A Thermal Runaway Early Warning Method for Lithium-Ion Batteries in Fast Charging Scenarios

Abstract

The present invention discloses a method for predicting thermal runaway of lithium-ion batteries for fast charging scenarios, including: constructing an electrochemical-thermal coupling model and a battery thermal runaway decomposition temperature rise model, fusing them to obtain a thermal runaway mechanism model, simulating thermal runaway data to form a first machine learning training data set; designing and conducting battery thermal runaway experiments based on real battery fast charging scenarios to obtain a second machine learning training data set composed of experimental data; constructing a thermal runaway warning model, including a first machine learning model and a second machine learning model, training the first machine learning model based on the deep learning training data set, and training the second machine learning model based on the machine learning training data set; obtaining multiple features under the battery fast charging condition, fusing the prediction results of the first machine learning model and the second machine learning model, and outputting the thermal runaway warning result of the lithium-ion battery.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage management, and specifically to a method for early warning of thermal runaway of lithium-ion batteries for fast charging scenarios. Background Art

[0002] Energy storage systems are divided into two types: fixed energy storage and mobile energy storage. Fixed energy storage is mainly used for energy storage power stations for life support and uninterruptible power supplies for some key facilities. Such energy storage systems can be placed indoors for a certain degree of heat preservation. Mobile energy storage is mainly for outdoor application scenarios, and is mainly used for general starting power supplies for vehicle low-temperature starting or dead battery starting, communication power supplies for communication equipment, and portable power supplies, etc.

[0003] Considering the diverse forms and wide distribution of energy storage systems, the safety of energy storage systems has become a major hidden danger. In this regard, by conducting research on the key technologies related to the fault diagnosis and safety early warning methods of lithium-ion batteries, exploring the variation laws of early physical parameters and the occasional occurrence of special behaviors (such as gas discharge and pressure change) during the whole life cycle and wide temperature range of energy storage lithium-ion batteries, and combining the analysis of characteristic signals before the trigger of battery thermal runaway by multi-sensors, a thermal runaway occurrence model of lithium-ion batteries is constructed, and an accurate early warning method for lithium battery thermal runaway is proposed, which has extremely crucial application value in reducing the explosion risk of lithium-ion batteries and extending the operation life of energy storage systems.

[0004] During the long-term use of lithium-ion batteries, aging and consistency degradation are inevitable phenomena. These degradation problems may lead to overheating and overcharging of single cells, and even trigger thermal runaway events. Thermal runaway may not only occur in a single battery cell, but also easily spread inside the energy storage cabinet and between battery packs. The occurrence of battery thermal runaway is very rapid and often difficult to predict in advance, which makes it difficult for energy storage system managers to obtain reliable early warning signals before the occurrence of thermal runaway, thus bringing serious safety hazards.

[0005] The main inducements for thermal runaway of lithium-ion batteries can be divided into: mechanical abuse, such as puncture or extrusion, which may cause internal short circuit and heat accumulation in the battery; thermal abuse, which may cause the melting of the internal separator of the battery; electrical abuse, such as fast charging, overcharging or over-discharging, which may form dendrites inside the battery and increase the risk of the separator being penetrated. These situations may all lead to internal short circuit of the battery, thus inducing thermal runaway.

[0006] Currently, researchers at home and abroad have explored various routes for battery thermal runaway warning strategies. Taking the evolution of relevant signals such as temperature, gas, impedance, voltage, and pressure as the research object, combined with advanced machine learning methods to conduct early warnings of thermal runaway. The mainstream research methods can be roughly divided into the following three categories: The first category is to monitor the changes in key physical parameters or characteristic information of the battery, and issue a warning after capturing the concerned information. According to the complexity or confidence level of the information, it can be divided into multi-level warnings; the second category is a data-driven thermal runaway warning method; the third category is a model-based thermal runaway warning method.

[0007] In the current research, the thermal runaway warning method mainly relies on the detection of a single signal, but this method is easily affected by noise interference, insufficient detection accuracy, and other faults. Therefore, in order to improve the reliability of thermal runaway warning, it is necessary to develop a warning strategy that comprehensively considers the synergistic effects of multiple signals. Whether it is a model-based method or a data-driven method, the core of early thermal runaway warning is to monitor the abnormal changes of characteristic parameters through the data collected by the BMS, so as to provide more time for dealing with thermal runaway problems. Summary of the Invention

[0008] Aiming at the deficiencies in the prior art that the thermal runaway warning strategy of lithium-ion batteries lacks the support of a mechanism model, relies on single or a small number of characteristics, has a narrow applicability, and the warning target is not clear enough, the present invention provides a thermal runaway warning method for lithium-ion batteries in a fast charging scenario. Facing the fast charging scenario, based on the fact that high-rate charging will cause the performance such as the capacity and output power of the battery to decay rapidly, and a large amount of heat generated during charging is difficult to dissipate evenly and effectively, which will also induce battery degradation and reduced safety, etc. Combining the fusion of multi-dimensional parameters and multi-source characteristic signals, early warning of lithium-ion battery thermal runaway is realized.

[0009] To achieve the above object, the present invention provides a thermal runaway warning method for lithium-ion batteries in a fast charging scenario, including the following steps:

[0010] Step 1, construct an electrochemical-thermal coupling model under the normal operating conditions of the battery and a battery thermal runaway decomposition and temperature rise model under the thermal runaway state;

[0011] Step 2, fuse the electrochemical-thermal coupling model and the battery thermal runaway decomposition and temperature rise model to obtain a thermal runaway mechanism model, and simulate the thermal runaway data of batteries in different states based on the thermal runaway mechanism model to form a first machine learning training data set, where the first machine learning training data set is a deep learning training data set;

[0012] Step 3, design and carry out a battery thermal runaway experiment based on the real battery fast charging scenario, and measure multi-dimensional characteristics during the experiment to obtain a second machine learning training data set composed of experimental data;

[0013] Step 4: Construct a thermal runaway warning model. The thermal runaway warning model includes multiple first machine learning models and multiple second machine learning models. Train the first machine learning models based on the first machine learning training dataset, and train the second machine learning models based on the second machine learning training dataset.

[0014] Step 5: Obtain the multi - element features of the battery under fast charging conditions and input them into the trained thermal runaway warning model. After fusing the prediction results of each first machine learning model and each second machine learning model, output the thermal runaway warning result of the lithium - ion battery.

[0015] In one embodiment, in Step 1, the electro - chemical - thermal coupling model adopts the form of combining the Newman model or the pseudo - two - dimensional model and the solid heat transfer model, and uses the hybrid power pulse characteristic working condition for parameter identification.

[0016] The battery thermal runaway decomposition temperature - rising model adopts the Hatchard decomposition temperature - rising model, and uses the adiabatic thermal runaway data of the battery tested by an adiabatic calorimeter for model parameter identification.

[0017] In one embodiment, in Step 2, the specific process of fusing the electro - chemical - thermal coupling model and the battery thermal runaway decomposition temperature - rising model to obtain a thermal runaway mechanism model is as follows:

[0018] Expand the boundary of the electro - chemical - thermal coupling model, and expand its applicable temperature boundary to the starting temperature point of the thermal runaway process of the Hatchard decomposition temperature - rising model, that is, use the internal physical parameters of the electro - chemical - thermal coupling model as the initial state parameters of the Hatchard decomposition temperature - rising model to conduct the simulation of the thermal runaway process.

[0019] In one embodiment, the characteristic data in the first machine learning training dataset includes the voltage, current, capacity, temperature, and internal physical parameters of the battery.

[0020] The characteristic data in the second machine learning training dataset includes the voltage, temperature, impedance, gas composition, and expansion force change characteristic signals of the battery.

[0021] In one embodiment, the input data of the first machine learning model includes the voltage, current, capacity, temperature, and internal physical parameters of the battery, and the output is the result of whether the battery has thermal runaway.

[0022] The input data of the second machine learning model includes the voltage, temperature, impedance, gas composition, and expansion force change characteristic signals of the battery, and the output is the result of whether the battery has thermal runaway.

[0023] In one embodiment, in the first machine learning model and the second machine learning model, temperature points with a temperature rise rate higher than 1 °C / s and lasting for more than 3 s are used as battery thermal runaway labels.

[0024] In one embodiment, step 5 is specifically as follows:

[0025] Under the fast charging condition of the battery, obtain its multiple features, and the multiple features include the voltage, current, capacity, temperature, internal physical parameters, impedance, gas composition, and expansion force change characteristic signals of the battery;

[0026] Respectively input the voltage, current, capacity, temperature, and internal physical parameters of the battery in the multiple features into each trained first machine learning model, and obtain the prediction results of each first machine learning model;

[0027] Respectively input the voltage, temperature, impedance, gas composition, and expansion force change characteristic signals of the battery in the multiple features into each trained second machine learning model, and obtain the prediction results of each second machine learning model;

[0028] Integrate the prediction results of each first machine learning model and each second machine learning model on whether the battery is thermally out of control, and take the item with a larger quantity as the final thermal runaway warning result of the lithium-ion battery for output.

[0029] Compared with the prior art, the present invention has the following beneficial technical effects:

[0030] 1. Compared with the method of relying only on single-dimensional features or artificially marked warning signals in the past, the present invention can realize the early and effective warning of lithium-ion battery thermal runaway by jointly extracting multiple features in the information-physical domain and constructing a model, and using an ensemble learning strategy to fuse the base models to construct a thermal runaway warning model;

[0031] 2. The present invention combines the mechanism model of the whole process of battery thermal runaway evolution, triggers from the underlying principle of fast-charge-induced thermal runaway, and forms an ensemble learning framework by fusing multiple base models trained with a thermal runaway simulation label data set and a thermal runaway experimental data set containing multiple features, greatly improving the robustness of the warning determination of thermal runaway;

[0032] 3. The present invention can perform the current thermal runaway warning determination of the battery based on the offline-trained thermal runaway warning model and in combination with the current fast-charging condition battery operation data containing multiple feature signals, and has strong practicality. Description of the Drawings

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the structures shown in these drawings.

[0034] Figure 1 It is a flowchart of the method for predicting thermal runaway of lithium-ion batteries for fast charging scenarios in the embodiments of the present invention;

[0035] Figure 2 It is a flowchart for constructing the thermal runaway mechanism model in the embodiments of the present invention;

[0036] Figure 3 It is a flowchart for extracting early multi-source features of information-physical domain joint thermal runaway in the embodiments of the present invention;

[0037] Figure 4 It is a flowchart for constructing and training the thermal runaway prediction model in the embodiments of the present invention;

[0038] Figure 5 It is a flowchart for early warning of thermal runaway for fast charging in the embodiments of the present invention.

[0039] The realization of the objectives, functional features and advantages of the present invention will be further described in conjunction with the embodiments and with reference to the drawings. Detailed implementation manners

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0041] It should be noted that all directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the drawings). If the specific posture changes, the directional indications will also change accordingly.

[0042] In addition, in the present invention, descriptions such as "first" and "second" are for descriptive purposes only, and should not be construed as indicating or implying their relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0043] In the present invention, unless otherwise clearly defined and limited, terms such as "connection" and "fixation" shall be understood in a broad sense. For example, "fixation" may be a fixed connection, a detachable connection, or integrated; it may be a mechanical connection, an electrical connection, a physical connection or a wireless communication connection; it may be directly connected, or indirectly connected through an intermediate medium, and may be the communication inside two components or the interaction relationship between two components, unless otherwise clearly limited. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0044] In addition, the technical solutions between various embodiments of the present invention can be combined with each other, but it must be based on the realization by those of ordinary skill in the art. When the combination of technical solutions is contradictory or cannot be realized, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0045] Aiming at the problems of the existing thermal runaway warning strategy for lithium-ion batteries lacking the support of a mechanism model, relying on single or a small number of features, narrow applicability, and unclear warning targets, this embodiment discloses a thermal runaway warning method for lithium-ion batteries in a fast charging scenario. By integrating a conventional electrochemical model, a classical thermal model, and a Hatchard electrochemical decomposition heating model after reaching a certain temperature threshold, a thermal runaway mechanism model capable of describing the whole process of thermal runaway caused by fast charging is constructed. At the same time, a fast charging-induced thermal runaway experiment in a real scenario is designed and carried out to parameterize and verify the constructed thermal runaway mechanism model, and in this process, a variety of characteristic signals are measured to obtain a physical experimental data set with multiple features. Subsequently, the first machine learning model is trained by the simulation data set obtained from the thermal runaway mechanism model, the second machine learning model is trained by the small-sample physical experimental data set, and these base models are combined to form an ensemble learning framework to complete the construction of the thermal runaway warning model. By introducing the multi-feature information in the early stage of thermal runaway and combining the ensemble learning method composed of multiple warning models, the early warning time of thermal runaway of lithium-ion batteries in the fast charging scenario is greatly improved, and the warning accuracy is also improved.

[0046] Reference Figure 1 , the thermal runaway warning method for lithium-ion batteries in a fast charging scenario in this embodiment specifically includes the following steps:

[0047] Step 1, construct an electrochemical-thermal coupling model under normal operating conditions of the battery and a battery thermal runaway decomposition and temperature rise model under thermal runaway conditions;

[0048] Step 2, fuse the electrochemical-thermal coupling model with the battery thermal runaway decomposition and temperature rise model to obtain a thermal runaway mechanism model, and simulate the thermal runaway data of batteries in different states based on the thermal runaway mechanism model to form a first machine learning training dataset, where the first machine learning training dataset is a deep learning training dataset;

[0049] Step 3, conduct a battery thermal runaway experiment based on the design of a real battery fast charging scenario, and measure multi-dimensional features during the experiment to obtain a second machine learning training dataset composed of experimental data;

[0050] Step 4, construct a thermal runaway warning model, the thermal runaway warning model includes multiple first machine learning models and multiple second machine learning models, and offline train the first machine learning models based on the first machine learning training dataset, and offline train the second machine learning models based on the second machine learning training dataset;

[0051] Step 5, obtain its multi-dimensional features under the battery fast charging condition, input them into the trained thermal runaway warning model, and after fusing the prediction results of each first machine learning model and each second machine learning model, output the thermal runaway warning result of the lithium-ion battery.

[0052] Current thermal runaway warning research and practical applications lack the construction of battery models in the early stage of thermal runaway development. In addition, the fusion of the underlying electrochemical model during normal battery operation and the thermal runaway process model during thermal runaway has not been effectively studied yet. For this reason, this embodiment realizes the full-process description of thermal runaway caused by fast charging through the construction of a mechanism model for the battery developing from a normal operating state in the early stage of thermal runaway to a thermal runaway state and the effective fusion of multiple models.

[0053] In this embodiment, the electrochemical-thermal coupling model adopts the form of combining the Newman model or the pseudo-two-dimensional model with the solid heat transfer model, and uses the hybrid power pulse characteristic working condition for parameter identification. The battery thermal runaway decomposition and temperature rise model adopts the Hatchard decomposition and temperature rise model, and uses the adiabatic thermal runaway data of the battery tested by an adiabatic calorimeter for model parameter identification. Among them, the parameter identification method can adopt conventional particle swarm optimization algorithms, nonlinear least squares methods or other intelligent optimization algorithms, etc. Before parameter identification, sensitivity analysis can also be used to identify the parameters that have the greatest impact on the model output, so as to determine which parameters need to be identified and their identification order.

[0054] Reference Figure 2, the fusion of the electrochemistry-thermal coupling model and the battery thermal runaway decomposition temperature rise model to obtain the thermal runaway mechanism model is specifically as follows: Expand the boundary of the electrochemistry-thermal coupling model considering internal physical parameters, and expand its applicable temperature boundary to the starting temperature point of the thermal runaway process of the Hatchard decomposition temperature rise model. That is, use the internal physical parameters (such as the concentration of each material component) of the electrochemistry-thermal coupling model as the initial state parameters of the Hatchard decomposition temperature rise model to simulate the thermal runaway process.

[0055] In this embodiment, the first machine learning training dataset obtained through simulation and the second machine learning training dataset obtained through actual experiments jointly complete the training of the thermal runaway warning model, realizing the information-physical domain joint, and greatly improving the robustness of the warning judgment of thermal runaway.

[0056] Reference Figure 3 , for the generation of the thermal runaway simulation label dataset, in this embodiment, a large number of thermal runaway simulation curves of batteries in various states are generated by using the Monte Carlo sampling method within the given variable parameter range, and the temperature points with a temperature rise rate dT / dt≥1℃ / s and lasting for more than 3s are used as the battery thermal runaway labels. Among them, the characteristic data in the thermal runaway simulation label dataset includes the voltage, current, capacity, temperature, and internal physical parameters of the battery. It should be noted that the Monte Carlo method is used in this embodiment to ensure the uniformity of the sample distribution when the number of samples is small, but in the specific application process, it is not limited to only using the Monte Carlo method for sampling, and direct random sampling and other sampling methods are also applicable. For the generation of the thermal runaway experimental label dataset, mainly collect multiple characteristic signals of the battery in the real-scene thermal runaway experiment. Among them, the characteristic data in the thermal runaway experimental label dataset includes the voltage, temperature, impedance, gas composition, and expansion force change characteristic signals of the battery.

[0057] After the thermal runaway simulation label dataset and the thermal runaway experimental label dataset are generated, the multi-dimensional features of the thermal runaway simulation label dataset and the thermal runaway experimental label dataset can be screened. In this embodiment, the Kullback-Leibler (K-L) test method is used to perform normal distribution and difference tests on the extracted features to screen out the features that can effectively correlate with the thermal runaway state, so as to form the deep learning training dataset and the machine learning training dataset finally used for the training of the warning model. Through this feature screening process, the effectiveness of the selected features in subsequent analysis is ensured, the algorithm efficiency is improved, and the accuracy and reliability of the model are improved. It should be noted that in the specific application process, it is not limited to only using the Kullback-Leibler (K-L) test method for feature screening, and Pearson and Spearman correlation coefficients and other feature screening methods can also be used.

[0058] The thermal runaway warning model in this embodiment consists of multiple first machine learning models and multiple second machine learning models. Among them, for the construction and training of the first machine learning model, the time series historical input data features in the first machine learning training dataset are extracted through a neural network and the non-linearity and long-term dependence relationships of the data are processed. A corresponding loss function is used to calculate the loss between the network output and the true label, and the model parameters are adjusted through the backpropagation algorithm to gradually reduce the loss. Since the feature dimensions of the thermal runaway experiment label data in the second machine learning training dataset are higher and the samples are fewer, a binary classification machine learning model is adopted in this embodiment, and the model evaluation performance is optimized by selecting appropriate optimizers (such as SGD, Adam, etc.) and adjusting model hyperparameters (such as learning rate, regularization parameter, depth of the tree, etc.).

[0059] After completing the training of all the first machine learning models and the second machine learning models, the prediction results of each base model can be combined and fused to obtain the final thermal runaway warning result of the lithium-ion battery. Among them, the fusion method can be simple (such as average, weighted average, voting), or complex (such as learned combination). The final ensemble model is evaluated using the test set, and metrics such as accuracy, precision, recall, and F1 score are calculated. According to the evaluation results, it may be necessary to return to the model training stage to further adjust the base model or the ensemble strategy to obtain better performance. It should be noted that common ensemble learning strategies, including but not limited to the three major families of algorithms and their sub-algorithms such as Bagging, Boosting, and Stacking, can be applied to the method of this embodiment.

[0060] Reference Figure 5 , the process of using the trained thermal runaway warning model to give a thermal runaway warning to the lithium-ion battery in this embodiment is specifically as follows:

[0061] First, under the fast charging condition of the battery, its multivariate features are obtained. The multivariate features include the voltage, current, capacity, temperature, internal physical parameters, impedance, gas composition, and expansion force change characteristic signals of the battery;

[0062] Then, the voltage, current, capacity, temperature, and internal physical parameters of the battery in the multivariate features are respectively input into each trained first machine learning model, and the prediction results of each first machine learning model are obtained;

[0063] At the same time, the voltage, temperature, impedance, gas composition, and expansion force change characteristic signals of the battery in the multivariate features are respectively input into each trained second machine learning model, and the prediction results of each second machine learning model are obtained;

[0064] Finally, the prediction results of each first machine learning model and each second machine learning model regarding whether the battery is thermally out of control are integrated, and the item with a larger quantity is used as the final lithium-ion battery thermal runaway warning result for output, that is, the prediction results of all models are fused using the voting method.

[0065] It should be noted that although the steps in this embodiment Figures 1 to 5 are shown in sequence according to the indication of the arrows, these steps do not necessarily have to be executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, Figures 1 to 5 at least a part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages do not necessarily have to be executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages does not necessarily have to be sequential, but can be executed alternately or in rotation with at least a part of other steps or sub-steps or stages of other steps.

[0066] The above are only the preferred embodiments of the present invention, and do not limit the protection scope of the present invention. Any equivalent structural transformation made under the inventive concept of the present invention by using the content of the specification and drawings of the present invention, or directly / indirectly applied to other related technical fields, is included in the protection scope of the present invention.

Claims

1. A lithium-ion battery thermal runaway early warning method for fast charging scenarios, characterized in that: The steps include: Step 1, constructing an electrochemical-thermal coupling model under normal operating conditions of the battery and a battery thermal runaway decomposition temperature rise model under thermal runaway conditions; Step 2, fusing the electrochemical-thermal coupling model with the battery thermal runaway decomposition and temperature rise model to obtain a thermal runaway mechanism model, and simulating thermal runaway data of batteries in different states based on the thermal runaway mechanism model to form a first machine learning training data set, wherein the first machine learning training data set is a deep learning training data set; Step 3: Design and conduct a battery thermal runaway experiment based on a real battery fast charging scenario, and measure multi-dimensional features during the experiment to obtain a second machine learning training data set consisting of experimental data; Step 4: construct a thermal runaway warning model, wherein the thermal runaway warning model includes a plurality of first machine learning models and a plurality of second machine learning models, and the first machine learning model is trained based on the first machine learning training data set, and the second machine learning model is trained based on the second machine learning training data set; Step 5, obtain its multivariate features under the battery fast charging condition, and input the trained thermal runaway warning model, and after fusing the prediction results of each of the first machine learning models and each of the second machine learning models, output the thermal runaway warning result of the lithium-ion battery.

2. The lithium-ion battery thermal runaway early warning method for fast charging scenarios according to claim 1, characterized in that: In step 1, the electrochemical-thermal coupling model adopts a combination of a Newman model or a pseudo two-dimensional model and a solid heat transfer model, and uses a mixed power pulse characteristic working condition for parameter identification; The battery thermal runaway decomposition and heating model adopts the Hatchard decomposition and heating model, and the parameters of the model are identified using the battery adiabatic thermal runaway data tested by an adiabatic calorimeter.

3. The lithium-ion battery thermal runaway early warning method for fast charging scenarios according to claim 2, characterized in that: In step 2, the electrochemical-thermal coupling model is integrated with the battery thermal runaway decomposition temperature rise model to obtain a thermal runaway mechanism model, specifically: The boundary of the electrochemical-thermal coupling model is expanded, and its applicable temperature boundary is expanded to the starting temperature point of the thermal runaway process of the Hatchard decomposition and heating model, that is, the internal physical parameters of the electrochemical-thermal coupling model are used as the initial state parameters of the Hatchard decomposition and heating model to simulate the thermal runaway process.

4. The lithium-ion battery thermal runaway early warning method for fast charging scenarios according to claim 1, 2 or 3, characterized in that: The feature data in the first machine learning training data set includes voltage, current, capacity, temperature and internal physical parameters of the battery; The characteristic data in the second machine learning training data set includes characteristic signals of battery voltage, temperature, impedance, gas composition, and expansion force change.

5. The lithium-ion battery thermal runaway early warning method for fast charging scenarios according to claim 4, characterized in that: The input data of the first machine learning model includes the voltage, current, capacity, temperature and internal physical parameters of the battery, and the output is the result of whether the battery is in thermal runaway; The input data of the second machine learning model includes the battery's voltage, temperature, impedance, gas composition, and expansion force change characteristic signals, and the output is the result of whether the battery is in thermal runaway.

6. The lithium-ion battery thermal runaway early warning method for fast charging scenarios according to claim 5, characterized in that: In both the first machine learning model and the second machine learning model, the temperature point at which the temperature rise rate is higher than 1°C / s and lasts for more than 3s is used as the battery thermal runaway label.

7. The lithium-ion battery thermal runaway early warning method for fast charging scenarios according to claim 5, characterized in that: Step 5 is as follows: Obtain the multi-dimensional characteristics of the battery under fast charging conditions. The multivariate characteristics include voltage, current, capacity, temperature, internal physical parameters, impedance, gas composition and expansion force change characteristic signals of the battery; Inputting the voltage, current, capacity, temperature and internal physical parameters of the battery in the multivariate features into each trained first machine learning model, and obtaining prediction results of each first machine learning model; Inputting the voltage, temperature, impedance, gas composition and expansion force change characteristic signals of the battery in the multivariate characteristics into the second machine learning models after each training is completed, and obtaining the prediction results of each of the second machine learning models; The prediction results of the first machine learning models and the second machine learning models on whether the battery is in thermal runaway are combined, and the item with a larger number is output as the final lithium-ion battery thermal runaway warning result.

Images