Early warning method and device for thermal runaway of lithium ion storage battery of electric bicycle

Abstract

The invention discloses an electric bicycle lithium ion storage battery thermal runaway early warning method and device, and belongs to the technical field of new energy batteries, and the method comprises the steps: determining key parameters according to the contribution degree of each parameter in battery operation data to the influence of the battery surface temperature change; constructing a battery surface temperature prediction model by taking the key parameters as input and the battery surface temperature change trend in the future time period as output, and predicting the battery surface temperature change trend in the future time period according to the key parameters of the target battery in the current preset time period; and performing thermal runaway early warning on the target battery according to the battery surface temperature change trend, the key parameter at the current moment and a preset key parameter risk threshold. According to the method, the temperature change trend of the surface of the battery is predicted through key parameters, then thermal runaway early warning evaluation is performed on the battery, and early warning can be accurately performed on the thermal runaway of the battery in a future time period through a small number of battery parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy batteries, and in particular to a method and device for predicting thermal runaway of lithium-ion batteries for electric bicycles. Background Art

[0002] Lithium-ion batteries are widely used in the energy storage field of electric bicycles. When they are externally abused, they may cause thermal runaway accidents. Battery thermal runaway is one of the main problems of battery safety. Thermal runaway may lead to fires and explosions, threatening people and property and causing serious safety hazards. Due to the compact internal space of the electric bicycle battery pack, once a thermal runaway accident occurs, the battery will generate a large amount of heat and release a large amount of toxic and harmful gases, seriously endangering the life and property safety of users.

[0003] In the prior art, thermal runaway detection technology is used to monitor the surface temperature and internal state of the battery in real time to predict thermal runaway. However, when thermal runaway is detected in this way, thermal runaway may be about to occur, and it is impossible to predict thermal runaway in advance. In addition, big data and cloud storage technologies are used, combined with machine learning algorithms to analyze a large amount of battery data to predict the thermal runaway time of the battery. However, this process requires collecting a large amount of battery data. Considering the cost and space factors of electric bicycles, collecting a large amount of battery data is not applicable to the field of electric bicycles.

[0004] Therefore, there is an urgent need for a method for predicting thermal runaway of lithium-ion batteries for electric bicycles to solve the problem that in the prior art, it is impossible to accurately predict the thermal runaway of the battery in the future time period through a small amount of battery data for electric bicycles. Summary of the Invention

[0005] In view of this, it is necessary to provide a method and device for predicting thermal runaway of lithium-ion batteries for electric bicycles, which can accurately predict the thermal runaway of the battery in the future time period of the electric bicycle through a small amount of battery data.

[0006] To solve the above technical problems, on the one hand, the present invention provides a method for predicting thermal runaway of lithium-ion batteries for electric bicycles, including: Determine key parameters according to the contribution degree of each parameter in the operation data of the lithium-ion battery of the electric bicycle to the change of the battery surface temperature; Taking the key parameters as the input and the change trend of the battery surface temperature in the future time period as the output, construct a battery surface temperature prediction model, and predict the change trend of the battery surface temperature in the future time period according to the key parameters of the target battery in the current preset time period; Perform thermal runaway warning on the target battery according to the change trend of the battery surface temperature, the key parameters at the current moment, and the preset key parameter risk threshold.

[0007] In a possible implementation manner, key parameters are determined according to the contribution degrees of various parameters in the battery operation data to the change of the battery surface temperature, including: Obtain the initial operation data during the battery operation process through charge and discharge cycle tests and thermoelectric abuse thermal runaway tests; Preprocess the initial operation data to obtain the battery operation data; Calculate the contribution degrees of various parameters in the operation data to the change of the battery surface temperature, and screen out the parameters with contribution degrees higher than the preset threshold as key parameters.

[0008] In a possible implementation manner, preprocessing the initial operation data to obtain the battery operation data includes: Fill in the missing data in the initial operation data according to the time domain property of the initial operation data to obtain the first initial operation data; Detect outliers in the first initial operation data, and replace the abnormal data according to the detection results to obtain the second initial operation data; Standardize different variables in the second initial operation data to obtain the battery operation data.

[0009] In a possible implementation manner, calculating the contribution degrees of various parameters in the operation data to the change of the battery surface temperature, and screening out the parameters with contribution degrees higher than the preset threshold as key parameters includes: Based on the Granger causality test, determine the contribution degrees of various parameters in the operation data to the change of the battery surface temperature; Based on the Shapley value, sort each parameter according to the contribution degree, and screen out the parameters with contribution degrees higher than the preset threshold as key parameters.

[0010] In a possible implementation manner, the operation data includes battery surface temperature data. Determining the contribution degrees of various parameters in the operation data to the change of the battery surface temperature includes: Based on a preset first regression model, predict the battery surface temperature at the next moment according to the historical data of the battery surface temperature to obtain the first surface predicted temperature; Based on a preset second regression model, predict the battery surface temperature at the next moment according to the historical data of each parameter and the battery surface temperature to obtain multiple second surface predicted temperatures; Successively calculate the residual variance between each second surface predicted temperature and the first surface predicted temperature, and determine the contribution degrees of various parameters to the change of the battery surface temperature according to the residual variance.

[0011] In a possible implementation manner, the battery surface temperature prediction model includes a feature interaction layer, a feature fusion layer, and a temperature prediction layer; The feature interaction layer is used to determine the dynamic range of the receptive field through high-dimensional channel indexing, and perform a convolution operation on the time-series data of the key parameters through a sliding window according to the dynamic range of the receptive field to obtain the time-series features of the key parameters; The feature fusion layer is used to embed position information for the time-series data at each time step after segmentation, map the time-series features to high-dimensional features, and fuse the high-dimensional features; The temperature prediction layer performs non-linear regression prediction based on the position information of the time-series features and the fused high-dimensional features to predict the temperature change trend of the battery surface.

[0012] In a possible implementation manner, the feature fusion layer includes a triangular position embedding layer, an encoder, and a decoder; The triangular position embedding layer is used to embed position information for the time-series features; The encoder is used to map the time-series features with embedded position information to a high-dimensional data space to obtain high-dimensional features, calculate the internal relationships between the high-dimensional features, and fuse the high-dimensional features according to the internal relationships; The decoder is used to reconstruct the output of the encoder.

[0013] In a possible implementation manner, thermal runaway warning for the target battery is performed according to the battery surface temperature change trend, the key parameters at the current moment, and a preset key parameter risk threshold, including: When the battery surface temperature change trend and the key parameters meet the preset first-level warning, a fault alarm measure is taken; When the battery surface temperature change trend and the key parameters meet the preset second-level warning, a thermal runaway warning is taken.

[0014] In a possible implementation manner, the setting process of the preset key parameter risk threshold includes: Obtain the change data of each key parameter and the battery surface temperature during the experiment of the battery, and draw a change trend graph of the key parameter and the battery surface temperature; Determine the key points of the battery surface temperature change; Determine the risk threshold of each key parameter according to the key points.

[0015] In a second aspect, the present invention also provides a battery thermal runaway warning device, including: A key parameter determination module, configured to determine key parameters according to the contribution degree of each parameter in the battery operation data to the battery surface temperature change; A temperature prediction module, which is used to take the key parameters as inputs and the battery surface temperature change trend in a future time period as the output, construct a battery surface temperature prediction model, and predict the battery surface temperature change trend in the future time period according to the key parameters of the target battery in the current preset time period; An early warning module, which is used to conduct a thermal runaway early warning on the target battery according to the battery surface temperature change trend, the key parameters at the current moment, and a preset key parameter risk threshold.

[0016] The beneficial effects of the present invention are as follows: First, analyze the contribution degree of each parameter in the battery operation data to the change of the battery surface temperature, and determine multiple key parameters according to the contribution degree of each parameter; then, take the key parameters as inputs and the battery surface temperature change trend in a future time period as the output, construct a battery surface temperature prediction model based on deep learning, and predict the battery surface temperature change trend in the future time period according to the key parameters of the target battery in the current preset time period obtained in real time. At this time, only a small number of key parameters need to be concerned, and these key parameters have a significant contribution to the change of the battery surface temperature, so the temperature change trend of the battery surface can be intelligently predicted according to the model, and the temperature change of the battery surface can be predicted intelligently; finally, conduct a thermal runaway early warning on the target battery according to the battery surface temperature change trend and the key parameters at the current moment. By analyzing the contribution degree of each parameter during the battery operation to the change of the battery temperature, the present invention determines the key parameters, intelligently predicts the temperature change trend of the battery surface by obtaining the real-time key parameters, and conducts a thermal runaway early warning assessment on the battery according to the temperature change trend and the key parameters, and can accurately conduct an early warning on the battery thermal runaway in a future time period through a small amount of battery data. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is a schematic flowchart of an embodiment of the method for thermal runaway early warning of an electric bicycle lithium-ion battery provided by the present invention; Figure 2 For the present invention Figure 1 It is a schematic flowchart of an embodiment of step S101 in the present invention; Figure 3 For the present invention Figure 2 It is a schematic flowchart of an embodiment of step S202 in the present invention; Figure 4 For the present invention Figure 2Flow diagram of an embodiment of step S203; Figure 5 This invention Figure 4 Flow diagram of an embodiment of step S401 in this invention; Figure 6 This invention Figure 1 Flow diagram of an embodiment of step S103 in this invention; Figure 7 Schematic structural diagram of an embodiment of the battery thermal runaway warning device provided by this invention. Detailed implementation manners

[0019] Next, the technical solutions in the embodiments of this invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of this invention. Obviously, the described embodiments are only a part of the embodiments of this invention, rather than all of the embodiments. Based on the embodiments in this invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of this invention.

[0020] In the description of the embodiments of this invention, unless otherwise specified, "a plurality of" means two or more.

[0021] The descriptions such as "first", "second", etc. involved in the embodiments of this invention are only for descriptive purposes, and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Therefore, the technical features defined with "first" and "second" may explicitly or implicitly include at least one of such features.

[0022] Referring to "embodiment" in this text means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of this invention. The appearance of this phrase at various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0023] This invention provides a method and device for warning thermal runaway of lithium-ion batteries for electric bicycles, which will be described separately below.

[0024] Figure 1 Flow diagram of an embodiment of the method for warning thermal runaway of lithium-ion batteries for electric bicycles provided by this invention. As Figure 1 shown, the method for warning thermal runaway of lithium-ion batteries for electric bicycles includes: S101. Determine key parameters according to the contribution degrees of various parameters in the operation data of the lithium-ion battery of the electric bicycle to the change of the battery surface temperature; It should be noted that by simulating the operation process of the lithium-ion battery of an electric bicycle through charge-discharge cycle experiments and thermal abuse and thermal runaway experiments, experimental data of various parameters are obtained, and these data are processed and analyzed; the methods for calculating the contribution degree of each parameter to the change in the battery surface temperature include, but are not limited to, using Granger causality test and Shapley value. In this embodiment, the contribution degrees of nineteen different parameters during the operation of the lithium-ion battery of an electric bicycle to the change in the battery surface temperature are obtained through experiments, and finally four most critical parameters are determined. The parameters during the battery operation process include capacity, specific capacity, charging capacity, charging specific capacity, battery surface temperature, ambient temperature, current, voltage, hydrogen concentration, etc. The four final determined critical parameters are battery surface temperature, ambient temperature, current voltage, and hydrogen concentration.

[0025] S102. Using the critical parameters as the input and the change trend of the battery surface temperature in the future time period as the output, construct a battery surface temperature prediction model, and predict the change trend of the battery surface temperature in the future time period according to the critical parameters of the target battery within the current preset time period; It should be noted that after determining the critical parameters, input the critical parameters into the trained model, and the change trend of the battery surface temperature in the future time period can be predicted. This model is a time series deep learning model, using the BtrNet network, but not limited to this network. Based on the deep learning algorithm, by learning the critical parameters, the change trend of the battery surface temperature at future moments is predicted.

[0026] Furthermore, it should be noted that after determining the critical parameters, install various sensors on the lithium-ion battery of the electric bicycle to detect the critical parameters of the vehicle-mounted battery. For example, integrate electrical signal, gas signal, and temperature signal sensors into an electrical-gas-temperature integrated sensor to collect hydrogen concentration, voltage and current, surface temperature, and ambient temperature of the lithium-ion battery of the electric bicycle.

[0027] S103. Conduct a thermal runaway early warning for the target battery according to the change trend of the battery surface temperature, the critical parameters at the current moment, and the preset critical parameter risk threshold.

[0028] It should be noted that by continuously conducting experiments to set the risk threshold of the critical parameters, the critical parameter data of the electric bicycle battery are obtained in real time through various sensors. According to these critical parameters, the change trend of the battery surface temperature in the future time period is predicted, and it is judged whether the change trend of the battery surface temperature and other critical parameters are within the risk threshold range, and a thermal runaway early warning is conducted for the battery within the risk threshold range.

[0029] Compared with the prior art, in this embodiment, by analyzing the contribution degrees of various parameters in the battery operation data to the change of the battery surface temperature, multiple key parameters are determined according to the contribution degrees of the parameters; then, taking the key parameters as the input and the change trend of the battery surface temperature in the future time period as the output, a deep learning-based battery surface temperature prediction model is constructed, and the change trend of the battery surface temperature in the future time period is predicted according to the key parameters of the target battery in the current preset time period obtained in real time. At this time, only a small number of key parameters need to be concerned, and these key parameters have a significant contribution to the change of the battery surface temperature. Then, the change trend of the battery surface temperature can be intelligently predicted according to the model, and the change of the battery surface temperature can be predicted intelligently; finally, a thermal runaway warning is given to the target battery according to the change trend of the battery surface temperature and the key parameters at the current moment. In this embodiment, by analyzing the contribution degrees of various parameters during battery operation to the change of the battery temperature, key parameters are determined, the change trend of the battery surface temperature is intelligently predicted by obtaining real-time key parameters, and a thermal runaway warning evaluation is performed on the battery according to the temperature change trend and the key parameters, so that the thermal runaway of the battery in the future time period can be warned intelligently and accurately through a small amount of battery data.

[0030] In some embodiments of the present invention, as Figure 2 shown, Figure 2 is a schematic flowchart of an embodiment of step S101 provided by the present invention, including: Figure 1 S201. Obtain initial operation data during the battery operation process through charge-discharge cycle tests and thermoelectric abuse thermal runaway tests; It should be noted that through experiments, normal charge-discharge experiments and thermal abuse thermal runaway experiments are carried out on the battery in different states, and a battery test system and a temperature measurement system are used to record the change of battery parameters during the experiment, including capacity, specific capacity, charge capacity, charge specific capacity, battery surface temperature, ambient temperature, current, voltage, and hydrogen concentration, etc.

[0031] S202. Preprocess the initial operation data to obtain battery operation data; It should be noted that the battery operation data obtained from the experiment is transmitted to an electronic terminal, and the operation data is processed and analyzed through a data processing tool on the electronic terminal, including missing data filling, abnormal data processing, and data standardization, etc.

[0032] S203. Calculate the contribution degrees of various parameters in the operation data to the change of the battery surface temperature, and screen out the parameters with contribution degrees higher than a preset threshold as key parameters.

[0033] Specifically, by calculating the contribution degrees of each parameter to the change of the battery surface temperature, the importance of each parameter is determined, and the key parameters that most affect the change of the battery surface temperature are screened out.

[0034] In this embodiment, by analyzing various operating parameters of the battery, multiple key parameters that have the greatest impact on the battery surface temperature are screened out, providing an accurate basis for predicting the battery surface temperature in the follow-up.

[0035] In some embodiments of the present invention, as Figure 3 shown, Figure 3 provided by the present invention, Figure 2 is a schematic flowchart of an embodiment of step S202, and S202 includes: S301. Fill in the missing data in the initial operating data according to the time domain of the initial operating data to obtain the first initial operating data; It should be noted that this embodiment has a strong time dependence on the initial operating data. The parameter value at a certain moment has a strong correlation with the parameter values at adjacent moments. Therefore, it is necessary to fill in the missing data, and the data filling methods include but are not limited to using the linear interpolation method.

[0036] Specifically, the formula for filling data by the linear interpolation method is: , where, is the missing value filled at moment, and are respectively the previous moment and the next moment of the missing moment, is the parameter value corresponding to moment, and is the parameter value corresponding to moment.

[0037] S302. Detect outliers in the first initial operating data, and replace the abnormal data according to the detection results to obtain the second initial operating data; It should be noted that the methods for detecting outliers in the operating data include but are not limited to prior knowledge and the interquartile range method.

[0038] Specifically, first, determine the normal value range of all operating data through prior knowledge. In this embodiment, it is stipulated that the temperature value range is [0, 600] °C, the voltage value range is [0, 50] V, the current value range is [-150, 150] A, and the hydrogen concentration range is [0, 500] ppm. Data outside the normal value range is considered abnormal data. Then, sort each parameter sequence from smallest to largest by the interquartile range (IQR) method, divide the obtained ordered data of each parameter into quartiles, and divide them into Q1, Q2, and Q3, which represent the 25th percentile, the 50th percentile, and the 75th percentile respectively. Calculate the interquartile range (IQR) and the upper and lower bound ranges, and regard the data outside the upper and lower bound ranges as outliers. The calculation formulas for the interquartile range (IQR) and the upper and lower bound ranges are as follows: ， ， ， Among them, is the interquartile range value, is the lower bound range, is the upper bound range, and are the first quartile and the third quartile of the quartile division respectively.

[0039] S303. Standardize different variables in the second initial operating data to obtain battery operating data.

[0040] It should be noted that since the initial operating data contains multiple different variables, the differences in units and orders of magnitude of different variables will cause the weight imbalance of the dataset. The methods for standardizing the operating data in this embodiment include, but are not limited to, using the Z-score method to standardize the dataset. After standardization, the mean of each column of features is 0 and the standard deviation is 1. The formula for standardizing the data is: , Among them, is the data after standardization, is the value of the parameter at a certain moment, μ is the mean of the parameter in this group of time series data, and σ is the standard deviation of the parameter in this group of time series data.

[0041] In this embodiment, by filling in missing data, processing abnormal data, and standardizing the operating data of the battery, the integrity, accuracy, and consistency of the battery operating data are improved, providing an accurate training dataset for the subsequent training of the model, and further improving the accuracy of the model.

[0042] In some embodiments of the present invention, as Figure 4 shown, Figure 4 is a schematic flowchart of an embodiment of step S203 provided by the present invention. Step S203 includes: Figure 2 S401. Based on the Granger causality test, determine the contribution degree of each parameter in the operation data to the change of the battery surface temperature; It should be noted that it is ensured that the historical time series data of all parameters and the battery surface temperature data are both stationary time series. By calculating the lag terms of each parameter, the prediction accuracy of the change of the battery surface temperature can be significantly improved to judge the causal contribution of each parameter to the change of the battery surface temperature.

[0043] S402. Based on the Shapley value, sort each parameter according to the contribution degree, and screen out the parameters with a contribution degree higher than the preset threshold as key parameters.

[0044] It should be noted that when multiple parameters all pass the causal contribution test, the Shapley value analysis method needs to be used for the parameters that pass the test to calculate the joint contribution degree of the time series of each parameter to the change of the battery surface temperature. In this embodiment, through the Shapley value analysis of each parameter, the battery surface temperature contributes the most to predicting the temperature of the battery in the future time step, followed by the ambient temperature, voltage and current, and hydrogen concentration respectively.

[0045] This embodiment identifies the contribution degree of each parameter in the battery operation data through the Granger causality test, sorts the contribution degree of each parameter to predicting the future battery surface temperature through the Shapley value analysis, and screens out the key parameters with significant contribution degree.

[0046] In some embodiments of the present invention, as Figure 5 shown, Figure 5 is a schematic flowchart of an embodiment of step S401 provided by the present invention. Step S401 includes: Figure 4 S501. Based on a preset first regression model, predict the battery surface temperature at the next moment according to the battery surface temperature, and obtain the first surface predicted temperature; It should be noted that through the first regression model, the battery surface temperature at the next moment is predicted in sequence according to the historical time series data of each parameter, and the reference result is determined according to the results of each first surface predicted temperature. In this embodiment, through calculation, it is obtained that the battery surface temperature has the greatest influence on predicting the battery surface temperature, so the parameter of the battery surface temperature is used as the reference parameter.

[0047] Specifically, the calculation formula of the first regression model is: , where, is the predicted temperature of the first surface, i.e., the predicted temperature of the battery surface at the next moment, is a constant term, is a regression coefficient, is white noise with zero mean, is the lag order. When the lag order = 1, the regression coefficient of the battery surface temperature is lower than 0.05, indicating that the battery surface temperature has a significant contribution to predicting the battery surface temperature on a short time scale.

[0048] S502. Based on a preset second regression model, predict the battery surface temperature at the next moment according to each parameter and the battery surface temperature to obtain the second predicted surface temperature; Specifically, the calculation formula of the second regression model is: , wherein, is the value of parameter Y at time t , is the value of parameter Y at time t-i , is the value of parameter X at time t-i , is a constant term, and are regression coefficients, is white noise with zero mean. When the lag order = 1, the regression coefficient of the ambient temperature is lower than 0.05, indicating that the temperature feature has a significant contribution to predicting the battery surface temperature on a short time scale. Combining with the fact that the battery surface temperature in the first regression model has a significant contribution to predicting the battery surface temperature on a short time scale, it shows that the temperature feature has a significant contribution to predicting the battery surface temperature on a short time scale. As the lag order increases, the regression coefficients of voltage, current and hydrogen concentration gradually decrease. When the lag order , all of them drop below the significance threshold, proving that voltage, current and hydrogen concentration all have a significant contribution to predicting the battery surface temperature on a long time scale.

[0049] S503. Successively calculate the residual variance between each second predicted surface temperature and the first predicted surface temperature. According to the residual variance, determine the contribution degree of each parameter to the change of the battery surface temperature.

[0050] In this embodiment, the benchmark parameters are determined by calculating the contribution degree of the parameters, and then the contribution degrees of other parameters are calculated successively, so that multiple key parameters that can most affect the change of the battery surface temperature can be screened out through the contribution degree.

[0051] In some embodiments of the present invention, the battery surface temperature prediction model includes a feature interaction layer, a feature fusion layer, and a temperature prediction layer; The feature interaction layer is used to determine the dynamic range of the receptive field through a high-dimensional channel index, and perform a convolution operation on the time-series data of the key parameters through a sliding window according to the dynamic range of the receptive field to obtain the time-series features of the key parameters; Specifically, the calculation formula for the dynamic range of the receptive field is: , where, is the dynamic range of the receptive field, represents taking the nearest odd operation, is the high-dimensional channel index function, is the output channel index.

[0052] Furthermore, the calculation formula for the time-series features is: , where, is the value of the output feature on the th output channel at time step , is the feature dimension, K is the dynamic range of the receptive field, is the convolution kernel weight, is the position index of the convolution kernel on the time axis, is the input feature dimension index, is the output channel index, is the th output channel bias term.

[0053] The feature fusion layer is used to embed position information for the time-series data at each time step after segmentation, map the time-series features to high-dimensional features, and fuse the high-dimensional features; In some embodiments of the present invention, the feature fusion layer includes a triangular position embedding layer, an encoder, and a decoder; The triangular position embedding layer is used to embed position information for the time-series features; Specifically, the calculation formula for the position information is: , , where, PE represents the position encoding, pos is the position of the element in the sequence, i is the dimension index of the encoding, d is the dimension of the position encoding, and the position embedding adds a vector representing position information to each feature in the input sequence to distinguish the position information at different time steps.

[0054] The encoder is used to map the temporal features embedded with position information into a high-dimensional data space to obtain high-dimensional features, calculate the internal relationships between the high-dimensional features, and fuse the high-dimensional features according to the internal relationships. Specifically, the encoder calculates the similarity between different time steps in the sequence through multi-head self-attention, performs internal interaction on the global information, and captures the temporal dependencies in the long sequence. The calculation formula is: , where, is the dependency relationship of the query matrix, key matrix, and value matrix, is the dimension of the key, Q is the query matrix, K is the key matrix, and V is the value matrix.

[0055] Furthermore, the encoder calculates the significant relationships between different features in the sequence through multi-head cross-attention, quantifies the interaction relationships between different features, and realizes feature fusion between scales. Multiple attention heads can capture the features of different subspaces from multiple angles, thereby further enhancing the expression ability of the input features. At the same time, in order to overcome the overfitting phenomenon that appears at the end of training, dropout technology is introduced in the cross-attention layer and the feed-forward neural network layer in the encoder to effectively prevent overfitting to extreme environmental temperature data or voltage jump effects during battery charge and discharge process changes, and improve the adaptability to different working condition scenarios.

[0056] The decoder is used to reconstruct the output of the encoder.

[0057] The temperature prediction layer performs non-linear regression prediction on the temperature change trend of the battery surface based on the position information of the temporal features and the fused high-dimensional features.

[0058] Specifically, the temperature prediction layer is implemented by a multi-layer perceptron (MLP) structure to perform non-linear regression prediction on the battery surface temperature at future time steps. The multi-layer perceptron (MLP) consists of multiple fully connected layers, each layer contains multiple neuron nodes, and non-linear mapping ability is introduced through activation functions. In this embodiment, the input of the multi-layer perceptron (MLP) is the high-dimensional feature vector output by the feature fusion layer, and the finally output is the predicted battery surface temperature value.

[0059] To better verify the performance of the battery surface temperature prediction model in this embodiment, the present invention adopts a variety of evaluation indicators, including the mean absolute error (MAE), the mean absolute percentage error (MAPE), and the coefficient of determination (R²). These indicators reflect the accuracy and reliability of the model prediction from different perspectives, providing an important reference basis for the optimization and practical application of the model. In this embodiment, through the model prediction experiments under different working conditions, the mean absolute error (MAE), the mean absolute percentage error (MAPE), and the coefficient of determination (R²) between the predicted temperature and the actual temperature are calculated. Table 1 shows the battery surface temperature prediction results under the constant current and constant voltage charging conditions at different charging rates provided by the present invention, and Table 2 shows the battery surface temperature prediction results under different temperature environments provided by the present invention; Table 1 Battery Surface Temperature Prediction Results under Constant Current and Constant Voltage Charging Conditions at Different Charging Rates

[0060] Table 2 Battery Surface Temperature Prediction Results under Different Temperature Environments

[0061] As can be seen from Table 1 and Table 2 above, by calculating the mean absolute error (MAE), the mean absolute percentage error (MAPE), and the coefficient of determination (R²) between the model temperature prediction value and the actual temperature value, it shows that the model can accurately predict the battery surface temperature under different working conditions.

[0062] Meanwhile, Table 3 shows the battery surface temperature prediction results of different battery models provided by the present invention.

[0063] Table 3 Battery Surface Temperature Prediction Results of Different Battery Models

[0064] Since the capacity of the lithium iron phosphate battery is relatively large, the heat generation during cycling at a current of 0.5 C is significantly higher than that of the ternary lithium battery, resulting in a significant increase in the mean absolute error of the network prediction results, and the decline of the mean absolute percentage error and R 2 of the model is relatively low. This shows that the temperature prediction model in this embodiment can still maintain a high accuracy when migrated to lithium-ion batteries of different models, and has excellent generalization ability.

[0065] In some embodiments of the present invention, the setting process of the preset key parameter risk threshold includes: Obtaining the change data of each key parameter and the battery surface temperature during the experiment, and drawing a change trend graph of the key parameter and the battery surface temperature; Determining the key points of the battery surface temperature change; Determining the risk threshold of each key parameter according to the key points.

[0066] It should be noted that by conducting normal charge and discharge tests and thermal abuse thermal runaway tests on the battery, the data of the battery surface temperature change and the change data of other key parameters under different working conditions are obtained, and the change trend diagrams of these key parameters are drawn. The inflection points of the battery surface temperature change are determined as key points in the change trend diagrams, and the risk thresholds of each key parameter are determined through these key points; in this embodiment, a first-level warning threshold interval and a second-level warning threshold interval are set for the key parameters through experiments, as shown in Table 4. Table 4 is the first-level warning threshold interval and the second-level warning threshold interval of the key parameters provided by the present invention. When the electric vehicle battery is within the corresponding risk level threshold range, the corresponding emergency plan is activated. During the first-level warning, that is, a fault alarm is issued, the battery is powered off, and it is reminded to replace the battery. During the second-level warning, at this time, it is a thermal runaway warning, the battery is powered off, fire extinguishing is prepared, and evacuation is reminded, etc.

[0067] Table 4 Warning Threshold Intervals of Key Parameters

[0068] In some embodiments of the present invention, as Figure 6 shown, Figure 6 is a schematic flow chart of an embodiment of step S103 provided by the present invention. S103 includes: Figure 1 S601. When the change trend of the battery surface temperature and the key parameters meet the preset first-level warning, take fault alarm measures; S602. When the change trend of the battery surface temperature and the key parameters meet the preset second-level warning, take thermal runaway warning.

[0069] To better implement the thermal runaway warning method for the lithium-ion battery of an electric bicycle in the embodiments of the present invention, correspondingly, on the basis of the thermal runaway warning method for the lithium-ion battery of an electric bicycle, as Figure 7 shown, the embodiment of the present invention further provides a battery thermal runaway warning device 700, including: A key parameter determination module 701, configured to determine key parameters according to the contribution degree of each parameter in the battery operation data to the change of the battery surface temperature; A temperature prediction module 702, configured to construct a battery surface temperature prediction model with the key parameters as inputs and the change trend of the battery surface temperature in a future time period as outputs, and predict the change trend of the battery surface temperature in a future time period according to the key parameters of the target battery within a current preset time period; A warning module 703, configured to perform thermal runaway warning on the target battery according to the change trend of the battery surface temperature, the key parameters at the current moment, and the preset key parameter risk thresholds.

[0070] ​The battery thermal runaway warning device 700 provided in the above embodiments can implement the technical solutions described in the embodiments of the method for warning thermal runaway of lithium-ion batteries of electric bicycles. For the specific implementation principles of the above modules or units, reference can be made to the corresponding content in the embodiments of the method for warning thermal runaway of lithium-ion batteries of electric bicycles, which will not be elaborated here.

[0071] In the embodiments of the present invention, the battery thermal runaway warning device can be an independent mobile device with edge computing capabilities. It is installed on an electric bicycle and can obtain the key parameters of the lithium-ion battery of the electric bicycle in real time through an electro-gas-temperature integrated sensor to predict the temperature change trend of the lithium-ion battery of the electric bicycle, so as to give a thermal runaway warning for the lithium-ion battery of the electric bicycle.

[0072] The method and device for warning thermal runaway of lithium-ion batteries of electric bicycles provided by the present invention have been introduced in detail above. Specific examples are used in this article to elaborate on the principles and implementation manners of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for warning of thermal runaway of a lithium-ion battery for an electric bicycle, characterized in that, Including: Determine key parameters according to the contribution degree of each parameter in the operation data of the lithium-ion battery of the electric bicycle to the change of the battery surface temperature; Taking the key parameters as the input and the change trend of the battery surface temperature in the future time period as the output, construct a battery surface temperature prediction model, and predict the change trend of the battery surface temperature in the future time period according to the key parameters of the target battery in the current preset time period; Perform thermal runaway early warning on the target battery according to the change trend of the battery surface temperature, the key parameters at the current moment, and the preset key parameter risk threshold.

2. The method for warning of thermal runaway of a lithium-ion battery for an electric bicycle according to claim 1, wherein Determine key parameters according to the contribution degree of each parameter in the battery operation data to the change of the battery surface temperature, including: Obtain the initial operation data during the battery operation process through charge and discharge cycle tests and thermoelectric abuse thermal runaway tests; Preprocess the initial operation data to obtain battery operation data; Calculate the contribution degree of each parameter in the operation data to the change of the battery surface temperature, and screen out the parameters with a contribution degree higher than the preset threshold as key parameters.

3. The method for warning of thermal runaway of a lithium-ion battery for an electric bicycle according to claim 2, characterized in that, Preprocess the initial operation data to obtain battery operation data, including: Fill in the missing data in the initial operation data according to the time domain characteristics of the initial operation data to obtain the first initial operation data; Perform outlier detection on the first initial operation data, and replace the abnormal data according to the detection results to obtain the second initial operation data; Perform standardization processing on different variables in the second initial operation data to obtain battery operation data.

4. The method for warning of thermal runaway of a lithium-ion battery for an electric bicycle according to claim 1, characterized in that, Calculate the contribution degree of each parameter in the operation data to the change of the battery surface temperature, and screen out the parameters with a contribution degree higher than the preset threshold as key parameters, including: Based on the Granger causality test, determine the contribution degree of each parameter in the operation data to the change of the battery surface temperature; Based on the Shapley value, sort each parameter according to the contribution degree, and screen out the parameters with a contribution degree higher than the preset threshold as key parameters.

5. The method for warning of thermal runaway of a lithium-ion battery of an electric bicycle according to claim 4, wherein, The operation data includes battery surface temperature data. Determining the contribution degree of each parameter in the operation data to the change of the battery surface temperature includes: Based on a preset first regression model, predict the battery surface temperature at the next moment according to the historical data of the battery surface temperature to obtain the first surface prediction temperature; Based on a preset second regression model, predict the battery surface temperature at the next moment according to the historical data of each parameter and the battery surface temperature to obtain multiple second surface prediction temperatures; Successively calculate the residual variance between each second surface prediction temperature and the first surface prediction temperature, and determine the contribution degree of each parameter to the change of the battery surface temperature according to the residual variance.

6. The method for warning of thermal runaway of a lithium-ion battery for an electric bicycle according to claim 1, characterized in that, The battery surface temperature prediction model includes a feature interaction layer, a feature fusion layer, and a temperature prediction layer; The feature interaction layer is used to determine the dynamic range of the receptive field through high-dimensional channel indexing, and perform a convolution operation on the time series data of the key parameters through a sliding window according to the dynamic range of the receptive field to obtain the time series features of the key parameters; The feature fusion layer is used to embed position information into the time series features, map the time series features with embedded position information to high-dimensional features, and fuse the high-dimensional features; The temperature prediction layer performs non-linear regression prediction based on the position information of the time series features and the fused high-dimensional features, and predicts the temperature change trend of the battery surface.

7. The method for warning of thermal runaway of a lithium-ion battery for an electric bicycle according to claim 6, wherein, The feature fusion layer includes a triangular position embedding layer, an encoder, and a decoder; The triangular position embedding layer is used to embed position information for the time series features; The encoder is used to map the time series features with embedded position information into a high-dimensional data space to obtain high-dimensional features, calculate the internal relationships between the high-dimensional features, and fuse the high-dimensional features according to the internal relationships; The decoder is used to reconstruct the output of the encoder.

8. The method for warning of thermal runaway of a lithium-ion battery for an electric bicycle according to claim 1, wherein, Performing thermal runaway warning on the target battery according to the battery surface temperature change trend, the key parameters at the current moment, and the preset key parameter risk threshold, including: When the battery surface temperature change trend and the key parameters meet the preset first-level warning, take fault alarm measures; When the battery surface temperature change trend and the key parameters meet the preset second-level warning, take thermal runaway warning.

9. The method for warning of thermal runaway of a lithium-ion battery of an electric bicycle according to claim 1, wherein The setting process of the preset key parameter risk threshold includes: Obtain the change data of each key parameter and the battery surface temperature during the experiment of the battery, and draw a change trend graph of the key parameter and the battery surface temperature; Determine the key points of the battery surface temperature change; Determine the risk threshold of each key parameter according to the key points.

10. A battery thermal runaway warning device, characterized in that, Including: A key parameter determination module, which is used to determine key parameters according to the contribution degree of each parameter in the battery operation data to the change of the battery surface temperature; A temperature prediction module, which is used to construct a battery surface temperature prediction model with the key parameters as the input and the battery surface temperature change trend in the future time period as the output, and predict the battery surface temperature change trend in the future time period according to the key parameters of the target battery within the current preset time period; An early warning module, which is used to perform thermal runaway warning on the target battery according to the battery surface temperature change trend, the key parameters at the current moment, and the preset key parameter risk threshold.

Images