term memory network is used to capture 、 、 、 The attention mechanism is used to detect disaster-sensitive features (such as sudden ) Dynamically assign higher weights to improve the reliability of decisions.

[0054] The long short-term memory network layer is designed as follows: Time step: 30 steps (30 seconds), The value must be greater than the first coefficient threshold for three consecutive times (sampling rate 1Hz → covering a 30-second window); Number of hidden layers: 2, balancing complexity and performance (experimental verification: the risk of overfitting increases when the number of layers is greater than 2); Number of hidden units: 64 / layer, input feature dimension 4→64 units can fully extract temporal patterns; Activation function: tanh (memory gate), standard LSTM configuration to avoid gradient disappearance.

[0055] The attention mechanism is designed as follows: Query vector Q: Q=H t W q +b q , focus on the current state; Key vector K: K=H {t-1} W k +b k , associated historical status; Attention weight α: α t =softmax(Q·K T / √d k ), weighted key features (d k =64 is the scaling factor); Weighted output: X' t =∑(α t,i ·H t,i ), disaster-sensitive characteristics (such as sudden ) to obtain high weight.

[0056] In one possible embodiment, a hybrid model of a long short-term memory network and an attention mechanism is trained using multimodal time series data of thermal runaway of lithium-ion batteries using a weighted cross entropy loss function and an Adam optimizer.

[0057] The training dataset is constructed as follows: Laboratory thermal runaway test: 200 sets of 18650 batteries (SOC=100%) overcharged to thermal runaway, 1200 samples; Actual operation data of energy storage power stations: marked with normal / warning status (provided by partners), 5,800 samples; Enhanced data: Add ±10% Gaussian noise / time offset, total sample size 7000.

[0058] The training parameters are configured as follows: Loss function: weighted cross entropy to address sample imbalance (normal: warning = 8:1); Optimizer: Adam (β1=0.9, β2=0.999); Learning rate: Initially 0.001, decaying by 50% every 10 rounds. Experimental verification: balancing convergence speed and stability; Batch size: 32, GPU memory limit (NVIDIA T4 16GB); Number of training rounds: 100 rounds, early stopping strategy: terminate if the validation set loss does not decrease for 5 consecutive rounds.

[0059] Specifically, the feature vector ,in, Indicates the voltage change rate in V / s, obtained based on the voltage signal. Indicates the temperature rise rate in °C / s, obtained based on the temperature signal. Represents the root mean square value of stress and strain, which is obtained based on the stress and strain signal. Figure 5 As shown in FIG, the feature vector is input into the trained hybrid model of long short-term memory network and attention mechanism for multi-source spatiotemporal feature fusion, and the hybrid model of long short-term memory network and attention mechanism performs classification and outputs the thermal runaway warning level.

[0060] In this embodiment, the thermal runaway warning levels specifically include: Observation level, triggered when a single modality data is abnormal; Action level: triggered when at least two modal data are abnormal and the thermal runaway sensitivity coefficient is greater than the second coefficient threshold.

[0061] Specifically, the second coefficient threshold is 0.2.

[0062] In one possible embodiment, the failure scenarios are shown in Table 1.

[0063] Table 1: Failure scenarios.

[0064] .

[0065] single May cause misjudgment, such as external vibration, but combined with sudden drop, steep increase and Drastic changes can improve the accuracy of early warnings.

[0066] The lithium-ion battery thermal runaway warning method provided in the embodiment of the present application uses a hybrid model of a long short-term memory network and an attention mechanism deployed in the cloud. The response delay is less than 200ms, and a warning is triggered 5-7 minutes before thermal runaway. The false alarm rate is less than 3%. It can realize the spatiotemporal feature fusion of multi-source data, clearly display the signal characteristics triggered by thermal runaway of the lithium-ion battery, and output the thermal runaway warning level based on the model.

[0067] The present invention provides a lithium-ion battery thermal runaway warning system, comprising: The sensing layer is used to synchronously obtain the low-frequency acoustic wave signal, temperature signal, voltage signal and stress and strain signal of the lithium-ion battery; The processing layer is used to perform noise reduction on the low-frequency sound wave signal and extract the time-frequency features of the low-frequency sound wave signal after noise reduction to obtain the characteristic frequency band energy corresponding to the early stage of thermal runaway events; The decision layer is used to calculate the thermal runaway sensitivity coefficient based on the energy of the characteristic frequency band. When the thermal runaway sensitivity coefficient is greater than the first coefficient threshold, a thermal runaway warning is triggered; and it is used to construct a feature vector based on the thermal runaway sensitivity coefficient, temperature signal, voltage signal and stress-strain signal, and input the feature vector into the trained long-short-term memory network and attention mechanism hybrid model for classification to obtain the thermal runaway warning level.

[0068] The lithium-ion battery thermal runaway warning system provided in the embodiment of the present application is as follows: Figure 1 As shown in the figure, the processing layer includes an embedded signal processing unit and a terminal processing unit. The embedded signal processing unit is developed based on the FPGA+ARM architecture, with a built-in dynamic noise suppression module and feature extraction algorithm, which can identify low-frequency acoustic wave signals in battery operation and perform pre-processing; the terminal processing unit is developed based on MATLAB, which can analyze and fit multi-dimensional temperature and stress-strain data, and perform coupling analysis in combination with low-frequency acoustic wave signals to construct a disaster signal feature mapping set during battery operation.

[0069] The lithium-ion battery thermal runaway warning system provided in this embodiment can execute the lithium-ion battery thermal runaway warning method provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects of the execution method.

[0070] An embodiment of the present application provides a computer device, including: Storage medium for storing computer programs; A processor is used to execute a computer program to implement the lithium-ion battery thermal runaway early warning method provided in any embodiment of the present application.

[0071] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the lithium-ion battery thermal runaway warning method provided in any embodiment of the present application is implemented.

[0072] An embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, the lithium-ion battery thermal runaway warning method provided in any embodiment of the present application is implemented.

[0073] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0074] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0075] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0077] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A lithium-ion battery thermal runaway early warning method, characterized in that: include: Synchronously acquire low-frequency acoustic wave signals, temperature signals, voltage signals, and stress-strain signals of lithium-ion batteries; Perform noise reduction on the low-frequency acoustic wave signal and extract the time-frequency features of the low-frequency acoustic wave signal after noise reduction to obtain the characteristic frequency band energy corresponding to the early stage of thermal runaway events; Calculate the thermal runaway sensitivity coefficient based on the energy of the characteristic frequency band. When the thermal runaway sensitivity coefficient is greater than the first coefficient threshold, a thermal runaway warning is triggered. Based on the thermal runaway sensitivity coefficient, temperature signal, voltage signal and stress-strain signal, a feature vector is constructed, and the feature vector is input into the trained long-short-term memory network and attention mechanism hybrid model for classification to obtain the thermal runaway warning level.

2. The lithium-ion battery thermal runaway early warning method according to claim 1, characterized in that: Synchronously acquiring low-frequency acoustic wave signals, temperature signals, voltage signals, and stress-strain signals of lithium-ion batteries includes: Through the non-invasive arrangement of acoustic sensing arrays, temperature sensing modules, voltage sensing modules and stress and strain sensing modules, the low-frequency acoustic wave signals, temperature signals, voltage signals and stress and strain signals of lithium-ion batteries are synchronously collected at a preset sampling rate.

3. The lithium-ion battery thermal runaway early warning method according to claim 2, characterized in that: The acoustic sensing array includes MEMS sensors arranged on two diagonal sides of the battery pack or battery module; The temperature sensing module includes a thermocouple and an infrared thermometer arranged on the surface of the single battery; The voltage sensing module includes a voltage probe connected to the output terminal of the battery module and a differential amplifier circuit with optical coupling isolation; The stress and strain sensing module includes a stress sensor arranged on the surface of the battery module and a single-bridge strain sensor arranged on the surface of the single battery.

4. The lithium-ion battery thermal runaway early warning method according to claim 1, characterized in that: Noise reduction processing for low-frequency sound wave signals includes: Constructing a transfer function based on the spatial arrangement of the acoustic sensing array; Based on the transfer function, a normalized least mean square adaptive filter is used to cancel the noise of the low-frequency sound wave signal. Among them, the noise transfer function for: ； in, represents the ambient noise spectrum collected by the reference sensor, Represents the target mixed signal spectrum collected by the main sensor.

5. The lithium-ion battery thermal runaway early warning method according to claim 1, characterized in that: The time-frequency feature extraction of the low-frequency acoustic wave signal after noise reduction is performed to obtain the characteristic frequency band energy corresponding to the early thermal runaway event, including: Perform Morlet wavelet transform on the low-frequency acoustic signal after noise reduction to extract the energy of the characteristic frequency band corresponding to the early stage of thermal runaway events; Among them, the characteristic frequency band energy corresponding to the early thermal runaway events includes the electrolyte boiling event energy, the diaphragm rupture event energy and the shell deformation event energy.

6. The lithium-ion battery thermal runaway early warning method according to claim 1, characterized in that: The calculation formula of thermal runaway sensitivity coefficient is: ； in, represents the thermal runaway sensitivity coefficient, represents the energy of the electrolyte boiling event, represents the energy of the diaphragm rupture event, represents the shell deformation event energy, Indicates the total sound energy in the low frequency band.

7. The lithium-ion battery thermal runaway early warning method according to claim 1, characterized in that: The hybrid model of long short-term memory network and attention mechanism includes: An input module for receiving a feature vector; The temporal feature extraction module is used to capture the temporal dependency of feature vectors through two layers of long short-term memory network units to obtain the hidden state sequence; The attention weighting module is used to calculate the weight distribution of historical hidden states based on the hidden state sequence and generate a weighted feature vector; Feature fusion module, used to reduce the dimension and perform nonlinear transformation on the weighted feature vector through the fully connected layer; The output module is used to generate a thermal runaway warning probability distribution through an activation function based on the weighted feature vector after dimensionality reduction and nonlinear transformation, and output a thermal runaway warning level according to the thermal runaway warning probability distribution.

8. The lithium-ion battery thermal runaway early warning method according to claim 1, characterized in that: The hybrid model of long short-term memory network and attention mechanism is trained using multimodal time series data of thermal runaway of lithium-ion batteries through weighted cross entropy loss function and Adam optimizer.

9. The lithium-ion battery thermal runaway early warning method according to claim 1, characterized in that: Thermal runaway warning levels include: Observation level, triggered when a single modality data is abnormal; Action level: triggered when at least two modal data are abnormal and the thermal runaway sensitivity coefficient is greater than the second coefficient threshold.

10. A lithium-ion battery thermal runaway warning system, characterized in that: include: The sensing layer is used to synchronously obtain the low-frequency acoustic wave signal, temperature signal, voltage signal and stress and strain signal of the lithium-ion battery; The processing layer is used to perform noise reduction on the low-frequency sound wave signal and extract the time-frequency features of the low-frequency sound wave signal after noise reduction to obtain the characteristic frequency band energy corresponding to the early stage of thermal runaway events; The decision layer is used to calculate the thermal runaway sensitivity coefficient based on the energy of the characteristic frequency band. When the thermal runaway sensitivity coefficient is greater than the first coefficient threshold, a thermal runaway warning is triggered; It is also used to construct a feature vector based on the thermal runaway sensitivity coefficient, temperature signal, voltage signal and stress-strain signal, and input the feature vector into the trained long-short-term memory network and attention mechanism hybrid model for classification to obtain the thermal runaway warning level.

Images