Battery detection method, prediction model training method, and related devices

Abstract

The application provides a battery detection method, a prediction model training method and related equipment, the method comprising: obtaining parameter information of a battery of a vehicle in a current time period, the parameter information comprising voltage, current, temperature and resistance of the battery; determining each first index value corresponding to the current time period according to the parameter information, and performing a preset operation on the same type of first index value to obtain each second index value; inputting each first index value and each second index value into a prediction model, and determining whether the battery has a thermal runaway risk according to the output information of the prediction model. In the application, each first index value including mean, variance, extreme value, kurtosis and skewness is determined through the temperature, voltage, resistance and current of the battery, each second index value is derived based on each first index value, that is, the thermal runaway of the battery is predicted through multiple parameters, and the detection accuracy of the thermal runaway of the battery is improved.

Description

Technical Field

[0001] This application relates to the field of battery technology, and in particular to a battery detection method, a prediction model training method, and related equipment. Background Technology

[0002] With the rise of new energy vehicles, batteries have become an indispensable core component, directly affecting the overall vehicle performance and safety.

[0003] During vehicle operation, the vehicle's battery may experience thermal runaway. Thermal runaway refers to the loss of vehicle control due to battery overheating. To ensure personal safety and minimize property damage, thermal runaway testing of the vehicle's battery is necessary.

[0004] In an exemplary technology, thermal runaway detection of a battery is achieved by monitoring the voltage, current, temperature, and resistance of the battery in a vehicle.

[0005] However, the above-mentioned methods for detecting thermal runaway use fewer parameters, resulting in lower detection accuracy. Summary of the Invention

[0006] This application provides a battery detection method, a prediction model training method, and related equipment to solve the problem of low detection accuracy of battery thermal runaway.

[0007] In a first aspect, this application provides a battery testing method, comprising:

[0008] Obtain parameter information of the vehicle's battery within the current time period, including battery voltage, current, temperature, and resistance;

[0009] Based on the parameter information, determine the first index values ​​corresponding to the current time period, and perform preset calculations on the first index values ​​of the same type to obtain the second index values. The first index values ​​include parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness.

[0010] Each of the first index values ​​and each of the second index values ​​are input into the prediction model, and the presence of thermal runaway risk in the battery is determined based on the output information of the prediction model.

[0011] In one embodiment, the step of determining each first indicator value corresponding to the current time period based on the parameter information, and performing a preset operation on the first indicator values ​​of the same type to obtain each second indicator value includes:

[0012] The current time period is divided into multiple sub-time periods, and the sub-parameter information corresponding to each sub-time period is determined according to the parameter information;

[0013] Determine the values ​​of each first index corresponding to the sub-parameter information based on the sub-parameter information;

[0014] The difference between first index values ​​of the same type in the adjacent sub-time periods is determined as the second index value.

[0015] In one embodiment, the step of obtaining the vehicle's battery parameter information within the current time period includes:

[0016] Obtain the operating parameter information corresponding to the battery within the current time period;

[0017] The abnormal parameters in the operating parameter information are processed to obtain the parameter information of the battery within the current time period.

[0018] In a second aspect, this application provides a prediction model training method for training the prediction model as described above, the prediction model training method comprising:

[0019] Multiple positive samples are obtained. The positive samples include the first parameter values ​​and the second parameter values ​​of batteries that have not experienced thermal runaway during a historical time period. The second parameter values ​​are obtained by processing the first parameter values ​​of the same type through a preset calculation. The first parameter values ​​include parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness.

[0020] Multiple negative samples are obtained. The negative samples include the values ​​of each third parameter and each fourth parameter of the battery that has experienced thermal runaway during the runaway period. The fourth parameter values ​​are obtained by processing the same type of third parameter values ​​through a preset operation. The third parameter values ​​include parameters determined by voltage, current, temperature and resistance.

[0021] The preset model is trained based on each of the positive samples and each of the negative samples to obtain the prediction model.

[0022] In one embodiment, before the step of training a preset model based on the positive samples and the negative samples to obtain a prediction model, the method further includes:

[0023] Configure the initial network parameters for the random forest model;

[0024] Based on the value range of each initial network parameter, a grid search is performed on the optimal network parameters of the random forest model to determine the target network parameters of the random forest model;

[0025] The initial network parameters of the random forest model are adjusted to the target network parameters to obtain the preset model.

[0026] In one embodiment, the step of training the preset model based on each of the positive samples and each of the negative samples includes:

[0027] Each of the positive samples and each of the negative samples is divided into a training set and a validation set;

[0028] Based on the training set, the preset model is trained to obtain the model to be determined, and the first accuracy of the model to be determined in predicting battery thermal runaway is obtained.

[0029] Based on the validation set, the model to be determined is validated to obtain the second accuracy of the model to be determined in predicting battery thermal runaway.

[0030] If the first accuracy rate is greater than the first preset threshold, and the difference between the first accuracy rate and the second accuracy rate is less than the preset difference, then the model to be determined is stored as a prediction model.

[0031] In one embodiment, after the step of validating the model to be determined based on the validation set, the method further includes:

[0032] When both the first accuracy and the second accuracy are less than the second preset threshold, the regularization parameter of the model to be determined is increased to obtain the first model to be trained, and the second preset threshold is less than the first preset threshold.

[0033] Add parameter values ​​to the samples in the training set to obtain the first training set. The added parameter values ​​are determined based on the existing parameter values ​​of the samples in the training set.

[0034] The prediction model is obtained by training the first model to be trained based on the first training set.

[0035] In one embodiment, after the step of validating the model to be determined based on the validation set, the method further includes:

[0036] When the first accuracy is greater than a first preset threshold, the first accuracy is greater than the second accuracy, and the difference between the first accuracy and the second accuracy is greater than or equal to a preset difference, the probability of the tree in the model to be determined is reduced to obtain a second model to be trained. The model to be determined is a random forest model, and the probability is used to indicate the probability that the tree uses the index in the sample.

[0037] The second model to be trained is trained using the training set to obtain the prediction model.

[0038] In one embodiment, after the step of validating the model to be determined based on the validation set, the method further includes:

[0039] When the first accuracy is greater than the first preset threshold, the first accuracy is greater than the second accuracy, and the difference between the first accuracy and the second accuracy is greater than or equal to a preset difference, the regularization parameter in the model to be determined is increased to obtain a third model to be trained.

[0040] The prediction model is obtained by training the third model to be trained based on the training set.

[0041] In one embodiment, after the step of validating the model to be determined based on the validation set, the method further includes:

[0042] When the first accuracy is greater than a first preset threshold, the first accuracy is greater than the second accuracy, and the difference between the first accuracy and the second accuracy is greater than or equal to a preset difference, positive samples and negative samples are added to the training set to obtain a second training set.

[0043] The prediction model is obtained by training the model to be determined based on the second training set.

[0044] Thirdly, this application provides a battery testing device, comprising:

[0045] The first acquisition module is used to acquire parameter information of the vehicle's battery in the current time period, including the battery's voltage, current, temperature and resistance.

[0046] The determination module is used to determine each first indicator value corresponding to the current time period based on the parameter information, and to perform preset calculations on the first indicator values ​​of the same type to obtain each second indicator value. The first indicator value includes parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness.

[0047] The input module is used to input each of the first index values ​​and each of the second index values ​​into the prediction model, and determine whether the battery has a risk of thermal runaway based on the output information of the prediction model.

[0048] Fourthly, this application provides a predictive model training apparatus, comprising:

[0049] The second acquisition module is used to acquire multiple positive samples. The positive samples include various first parameter values ​​and various second parameter values ​​of batteries that have not experienced thermal runaway during a historical time period. The second parameter values ​​are obtained by processing the same type of first parameter values ​​through a preset calculation. The first parameter values ​​include parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness.

[0050] The third acquisition module is used to acquire multiple negative samples. The negative samples include the values ​​of each third parameter and each fourth parameter of the battery that has experienced thermal runaway during the runaway period. The fourth parameter values ​​are obtained by processing the same type of third parameter values ​​through a preset operation. The third parameter values ​​include parameters determined by voltage, current, temperature and resistance.

[0051] The training module is used to train the preset model based on each of the positive samples and each of the negative samples to obtain the prediction model.

[0052] Fifthly, this application provides a terminal device, including: a memory and a processor;

[0053] The memory stores computer-executed instructions;

[0054] The processor executes computer execution instructions stored in the memory, causing the battery detection device to perform the battery detection method or the prediction model training method as described above.

[0055] Sixthly, this application provides a vehicle, including: a memory and a processor;

[0056] The memory stores computer-executed instructions;

[0057] The processor executes computer execution instructions stored in the memory, causing the battery detection device to perform the battery detection method or the prediction model training method as described above.

[0058] In a seventh aspect, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the battery detection method or the prediction model training method described above.

[0059] The battery detection method, prediction model training method, and related equipment provided in this application acquire parameter information such as voltage, current, temperature, and resistance of a vehicle's battery within a current time period. Based on the parameter information, the application determines various first indicator values ​​corresponding to the current time period. Then, it performs preset calculations on similar first indicator values ​​to obtain various second indicator values. Inputting these first and second indicator values ​​into the prediction model determines whether the battery has a risk of thermal runaway. In this application, various first indicator values, including mean, variance, extreme values, kurtosis, and skewness, are determined based on the battery's temperature, voltage, resistance, and current. Various second indicator values ​​are then derived from these first indicator values. In other words, multiple parameters are used to predict battery thermal runaway, improving the accuracy of battery thermal runaway detection. Attached Figure Description

[0060] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0061] Figure 1 This is a flowchart illustrating the first embodiment of the battery testing method of this application;

[0062] Figure 2 This is a flowchart illustrating the second embodiment of the battery testing method of this application;

[0063] Figure 3 This is a diagram illustrating the division of the time periods covered in this application;

[0064] Figure 4 This is a flowchart illustrating the first embodiment of the prediction model training method of this application;

[0065] Figure 5 This is a flowchart illustrating the second embodiment of the prediction model training method of this application;

[0066] Figure 6 This is a flowchart illustrating the third embodiment of the prediction model training method of this application;

[0067] Figure 7 This is a schematic diagram of the functional modules of the battery testing device of this application;

[0068] Figure 8 This is a schematic diagram of the functional modules of the prediction model training device of this application;

[0069] Figure 9 This is a schematic diagram of the hardware structure of the terminal equipment / vehicle in this application.

[0070] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0071] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0072] With the rise of new energy vehicles, batteries have become an indispensable core component, directly affecting the overall vehicle performance and safety.

[0073] During vehicle operation, the vehicle's battery may experience thermal runaway. Thermal runaway refers to the loss of vehicle control due to battery overheating. To ensure personal safety and minimize property damage, thermal runaway testing of the vehicle's battery is necessary.

[0074] In an exemplary technology, thermal runaway detection of a battery is achieved by monitoring the voltage, current, temperature, and resistance of the battery in a vehicle.

[0075] The inventors of this patent discovered that the detection method for thermal runaway uses fewer parameters and has a lower detection accuracy.

[0076] The inventor of this patent therefore conceived of determining various first index values, including mean, variance, extreme values, kurtosis, and skewness, based on the temperature, voltage, resistance, and current of the battery. Then, various second index values ​​are derived based on each of the first index values. In other words, the thermal runaway of the battery is predicted through multiple parameters, thereby improving the detection accuracy of battery thermal runaway.

[0077] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0078] Reference Figure 1 , Figure 1 This is a first embodiment of the battery testing method of this application. The battery testing method includes the following steps:

[0079] Step S101: Obtain the parameter information of the vehicle's battery in the current time period. The parameter information includes the battery's voltage, current, temperature, and resistance.

[0080] In this embodiment, the executing entity is a battery detection device, which can be any terminal device or a vehicle. For ease of description, the term "device" will be used to refer to the battery detection device below.

[0081] During vehicle operation, the device acquires parameter information of the vehicle's battery within the current time period. The current time period is a recent period of time for the vehicle; for example, the device acquires the past 2 hours as the current time period, meaning the interval between the current time point and the historical time point is 2 hours.

[0082] The parameter information includes the parameters themselves and their acquisition times. Parameters include battery voltage, current, temperature, and the battery itself. For example, if the battery current is collected every 10 seconds, and the current time period is 2 hours, the parameter information includes 12 collected current readings, each with a corresponding acquisition time. The battery acquisition interval is not limited to the aforementioned 10 seconds. The resistance can be the battery's insulation resistance, and the battery includes, but is not limited to, lithium batteries. The device can obtain parameter information for the current time period uploaded by the vehicle from the Internet of Things (IoT), or it can directly obtain parameter information for the current time period from the vehicle.

[0083] Step S102: Determine the first index values ​​corresponding to the current time period based on the parameter information, and perform preset calculations on the first index values ​​of the same type to obtain the second index values. The first index values ​​include parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness.

[0084] After obtaining the parameter information for the current time period, the device determines the corresponding first index values ​​for the current time period based on the parameter information. Each first index value includes parameters determined by voltage, current, temperature, and resistance, and the parameters include at least one of mean, variance, extreme values, kurtosis, and skewness.

[0085] For example, if the parameters include voltage, current, resistance, and temperature, then the indicators are divided into voltage indicators, current indicators, resistance indicators, and temperature indicators. The device retrieves the parameters corresponding to each type of indicator from the parameter information; for example, it retrieves the various voltages under the voltage indicator. The device then calculates the indicator values ​​for each indicator of the same type, and these indicator values ​​are the first indicator values. For example, the device calculates at least one of the following for current: mean, variance, extreme values, kurtosis, and skewness; it calculates at least one of the following for voltage: mean, variance, extreme values, kurtosis, and skewness; it calculates at least one of the following for resistance: mean, variance, extreme values, kurtosis, and skewness; and it calculates at least one of the following for temperature: mean, variance, extreme values, kurtosis, and skewness.

[0086] After determining each first index value, multiple second index values ​​are determined based on each first index value. For example, the device acquires multiple first index values ​​of the same type, and then performs preset calculations on these first index values ​​to obtain multiple second index values. For instance, the device calculates the skewness of currents adjacent to each other at any two time points to obtain multiple skewnesses, and then subtracts the skewnesses from each pair of adjacent skewnesses to obtain multiple skewness difference values, which are then used as the second index values.

[0087] Step S103: Input each first index value and each second index value into the prediction model, and determine whether the battery has a risk of thermal runaway based on the output information of the prediction model.

[0088] The device incorporates a predictive model. It inputs various first and second indicator values ​​into the predictive model and determines whether the battery faces a risk of thermal runaway based on the model's output. For example, the predictive model outputs the probability of thermal runaway. If the probability is greater than a preset threshold, a risk of thermal runaway is identified; if it is less than or equal to the threshold, no risk is identified. If a risk of thermal runaway is identified, a warning message is sent to the vehicle, causing it to stop and wait for the battery temperature to decrease before resuming operation.

[0089] In this embodiment, the vehicle's battery parameters such as voltage, current, temperature, and resistance are acquired within the current time period. Based on these parameters, first indicator values ​​corresponding to the current time period are determined. Then, pre-calculated calculations are performed on similar first indicator values ​​to obtain second indicator values. Inputting these first and second indicator values ​​into a prediction model determines whether the battery is at risk of thermal runaway. In this application, first indicator values, including mean, variance, extreme values, kurtosis, and skewness, are determined using the battery's temperature, voltage, resistance, and current. Second indicator values ​​are then derived from these first indicator values. This multi-parameter approach improves the accuracy of battery thermal runaway detection.

[0090] Reference Figure 2 , Figure 2 This is a second embodiment of the battery testing method of this application. Based on the first embodiment, step S102 includes:

[0091] Step S201: Divide the current time period into multiple sub-time periods, and determine the sub-parameter information corresponding to each sub-time period based on the parameter information.

[0092] In this embodiment, to increase the number of parameters, the device divides the current time period into multiple sub-time periods. For example, refer to... Figure 3 The current time is 2023-10-10 10:20:30. Dividing the current time period into segments with an interval of M = 60 minutes, we get N-1, N-2, N-3, N-4, N-5, N-6, and N-7, where N = 7, indicating a total of 7 sub-time periods. Alternatively, we can divide the current time period into segments with an interval of M = 30 minutes, resulting in N-1, N-2, N-3, N-4, N-5, N-6, N-7, and N-8, where N = 8, indicating a total of 8 sub-time periods.

[0093] After determining each sub-time period, the device determines the sub-parameter information corresponding to each sub-time period based on the parameter information, that is, the current, resistance, voltage and temperature of the battery collected in the sub-time period are used as the sub-parameter information.

[0094] Step S202: Determine the values ​​of each first index corresponding to the sub-parameter information based on the sub-parameter information.

[0095] After determining the sub-parameter information for each sub-time period, the device determines the corresponding first index values ​​based on the sub-parameter information. For example, if the sub-time period is from 9:20:30 on October 10, 2023 to 10:20:30 on October 10, 2023, the device calculates at least one of the following: mean, variance, extreme values, kurtosis, and skewness of the current at temperature; at least one of the following: mean, variance, extreme values, kurtosis, and skewness of the voltage at temperature; at least one of the following: mean, variance, extreme values, kurtosis, and skewness of the resistance at resistance; and at least one of the following: mean, variance, extreme values, kurtosis, and skewness of the temperature at temperature. All of these calculated parameters are used as first index values. In this way, the device obtains the corresponding first index values ​​for each sub-time period.

[0096] Step S203: Determine the difference between the first index values ​​of the same type in adjacent sub-time periods, and use it as the second index value.

[0097] After determining the first index values ​​corresponding to each sub-time period, the device sequentially calculates the difference of the same type of first index values ​​for adjacent sub-time periods to obtain the second index values.

[0098] For example, if each sub-time period includes N-1, N-2, and N-3, then the differences between the extreme voltage values, extreme current values, extreme resistance values, and extreme temperature values ​​of N-1 and N-2 are used as the second index value; similarly, the differences between the extreme voltage values, extreme current values, extreme resistance values, and extreme temperature values ​​of N-2 and N-3 are also used as the second index value. Furthermore, the differences in kurtosis, variance, and skewness between adjacent sub-time periods can also be calculated.

[0099] In this embodiment, the device divides the current time period into multiple sub-time periods, determines a first index value based on the sub-parameter information of the sub-time periods, and determines a second index value by the difference between the first index values ​​of adjacent sub-time periods, thereby deriving more parameters for predicting battery thermal runaway and improving the prediction accuracy of battery thermal runaway.

[0100] In one embodiment, the parameter information for the current time period is obtained by preprocessing the collected battery data. Specifically, the device acquires the operating parameter information corresponding to the battery within the current time period, processes any abnormal parameters in the operating parameter information, and thus obtains the battery's parameter information for the current time period.

[0101] For example, abnormal parameters include missing values ​​and outliers. Missing values ​​in the runtime parameter information are handled as follows: Parameters with entirely empty rows are deleted; other parameters with a relatively low percentage of missing values ​​are filled with the mean, mode, or median. Outliers in the runtime parameter information are modified, for example, by replacing them with the mean, mode, or median.

[0102] In this embodiment, abnormal parameters in the battery operating parameters collected during the current time period are processed to ensure that the values ​​of each first indicator and each second indicator are true and accurate, thereby improving the accuracy of battery thermal runaway prediction.

[0103] Reference Figure 4 , Figure 4 This is a first embodiment of the prediction model training method of this application. The prediction model training method includes the following steps:

[0104] Step S401: Obtain multiple positive samples. The positive samples include the first parameter values ​​and second parameter values ​​of batteries that have not experienced thermal runaway during historical time periods. The second parameter values ​​are obtained by performing preset calculations on the same type of first parameter values. The first parameter values ​​include parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness.

[0105] In this embodiment, the execution entity is a prediction model training device. For ease of description, the term "training device" will be used to refer to the prediction model training device below. The training device can be a server, a vehicle, or other terminal device capable of model training.

[0106] The training device acquires multiple positive samples, which are determined by the parameters of batteries that have not experienced thermal runaway. The positive samples include various first parameter values ​​and various second parameter values ​​of batteries that have not experienced thermal runaway over a historical time period. The second parameter values ​​are obtained by performing a preset operation on the same type of first parameter values. The preset operation is, for example, subtraction. The first parameter values ​​include parameters determined by voltage, current, resistance, and temperature, and these parameters include at least one of mean, variance, kurtosis, and skewness.

[0107] In one example, the positive sample is obtained by manually processing the parameters of a battery that has not experienced thermal runaway.

[0108] In another example, the training device processes the parameters of batteries that have not experienced thermal runaway to obtain positive samples.

[0109] Specifically, in this embodiment, to increase the number of parameters, the training device divides the historical time period into multiple sub-time periods. For example, refer to... Figure 3 The current time is 2023-10-10 10:20:30. Dividing the historical time period into intervals of M = 60 minutes, we get N-1, N-2, N-3, N-4, N-5, N-6, and N-7, where N = 7, indicating a total of 7 sub-time periods. Alternatively, we can divide the historical time period into intervals of M = 30 minutes, resulting in N-1, N-2, N-3, N-4, N-5, N-6, N-7, and N-8, a total of 8 sub-time periods, where N = 8, indicating a total of 8 sub-time periods.

[0110] After determining each sub-time period, the training device determines the sub-parameter information corresponding to each sub-time period based on the parameter information, that is, the current, resistance, voltage and temperature of the detection battery in the sub-time period are used as the sub-parameter information.

[0111] After determining the sub-parameter information for each sub-time period, the corresponding first parameter values ​​are determined based on the sub-parameter information. For example, a sub-time period is from 9:20:30 on October 10, 2023 to 10:20:30 on October 10, 2023. The system calculates at least one of the following parameters for the temperature: mean, variance, extreme values, kurtosis, and skewness; at least one of the following parameters for the voltage: mean, variance, extreme values, kurtosis, and skewness; at least one of the following parameters for the resistance: mean, variance, extreme values, kurtosis, and skewness; and at least one of the following parameters for the temperature: mean, variance, extreme values, kurtosis, and skewness. All of these calculated parameters are used as first parameter values. In this way, the training device obtains the corresponding first parameter values ​​for each sub-time period.

[0112] After determining the first parameter values ​​corresponding to each sub-time period, the training device sequentially calculates the difference of the same type of first parameter values ​​for adjacent sub-time periods to obtain the second parameter values.

[0113] For example, if each sub-time period includes N-1, N-2, and N-3, then the differences between the extreme voltage values, extreme current values, extreme resistance values, and extreme temperature values ​​of N-1 and N-2 are used as the second parameter values; similarly, the differences between the extreme voltage values, extreme current values, extreme resistance values, and extreme temperature values ​​of N-2 and N-3 are also used as the second parameter values. Furthermore, the differences in kurtosis, variance, and skewness between adjacent sub-time periods can also be calculated.

[0114] Step S402: Obtain multiple negative samples. The negative samples include the values ​​of each third parameter and each fourth parameter of the battery that has experienced thermal runaway during the runaway period. The fourth parameter values ​​are obtained by performing preset calculations on the same type of third parameter values. The third parameter values ​​include parameters determined by voltage, current, temperature and resistance.

[0115] The training device acquires multiple negative samples. These negative samples include the values ​​of each third parameter and each fourth parameter of a battery that experienced thermal runaway during the runaway period. The fourth parameter values ​​are obtained by performing preset calculations on the same type of three parameter values. The third parameter values ​​include parameters determined by voltage, current, temperature, and resistance, and these parameters include at least one of extreme values, kurtosis, skewness, mean, and variance. Furthermore, the runaway period is determined based on the time point when the battery experienced thermal runaway. For example, if the acquisition period is 2 hours, the runaway period is the interval between the battery's thermal runaway time point and a historical time point, with the historical time point being earlier than the thermal runaway time point. Additionally, the ratio between the positive and negative samples acquired by the device is a target ratio, for example, 1:1.

[0116] In one example, negative samples are obtained by manually processing the parameters of batteries that have not experienced thermal runaway.

[0117] In another example, the training device processes the parameters of the battery during the runaway period to obtain negative samples. The process for obtaining negative samples is the same as that for obtaining positive samples, and will not be repeated here.

[0118] Step S403: Train the preset model based on each positive sample and each negative sample to obtain the prediction model.

[0119] After obtaining multiple positive samples and multiple negative samples, the device trains a preset model based on the positive and negative samples to obtain a prediction model.

[0120] In this embodiment, the training device acquires positive and negative samples, and then trains a prediction model based on each positive and negative sample to accurately predict whether the battery in the vehicle will experience thermal runaway.

[0121] Reference Figure 5 , Figure 5 This is a second embodiment of the prediction model training method of this application. Based on the first embodiment, before step S403, it further includes:

[0122] Step S501: Configure the initial network parameters of the random forest model.

[0123] In this embodiment, the prediction model is obtained by training a random forest model. First, the initial network parameters of the random forest model are configured. These initial network parameters include the number of trees, the maximum tree depth, the maximum number of tree partitions, the minimum information gain, and the minimum number of samples.

[0124] Step S502: Based on the value range of each initial network parameter, perform a grid search on the optimal network parameters of the random forest model to determine the target network parameters of the random forest model.

[0125] Each initial network parameter has a corresponding range of values, which is a possible range of values. For example, the number of trees can be 20, 100, or 200; the maximum depth of the trees can be none, 10, 20, or 30; the minimum number of samples required for a node split can be 2, 5, or 10; and the minimum number of samples required for a leaf node can be 1, 2, or 4.

[0126] The device performs a grid search on the optimal network parameters of the random forest model based on the value range of each initial network parameter. Grid search is a method for searching for the best combination of parameters within a given range of values; that is, the network search is used to fine-tune the network within a range to determine the optimal combination of network parameters. The optimal individual network parameters can be obtained through grid search and used as the target network parameters.

[0127] Step S503: Adjust the initial network parameters of the random forest model to the target network parameters to obtain the preset model.

[0128] After obtaining the parameters of each target network, the device adjusts the initial network parameters of the random forest model to the target network parameters, thus obtaining the preset model.

[0129] In this embodiment, a preset model can be obtained by performing a grid search on the initial network parameters of the random forest model, and then a prediction model with high accuracy can be trained using the preset model.

[0130] Reference Figure 6 , Figure 6 This is a third embodiment of the prediction model training method of this application. Based on the first or second embodiment, step S403 includes:

[0131] Step S601: Divide each positive sample and each negative sample into a training set and a validation set.

[0132] In this embodiment, the training device divides each positive sample and each negative sample into a training set and a validation set. The ratio between the number of positive samples and the number of negative samples in the training set is a target ratio, such as 1:1. Alternatively, positive samples and each negative sample can be divided into a training set, a validation set, and a test set. The ratio of samples in the training set, validation set, and test set is, for example, 6:2:2.

[0133] Step S602: Train the preset model according to the training set to obtain the model to be determined, and obtain the first accuracy of the model to be determined in predicting battery thermal runaway.

[0134] After obtaining the training set, the preset model is trained based on each sample in the training set to obtain the model to be determined. The preset model can be the random forest model mentioned above, or other models such as neural networks.

[0135] After obtaining the model to be determined, it is tested on a test set to obtain the prediction accuracy of the model. This prediction accuracy is defined as the first accuracy. First accuracy = TP / (TP+FP), where TP refers to the number of positive samples that are labeled as positive, and FP refers to the number of negative samples (false samples) that are labeled as positive.

[0136] Step S603: Validate the model to be determined based on the validation set to obtain the second accuracy of the model to be determined in predicting battery thermal runaway.

[0137] The training device then validates the model to be determined using samples from the validation set, obtaining a second accuracy rate for predicting battery runaway. Second accuracy rate = TP / (TP+FP).

[0138] Step S604: If the first accuracy rate is greater than the first preset threshold and the difference between the first accuracy rate and the second accuracy rate is less than the preset difference, then the model to be determined is stored as a prediction model.

[0139] After obtaining the first accuracy and the second accuracy, the comparison between the first accuracy and the second accuracy can determine whether the model to be trained has completed training.

[0140] For example, when both the first accuracy and the second accuracy are greater than the first preset threshold, and the difference between the first accuracy and the second accuracy is less than a preset difference, then the training of the model to be determined is complete, and the model to be determined is stored as a prediction model. The first preset threshold can be any suitable value, for example, the first preset threshold is 90%.

[0141] If both the first accuracy and the second accuracy are less than a second preset threshold, the model to be determined is considered underfitting and needs to be retrained. The second preset threshold is less than the first preset threshold; for example, if the second preset threshold is 40%, and the first accuracy is 30% and the second accuracy is 29%, the trained model is underfitting. To address this, the device increases the regularization parameter of the model to be determined to obtain a first training model. Increasing the regularization parameter allows the first training model to learn more features. The training device adds parameter values ​​to the samples in the training set to obtain a first training set, that is, adds or derives more features to the samples to provide more features for the model to learn. The training device trains the first training model using the first training set to obtain a prediction model. This avoids obtaining an underfitting prediction model and ensures a high prediction accuracy. The added parameter values ​​are determined based on the existing parameter values ​​of the samples in the training set.

[0142] Furthermore, if the first accuracy rate is greater than a first preset threshold, the first accuracy rate is greater than a second accuracy rate, and the difference between the first accuracy rate and the second accuracy rate is greater than a preset difference, then it can be determined that the trained model has an overfitting problem. For example, if the first preset threshold is 80%, and the first accuracy rate is 90%, the second accuracy rate is 70%, and the difference between the first accuracy rate and the second accuracy rate is 20%, which is greater than the preset difference of 10%, then it is determined that the accuracy rate on the validation set is much lower than the accuracy rate on the training set. In this case, the trained model has overfitted.

[0143] In one example, the training device reduces the probability of trees in the initial model (a random forest model) to obtain a second training model. The tree probability indicates the likelihood of a tree using an indicator from the samples; that is, the probability refers to the randomness of the tree's feature usage. For example, changes in `max_features` (the maximum number of features) can cause changes in probability. By reducing `max_features` to be less than the total number of features, the training device lowers the probability, which helps reduce the risk of overfitting. The training device then trains the second training model on the training set to obtain the prediction model.

[0144] In another example, the training device adds regularization parameters to the model to be determined to avoid overfitting. For example, L1 and L2 regularization parameters are added to the model library to limit the size of the model's parameters and prevent the model from becoming overly dependent on certain features. The model with added regularization parameters is the third model to be trained, and the training device trains the third model to be trained based on the training set to obtain the prediction model.

[0145] In another example, overfitting is prevented by stopping training. For instance, after several rounds of training, if the model's accuracy on the validation set begins to decline, training is stopped, resulting in a predictive model.

[0146] In another example, when the model to be determined exhibits overfitting, positive and negative samples are added to the training set to obtain a second training set. This increases the amount of data used to train the model, improving its generalization ability and preventing overfitting. The training device then uses the second training set to train the model to be determined to obtain a prediction model.

[0147] In this embodiment, the device divides each sample to obtain a training set and a validation set, then trains a preset model using the training set to obtain a model to be determined, and determines whether the model to be determined is overfitting or underfitting based on the first accuracy of the model to be determined and the second accuracy based on the validation set, thereby training a prediction model with high accuracy in predicting thermal runaway.

[0148] This application also provides a battery testing device, with reference to... Figure 7 The battery testing device 700 includes:

[0149] The first acquisition module 710 is used to acquire parameter information of the vehicle's battery in the current time period, including the battery's voltage, current, temperature and resistance.

[0150] The determination module 720 is used to determine the values ​​of each first indicator corresponding to the current time period based on the parameter information, and to perform preset calculations on the first indicator values ​​of the same type to obtain the values ​​of each second indicator. The first indicator values ​​include parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness.

[0151] The input module 730 is used to input each first index value and each second index value into the prediction model, and to determine whether the battery has a risk of thermal runaway based on the output information of the prediction model.

[0152] In one embodiment, the battery detection device 700 is specifically used for:

[0153] Divide the current time period into multiple sub-time periods, and determine the sub-parameter information corresponding to each sub-time period based on the parameter information;

[0154] Determine the values ​​of each first index corresponding to the sub-parameter information based on the sub-parameter information;

[0155] The difference between the first indicator values ​​of the same type in adjacent sub-time periods is determined as the second indicator value.

[0156] In one embodiment, the battery detection device 700 is specifically used for:

[0157] Obtain the operating parameter information of the battery within the current time period;

[0158] The abnormal parameters in the operating parameter information are processed to obtain the battery parameter information within the current time period.

[0159] This application also provides a predictive model training device, referring to... Figure 8 The prediction model training device 800 includes:

[0160] The second acquisition module 810 is used to acquire multiple positive samples. The positive samples include the first parameter values ​​and the second parameter values ​​of batteries that have not experienced thermal runaway during a historical time period. The second parameter values ​​are obtained by processing the first parameter values ​​of the same type through a preset calculation. The first parameter values ​​include parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness.

[0161] The third acquisition module 820 is used to acquire multiple negative samples. The negative samples include the values ​​of each third parameter and each fourth parameter of the battery that has experienced thermal runaway during the runaway period. The fourth parameter values ​​are obtained by processing the same type of third parameter values ​​through preset calculations. The third parameter values ​​include parameters determined by voltage, current, temperature and resistance.

[0162] The training module 830 is used to train the preset model based on each positive sample and each negative sample to obtain the prediction model.

[0163] In one embodiment, the prediction model training device 800 is specifically used for:

[0164] Configure the initial network parameters for the random forest model;

[0165] Based on the value range of each initial network parameter, a grid search is performed on the optimal network parameters of the random forest model to determine the target network parameters of the random forest model.

[0166] The initial network parameters of the random forest model are adjusted to the target network parameters to obtain the preset model.

[0167] In one embodiment, the prediction model training device 800 is specifically used for:

[0168] Each positive sample and each negative sample is divided into a training set and a validation set;

[0169] Based on the training set, the preset model is trained to obtain the model to be determined, and the first accuracy of the model to be determined in predicting battery thermal runaway is obtained.

[0170] Based on the validation set, the model to be determined is validated, and the second accuracy of the model to be determined in predicting battery thermal runaway is obtained.

[0171] If the first accuracy rate is greater than the first preset threshold, and the difference between the first accuracy rate and the second accuracy rate is less than the preset difference, then the model to be determined is stored as a prediction model.

[0172] In one embodiment, the prediction model training device 800 is specifically used for:

[0173] When both the first accuracy and the second accuracy are less than the second preset threshold, the regularization parameter of the model to be determined is increased to obtain the first model to be trained, and the second preset threshold is less than the first preset threshold.

[0174] Add parameter values ​​to the samples in the training set to obtain the first training set. The added parameter values ​​are determined based on the existing parameter values ​​of the samples in the training set.

[0175] The prediction model is obtained by training the first model to be trained based on the first training set.

[0176] In one embodiment, the prediction model training device 800 is specifically used for:

[0177] When the first accuracy is greater than the first preset threshold, the first accuracy is greater than the second accuracy, and the difference between the first accuracy and the second accuracy is greater than or equal to the preset difference, the probability of the tree in the model to be determined is reduced to obtain the second model to be trained. The model to be determined is a random forest model, and the probability is used to indicate the probability of the tree using the index in the sample.

[0178] The prediction model is obtained by training the second model to be trained based on the training set.

[0179] In one embodiment, the prediction model training device 800 is specifically used for:

[0180] When the first accuracy is greater than the first preset threshold, the first accuracy is greater than the second accuracy, and the difference between the first accuracy and the second accuracy is greater than or equal to the preset difference, the regularization parameter in the model to be determined is increased to obtain the third model to be trained.

[0181] The prediction model is obtained by training the third model to be trained based on the training set.

[0182] In one embodiment, the prediction model training device 800 is specifically used for:

[0183] When the first accuracy is greater than the first preset threshold, the first accuracy is greater than the second accuracy, and the difference between the first accuracy and the second accuracy is greater than or equal to the preset difference, positive samples and negative samples are added to the training set to obtain the second training set.

[0184] The prediction model is obtained by training the model to be determined based on the second training set.

[0185] Figure 9 This is a schematic diagram of the hardware structure of a terminal device / vehicle according to an exemplary embodiment.

[0186] The terminal device / vehicle 900 may include: a processor 901, such as a CPU, a memory 902, and a transceiver 903. Those skilled in the art will understand that... Figure 9 The structure shown does not constitute a limitation on the terminal device / vehicle and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. Memory 902 may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0187] The processor 901 can call computer programs or computer execution instructions stored in the memory 902 to complete all or part of the steps of the above-mentioned battery detection method or prediction model training method.

[0188] Transceiver 903 is used to receive information sent by external devices and to send information to external devices.

[0189] A non-transitory computer-readable storage medium, wherein instructions (computer-executable instructions) in the storage medium, when executed by a processor of a terminal device / vehicle, enable the terminal device / vehicle to perform the aforementioned battery detection method or prediction model training method.

[0190] A computer program product includes a computer program that, when executed by a processor of a terminal device / vehicle, enables the terminal device / vehicle to perform the aforementioned battery detection method or prediction model training method.

[0191] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0192] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A battery testing method, characterized in that, include: Obtain parameter information of the vehicle's battery within the current time period, including battery voltage, current, temperature, and resistance; The current time period is divided into multiple sub-time periods, and sub-parameter information corresponding to each sub-time period is determined according to the parameter information; each first index value corresponding to the sub-parameter information is determined according to the sub-parameter information; the difference between the same type of first index values ​​of adjacent sub-time periods is determined as a second index value, wherein the first index value includes parameters determined by voltage, current, temperature and resistance, and the parameters include at least one of mean, variance, extreme value, kurtosis and skewness; Each of the first index values ​​and each of the second index values ​​are input into the prediction model, and the presence of thermal runaway risk in the battery is determined based on the output information of the prediction model.

2. The battery testing method according to claim 1, characterized in that, The step of obtaining the vehicle's battery parameter information within the current time period includes: Obtain the operating parameter information corresponding to the battery within the current time period; The abnormal parameters in the operating parameter information are processed to obtain the parameter information of the battery within the current time period.

3. A method for training a prediction model, characterized in that, A method for training a prediction model in a battery detection method as described in claim 1 or 2, the prediction model training method comprising: Multiple positive samples are obtained. The positive samples include the first parameter values ​​and the second parameter values ​​of batteries that have not experienced thermal runaway during a historical time period. The second parameter values ​​are obtained by processing the first parameter values ​​of the same type through a preset calculation. The first parameter values ​​include parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness. Multiple negative samples are obtained. The negative samples include the values ​​of each third parameter and each fourth parameter of the battery that has experienced thermal runaway during the runaway period. The fourth parameter values ​​are obtained by processing the same type of third parameter values ​​through a preset operation. The third parameter values ​​include parameters determined by voltage, current, temperature and resistance. The preset model is trained based on each of the positive samples and each of the negative samples to obtain the prediction model.

4. The prediction model training method according to claim 3, characterized in that, Before the step of training the preset model based on each of the positive samples and each of the negative samples to obtain the prediction model, the method further includes: Configure the initial network parameters for the random forest model; Based on the value range of each initial network parameter, a grid search is performed on the optimal network parameters of the random forest model to determine the target network parameters of the random forest model; The initial network parameters of the random forest model are adjusted to the target network parameters to obtain the preset model.

5. The prediction model training method according to claim 3, characterized in that, The step of training the preset model based on each of the positive samples and each of the negative samples includes: Each of the positive samples and each of the negative samples is divided into a training set and a validation set; Based on the training set, the preset model is trained to obtain the model to be determined, and the first accuracy of the model to be determined in predicting battery thermal runaway is obtained. Based on the validation set, the model to be determined is validated to obtain the second accuracy of the model to be determined in predicting battery thermal runaway. If the first accuracy rate is greater than the first preset threshold, and the difference between the first accuracy rate and the second accuracy rate is less than the preset difference, then the model to be determined is stored as a prediction model.

6. The prediction model training method according to claim 5, characterized in that, After the step of validating the model to be determined based on the validation set, the method further includes: When both the first accuracy and the second accuracy are less than the second preset threshold, the regularization parameter of the model to be determined is increased to obtain the first model to be trained, and the second preset threshold is less than the first preset threshold. Add parameter values ​​to the samples in the training set to obtain the first training set. The added parameter values ​​are determined based on the existing parameter values ​​of the samples in the training set. The prediction model is obtained by training the first model to be trained based on the first training set.

7. The prediction model training method according to claim 5, characterized in that, After the step of validating the model to be determined based on the validation set, the method further includes: When the first accuracy is greater than a first preset threshold, the first accuracy is greater than the second accuracy, and the difference between the first accuracy and the second accuracy is greater than or equal to a preset difference, the probability of the tree in the model to be determined is reduced to obtain a second model to be trained. The model to be determined is a random forest model, and the probability is used to indicate the probability that the tree uses the index in the sample. The second model to be trained is trained using the training set to obtain the prediction model.

8. The prediction model training method according to claim 5, characterized in that, After the step of validating the model to be determined based on the validation set, the method further includes: When the first accuracy is greater than the first preset threshold, the first accuracy is greater than the second accuracy, and the difference between the first accuracy and the second accuracy is greater than or equal to a preset difference, the regularization parameter in the model to be determined is increased to obtain a third model to be trained. The prediction model is obtained by training the third model to be trained based on the training set.

9. The prediction model training method according to claim 5, characterized in that, After the step of validating the model to be determined based on the validation set, the method further includes: When the first accuracy is greater than a first preset threshold, the first accuracy is greater than the second accuracy, and the difference between the first accuracy and the second accuracy is greater than or equal to a preset difference, positive samples and negative samples are added to the training set to obtain a second training set. The prediction model is obtained by training the model to be determined based on the second training set.

10. A battery testing device, characterized in that, include: The first acquisition module is used to acquire parameter information of the vehicle's battery in the current time period, including the battery's voltage, current, temperature and resistance. The determination module is used to divide the current time period into multiple sub-time periods, and determine the sub-parameter information corresponding to each sub-time period according to the parameter information; determine each first index value corresponding to the sub-parameter information according to the sub-parameter information; determine the difference between the first index values ​​of the same type in the sub-time periods that are adjacent in time as a second index value, wherein the first index value includes parameters determined by voltage, current, temperature and resistance, and the parameters include at least one of mean, variance, extreme value, kurtosis and skewness; The input module is used to input each of the first index values ​​and each of the second index values ​​into the prediction model, and determine whether the battery has a risk of thermal runaway based on the output information of the prediction model.

11. A predictive model training device, characterized in that, For training a prediction model in a battery detection device as described in claim 10, the prediction model training device comprises: The second acquisition module is used to acquire multiple positive samples. The positive samples include various first parameter values ​​and various second parameter values ​​of batteries that have not experienced thermal runaway during a historical time period. The second parameter values ​​are obtained by processing the same type of first parameter values ​​through a preset calculation. The first parameter values ​​include parameters determined by voltage, current, temperature and resistance. The parameters include at least one of mean, variance, extreme value, kurtosis and skewness. The third acquisition module is used to acquire multiple negative samples. The negative samples include the values ​​of each third parameter and each fourth parameter of the battery that has experienced thermal runaway during the runaway period. The fourth parameter values ​​are obtained by processing the same type of third parameter values ​​through a preset operation. The third parameter values ​​include parameters determined by voltage, current, temperature and resistance. The training module is used to train the preset model based on each of the positive samples and each of the negative samples to obtain the prediction model.

12. A terminal device, characterized in that, include: Memory and processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the battery detection device to perform the battery detection method as described in any one of claims 1 to 2 or the prediction model training method as described in any one of claims 3 to 9.

13. A vehicle, characterized in that, include: Memory and processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the battery detection device to perform the battery detection method as described in any one of claims 1 to 2 or the prediction model training method as described in any one of claims 3 to 9.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the battery detection method as described in any one of claims 1 to 2 or the prediction model training method as described in any one of claims 3 to 9.

Images