Method for constructing threshold prediction model for lithium battery thermal runaway early warning, lithium battery thermal runaway early warning method, system, equipment and computer program product
By constructing a threshold prediction model for lithium battery thermal runaway early warning and using a linear regression model to dynamically adjust the early warning threshold, the problem of failing to effectively consider battery aging factors in existing technologies is solved, thus improving the accuracy of lithium battery thermal runaway early warning.
Patent Information
- Application Number
- CN202511103507.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-07
- Publication Date
- 2025-11-21
AI Technical Summary
Existing lithium battery thermal runaway early warning methods fail to effectively consider battery aging factors, leading to false alarms or missed alarms and low accuracy.
A threshold prediction model for early warning of thermal runaway in lithium batteries is constructed. By collecting multiple state parameters and thermal runaway characteristic data under different aging conditions, a linear regression model is trained, and the early warning threshold for thermal runaway is dynamically adjusted.
It improves the accuracy of lithium battery thermal runaway early warning and avoids misjudgments caused by normal battery aging.
Smart Images
Figure CN120993232A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of power batteries, in particular to a method for constructing a threshold prediction model for lithium battery thermal runaway early warning, a lithium battery thermal runaway early warning method, a lithium battery thermal runaway early warning system, a device for lithium battery thermal runaway early warning, and a computer program product. BACKGROUND
[0002] Lithium battery thermal runaway early warning has always been a hot and difficult point in the related field. The commonly used method in the prior art is to collect various sensor data such as voltage, temperature, current, and gas related data of the battery management system (BMS), and preset thermal runaway thresholds such as voltage difference, temperature rise rate, and sensitive gas accumulation concentration, and when the collected real-time sensor data triggers the preset thermal runaway threshold, thermal runaway early warning or alarm is performed.
[0003] However, even if thermal runaway does not occur, the lithium battery will cause the voltage difference to increase, the temperature to rise, and the sensitive gas concentration to rise during normal aging due to the increase in internal resistance and the decrease in capacity, etc. If the above-mentioned method of the prior art is used for thermal runaway early warning, false positives or false negatives will occur. That is, due to the lack of consideration of battery aging factors, the accuracy of the traditional lithium battery thermal runaway early warning method needs to be improved. SUMMARY
[0004] The purpose of the present application is to provide a method for constructing a threshold prediction model for lithium battery thermal runaway early warning, a lithium battery thermal runaway early warning method, a lithium battery thermal runaway early warning system, a device for lithium battery thermal runaway early warning, and a computer program product, so as to at least partially solve the above-mentioned problems existing in the prior art.
[0005] According to a first aspect of the present application, a method for constructing a threshold prediction model for lithium battery thermal runaway early warning is provided, which comprises:
[0006] Collecting a plurality of lithium batteries of a predetermined model in different aging states, a plurality of state parameters capable of representing the aging state of the lithium battery, and a thermal runaway characteristic threshold of the corresponding thermal runaway characteristic data of the lithium battery when thermal runaway occurs;
[0007] Obtaining a corresponding thermal runaway early warning threshold based on the thermal runaway characteristic threshold; and
[0008] Constructing a data set with the collected plurality of state parameters as input data and the thermal runaway early warning threshold as output data, and training the threshold prediction model with the data set.
[0009] In an example embodiment, the data set is divided into a training data set and a validation data set, the threshold prediction model is trained using the training data set, and the trained threshold prediction model is validated using the validation data set until a deviation between a thermal runaway early warning threshold output by the trained threshold prediction model and a corresponding thermal runaway early warning threshold in the validation data set is less than a preset threshold, obtaining a trained threshold prediction model.
[0010] In an example embodiment, the plurality of state parameters include a capacity fade rate and / or an internal resistance growth rate.
[0011] In an example embodiment, the thermal runaway characteristic data is a self-generated heat temperature or a characteristic gas release rate of the lithium battery.
[0012] In an example embodiment, the plurality of state parameters capable of representing the aging state of the lithium battery are obtained by charging and discharging a plurality of lithium batteries of the predetermined model at a 1C charging and discharging rate and using a constant current and constant voltage method to cycle to a corresponding SOH.
[0013] In an example embodiment, the thermal runaway characteristic threshold is obtained by adjusting the SOC of a plurality of lithium batteries of the predetermined model in different aging states to 100% and triggering thermal runaway.
[0014] In an example embodiment, the thermal runaway early warning threshold is the thermal runaway characteristic threshold minus a predetermined margin value.
[0015] In an example embodiment, the threshold prediction model is a linear regression model. In particular, the threshold prediction model is:
[0016] Th W = Th R ×(1+K1×P1+K2×P2),
[0017] wherein Th W is a thermal runaway early warning threshold, Th R is a thermal runaway reference threshold, P1 and P2 are the state parameters, respectively, and K1 and K2 are aging coefficients to be fitted, wherein the thermal runaway reference threshold is a thermal runaway early warning threshold corresponding to a fresh lithium battery with an SOH of 100% when thermal runaway occurs at 100% SOC.
[0018] According to a second aspect of the present application, a lithium battery thermal runaway early warning method is provided, which is performed using a threshold prediction model constructed by the method for constructing a threshold prediction model for lithium battery thermal runaway early warning according to the first aspect of the present application, and the lithium battery thermal runaway early warning method comprises:
[0019] acquire the plurality of state parameters of the lithium battery of the predetermined model, which can represent the aging state of the lithium battery, and the thermal runaway characteristic data value corresponding to the thermal runaway characteristic data;
[0020] predict a thermal runaway early warning threshold based on the plurality of state parameters using the threshold prediction model; and
[0021] when the thermal runaway characteristic data value is greater than or equal to the predicted thermal runaway early warning threshold, perform thermal runaway early warning.
[0022] According to a third aspect of the present application, a lithium battery thermal runaway early warning system is provided, comprising:
[0023] a data acquisition module configured to acquire a plurality of state parameters of a lithium battery of a predetermined model, which can represent the aging state of the lithium battery, and a thermal runaway characteristic data value corresponding to the thermal runaway characteristic data;
[0024] a prediction module configured to call a threshold prediction model constructed by the method of constructing a lithium battery thermal runaway early warning threshold prediction model according to the first aspect of the present application to predict a thermal runaway early warning threshold based on the plurality of state parameters; and
[0025] an early warning module configured to compare the thermal runaway characteristic data value with the predicted thermal runaway early warning threshold, and when the thermal runaway characteristic data value is greater than or equal to the predicted thermal runaway early warning threshold, perform thermal runaway early warning.
[0026] According to a fourth aspect of the present application, a device for lithium battery thermal runaway early warning is provided, comprising:
[0027] at least one processor; and
[0028] a memory having program instructions stored therein, which can be executed by the at least one processor, and when the program instructions are executed by the at least one processor, implement the method according to the first aspect or the second aspect of the present application.
[0029] According to a fifth aspect of the present application, a computer program product, for example a computer readable program carrier, contains or stores computer program instructions, which are executed by a processor to at least assist in implementing the steps of the method according to the first aspect or the second aspect of the present application.
[0030] The present application according to the above aspects of the present application has the beneficial effect that as the lithium battery ages, the early warning threshold of thermal runaway can be dynamically adjusted accordingly, avoiding false positives of thermal runaway due to normal aging of the battery, and thus improving the accuracy of lithium battery thermal runaway early warning. BRIEF DESCRIPTION OF DRAWINGS
[0031] The principles, features and advantages of the present application can be better understood by referring to the following detailed description of the application in conjunction with the accompanying drawings in which:
[0032] Figure 1 A flow chart schematically illustrating main steps of a method of constructing a threshold prediction model for lithium battery thermal runaway early warning according to an example embodiment of the present application;
[0033] Figure 2 A flow chart schematically illustrating main steps of a method of constructing a threshold prediction model for lithium battery thermal runaway early warning according to an example embodiment of the present application;
[0034] Figure 3 A schematic block diagram of a lithium battery thermal runaway early warning system according to an example embodiment of the present application is shown;
[0035] Figure 4 A schematic block diagram of a device for lithium battery thermal runaway early warning according to an example embodiment of the present application is shown. DETAILED DESCRIPTION
[0036] In order to make the technical problems to be solved by the present application, the technical solutions and the beneficial technical effects more clearly, the present application will be further described in detail below in conjunction with the drawings and multiple example embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the protection scope of the present application.
[0037] Figure 1 A flow chart schematically illustrating main steps of a method of constructing a threshold prediction model for lithium battery thermal runaway early warning according to an example embodiment of the present application. As shown in Figure 1 the method of constructing a threshold prediction model for lithium battery thermal runaway early warning according to an example embodiment of the present application mainly includes the following steps:
[0038] S101: collecting a plurality of lithium batteries of a predetermined model in different aging states, a plurality of state parameters capable of representing the aging state of the lithium batteries and corresponding thermal runaway characteristic threshold values of the thermal runaway characteristic data of the lithium batteries when thermal runaway occurs;
[0039] S102: obtaining corresponding thermal runaway early warning threshold values based on the thermal runaway characteristic threshold values; and
[0040] S103: constructing a data set by taking the collected plurality of state parameters as input data and taking the thermal runaway early warning threshold values as output data, and training the threshold prediction model using the data set.
[0041] In step S101, the plurality of state parameters can include, for example, the capacity fade rate AC and the internal resistance growth rate AR of the battery. The capacity fade rate AC (= 100% SOH - corresponding SOH) can be directly obtained by charging and discharging a plurality of lithium batteries of a predetermined model at a 1C charging and discharging rate and using a constant current constant voltage (CCCV) method to cycle to the corresponding SOH, and the internal resistance growth rate AR of the battery can be obtained by continuous acquisition, calculation and recording by the BMS (for example, obtained by calculation of the voltage and current values in a certain discharge interval by the BMS). For example, for a plurality of lithium batteries of a certain model, using the CCCV charging and discharging, each battery is charged from 0% to 100% in 1 hour, and then discharged from 100% to 0% in 1 hour, and such charging and discharging operation is cycled until each battery reaches 90% SOH, 70% SOH, etc., thereby obtaining the capacity fade rate AC of the corresponding battery as 10% and 30%, and recording the internal resistance growth rate AR of the corresponding battery as 10% and 28%. Then, each battery in the aging state is charged to 100% SOC, and then a thermal runaway is simulated and triggered, i.e. a thermal runaway experiment is performed on the battery, and the value of the thermal runaway characteristic data at the time of thermal runaway is recorded as the thermal runaway characteristic threshold. The thermal runaway experiment can use methods such as overcharging, overdischarging, needle puncture, extrusion, high temperature, internal short circuit, external short circuit, etc., and in different thermal runaway experiments, the humidity, temperature, power, etc. of the external conditions of the lithium battery in the different SOH states are kept consistent.
[0042] The thermal runaway characteristic data is data that can be used to mark the battery state characteristics at the time of thermal runaway, which can be, for example, the self-generated heat temperature T of the battery (obtained by using a corresponding temperature sensor), or the characteristic gas release rate of the battery, such as the CO release rate Q (obtained by the CO accumulation in a gas sensor), and correspondingly, the thermal runaway characteristic threshold is, for example, the self-generated heat temperature T0 or the CO release rate Q0 of the corresponding battery at the time of thermal runaway. In addition, a set of reference data is also recorded by recording the thermal runaway characteristic threshold of a fresh battery of 100% SOH triggering thermal runaway at 100% SOC.
[0043] In step S102, since early warning is needed before the battery actually reaches thermal runaway, the actual thermal runaway warning threshold should occur earlier than the thermal runaway characteristic threshold, and generally be lower than the thermal runaway characteristic threshold, so that the warning is triggered when the value of the thermal runaway characteristic data is equal to or slightly greater than the thermal runaway warning threshold but less than the thermal runaway characteristic threshold. Therefore, the thermal runaway warning threshold is the thermal runaway characteristic threshold minus a predetermined margin value. The predetermined margin value is different depending on the thermal runaway characteristic data. For example, for the self-generated heat temperature T, the thermal runaway warning threshold T W= thermal runaway characteristic threshold T0-5°C. In addition, for characteristic data such as the CO release rate Q, a corresponding suitable predetermined margin value can be set according to the specific battery model and the variation characteristics of the characteristic data. Among them, the corresponding thermal runaway early warning threshold T W as the thermal runaway reference threshold T R .
[0044] Taking the thermal runaway characteristic data of the self-heat generation temperature T as an example, the data collected in the above-mentioned examples are shown in the following table:
[0045] Table 1
[0046] SOH 100％ 90％ 70％ SOC 100％ 100％ 100％ ΔR 0 10％ 28％ ΔC 0 10％ 30％ [T0] 82℃ 93℃ 113℃ [TECHNICAL FIELD] W ]] 77°C (T R )]]> 88℃ 108℃
[0047] Of course, the collected state parameters are not limited to the two items of capacity decay rate AC and internal resistance growth rate AR, and other characteristic parameters can be additionally or alternatively included, which are not listed here. In addition, the amount of collected data should be appropriately more than that listed in the table, in order to enrich the diversity of the collected data, and also obvious deviation data sets should be removed, so that the data set can be better constructed in the subsequent steps to train the threshold prediction model.
[0048] In step S103, for example, the two state parameters of the capacity decay rate AC and the internal resistance growth rate AR of the above-mentioned battery are taken as input data, and the thermal runaway early warning threshold T W is constructed as output data, and the threshold prediction model is trained with the data set. The threshold prediction model can be a linear regression model. For example, the threshold prediction model can be:
[0049] Th W = Th R × (1 + K1 × P1 + K2 × P2),
[0050] where Th W is the thermal runaway early warning threshold, Th R is the thermal runaway reference threshold, P1 and P2 are state parameters, and K1 and K2 are aging coefficients to be fitted.
[0051] Similarly, taking the example in the above table of the thermal runaway characteristic data of the self-heat generation temperature T, the corresponding threshold prediction model can be:
[0052] T W = T R × (1 + K1 × AR + K2 × AC),
[0053] Similarly, when the thermal runaway characteristic data is the CO release rate Q, the corresponding threshold prediction model can be:
[0054] QW = Q R = Q
[0055] A threshold prediction model can be constructed by selecting one thermal runaway feature data. However, in actual use, in order to be safe, a plurality of thermal runaway feature data can be selected to construct corresponding threshold prediction models, so as to be able to make thermal runaway early warning based on multiple aspects.
[0056] In the training of the threshold prediction model, the data set is divided into a training data set and a validation data set. The training data set is used to train the threshold prediction model and adjust the parameters of the threshold prediction model, i.e., the aging coefficients K1 and K2. The trained threshold prediction model is verified by using the validation data set, until the deviation between the thermal runaway early warning threshold output by the trained threshold prediction model and the corresponding thermal runaway early warning threshold in the validation data set is less than a preset threshold, and the trained threshold prediction model is obtained.
[0057] For example, 80% of the data in the plurality of data sets can be randomly selected as the training data set, and the remaining 20% can be used as the validation data set. Then, the training data set is input into the threshold prediction model, thereby training the threshold prediction model and adjusting the parameters of the threshold prediction model. After training the training data set, the trained threshold prediction model is verified by using the validation data set. Specifically, the input data in the validation data set is input into the threshold prediction model to obtain the output data output by the model, and then the deviation between the output data output by the model and the corresponding output data in the validation data set is calculated. If the deviation is less than or equal to the preset threshold, it can be considered that the trained threshold prediction model has converged, and the training of the threshold prediction model is completed. If there is a deviation greater than the preset threshold, the training round can be increased, i.e., the training data set is input into the threshold prediction model again to train the model, and then the trained model is verified by using the validation data set again, until the deviation between the thermal runaway early warning threshold output by the trained threshold prediction model and the corresponding thermal runaway early warning threshold in the validation data set is less than the preset threshold, thereby obtaining the trained threshold prediction model. Through the above training, the threshold prediction model can effectively converge, and can effectively predict the thermal runaway early warning threshold of the battery.
[0058] The above data division method and training method are only exemplary. For example, if it is found that the model does not converge after one round of training, the validation data set can be merged back into the training data set, and a new validation data set can be divided again, and a new round of training and verification of the model can be performed until the trained model converges. In addition, other known methods are also feasible.
[0059] After obtaining the trained threshold prediction model, the threshold prediction model can be used to actually predict the thermal runaway early warning threshold of the lithium battery of the corresponding model. As shown in Figure 2 The lithium battery thermal runaway early warning method according to one example embodiment of the present application mainly includes the following steps:
[0060] S201: Obtain the plurality of state parameters of the lithium battery of the predetermined model, which can represent the aging state of the lithium battery, and the thermal runaway feature data value corresponding to the thermal runaway feature data;
[0061] S202: Use the trained threshold prediction model to predict the thermal runaway early warning threshold based on the plurality of state parameters; and
[0062] S203: When the thermal runaway feature data value is greater than or equal to the predicted thermal runaway early warning threshold, thermal runaway early warning is performed.
[0063] Still taking the above example, in step S201, the capacity decay rate ΔC, the internal resistance growth rate ΔR of the lithium battery of the predetermined model, and the thermal runaway feature data value corresponding to the thermal runaway feature data based on which the threshold prediction model is trained, such as the battery self-generated heat temperature T and / or the CO release rate Q, can be obtained through the BMS. Then in step S202, the plurality of state parameters are input into the trained threshold prediction model, so that it can predict the corresponding thermal runaway early warning threshold. Next, in step S203, it is determined whether to perform thermal runaway early warning. In one example, for example, ΔC = 18%, ΔR = 15% is collected, and the fitting result of the trained model is K1 = 1 and K2 = 0.4, then the thermal runaway early warning threshold output by the threshold prediction model is:
[0064] T W = 77 x (1 + 1 x 15% + 0.4 x 18%) = 94℃,
[0065] Therefore, when the battery self-generated heat temperature as the thermal runaway feature data value is equal to or greater than 94℃, thermal runaway early warning is performed.
[0066] When the thermal runaway feature data is the CO release rate Q and / or other data, the logic is basically the same.
[0067] Since the plurality of state parameters are associated with the aging state of the battery, by means of the threshold prediction model, the thermal runaway early warning threshold can be appropriately adjusted in combination with the aging state of the battery, thereby avoiding false positives caused by normal aging of the battery and improving the accuracy of thermal runaway early warning.
[0068] Figure 3 A schematic block diagram of a lithium battery thermal runaway early warning system 1 according to one example embodiment of the present application is shown. As shown in Figure 1As shown, the lithium battery thermal runaway early warning system 1 according to one exemplary embodiment of the present application comprises:
[0069] a data acquisition module 11 configured to acquire a plurality of state parameters of a predetermined model of lithium battery, which can represent the aging state of the lithium battery, and thermal runaway feature data values corresponding to thermal runaway feature data;
[0070] a prediction module 12 configured to predict a thermal runaway early warning threshold based on the plurality of state parameters by calling a threshold prediction model constructed by the method of constructing a lithium battery thermal runaway early warning threshold prediction model according to the above method;
[0071] an early warning module 13 configured to compare the thermal runaway feature data values with the predicted thermal runaway early warning threshold, and when the thermal runaway feature data values are greater than or equal to the predicted thermal runaway early warning threshold, a thermal runaway early warning is made.
[0072] Figure 4 A schematic block diagram of the device 2 for lithium battery thermal runaway early warning according to one exemplary embodiment of the present application is shown. As Figure 4 shown, the device 2 for lithium battery thermal runaway early warning according to one exemplary embodiment of the present application comprises:
[0073] at least one processor 21; and
[0074] a memory 22, in which program instructions capable of being executed by the at least one processor 21 are stored, which when executed by the at least one processor 21 realize the method of constructing a lithium battery thermal runaway early warning threshold prediction model according to one exemplary embodiment of the present application or the lithium battery thermal runaway early warning method according to one exemplary embodiment of the present application.
[0075] Although specific implementations of the present application are described in detail herein, they are only given for the purpose of explanation and should not be considered as limiting the scope of the present application. Various substitutions, modifications and alterations can be conceived without departing from the spirit and scope of the present application.
Claims
1. A method for constructing a threshold prediction model for early warning of thermal runaway in lithium batteries, comprising: Collect multiple state parameters that characterize the aging state of the lithium batteries and the corresponding thermal runaway characteristic data of the lithium batteries at the thermal runaway characteristic threshold when thermal runaway occurs from multiple lithium batteries of a predetermined model under different aging conditions. The corresponding thermal runaway early warning threshold is obtained based on the thermal runaway characteristic threshold; and A dataset is constructed using multiple collected state parameters as input data and thermal runaway early warning threshold as output data. The threshold prediction model is then trained using the dataset.
2. The method for constructing a threshold prediction model for early warning of thermal runaway in lithium batteries according to claim 1, wherein, The dataset is divided into a training dataset and a validation dataset. The threshold prediction model is trained using the training dataset and validated using the validation dataset until the deviation between the thermal runaway warning threshold output by the trained threshold prediction model and the corresponding thermal runaway warning threshold in the validation dataset is less than a preset threshold, thus obtaining the trained threshold prediction model.
3. The method for constructing a threshold prediction model for early warning of thermal runaway in lithium batteries according to claim 1 or 2, wherein, The multiple state parameters include capacity decay rate and / or internal resistance growth rate; and / or The thermal runaway characteristic data refers to the self-generated heat temperature or characteristic gas release rate of the lithium battery.
4. The method for constructing a threshold prediction model for early warning of thermal runaway in lithium batteries according to any one of claims 1 to 3, wherein, By cycling multiple lithium batteries of the predetermined model at a charge-discharge rate of 1C using a constant current and constant voltage method to the corresponding state of harmonic equilibrium (SOH), the plurality of state parameters characterizing the aging state of the corresponding lithium batteries are obtained; and / or The thermal runaway characteristic threshold is obtained by adjusting the SOC of multiple lithium batteries of the predetermined model under different aging conditions to 100% and triggering thermal runaway.
5. The method for constructing a threshold prediction model for early warning of thermal runaway in lithium batteries according to any one of claims 1 to 4, wherein, The thermal runaway early warning threshold is the thermal runaway characteristic threshold minus a predetermined margin value.
6. The method for constructing a threshold prediction model for early warning of thermal runaway in lithium batteries according to any one of claims 1 to 5, wherein, The threshold prediction model is a linear regression model, specifically: Th W =Th R ×(1+K1×P1+K2×P2), Among them Th W The threshold for thermal runaway warning, Th R The thermal runaway baseline threshold is defined as follows: P1 and P2 are the state parameters, and K1 and K2 are the aging coefficients to be fitted. The thermal runaway baseline threshold is the thermal runaway warning threshold corresponding to the thermal runaway of a fresh lithium battery with a state of 100% SOH at 100% SOC.
7. A method for early warning of thermal runaway in a lithium battery, wherein the method is executed using a threshold prediction model constructed by the method for constructing a threshold prediction model for early warning of thermal runaway in a lithium battery according to any one of claims 1 to 6, the method comprising: Obtain the multiple state parameters that characterize the aging state of the lithium battery of the predetermined model and the thermal runaway characteristic data values corresponding to the thermal runaway characteristic data; The threshold prediction model is used to predict the thermal runaway early warning threshold based on the multiple state parameters; and A thermal runaway warning is issued when the thermal runaway characteristic data value is greater than or equal to the predicted thermal runaway warning threshold.
8. A lithium battery thermal runaway early warning system (1), comprising: The data acquisition module (11) is configured to acquire multiple state parameters of a predetermined type of lithium battery that can characterize the aging state of the lithium battery and thermal runaway characteristic data values corresponding to thermal runaway characteristic data. Prediction module (12), which is configured to invoke a threshold prediction model constructed by the method of constructing a threshold prediction model for lithium battery thermal runaway warning according to any one of claims 1 to 6 to predict the thermal runaway warning threshold based on the plurality of state parameters; and The early warning module (13) is configured to compare the thermal runaway feature data value with the predicted thermal runaway early warning threshold, and to issue a thermal runaway early warning when the thermal runaway feature data value is greater than or equal to the predicted thermal runaway early warning threshold.
9. A device (2) for early warning of thermal runaway in lithium batteries, comprising: At least one processor (21); and A memory (22) storing program instructions executable by the at least one processor (21), which, when executed by the at least one processor (21), implement the method according to any one of claims 1 to 7.
10. A computer program product, such as a computer-readable program carrier, comprising or storing computer program instructions that, when executed by a processor, at least auxiliaryly implement the steps of the method according to any one of claims 1 to 7.