Vehicle battery thermal runaway early warning and thermal runaway device

Abstract

The invention provides a vehicle battery thermal runaway early warning and thermal runaway device, and the method comprises the steps: obtaining an operation data set of a to-be-detected battery in a preset time period; the operation data set is input into a battery thermal runaway prediction model to predict the error range of each data sample in the operation data set in the preset time period, and the battery thermal runaway prediction model is used for representing the relation between the error range and the data samples; and calculating a difference value between each data sample and the corresponding target data, and when any difference value is not in the corresponding error range, triggering thermal runaway early warning. According to the application, the running state of the vehicle battery in the preset time period is monitored and the thermal runaway risk is early warned, so that the running safety of the battery is ensured.

Description

Technical Field

[0001] The present application relates to the field of vehicle technology, and in particular to a vehicle battery thermal runaway warning and thermal runaway device. Background Art

[0002] With the development of electric vehicles, negative impacts are gradually becoming apparent. High prices, limited range, short cycle life, and limited safety and reliability are among their most prominent shortcomings. Currently, the dual demands for fast charging and long lifespan of lithium-ion batteries are growing. However, fast charging often accelerates battery aging, particularly the formation of lithium metal plating on the negative electrode, leading to a significant decrease in capacity and performance. Due to their design, lithium-ion batteries for electric vehicles operate optimally between 20 and 40°C. Temperatures that are too high or too low can affect battery performance and pose safety risks. The charging and discharging process of power batteries generates heat, which in turn causes the temperature to rise. If the temperature in flammable areas reaches a certain level, reaching the ignition point of the material, thermal runaway can occur in the power battery. Thermal runaway refers to a chain reaction of exothermic reactions in battery cells that causes the battery temperature to rise uncontrollably, eventually leading to explosion and fire.

[0003] In addition, when the battery is in abnormal operating conditions, such as overcharging, over-discharging, or used in a high-temperature environment, the electrode material will undergo corresponding side reactions with the electrolyte and produce a large amount of gas. At the same time, the battery shell will continue to expand as the internal air pressure increases. When the battery cannot expand to break through the critical volume state, the battery will explode and catch fire, causing serious safety problems. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a vehicle battery thermal runaway warning and thermal runaway device to overcome at least one of the above-mentioned defects.

[0005] In a first aspect, an embodiment of the present application provides a vehicle battery thermal runaway warning, the method comprising: obtaining an operating data set of a battery to be tested within a preset time period; inputting the operating data set into a battery thermal runaway prediction model to predict the error range of each data sample in the operating data set within the preset time period, wherein the battery thermal runaway prediction model is used to characterize the relationship between the error range and the data sample; calculating the difference between each data sample and the corresponding target data, and triggering a thermal runaway warning when any difference is not within the corresponding error range. In an optional embodiment of the present application, the operating data set is obtained in the following manner: when it is detected that the battery to be tested is placed in a test platform, the battery is controlled to be powered on, the test platform includes a controllable temperature environment box, a multi-channel data acquisition instrument and an overcharge control module, the controllable temperature environment box is used to adjust the temperature of the battery to be tested in the test platform, the multi-channel data acquisition instrument is used to collect the operating data set of the battery to be tested during the test, and send the operating data set to the processor, the overcharge control module is used to trigger battery thermal runaway by adjusting the charging current or voltage of the battery to be tested; determine the test items, the test items include temperature test, current test and voltage test; start the test according to the test items, and determine the data received in real time from the multi-channel data acquisition instrument as the operating data set of the battery to be tested.

[0006] In an optional embodiment of the present application, the battery thermal runaway prediction model includes an input layer, a one-dimensional convolution feature extraction layer, an attention mechanism module, a residual structure module, a long short-term memory network layer and an output layer, wherein the input layer is used to receive an operating data set collected within a preset time period after preprocessing to form a feature matrix; the one-dimensional convolution feature extraction layer is used to perform a sliding convolution operation on the feature matrix to generate a three-dimensional feature matrix image containing multiple channels; the attention mechanism module is used to compress the three-dimensional feature matrix image through global average pooling, two fully connected layers and an activation function, determine the channel weight vector, and recalibrate the three-dimensional feature matrix image according to the channel weight; the residual structure module is used to perform a one-dimensional convolution operation on the recalibrated three-dimensional feature matrix image, and generate a residual output through cross-channel adjustment and feature addition; the long short-term memory network layer is used to perform time series feature extraction on the residual output and output a hidden state vector; the output layer is used to map the hidden state vector to the error range of each operating data within the preset time period.

[0007] In an optional embodiment of the present application, the residual structure module is configured to: perform a one-dimensional convolution operation on the recalibrated three-dimensional feature matrix image to output a first feature map; adjust the number of channels and spatial size of the three-dimensional feature matrix image through a preset number of convolution kernels to output a second feature map; add the first feature map and the second feature map element by element to generate a residual output.

[0008] In an optional embodiment of the present application, the attention mechanism module is configured to: perform global average pooling on the input three-dimensional feature matrix image to generate a channel descriptor vector; integrate channel features through two-level fully connected layers, the first-level fully connected layer scales the number of channels, and the second-level fully connected layer generates a channel weight vector; use the activation function to generate a channel weight vector, and recalibrate the input feature tensor according to the channel weight.

[0009] In an optional embodiment of the present application, each data sample in the operating data set includes a timestamp and the surface temperature, output voltage and battery expansion pressure signal of the battery to be tested detected at the timestamp. The operating data set describes the continuous process of the battery from normal cycle to overcharge to thermal runaway in a time series.

[0010] In an optional embodiment of the present application, the initial battery thermal runaway prediction model is trained in the following manner: a training sample set is obtained, the training sample set includes multiple training samples, each training sample includes sample operation data, and an error range of each sample operation data within the preset time period; the sample operation data is used as the input of the initial battery thermal runaway prediction model, and the error range of each sample operation data within the preset time period is used as the output of the initial battery thermal runaway prediction model to train the initial battery thermal runaway prediction model.

[0011] In a second aspect, an embodiment of the present application also provides a vehicle battery thermal runaway warning device, the device comprising: an operating data acquisition module, for acquiring an operating data set of a battery to be tested within a preset time period; an error range prediction module, for inputting the operating data set into a battery thermal runaway prediction model to predict the error range of each data sample in the operating data set within the preset time period, wherein the battery thermal runaway prediction model is used to characterize the relationship between the error range and the data sample; a thermal runaway warning module, for calculating the difference between each data sample and the corresponding target data, and triggering a thermal runaway warning when any difference is not within the corresponding error range. In a third aspect, an embodiment of the present application further provides an electronic device comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the method described above are performed.

[0012] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described above are executed.

[0013] The embodiment of the present application provides a vehicle battery thermal runaway warning and thermal runaway device, wherein the method includes: obtaining an operating data set of the battery to be tested within a preset time period; inputting the operating data set into a battery thermal runaway prediction model to predict the error range of each data sample in the operating data set within the preset time period, wherein the battery thermal runaway prediction model is used to characterize the relationship between the error range and the data sample; calculating the difference between each data sample and the corresponding target data, and triggering a thermal runaway warning when any difference is not within the corresponding error range. Through this application, accurate monitoring of the operating status of the vehicle battery within a preset time period and thermal runaway risk warning can be achieved, effectively solving the thermal runaway risks caused by abnormal operating conditions during the charging and discharging process of the lithium-ion battery, and ensuring the safety of battery operation.

[0014] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 A flow chart of a vehicle battery thermal runaway warning method provided in an embodiment of the present application; Figure 2 A flowchart for obtaining an operating data set provided in an embodiment of the present application; Figure 3 A schematic diagram of a battery thermal runaway test platform test provided in an embodiment of the present application; Figure 4 A schematic diagram of the structure of a battery thermal runaway prediction model provided in an embodiment of the present application; Figure 5 A schematic diagram of the structure of a vehicle battery thermal runaway warning device provided in an embodiment of the present application; Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for which protection is claimed, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, each other embodiment obtained by those skilled in the art without making creative work falls within the scope of protection of the present application.

[0018] First, the application scenarios to which this application is applicable are introduced. This application can be applied in the field of vehicle technology.

[0019] The core component of new energy vehicles is the power battery, the heart of all mobile devices. It serves as the fundamental unit for energy storage and conversion. Its technological advancement will become a key supporter of the global automotive industry's electrification transformation. According to data released by SNE Research, China's power battery production capacity will exceed 1TWh by 2024, entering the terawatt-hour (1TWh = 1000GWh) era. Currently, demand in the international power battery market is strong.

[0020] High-energy-density batteries can provide electric vehicles with longer driving range, solve mileage anxiety, and extend the battery life of electronic devices. Traditional lithium-ion batteries are limited by the energy density bottleneck of liquid electrolytes (150-260wh / kg) and safety hazards, while the energy density of solid-state batteries can easily reach 300-500Wh / kg or even higher. In addition, in theory, solid-state electrolytes are non-flammable and have no leakage risk. Unlike ordinary batteries, they will not cause safety problems due to electrolyte combustion and leakage, greatly reducing the probability of thermal runaway. Solid-state batteries have demonstrated unparalleled advantages and development potential in promoting the popularization of new energy vehicles and for individual consumers seeking more reliable and long-lasting electronic products. It is not only a technological advancement, but also a bridge to a greener and more efficient future, heralding the advent of a new energy era.

[0021] Research has found that with the development of electric vehicles, their shortcomings, such as high prices, short range, limited lifespan, and poor safety and reliability, have become increasingly prominent. Demand for fast charging and long lifespans for lithium-ion batteries is increasing, but fast charging can accelerate battery aging, particularly by forming lithium metal plating on the negative electrode, reducing battery capacity and performance. The optimal operating temperature for lithium-ion power batteries is 20-40°C. Temperatures too high or too low can affect performance and pose safety risks. The heat generated by charging and discharging causes the temperature to rise, and reaching the ignition point can trigger thermal runaway, a chain reaction of heat release from battery cells that causes the temperature to rise uncontrollably, ultimately leading to explosion and fire.

[0022] In addition, abnormal operating conditions such as overcharging, over-discharging or use in high-temperature environments will cause the electrode material to react with the electrolyte to produce gas, and the battery shell will expand. If it cannot break through the critical volume state, it may explode and catch fire, causing serious safety problems.

[0023] Based on this, the vehicle battery thermal runaway warning method provided in the embodiment of the present application obtains the operating data set of the battery to be tested within a preset time period and inputs it into the battery thermal runaway prediction model, predicts the error range of each data sample, and then calculates the difference between the data sample and the target data. When the difference exceeds the error range, the thermal runaway warning is triggered, thereby realizing accurate monitoring of the vehicle battery operating status and thermal runaway risk warning, effectively solving the thermal runaway risks caused by abnormal operating conditions during the charging and discharging process of the lithium-ion battery, and ensuring the safe operation of the battery.

[0024] See also Figure 1 , Figure 1 This is a flow chart of the vehicle battery thermal runaway warning method provided in the embodiment of the present application. Figure 1 As shown in , the vehicle battery thermal runaway warning method provided by the embodiment of the present application includes: S101: Acquire an operating data set of a battery to be tested within a preset time period.

[0025] In an optional embodiment, the data types in the operating data set include but are not limited to battery characteristic parameters such as lithium battery surface temperature, output voltage, and battery expansion pressure signal.

[0026] This application places the battery under test on a test platform and conducts tests using normal cycling, increased temperature, increased voltage, or increased current to induce overcharge-induced thermal runaway. During the test, data samples such as the lithium battery surface temperature, output voltage, and battery expansion pressure signal are collected as an operational dataset. This dataset describes the continuous process of the battery from normal cycling to overcharge and then to thermal runaway in a time series. By measuring the battery expansion pressure signal, the internal volume changes of the battery are effectively mapped, and the continuity can be determined based on the timestamp of the dataset.

[0027] See also Figure 2 , Figure 2This is a flowchart of obtaining a running data set provided in the embodiment of the present application. Figure 2 As shown in , the running dataset is obtained by: S201 : When it is detected that a battery to be tested is placed on a test platform, the battery is controlled to be powered on.

[0028] See also Figure 3 , Figure 3 Schematic diagram of the battery thermal runaway test platform test provided in an embodiment of the present application.

[0029] like Figure 3 As shown in the figure, the battery under test is fixed in the battery test chamber using a fixture and mica board. The battery's status can be observed through an observation window to ensure the battery's stability and safety during the experiment. The mica board also provides a certain degree of insulation and heat insulation. The exhaust channel is used to promptly discharge the gas generated inside the battery in the event of abnormal conditions such as thermal runaway. This prevents excessive pressure in the chamber from causing serious consequences such as explosion, thus ensuring the safety of the experimental environment.

[0030] A paperless data acquisition instrument (multi-channel data acquisition instrument) is used to collect operational data sets, including but not limited to pressure, temperature, voltage, capacity, and current. This instrument features high-precision, multi-channel simultaneous data acquisition capabilities, enabling real-time and accurate acquisition of key battery parameters during the experiment. The instrument sends the collected operational data sets to a processor (connected to a computer for data analysis) for analysis using professional data analysis software. Through in-depth mining and analysis of this data, battery performance can be assessed and thermal runaway risk can be predicted, providing important data support and technical basis for battery research, development, and safe use.

[0031] In this battery experiment setup, the battery to be tested is fixed in the battery experiment box by a fixture and a mica plate. The exact position of the battery in the experiment box is determined according to the shape, size and experimental requirements of the battery. The design of the fixture will match the shape of the battery to ensure that the battery can be accurately placed in the predetermined position. In addition, the fixture usually has adjustable clamping devices such as bolts, nuts, and clips. By rotating the bolts or operating the clips, the fixture applies appropriate pressure to the battery, firmly fixing the battery to prevent the battery from moving or shaking during the experiment. For example, for square batteries, the fixture may clamp from both sides of the battery; for cylindrical batteries, the fixture may use a ring structure to hold and clamp it.

[0032] Mica sheets have excellent insulating properties. Placing them between the battery and the fixture, or between the battery and the experimental chamber, can prevent electrical short circuits between the battery and other components. For example, placing a mica sheet where the battery's electrodes meet the fixture prevents the fixture from conducting electricity and potentially affecting the battery's normal operation or causing safety issues. Mica sheets also have a certain degree of flexibility and high-temperature resistance. During the battery fixing process, they act as a buffer, reducing pressure damage from the fixture on the battery surface. Furthermore, if the battery overheats during the experiment, the mica sheet can withstand the temperature and protect surrounding components from the high temperature.

[0033] The shape and size of the mica sheet can be customized to suit the structure of the battery and fixture. This helps further define the battery's position and ensure its stability within the experimental chamber. For example, the mica sheet can be designed to fit snugly around the bottom or side of the battery, nestling between the fixture and the battery for greater stability.

[0034] In an optional embodiment, the test platform includes a controllable temperature environment chamber, a multi-channel data acquisition instrument, and an overcharge control module. The controllable temperature environment chamber is used to adjust the temperature of the battery under test in the test platform. The multi-channel data acquisition instrument is used to collect an operating data set of the battery under test during the test and send the operating data set to the processor. The overcharge control module is used to trigger battery thermal runaway by adjusting the charging current or voltage of the battery under test. Here, upon detecting the presence of a battery under test in the test platform, the battery is powered on. The test platform comprises a temperature-controlled environmental chamber, a multi-channel data acquisition instrument, and an overcharge control module. The temperature-controlled environmental chamber regulates the battery temperature, the multi-channel data acquisition instrument collects battery operating data and transmits it to the processor, and the overcharge control module triggers thermal runaway.

[0035] S202: Perform testing according to the predicted test items.

[0036] Predictive test items include temperature test, current test and voltage test; Tests are performed based on predicted test items, including temperature tests, current tests, and voltage tests, to comprehensively evaluate the battery's performance under different operating conditions.

[0037] S203 , determining the data received in real time from the multi-channel data acquisition instrument as an operating data set of the battery to be tested.

[0038] The data received in real time from the multi-channel data acquisition instrument is determined as the operating data set of the battery under test. Each data sample in this operating data set includes a timestamp and the surface temperature, output voltage and battery expansion pressure signal of the battery under test detected at the timestamp. The operating data set describes the continuous process of the battery from normal cycle to overcharge to thermal runaway in a time series.

[0039] The accuracy and completeness of the operating data set are ensured, providing a data basis for subsequent thermal runaway predictions. Through the controllable temperature environment chamber and overcharge control module, various operating conditions that the battery may encounter in actual use can be simulated, improving the comprehensiveness and accuracy of the test.

[0040] S102: Input the operating data set into a battery thermal runaway prediction model to predict the error range of each data sample in the operating data set within a preset time period.

[0041] Among them, the battery thermal runaway prediction model is used to characterize the relationship between the error range and the data sample.

[0042] Among them, the battery thermal runaway prediction model is built based on the attention mechanism-residual structure-long short-term memory network (SE-Res-LSTM), which is used to characterize the relationship between the error range and data samples.

[0043] Based on the dataset, this application designs a battery expansion pressure regression prediction algorithm based on the attention mechanism-residual structure-long short-term memory network (SE-Res-LSTM), and uses the prediction error to perform real-time detection of battery overcharge thermal runaway.

[0044] By learning the mapping relationship between errors and operating status in historical data, the model can dynamically predict the error range of each data sample within a preset time period, achieve accurate prediction of battery operating status, and improve the timeliness and accuracy of thermal runaway warning. By introducing the attention mechanism and residual structure, the model can better capture the key features in battery operating data and improve prediction accuracy.

[0045] S103: Calculate the difference between each data sample and the corresponding target data. When any difference is not within the corresponding error range, trigger a thermal runaway warning.

[0046] In this step, the difference between the current data sample and the corresponding target data (such as data under normal circulation state) is calculated in real time, and the difference is compared with the error range predicted by the model. If the difference exceeds the error range, it is determined to be a thermal runaway risk. If the difference does not exceed the error range, it is determined that there is no thermal runaway risk and the test continues.

[0047] When the risk of thermal runaway is detected, a thermal runaway warning is immediately triggered so that appropriate emergency response measures can be taken.

[0048] This application uses statistical analysis to conduct an in-depth analysis of the processed experimental data to determine the effectiveness of the above method, and further determine the dynamic characteristics of the battery chemical changes during the charging and discharging process. At the same time, the effectiveness of the above method is repeatedly verified. If the final result meets the set prediction threshold, it is determined to be applied. If improvement is required, the above steps are repeated.

[0049] The effectiveness of the SE-Res-LSTM model in the battery expansion pressure prediction task was verified through experiments, and the model prediction effect was evaluated using indicators such as mean absolute percentage error (MAPE), root mean square error (RMSE) and coefficient of determination (R2).

[0050] In a first optional embodiment, the mean absolute percentage error (MAPE) is evaluated as follows. First, for each sample data point i (there are n samples in total), the difference between the true value yi and the predicted value y^i is calculated, and then this difference is divided by the true value yi to obtain the absolute percentage error of the predicted value relative to the true value of each sample point.

[0051] Next, the absolute percentage errors of all sample points are summed.

[0052] Finally, divide the sum by the number of samples, n, and multiply by 100% to get the mean absolute percentage error (MAPE). A smaller MAPE indicates a smaller relative error between the model's predicted value and the true value, and a higher prediction accuracy.

[0053] In a second optional embodiment, the root mean square error (RMSE) is evaluated as follows: for each sample data point i (a total of n samples), the difference between the true value yi and the predicted value y^i is first calculated, and the difference is squared.

[0054] Then, sum the squared differences of all sample points.

[0055] Next, divide the sum by the number of samples n to get the mean squared difference.

[0056] Finally, we take the square root of the mean squared error to get the root mean square error (RMSE). RMSE reflects the average error between the model's predicted value and the true value. The smaller the RMSE, the closer the model's predicted value is to the true value, and the better the prediction effect.

[0057] In a third optional embodiment, the coefficient of determination (R2) is evaluated as follows: in the numerator, the square of the difference between the true value yi and the predicted value y^i of each sample is calculated, and the squared differences of all samples are summed to obtain the sum of the squares of the errors between the predicted values ​​and the true values.

[0058] Denominator: Calculate the square of the difference between the true value yi of each sample and the mean yˉ of the true value, and sum the squared differences of all samples to obtain the total sum of squared deviations.

[0059] The coefficient of determination (R²) is calculated by subtracting the sum of squared errors in the numerator from 1 and dividing it by the sum of squared deviations in the denominator. R² ranges from 0 to 1. The closer it is to 1, the better the model fits the data, meaning the model explains a higher proportion of the variation in the dependent variable.

[0060] This application realizes real-time monitoring and early warning of battery thermal runaway risks, effectively reducing the probability of thermal runaway accidents. Through the dynamic prediction error range, it can adapt to the changes in the operating status of the battery under different working conditions, thereby improving the flexibility and reliability of the early warning.

[0061] Specifically, this application presents a lithium-ion battery thermal runaway early warning method and technology based on prediction error, which can effectively improve the thermal runaway safety performance of battery cells and fundamentally resolve the potential safety hazards of thermal runaway. By using statistical analysis to deeply explore the processed data, the specific impact of different charging strategies on battery aging is determined, revealing the dynamic characteristics of the chemical changes within the battery during the charging process. This improves the safety performance of power batteries, adapts to the rapid development of electric vehicle power battery systems, and greatly facilitates the rapid development of electric vehicle technology. It also greatly improves product consistency, reduces product testing steps in the battery production process, simplifies some manufacturing processes, and effectively improves production efficiency.

[0062] This application implements an optimized SE-Res-LSTM algorithm model through an attention mechanism (squeeze-and-excitation network, SE) and a residual network (ResNet). This enables the model to adaptively adjust the weights between feature channels, improving the model's focus on and representation of important features, helping the model capture the relationship between battery surface temperature, voltage, and expansion pressure.

[0063] Residual connections, by introducing an identity mapping, help the model learn linear relationships and deeper feature representations within the original information. Furthermore, LSTM, a neural network structure suitable for processing sequential data, can capture dependencies between features and effectively perform regression predictions on inflation pressure signals.

[0064] The specific process is to preprocess the multi-dimensional original features of the battery and flatten the data to form a feature matrix image, which contains all the feature information. A single sample is a component image of the image. The local features of the image are extracted through a layer of 1-D convolutional network, and a three-dimensional feature matrix image with multiple channels is formed after sliding convolution. Each element in the feature map image of different channels corresponds to the feature of a sub-interval in the original signal, which makes the images correlated. This correlation represents a common feature worthy of attention. By embedding the SE-Net module, each input feature map is compressed into a descriptor using average pooling to obtain a one-dimensional vector, thereby obtaining a global receptive field. In order to focus on the correlation between different channels, it is necessary to integrate the channel information through two fully connected layers to obtain a more comprehensive channel correlation information.

[0065] See also Figure 4 , Figure 4 This is a schematic diagram of the structure of the battery thermal runaway prediction model provided in an embodiment of the present application.

[0066] like Figure 4 As shown in the figure, the battery thermal runaway prediction model of the present application includes an input layer (Input Layer, corresponding to "I"), a one-dimensional convolutional feature extraction layer (CONV1, CONV2), an attention mechanism module (SEnet), a residual structure module (not shown in the figure), a long short-term memory network layer (LSTM) and an output layer (FC3).

[0067] The input layer is used to receive the pre-processed operating data set (battery characteristics) collected within a preset time period to form a feature matrix; The input layer ensures the standardization and consistency of the model input data, providing a basis for subsequent feature extraction and prediction.

[0068] The one-dimensional convolutional feature extraction layer (CONV1, CONV2) is used to perform sliding convolution operations on the feature matrix to generate a three-dimensional feature matrix image with multiple channels; The one-dimensional convolutional feature extraction layer extracts local features of the data through convolution operations, enhancing the model's ability to perceive the spatial structure of the data.

[0069] The attention mechanism module (SEnet) is used to compress the 3D feature matrix image through global average pooling, two fully connected layers, and an activation function, determine the channel weight vector, and recalibrate the 3D feature matrix image according to the channel weights; Here, the attention mechanism module is configured as: Perform global average pooling on the input three-dimensional feature matrix image to generate a channel descriptor vector; Channel features are integrated through two-level fully connected layers. The first-level fully connected layer scales the number of channels, and the second-level fully connected layer generates the channel weight vector. The activation function is used to generate a channel weight vector and the input feature tensor is recalibrated according to the channel weight.

[0070] Attention mechanism module Through the attention mechanism, the model can adaptively adjust the weights between feature channels, improve the attention and representation ability of important features, and help capture key features in battery operation data.

[0071] The residual structure module is used to perform a one-dimensional convolution operation on the recalibrated three-dimensional feature matrix image and generate a residual output by cross-channel adjustment and feature addition; The residual structure module is configured as follows: Perform a one-dimensional convolution operation on the recalibrated three-dimensional feature matrix image and output the first feature map; Adjust the number of channels and spatial size of the three-dimensional feature matrix image through a preset number of convolution kernels, and output a second feature map; The first feature map is added to the second feature map element-wise to generate the residual output.

[0072] The residual structure introduces an identity mapping, which helps the model learn linear relationships and deeper feature representations in the original information, thereby improving the training stability and prediction accuracy of the model.

[0073] The long short-term memory network layer (LSTM) is used to extract temporal features from the residual output and output the hidden state vector; The LSTM network (Long Short-Term Memory) can capture the dependencies between features and effectively perform regression predictions on time series data such as expansion pressure signals, thereby improving the model's ability to process time series data.

[0074] The output layer (FC3) is used to map the hidden state vector to the error range of each running data within a preset time period.

[0075] The output layer enables the model to predict the error range of the battery operating status, providing a quantitative basis for thermal runaway warning.

[0076] The other components are as follows: The Rectified Linear Unit (RELU) linear rectifier function introduces nonlinear characteristics, enabling the model to learn more complex patterns, enhancing the model's expressiveness and accelerating the model's training convergence.

[0077] GAP global average pooling performs spatial average pooling on feature maps to generate channel descriptor vectors, reducing the number of parameters and model complexity while retaining important channel information.

[0078] FC1 and FC2, fully connected layers, perform linear transformation on the input features to achieve further integration and transformation of features, adjust the dimension and distribution of features, and prepare for subsequent activation function operations and feature recalibration.

[0079] × (multiplication symbol): Located at the intersection of branches above the LSTM layer.

[0080] Meaning: Indicates element-wise multiplication.

[0081] Function: Weight the channel weight vector output by the attention mechanism module (SEnet) and the three-dimensional feature matrix output by the residual structure module channel by channel.

[0082] Specific operation: Multiply each channel of the feature matrix by the corresponding weight coefficient to achieve recalibration of the feature channel (i.e., enhance important channels and suppress irrelevant channels).

[0083] + (addition symbol): Located inside the residual structure module.

[0084] Meaning: Represents element-by-element addition in residual connection.

[0085] Function: Add the feature map after the convolution operation in the residual module (the first feature map) and the identity map (or the adjusted original feature map, that is, the second feature map) element by element.

[0086] Specific operation: Residual output = first feature map + second feature map, which is used to alleviate the gradient disappearance problem and promote information flow.

[0087] The battery thermal runaway prediction model of the present application introduces the attention mechanism and residual structure, and the model can more accurately capture the key features in the battery operation data, thereby improving the prediction accuracy. The residual structure introduces the identity mapping, which helps the model learn the linear relationship and deeper feature representation in the original information, thereby improving the training stability and generalization ability of the model. Moreover, the model can predict the error range of the battery operating status in real time, providing a quantitative basis for thermal runaway warning, and realizing real-time monitoring and warning of battery thermal runaway risks.

[0088] The battery expansion pressure regression prediction algorithm based on the attention mechanism-residual structure-long short-term memory network (SE-Res-LSTM) in this application uses prediction errors to perform real-time detection of battery overcharge thermal runaway, improves the accuracy and efficiency of the battery cell material design process, improves the safety and pass rate of power battery manufacturing, greatly improves product quality, and can directly guide the design improvement of electric vehicles and power batteries from a mechanistic perspective, which helps to improve the life and safety of electric vehicles and batteries.

[0089] When embedding the SE architecture, the model does not directly connect it to the second 1-D convolutional layer, but rather to the first layer. After global pooling, two fully connected layers scale the number of channels to match the output of the second convolutional layer. This is expected to better learn the correlations between channels.

[0090] Because convolution increases the number of feature channels and reduces the spatial size, the image must pass through a convolutional layer with a kernel size of 1 and 64 kernels to achieve cross-channel information exchange. Furthermore, padding is performed on the image to ensure that the spatial size of the image is consistent for addition. After addition, the sum image in the original feature image is passed to the LSTM layer for training to learn the relationship between features. The hidden state vector output from the final step is input into the fully connected layer and the regression layer to obtain the output. Table 1 below shows the detailed parameters of this SE-Res-LSTM.

[0091] Table 1

[0092] The process of the above-mentioned lithium-ion battery thermal runaway detection method can be briefly described as follows: the battery global data set is divided into a battery normal cycle data set and a battery thermal runaway data set based on the battery voltage of 3.65 V; the normal cycle data set is used to train the SE-Res-LSTM; the trained model is used to predict the global data set to obtain a prediction error data set; and the current state of the battery is judged to be overcharged and thermal runaway based on the trained prediction model.

[0093] In an optional embodiment, the initial battery thermal runaway prediction model is trained in the following manner: Get the training sample set.

[0094] Here, the training sample set can be obtained from historical battery operation data, experimental test data, or simulation data. These data should cover the operation of the battery under different operating conditions (such as temperature, current, voltage, etc.).

[0095] The training sample set includes multiple training samples, each of which includes sample operating data and an error range for each sample operating data within a preset time period. Each training sample includes two pieces of information: the sample operating data and the corresponding error range. The sample operating data can be time-series data such as battery voltage, current, and temperature within a preset time period. The error range is determined based on historical data or expert experience and is used to measure the accuracy of the model's prediction results.

[0096] The sample operation data is used as the input of the initial battery thermal runaway prediction model, and the error range of each sample operation data within a preset time period is used as the output of the initial battery thermal runaway prediction model to train the initial battery thermal runaway prediction model.

[0097] In this step, the sample operating data is used as input to the initial battery thermal runaway prediction model. Before entering the model, this data may need to be preprocessed, such as normalization and filtering, to improve the model training effect.

[0098] The error range of each sample running data within a preset time period is used as the model output. The goal of the model is to learn the mapping relationship between input data and output error range so that the error range of new data can be accurately estimated in subsequent predictions.

[0099] The initial battery thermal runaway prediction model is trained using a training sample set. During training, the model continuously adjusts its internal parameters (such as weights and biases) to minimize the difference between the predicted error range and the actual error range. Optimization algorithms such as gradient descent, stochastic gradient descent, and Adam can be used to accelerate the model training process and improve training results.

[0100] This application trains an initial battery thermal runaway prediction model by obtaining a training sample set containing sample operating data and corresponding error ranges, using the sample operating data as model input and the error ranges as model output. This process aims to enable the model to learn the mapping relationship between input data and output error ranges, so that the error ranges of new data can be accurately estimated in subsequent predictions, thereby achieving accurate early warning of battery thermal runaway risks.

[0101] Through training, the model can learn the complex relationship between battery operating data and error range, thereby improving the accuracy of predicting battery thermal runaway risks. Using a diverse set of training samples for training can make the model have better generalization capabilities and be able to adapt to battery operating data under different working conditions. By selecting appropriate optimization algorithms and adjusting model parameters, the performance of the model can be further optimized, and the training speed and prediction efficiency can be improved. The trained model can be applied to the actual battery thermal runaway early warning system, providing reliable technical support for the safe operation of electric vehicles.

[0102] The vehicle battery thermal runaway warning method provided in the embodiment of the present application obtains an operating data set of the battery to be tested within a preset time period and inputs it into a battery thermal runaway prediction model, predicts the error range of each data sample, calculates the difference between the data sample and the target data, and triggers a thermal runaway warning when the difference exceeds the error range, thereby achieving accurate monitoring of the vehicle battery operating status and thermal runaway risk warning, effectively solving the thermal runaway risks caused by abnormal operating conditions during the charging and discharging process of the lithium-ion battery, and ensuring the safe operation of the battery.

[0103] The above-mentioned method of the present application can shorten the cycle from product conception to production, reduce errors and lower costs, and at the same time form an empirical analysis model that can be applied to the design and optimization of other related manufacturing fields.

[0104] In Example 1, this application compared the SE-Res-LSTM model with a related research application model to test its regression prediction performance. Furthermore, to verify the effectiveness of the improvements made to the LSTM in this work, the SE and Res modules were removed to better understand the impact of each module on model performance. All models were trained using the same training set, and to provide an intuitive understanding of the prediction performance, a normal cycle dataset was used as the test set.

[0105] In Example 2, this application evaluates the role of these two modules in predicting battery expansion pressure by removing the residual structure and channel attention mechanism respectively. It can be seen that SE-LSTM and Res-LSTM are superior to LSTM and CNN-LSTM, which are widely used in related studies, in terms of overall prediction effect. SE-LSTM, because it can pay attention to the connection between different channels, strengthens the model's ability to capture nonlinear changes to a certain extent, and can achieve better prediction results at the inflection point of battery expansion pressure change and the pressure change stage when charging is stopped. However, the prediction performance during the battery static process is not as stable as SE-Res-LSTM. On the contrary, although the residual structure can enhance the model's capture of linear changes, the expansion pressure changes caused by battery charging and discharging involve multiple factors, and the prediction effect of Res-LSTM in the pressure change stage is not as good as SE-LSTM.

[0106] Table 2

[0107] Table 2 compares the evaluation metrics of the prediction results from the comparative experiments. It can be seen that the model proposed in this work achieves the highest prediction accuracy. The CNN-LSTM model improves prediction accuracy by 0.1394% compared to the LSTM, demonstrating that the CNN is capable of extracting high-level features from battery operating data. The introduction of the residual structure improves the model's prediction accuracy by 0.2521%, demonstrating that this structure helps the model handle both high-level features and nonlinear relationships between features. However, compared to the SE-LSTM, which retains the channel attention mechanism, its prediction performance still differs by 0.1868%, indicating that the nonlinear relationships between battery features are richer, necessitating the introduction of the channel attention mechanism. The SE-Res-LSTM, which incorporates residual connections and the channel attention mechanism, achieves the highest prediction accuracy.

[0108] In Example 3, after training on a dataset of normal battery cycles, the SE-Res-LSTM model demonstrated excellent performance in predicting battery expansion pressure signals under normal cycling conditions. Next, the trained network will be fed a dataset of battery overcharge and thermal runaway events to analyze its predictive capabilities in overcharge and thermal runaway scenarios, preliminarily verifying the feasibility of the prediction error threshold determination method.

[0109] Experiments have shown that within a short period of time after a lithium-ion battery is overcharged, the SE-Res-LSTM model, trained on normal battery cycle data, performs poorly at predicting the expansion pressure of lithium-ion batteries. When overcharge occurs, the SE-Res-LSTM's predicted battery expansion pressure decreases rapidly, while the true value increases rapidly. These combined effects rapidly increase the prediction error during the overcharge thermal runaway process. This change supports the idea of ​​using a residual monitor to detect overcharge thermal runaway.

[0110] This application has built a lithium-ion battery thermal runaway test platform, and adopted normal cycle and overcharge-induced thermal runaway methods to collect data sets such as lithium battery surface temperature, output voltage and battery expansion pressure signal. This data set describes the continuous process of the battery from normal cycle to overcharge to thermal runaway in time series, and effectively maps the internal volume change of the battery by measuring the battery expansion pressure signal. Further, based on the data set, a battery expansion pressure regression prediction algorithm based on the attention mechanism-residual structure-long short-term memory network (SE-Res-LSTM) is designed, and the prediction error is used to perform real-time detection of battery overcharge thermal runaway. This application not only provides a feasible solution for thermal runaway safety warning, but also helps to optimize the battery management system of future electric vehicles, contributes to battery utilization efficiency and durability, and significantly improves the timeliness and accuracy of warnings.

[0111] Based on the same inventive concept, the embodiment of the present application also provides a vehicle battery thermal runaway warning device corresponding to the vehicle battery thermal runaway warning method. Since the principle of solving the problem by the device in the embodiment of the present application is similar to the above-mentioned vehicle battery thermal runaway warning method in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0112] See also Figure 5 , Figure 5 This is a schematic diagram of the structure of the vehicle battery thermal runaway warning device provided in the embodiment of the present application. Figure 5 As shown in , the vehicle battery thermal runaway warning device 500 includes: An operating data acquisition module 501 is used to acquire an operating data set of the battery under test within a preset time period; an error range prediction module 502, configured to input the operating data set into a battery thermal runaway prediction model to predict an error range for each data sample in the operating data set within the preset time period, wherein the battery thermal runaway prediction model is configured to characterize a relationship between the error range and the data sample; The thermal runaway warning module 503 is used to calculate the difference between each data sample and the corresponding target data, and trigger a thermal runaway warning when any difference is not within the corresponding error range.

[0113] See also Figure 6 , Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present application. Figure 6 As shown in FIG, the electronic device 300 includes a processor 310 , a memory 320 and a bus 330 .

[0114] The memory 320 stores machine-readable instructions executable by the processor 310. When the electronic device 300 is running, the processor 310 communicates with the memory 320 via the bus 330. When the machine-readable instructions are executed by the processor 310, the above-mentioned Figure 1 The steps of the vehicle battery thermal runaway warning method in the illustrated method embodiment and the specific implementation thereof can be found in the method embodiment and will not be described in detail here.

[0115] The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the computer program can execute the above-mentioned Figure 1 The steps of the vehicle battery thermal runaway warning method in the illustrated method embodiment and the specific implementation thereof can be found in the method embodiment and will not be described in detail here.

[0116] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0117] In the several embodiments provided in this application, it should be understood that the disclosed systems, 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 merely a logical function division. There may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some communication interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0118] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0119] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0120] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0121] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The scope of protection of the present application is not limited thereto. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-mentioned embodiments within the technical scope disclosed in the present application, or perform equivalent replacements for some of the technical features thereof. These modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A vehicle battery thermal runaway early warning method, characterized in that: include: Obtaining an operating data set of the battery to be tested within a preset time period; Inputting the operating data set into a battery thermal runaway prediction model to predict an error range for each data sample in the operating data set within the preset time period, wherein the battery thermal runaway prediction model is used to characterize the relationship between the error range and the data sample; The difference between each data sample and the corresponding target data is calculated. When any difference is not within the corresponding error range, a thermal runaway warning is triggered.

2. The method according to claim 1, characterized in that The running dataset is obtained by: When a battery under test is detected to be placed in a test platform, the battery is controlled to be powered on. The test platform includes a temperature-controlled environmental chamber, a multi-channel data acquisition instrument, and an overcharge control module. The temperature-controlled environmental chamber is used to adjust the temperature of the battery under test in the test platform. The multi-channel data acquisition instrument is used to collect an operating data set of the battery under test during the test and send the operating data set to a processor. The overcharge control module is used to trigger battery thermal runaway by adjusting the charging current or voltage of the battery under test. Determine test items, the test items including temperature test, current test and voltage test; The test is started according to the test items, and the data received in real time from the multi-channel data acquisition instrument is determined as the operating data set of the battery to be tested.

3. The method according to claim 1, characterized in that The battery thermal runaway prediction model includes an input layer, a one-dimensional convolutional feature extraction layer, an attention mechanism module, a residual structure module, a long short-term memory network layer, and an output layer. The input layer is used to receive the pre-processed running data set collected within a preset time period to form a feature matrix; The one-dimensional convolution feature extraction layer is used to perform a sliding convolution operation on the feature matrix to generate a three-dimensional feature matrix image containing multiple channels; The attention mechanism module is used to compress the three-dimensional feature matrix image through global average pooling, two fully connected layers and an activation function, determine a channel weight vector, and recalibrate the three-dimensional feature matrix image according to the channel weight; The residual structure module is used to perform a one-dimensional convolution operation on the recalibrated three-dimensional feature matrix image, and generate a residual output by cross-channel adjustment and feature addition; The long short-term memory network layer is used to extract time series features from the residual output and output a hidden state vector; The output layer is used to map the hidden state vector to the error range of each operating data within the preset time period.

4. The method according to claim 3, characterized in that The residual structure module is configured as follows: Performing a one-dimensional convolution operation on the recalibrated three-dimensional feature matrix image to output a first feature map; Adjusting the number of channels and spatial size of the three-dimensional feature matrix image through a preset number of convolution kernels, and outputting a second feature map; The first feature map is added to the second feature map element-wise to generate the residual output.

5. The method according to claim 3, characterized in that The attention mechanism module is configured as follows: Perform global average pooling on the input three-dimensional feature matrix image to generate a channel descriptor vector; Channel features are integrated through two-level fully connected layers. The first-level fully connected layer scales the number of channels, and the second-level fully connected layer generates the channel weight vector. The activation function is used to generate a channel weight vector and the input feature tensor is recalibrated according to the channel weight.

6. The method according to claim 1, characterized in that Each data sample in the operating data set includes a timestamp and the surface temperature, output voltage, and battery expansion pressure signal of the battery to be tested detected at the timestamp. The operating data set describes the continuous process of the battery from normal cycle to overcharge to thermal runaway in a time series.

7. The method according to claim 1, characterized in that The initial battery thermal runaway prediction model is trained in the following way: Acquire a training sample set, wherein the training sample set includes a plurality of training samples, each training sample includes sample operation data and an error range of each sample operation data within the preset time period; The sample operating data is used as the input of the initial battery thermal runaway prediction model, and the error range of each sample operating data within the preset time period is used as the output of the initial battery thermal runaway prediction model to train the initial battery thermal runaway prediction model.

8. A vehicle battery thermal runaway warning device, characterized in that: include: An operating data acquisition module is used to obtain an operating data set of the battery to be tested within a preset time period; an error range prediction module, configured to input the operating data set into a battery thermal runaway prediction model to predict an error range for each data sample in the operating data set within the preset time period, wherein the battery thermal runaway prediction model is configured to characterize a relationship between the error range and the data sample; The thermal runaway warning module is used to calculate the difference between each data sample and the corresponding target data. When any difference is not within the corresponding error range, a thermal runaway warning is triggered.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and the processor executes the machine-readable instructions to perform the steps of any one of the methods described in claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are executed.

Images