Intelligent early warning method for phase change thermal runaway of lithium-ion batteries

Abstract

The present invention relates to the field of thermal runaway warning technology, specifically to an intelligent warning method for thermal runaway of lithium-ion battery phase change. The specific steps are: S1: Collect time series data related to the temperature of the lithium-ion battery, normalize the data, and extract features; S2, obtain the real-time temperature and temperature change index of the lithium-ion battery, and build a CNN-LSTM prediction model based on the multi-head attention mechanism optimized by the dung beetle algorithm; S3, set the lithium-ion battery thermal runaway warning threshold, set the lithium-ion battery thermal runaway warning change index threshold range; S4, judge whether the lithium-ion battery will have thermal runaway through the real-time temperature, predicted temperature and temperature change index of the target battery, and if so, issue a thermal runaway warning. The present invention warns of thermal runaway of lithium-ion battery phase change by setting multiple warning modes, thereby improving the accuracy and timeliness of the warning, thereby ensuring the safety of lithium-ion batteries.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal runaway early warning, and in particular to an intelligent early warning method for phase-change thermal runaway of a lithium-ion battery. Background Art

[0002] The phase change process in lithium-ion batteries mainly includes: Electrode material phase change: The positive electrode material will undergo phase change during the process of lithium ion insertion and extraction. The negative electrode material will also undergo phase change during the process of lithium ion adsorption or desorption. These phase change processes will cause volume changes and changes in crystal structure. Electrolyte phase change: The electrolyte usually uses a liquid organic solvent, which will undergo a liquid-solid phase change at low temperatures to form a solid electrolyte. The solid electrolyte phase change process will cause changes in ionic conductivity and affect battery performance. Interface phase change: The electrolyte interface will form interface layers such as solid electrolyte interface membranes, and their structure and composition will change continuously with charging and discharging. The phase change of the SEI layer will affect the electrode surface reaction kinetics and battery impedance characteristics. Precipitate phase change: Under conditions such as high voltage and high temperature, the electrolyte and electrode materials will undergo decomposition reactions to generate some sediments. The phase change of these precipitates will block the electrode pores and increase the battery impedance.

[0003] These phase change processes are usually accompanied by thermal effects, causing local or overall temperature changes. At the same time, temperature changes will in turn affect the phase change behavior, forming a complex temperature-phase change coupling process. When the temperature rises to a certain range, thermal runaway of lithium-ion batteries may occur. When thermal runaway occurs in lithium-ion batteries, the battery temperature may rise sharply, chemical reactions may occur inside the battery, electrolyte decomposition may release flammable gases, the battery may swell and deform, or even explode. Fires and explosions may cause property damage, release toxic and hazardous substances to harm the environment, and even threaten the personal safety of users. Early warning of intelligent phase change thermal runaway of lithium-ion batteries through real-time temperature monitoring is not enough to ensure the safe use of lithium-ion batteries. It is difficult to accurately predict when the battery will experience thermal runaway, and there is uncertainty in the early warning. Summary of the invention

[0004] Based on this, it is necessary to propose an intelligent early warning method for phase change thermal runaway of lithium-ion batteries in response to the problems of insufficient accuracy and reliability of detection methods, difficulty in accurately predicting when phase change thermal runaway of the battery will occur, and uncertainty in early warning.

[0005] The present invention is implemented by the following technical solution: The intelligent early warning method for phase change thermal runaway of lithium-ion batteries comprises the following steps:

[0006] S1: Collect time series data related to lithium-ion battery temperature, including charging current, discharging current, and battery voltage, normalize and preprocess the data, and extract features, including sliding window features, differential features, and frequency domain features;

[0007] S2: Obtain the real-time temperature, temperature change index and predicted temperature of the lithium-ion battery at the next moment. The specific method is as follows:

[0008] S21, using a temperature sensor to monitor the temperature of the lithium-ion battery in real time;

[0009] S22, calculating a temperature change index of the lithium-ion battery within a preset time period according to an increase or decrease in the temperature of the lithium-ion battery;

[0010] S23: Construct a CNN-LSTM prediction model based on the multi-head attention mechanism optimized by the dung beetle algorithm to predict the target battery temperature data;

[0011] S3. Set two lithium-ion battery phase change thermal runaway warning thresholds, including:

[0012] S31, setting a lithium-ion battery phase change thermal runaway warning threshold according to the performance parameters of the target lithium-ion battery and various side reactions induced when the phase change thermal runaway phenomenon occurs;

[0013] S32, by modeling and simulating the phase change thermal runaway of the lithium-ion battery, obtaining the normal temperature change index range of the lithium-ion battery under various normal operating conditions and the abnormal temperature change index range when the phase change thermal runaway phenomenon occurs, and setting the lithium-ion battery phase change thermal runaway warning change index threshold range according to the two temperature change indices;

[0014] S4. Determine whether the lithium-ion battery will experience phase change thermal runaway through the real-time temperature, predicted temperature and temperature change index of the target battery, and issue a thermal runaway warning if so.

[0015] By setting multiple warning modes to warn of phase change thermal runaway of lithium-ion batteries, the comprehensiveness of the warning can be improved, the reliability of the warning can be enhanced, the sensitivity of the warning can be optimized, the feedback information of the warning can be enriched, and the intelligence level of the warning can be improved, thereby ensuring the safety of lithium-ion batteries.

[0016] Furthermore, in S1, the formula for data normalization is:

[0017] In the formula, X is the original data value, X min is the minimum value in the data set, X max is the maximum value in the data set, and the data is divided into training set, validation set and test set.

[0018] Furthermore, in S1, the formula of sliding window feature is expressed as:

[0019] in, represents the sliding window feature at time t, Xt-w+1 , X t-w+2 , ..., X t Represents w data points from time t-w+1 to time t;

[0020] The formula of differential features is expressed as:

[0021] in, Represents the differential feature at time t, that is, the difference between the current time and the previous time;

[0022] The frequency domain features are obtained as follows:

[0023] in, It represents the frequency domain characteristics at time t, and f() is the fast Fourier transform.

[0024] Further, in S22, the lithium-ion battery operating state includes the lithium-ion battery temperature rising, stable and falling states, and the calculation formula for the temperature growth index e1 of the lithium-ion battery in the temperature rising state is as follows:

[0025] In the formula, It represents the change of battery temperature after time t. It represents the maximum temperature rise of the battery, t is the discharge time, m is the mass of the battery, c is the specific heat capacity of the battery, h is the convection heat transfer coefficient of the battery, and A is the surface area of ​​the battery.

[0026] The calculation formula for the change in battery temperature relative to the ambient temperature when the lithium-ion battery is in a stable temperature state and the lithium-ion battery is in a continuous stable output state is as follows:

[0027] In the formula, is the change in battery temperature relative to the ambient temperature, I is the discharge current of the battery, R is the internal resistance of the battery, m is the mass of the battery, and c is the specific heat capacity of the battery;

[0028] The calculation formula for the temperature T of the lithium-ion battery in the standby state in the lithium-ion battery temperature stable state is as follows:

[0029] Where T is the actual temperature of the battery, is the ambient temperature, It is the fluctuation range of battery temperature relative to normal temperature.

[0030] The calculation formula for the temperature drop index e2 of the lithium-ion battery when the temperature drops is as follows:

[0031] In the formula, is the change in battery temperature after time t2, is the maximum temperature drop when the battery stops discharging, t2 is the time after the discharge stops, m is the mass of the battery, c is the specific heat capacity of the battery, h2 is the convection heat transfer coefficient of the battery, and A is the surface area of ​​the battery.

[0032] Furthermore, in S23, the specific steps of predicting the target battery temperature data are as follows:

[0033] S231: Extract features through convolutional neural networks, model time dependencies through long-term and short-term neural networks, and use multi-head attention mechanisms to enhance the model’s ability to capture key features.

[0034] S232: Use the dung beetle algorithm to optimize the hyperparameters of the CNN-LSTM model;

[0035] S233: Use the training set to train the optimized CNN-LSTM model, and use an independent validation set to evaluate the performance of the model. The evaluation criteria include mean square error and root mean square error.

[0036] S234: Based on the performance of the model on the validation set, repeat the training and evaluation until the model performance reaches the optimal level.

[0037] Furthermore, in S231, the convolutional neural network feature extraction method is:

[0038] in, Represents the features extracted by CNN;

[0039] The application formula of the multi-head attention mechanism is:

[0040]

[0041] Among them, represents the attention weight of the i-th attention head at time t, represents the weight vector of the i-th attention head, Feature representation representing the aggregation of attention mechanisms;

[0042] The long-term short-term neural network part models time dependence in the following way:

[0043] in, and Respectively represent the hidden state and cell state of LSTM at time t;

[0044] The final output is calculated as:

[0045] in, represents the predicted temperature of the model at time t.

[0046] Further, in S232, the dung beetle algorithm optimizes the hyperparameters of the CNN-LSTM model. The dung beetle algorithm simulates the foraging behavior of the dung beetle colony and iteratively optimizes the hyperparameter combination to minimize the MSE loss function on the validation set. The formula is:

[0047] Where N represents the number of samples in the validation set. represents the predicted temperature of the model, Indicates the true temperature.

[0048] Furthermore, the core steps of the dung beetle algorithm are as follows:

[0049] a. Initialize the position and direction of the dung beetle colony;

[0050] b. Calculate the loss function value of each dung beetle and determine the optimal individual position;

[0051] c. Update the position and direction of each dung beetle by:

[0052]

[0053] in, represents the position of the i-th dung beetle in the j-th dimension, represents the speed of the i-th dung beetle in the j-th dimension, represents the optimal position of the i-th dung beetle in the j-th dimension, represents the best position of the entire group in the jth dimension, w represents the inertia weight, and is the acceleration constant, and Represents a random number;

[0054] d. Determine whether each dung beetle is beyond the boundary, and if so, expand the search range;

[0055] e. Update the current optimal solution;

[0056] f. Repeat steps be until the maximum number of iterations is reached and output the hyperparameters of the CNN-LSTM model.

[0057] Furthermore, the calculation formula of the mean square error is as follows:

[0058] Where N represents the number of samples in the validation set. represents the predicted temperature of the model, Indicates the true temperature;

[0059] The formula for calculating the root mean square error is as follows:

[0060] Here, RMSE is the square root of MSE and represents the standard deviation of the prediction error.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] The present invention can better distinguish normal temperature changes from abnormal temperature changes by judging whether the temperature change index is normal, and can timely discover abnormal temperature changes that may cause thermal runaway, thereby reducing the false alarm rate of the thermal runaway warning system and improving the accuracy of the warning. By constructing a CNN-LSTM prediction model based on a multi-head attention mechanism optimized by the dung beetle algorithm, a lithium-ion battery temperature prediction model is constructed using CNN-LSTM, and the model parameters are optimized and trained using the dung beetle algorithm. In the test phase, a new data sample is input, and the predicted value of the battery temperature is obtained through model prediction. Compared with other prediction methods, it has higher prediction accuracy, lower calculation complexity, shorter training time, and better practicality. By setting a variety of lithium-ion battery phase change thermal runaway warning thresholds, the timeliness of the warning can be improved. Compared with relying only on the current temperature value for warning, the method based on temperature prediction can more timely discover potential thermal runaway risks. The probability of thermal runaway can be reduced, and timely warning can allow the battery management system to take cooling or other protective measures in time, thereby avoiding the temperature from continuing to rise to the critical point that causes thermal runaway, greatly reducing the probability of thermal runaway accidents, and improving the safety of battery use. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a step diagram of the intelligent early warning method for phase change thermal runaway of a lithium-ion battery in Example 1 of the present invention;

[0064] Figure 2 This is a flow chart of constructing a lithium-ion battery temperature prediction model in Example 1 of the present invention;

[0065] Figure 3 This is a flowchart of optimizing the CNN-LSTM prediction model using the dung beetle algorithm in Example 1 of the present invention. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] It should be noted that when a component is referred to as being "mounted on" another component, it may be directly on the other component or there may be a central component. When a component is considered to be "set on" another component, it may be directly set on the other component or there may be a central component at the same time. When a component is considered to be "fixed to" another component, it may be directly fixed on the other component or there may be a central component at the same time.

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "or / and" used herein includes any and all combinations of one or more of the related listed items.

[0069] Example 1: Please refer to Figure 1 This embodiment provides an intelligent early warning method for phase change thermal runaway of a lithium-ion battery, comprising the following steps:

[0070] S1: Collect time series data related to lithium-ion battery temperature, including charging current, discharging current, and battery voltage, normalize and preprocess the data, and extract features, including sliding window features, differential features, and frequency domain features.

[0071] The formula for data normalization is:

[0072] In the formula, X is the original data value, X min is the minimum value in the data set, X max is the maximum value in the data set, and the data is divided into training set, validation set and test set.

[0073] Considering that the change of battery temperature is a continuous time series process, some time series features can be extracted using the sliding window method:

[0074] Sliding average: At the current time step t, calculate the average value of the eigenvalues ​​in the previous n time steps. The calculation formula is:

[0075]

[0076] Sliding standard deviation: Calculate the standard deviation of the eigenvalues ​​in the first n time steps. The calculation formula is:

[0077]

[0078] The sliding window features obtained in this way can better characterize the dynamic characteristics of temperature changes.

[0079] The formula of sliding window feature is expressed as:

[0080] in, represents the sliding window feature at time t, X t-w+1 , X t-w+2 , ..., X t Represents w data points from time t-w+1 to time t;

[0081] The formula of differential features is expressed as:

[0082] in, Represents the differential feature at time t, that is, the difference between the current time and the previous time;

[0083] The frequency domain features are obtained as follows:

[0084] in, It represents the frequency domain characteristics at time t, and f() is the fast Fourier transform.

[0085] By performing fast Fourier transform on the temperature time series, the amplitude spectrum and phase spectrum in the frequency domain can be obtained. The frequency domain features can capture the periodic pattern of temperature changes and help improve the prediction effect.

[0086] Combining the above three feature engineering methods can effectively characterize the dynamic pattern of battery temperature changes and further improve the accuracy of temperature prediction.

[0087] S2: Obtain the real-time temperature, temperature change index and predicted temperature of the lithium-ion battery at the next moment. The specific method is as follows:

[0088] S21. Use a temperature sensor to monitor the temperature of the lithium-ion battery in real time.

[0089] S22. Calculate a temperature variation index of the lithium-ion battery within a certain period of time according to an increase or decrease in the temperature of the lithium-ion battery.

[0090] The operating state of the lithium-ion battery includes the state of lithium-ion battery temperature rising, stable and falling. The calculation formula for the temperature growth index e1 of the lithium-ion battery in the state of lithium-ion battery temperature rising is as follows:

[0091] In the formula, It represents the change of battery temperature after time t. It represents the maximum temperature rise of the battery, t is the discharge time, m is the mass of the battery, c is the specific heat capacity of the battery, h is the convection heat transfer coefficient of the battery, and A is the surface area of ​​the battery.

[0092] The calculation formula for the change in battery temperature relative to the ambient temperature when the lithium-ion battery is in a stable temperature state and the lithium-ion battery is in a continuous stable output state is as follows:

[0093] In the formula, is the change in battery temperature relative to the ambient temperature, I is the discharge current of the battery, R is the internal resistance of the battery, m is the mass of the battery, and c is the specific heat capacity of the battery;

[0094] The calculation formula for the temperature T of the lithium-ion battery in the standby state in the lithium-ion battery temperature stable state is as follows:

[0095] Where T is the actual temperature of the battery, is the ambient temperature, It is the fluctuation range of battery temperature relative to normal temperature.

[0096] The calculation formula for the temperature drop index e2 of the lithium-ion battery when the temperature drops is as follows:

[0097] In the formula, is the change in battery temperature after time t2, is the maximum temperature drop when the battery stops discharging, t2 is the time after the discharge stops, m is the mass of the battery, c is the specific heat capacity of the battery, h2 is the convection heat transfer coefficient of the battery, and A is the surface area of ​​the battery.

[0098] In this embodiment, by judging whether the temperature change index exceeds the preset threshold range, normal temperature changes and abnormal temperature changes can be better distinguished, and abnormal temperature changes that may lead to thermal runaway can be discovered in time, thereby reducing the false alarm rate of the thermal runaway warning system and improving the accuracy of the warning.

[0099] S23: Construct a CNN-LSTM prediction model based on the multi-head attention mechanism optimized by the dung beetle algorithm to predict the target battery temperature data.

[0100] The specific prediction steps are:

[0101] S231: The convolutional neural network is used to extract features, the long-term and short-term neural network is used to model time dependencies, and the multi-head attention mechanism is used to enhance the model's ability to capture key features.

[0102] The convolutional neural network feature extraction method is:

[0103] in, Represents the features extracted by CNN;

[0104] The application formula of the multi-head attention mechanism is:

[0105]

[0106] Among them, represents the attention weight of the i-th attention head at time t, represents the weight vector of the i-th attention head, Feature representation representing the aggregation of attention mechanisms;

[0107] The long-term short-term neural network part models time dependence in the following way:

[0108] in, and Respectively represent the hidden state and cell state of LSTM at time t;

[0109] The final output is calculated as:

[0110] in, represents the predicted temperature of the model at time t.

[0111] In this embodiment, the convolutional neural network layer can extract local features of the input data and has a strong modeling capability for time series data. The long short-term memory network layer can effectively model the long-term dependencies of time series data. The multi-head attention mechanism can capture key feature information in the input sequence.

[0112] S232: Use the Dung Beetle Algorithm to optimize the hyperparameters of the CNN-LSTM model.

[0113] The dung beetle algorithm simulates the social behavior of dung beetles in the real world. It iteratively optimizes the hyperparameter combination to minimize the MSE loss function on the validation set and updates the position of each dung beetle to find the global optimal solution. The formula is:

[0114] Where N represents the number of samples in the validation set. represents the predicted temperature of the model, Indicates the true temperature;

[0115] The core steps of the dung beetle algorithm are as follows:

[0116] a. Initialize the position and direction of the dung beetle colony;

[0117] b. Calculate the loss function value of each dung beetle and determine the optimal individual position;

[0118] c. Update the position and direction of each dung beetle by:

[0119]

[0120] in, represents the position of the i-th dung beetle in the j-th dimension, represents the speed of the i-th dung beetle in the j-th dimension, represents the optimal position of the i-th dung beetle in the j-th dimension, represents the best position of the entire group in the jth dimension, w represents the inertia weight, and is the acceleration constant, and Represents a random number;

[0121] d. Determine whether each dung beetle is beyond the boundary, and if so, expand the search range;

[0122] e. Update the current optimal solution;

[0123] f. Repeat steps be until the maximum number of iterations is reached and output the hyperparameters of the CNN-LSTM model.

[0124] S233: Use the training set to train the optimized CNN-LSTM model, and use an independent validation set to evaluate the performance of the model. The evaluation criteria include mean square error and root mean square error.

[0125] The formula for calculating the mean square error is as follows:

[0126] Where N represents the number of samples in the validation set. represents the predicted temperature of the model, Indicates the true temperature;

[0127] The formula for calculating the root mean square error is as follows:

[0128] Here, RMSE is the square root of MSE and represents the standard deviation of the prediction error.

[0129] S234: Based on the performance of the model on the validation set, repeat the training and evaluation until the model performance reaches the optimal level.

[0130] In this embodiment, CNN-LSTM is used to build a lithium-ion battery temperature prediction model, and the dung beetle algorithm is used to optimize the model parameters. In the test phase, a new data sample is input, and the predicted value of the battery temperature is obtained through model prediction. Compared with other prediction methods, it has higher prediction accuracy, lower computational complexity, shorter training time, and better practicality.

[0131] S3. Set two lithium-ion battery phase change thermal runaway warning thresholds, including:

[0132] S31. Setting a warning threshold according to the performance parameters of the target lithium-ion battery and various side reactions induced when phase change thermal runaway occurs.

[0133] In this embodiment, the lithium-ion battery phase change thermal runaway warning threshold is obtained in the following manner:

[0134] S311. Determine key performance parameters and their critical values, including:

[0135] When the battery temperature rises to a certain threshold, it will trigger a thermal runaway reaction; voltage that is too high or too low will cause electrolyte decomposition, metal lithium deposition, etc.; excessive current will cause local overheating, electrolyte decomposition, etc.; overcharging and over-discharging will generate a lot of heat and harmful gases;

[0136] S312. Based on experimental data and literature analysis, determine the critical values ​​of various key parameters, including:

[0137] Temperature critical value: generally around 80-90℃; voltage critical value: 2.5-4.2V; current critical value: 1-2A; charge and discharge state critical value: more than 90% of charging capacity, less than 20% of discharge capacity.

[0138] S313. Side reactions induced by thermal runaway of lithium-ion battery phase change include:

[0139] Electrolyte decomposition: High temperature will cause the electrolyte to decompose and produce flammable gases, which will further increase the internal temperature of the battery;

[0140] Decomposition of electrode materials: Positive and negative electrode materials will undergo chemical decomposition at high temperatures, and the decomposition products will also promote further temperature rise;

[0141] Internal short circuit: The temperature rise causes the internal diaphragm to be damaged, resulting in an internal short circuit. The short circuit will generate a lot of heat, causing the temperature to rise sharply;

[0142] Chain thermal reaction: Once the temperature exceeds the critical value, a chain thermal reaction will occur, and the temperature will quickly rise to over 500°C, which can easily cause the battery to explode;

[0143] Thermal expansion deformation: The rapid rise in temperature inside the battery will cause the thermal expansion and deformation of the material, which will further aggravate the internal short circuit and temperature rise;

[0144] Increased pressure: High-temperature decomposition reactions will produce a large amount of gas, causing the internal pressure of the battery to increase. The increased pressure will also cause mechanical deformation of the battery or even explosion.

[0145] S32. By modeling and simulating the phase change thermal runaway of lithium-ion batteries, the normal temperature change index range of the lithium-ion battery under various normal operating conditions and the abnormal temperature change index range when the phase change thermal runaway phenomenon occurs are obtained, and the threshold range of the lithium-ion battery phase change thermal runaway warning change index is set according to the two temperature change indices.

[0146] For example, in the process of lithium-ion battery discharge causing temperature growth, based on the study of temperature growth during lithium-ion battery discharge, the following reference data on the range of exponential growth can be obtained:

[0147] Table 1: Temperature growth index of lithium-ion batteries under normal conditions

[0148]

[0149] In this embodiment, the mass is 2KG, the specific heat capacity is 1200J / (KG*℃), the convection heat transfer coefficient is 7.5, and the surface area is 480cm 2 The thermal runaway test was conducted on a lithium-ion battery with a maximum temperature rise of 40°C. The experimental results are as follows: At an ambient temperature of about 25°C and a discharge rate of 0.5C-1C, the temperature change of the lithium-ion battery after 0.5h under normal conditions is about 5°C~11°C, and the temperature growth exponent e1 is usually between 1.2-1.5. That is, the temperature growth over time is roughly exponential, and the exponential coefficient is between 1.2-1.5.

[0150] Temperature growth index range during high rate discharge: During high rate discharge at 2C-3C, the temperature change of lithium-ion batteries after 0.5h under normal conditions is about 10℃~17℃, and the temperature growth index e1 can reach 1.5-2.0. Under ultra-high rate discharge of 5C and above, the temperature growth index can exceed 2.0.

[0151] Temperature growth index range in low temperature environment: In low temperature environment below 0℃, due to the increase of internal resistance, the temperature growth index will be larger. For example, in -10℃ environment, the temperature growth index e1 of 0.5C discharge can reach 1.5-2.0.

[0152] In this embodiment, the temperature growth index e1 of the lithium-ion battery during discharge can be set to between 1.2 and 2.0.

[0153] For the case where the temperature of a lithium-ion battery decreases after it stops discharging, based on research data, the following reference data on the range of exponential reduction can be obtained:

[0154] Table 2: Temperature reduction index table of lithium-ion batteries under normal conditions

[0155]

[0156] When the ambient temperature is around 25℃, after stopping the 0.5C-1C discharge, the temperature change of the lithium-ion battery in normal state after 0.5h is about 0℃~8℃, and the temperature reduction index e2 is usually between 0.8-1.0. That is, the temperature decrease over time is roughly in the form of an exponential, and the exponential coefficient is between 0.8-1.0.

[0157] Temperature reduction index range after high rate discharge: For 2C-3C high rate discharge and stop, the temperature reduction index e2 can reach 0.7-0.9. For 5C and above ultra-high rate discharge and stop, the temperature reduction index can be reduced to 0.6-0.8.

[0158] Temperature reduction index range in low temperature environment: In low temperature environment below 0℃, due to poor heat dissipation conditions, the temperature reduction index will be greater. For example, in a -10℃ environment, after stopping 0.5C discharge, the temperature reduction index e2 can reach 0.8-1.0.

[0159] In this embodiment, the temperature change of the lithium-ion battery after 0.5 hours under normal conditions is about 0° C. to 18° C., and the temperature reduction index e2 during the lithium-ion battery stopping discharge process can be set to between 0.6 and 1.0.

[0160] S4. Determine whether the lithium-ion battery will experience thermal runaway based on the real-time temperature, predicted temperature and temperature change index of the target battery, and issue a thermal runaway warning if so.

[0161] When the predicted temperature of the lithium-ion battery reaches the phase change thermal runaway warning threshold or the temperature change index of the lithium-ion battery exceeds the normal range, a warning ring can be used as a warning method. At the same time, the cause of the lithium-ion battery phase change thermal runaway warning can be broadcasted by voice, such as the predicted temperature reaching the phase change thermal runaway warning threshold, the temperature change index is abnormal, etc.

[0162] In this embodiment, by setting multiple warning modes to warn of phase change thermal runaway of lithium-ion batteries, the comprehensiveness of the warning can be improved, the reliability of the warning can be enhanced, the sensitivity of the warning can be optimized, the feedback information of the warning can be enriched, and the intelligence level of the warning can be improved, thereby ensuring the safety of lithium-ion batteries.

[0163] In summary, this embodiment provides an intelligent early warning method for phase change thermal runaway of a lithium-ion battery. By judging whether the temperature change index is normal, it can better distinguish between normal temperature changes and abnormal temperature changes, and can timely discover abnormal temperature changes that may cause phase change thermal runaway, thereby reducing the false alarm rate of the phase change thermal runaway early warning system and improving the accuracy of the early warning. By constructing a CNN-LSTM prediction model based on a multi-head attention mechanism optimized by the dung beetle algorithm, a lithium-ion battery temperature prediction model is constructed using CNN-LSTM, and the model parameters are optimized and trained using the dung beetle algorithm. In the test phase, a new data sample is input, and the predicted value of the battery temperature is obtained through model prediction. Compared with other prediction methods, it has higher prediction accuracy, lower computational complexity, shorter training time, and better practicality. By setting a variety of lithium-ion battery phase change thermal runaway warning thresholds, the timeliness of the early warning can be improved. Compared with relying only on the current temperature value for early warning, the temperature prediction-based method can more timely discover potential thermal runaway risks. It can reduce the probability of thermal runaway. Timely warning can allow the battery management system to take cooling or other protective measures in time, so as to avoid the temperature from rising continuously to the critical point that causes thermal runaway, greatly reduce the probability of thermal runaway accidents, and improve the safety of battery use. It can optimize the charging and discharging strategy. Temperature prediction can help predict the temperature changes of the battery during the charging and discharging process, so as to adjust the charging and discharging current, duration and other parameters to avoid excessive temperature. Under the premise of ensuring performance, the risk of thermal runaway can be minimized. It can extend the service life of the battery. High temperature will accelerate the internal chemical reaction of the battery, causing the battery capacity and power to decay. Through temperature prediction, the use of the battery under high temperature conditions can be better controlled, thereby reducing the thermal aging of the battery and extending its service life. It can improve the battery use experience. When the user uses the battery, he can get early warning about future temperature changes, which helps the user to arrange the battery use plan reasonably, avoid unexpected downtime, and improve the reliability of battery use and the user experience.

[0164] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0165] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. Intelligent early warning method for phase change thermal runaway of lithium-ion batteries, It is characterized in that The steps include: S1: Collect time series data related to lithium-ion battery temperature, including charging current, discharging current, and battery voltage, normalize and preprocess the data, and extract features, including sliding window features, differential features, and frequency domain features; S2: Obtain the real-time temperature, temperature change index and predicted temperature of the lithium-ion battery at the next moment. The specific method is as follows: S21, using a temperature sensor to monitor the temperature of the lithium-ion battery in real time; S22, calculating a temperature change index of the lithium-ion battery within a preset time period according to an increase or decrease in the temperature of the lithium-ion battery; S23: Construct a CNN-LSTM prediction model based on the multi-head attention mechanism optimized by the dung beetle algorithm to predict the target battery temperature data; S3. Setting two lithium-ion battery phase change thermal runaway warning thresholds, including: S31, setting a lithium-ion battery phase change thermal runaway warning threshold according to the performance parameters of the target lithium-ion battery and various side reactions induced when the phase change thermal runaway phenomenon occurs; S32, by modeling and simulating the phase change thermal runaway of the lithium-ion battery, obtaining the normal temperature change index range of the lithium-ion battery under various normal operating conditions and the abnormal temperature change index range when the phase change thermal runaway phenomenon occurs, and setting the lithium-ion battery phase change thermal runaway warning change index threshold range according to the two temperature change indices; S4. Determine whether the lithium-ion battery will experience phase change thermal runaway through the real-time temperature, predicted temperature and temperature change index of the target battery, and issue a thermal runaway warning if so.

2. The intelligent early warning method for phase change thermal runaway of a lithium-ion battery according to claim 1, It is characterized in that In S1, the formula for data normalization is: In the formula, X is the original data value, X min is the minimum value in the data set, X max is the maximum value in the data set, and the data is divided into training set, validation set and test set.

3. The intelligent early warning method for phase change thermal runaway of a lithium-ion battery according to claim 1, It is characterized in that In S1, the formula of sliding window features is expressed as: in, represents the sliding window feature at time t, X t-w+1 , X t-w+2 , ..., X t Represents w data points from time t-w+1 to time t; The formula of differential features is expressed as: in, Represents the differential feature at time t, that is, the difference between the current time and the previous time; The frequency domain features are obtained as follows: in, It represents the frequency domain characteristics at time t, and f() is the fast Fourier transform.

4. The intelligent early warning method for phase change thermal runaway of a lithium-ion battery according to claim 1, It is characterized in that In S22, the lithium-ion battery operating state includes the lithium-ion battery temperature rising, stable and falling states, and the temperature growth index e of the lithium-ion battery in the lithium-ion battery temperature rising state is 1 The calculation formula is as follows: In the formula, It represents the change of battery temperature after time t. represents the maximum temperature rise of the battery, t is the discharge time, m is the mass of the battery, c is the specific heat capacity of the battery, h is the convection heat transfer coefficient of the battery, and A is the surface area of ​​the battery; The calculation formula for the change in battery temperature relative to the ambient temperature when the lithium-ion battery is in a stable temperature state and the lithium-ion battery is in a continuous stable output state is as follows: In the formula, is the change in battery temperature relative to the ambient temperature, I is the discharge current of the battery, R is the internal resistance of the battery, m is the mass of the battery, and c is the specific heat capacity of the battery; The calculation formula for the temperature T of the lithium-ion battery in the standby state in the lithium-ion battery temperature stable state is as follows: Where T is the actual temperature of the battery, is the ambient temperature, It is the fluctuation range of battery temperature relative to normal temperature; The temperature drop index e of the lithium-ion battery when the temperature drops 2 The calculation formula is as follows: In the formula, is the elapsed time t 2 The change in battery temperature after is the maximum temperature drop when the battery stops discharging, t 2 is the time after discharge stops, m is the mass of the battery, c is the specific heat capacity of the battery, h 2 is the convective heat transfer coefficient of the battery, and A is the surface area of ​​the battery.

5. The intelligent early warning method for phase change thermal runaway of a lithium-ion battery according to claim 1, It is characterized in that In S23, the specific steps of predicting the target battery temperature data are: S231: Extract features through convolutional neural networks, model time dependencies through long-term and short-term neural networks, and use multi-head attention mechanisms to enhance the model’s ability to capture key features. S232: Use the dung beetle algorithm to optimize the hyperparameters of the CNN-LSTM model; S233: Use the training set to train the optimized CNN-LSTM model, and use an independent validation set to evaluate the performance of the model. The evaluation criteria include mean square error and root mean square error. S234: Based on the performance of the model on the validation set, repeat the training and evaluation until the model performance reaches the optimal level.

6. The intelligent early warning method for phase change thermal runaway of a lithium-ion battery according to claim 5, It is characterized in that In S231, the convolutional neural network feature extraction method is: in, Represents the features extracted by CNN; The application formula of the multi-head attention mechanism is: Among them, represents the attention weight of the i-th attention head at time t, represents the weight vector of the i-th attention head, Feature representation representing the aggregation of attention mechanisms; The long-term short-term neural network part models time dependence in the following way: in, and Respectively represent the hidden state and cell state of LSTM at time t; The final output is calculated as: in, represents the predicted temperature of the model at time t.

7. The intelligent early warning method for phase change thermal runaway of a lithium-ion battery according to claim 5, It is characterized in that In S232, the dung beetle algorithm optimizes the hyperparameters of the CNN-LSTM model. The dung beetle algorithm simulates the foraging behavior of the dung beetle group and iteratively optimizes the hyperparameter combination to minimize the MSE loss function on the validation set. The formula is: Where N represents the number of samples in the validation set. represents the predicted temperature of the model, Indicates the true temperature.

8. The intelligent early warning method for phase change thermal runaway of a lithium-ion battery according to claim 7, It is characterized in that The core steps of the dung beetle algorithm are as follows: a. Initialize the position and direction of the dung beetle colony; b. Calculate the loss function value of each dung beetle and determine the optimal individual position; c. Update the position and direction of each dung beetle by: in, represents the position of the i-th dung beetle in the j-th dimension, represents the speed of the i-th dung beetle in the j-th dimension, represents the optimal position of the i-th dung beetle in the j-th dimension, represents the best position of the entire group in the jth dimension, w represents the inertia weight, and is the acceleration constant, and Represents a random number; d. Determine whether each dung beetle is beyond the boundary, and if so, expand the search range; e. Update the current optimal solution; f. Repeat steps be until the maximum number of iterations is reached and output the hyperparameters of the CNN-LSTM model.

9. The intelligent early warning method for phase change thermal runaway of a lithium-ion battery according to claim 5, It is characterized in that In S233, the calculation formula of the mean square error is as follows: Where N represents the number of samples in the validation set. represents the predicted temperature of the model, Indicates the true temperature; The formula for calculating the root mean square error is as follows: Here, RMSE is the square root of MSE and represents the standard deviation of the prediction error.

Images