A training method, system and detection method for anomaly recognition model

Abstract

The present invention provides a training method, system and detection method for an abnormality recognition model, including: based on a big data platform, obtaining the operating data of each battery cell of each vehicle at preset intervals, wherein the operating data includes voltage data and operating condition characteristics; obtaining a first data set and multiple vehicle characteristic data based on the operating data; performing correlation analysis on each of the vehicle characteristic data and the first data set, and selecting vehicle characteristic data with a correlation greater than a preset threshold as a first training set; based on the first training set, iteratively training an abnormality recognition model under the constraint of a loss function to obtain a trained abnormality recognition model. The present invention obtains training data based on a big data platform, and selects operating condition characteristics with a high correlation with voltage data as input features, which effectively improves the recognition accuracy of the abnormality recognition model obtained by training, can realize the refined prediction of early abnormalities of vehicle batteries, and eliminates safety hazards of vehicles.

Description

Technical Field

[0001] The present invention relates to the automotive field, and in particular to a training method, system and detection method of an abnormality recognition model. Background Art

[0002] At present, with the increasing promotion of new energy vehicles in my country, the number of vehicles and mileage have increased significantly, and the key indicator of vehicle batteries, vehicle battery fault warning, has also received more and more attention. The current vehicle fault alarm software is operated by a single-chip microcomputer, but the operating resources and storage resources of the single-chip microcomputer are very limited, and it is difficult to meet the ability to record and analyze long-term battery data and discover early abnormalities of the battery, so other solutions need to be found.

[0003] The current vehicle battery failure warning method has the following shortcomings:

[0004] (1) The current vehicle battery anomaly detection algorithm is run on a single-chip microcomputer, which has very limited operating and storage resources. It can only perform calculations based on a few pieces of stored battery history information, which makes it difficult to record and analyze long-term battery data and detect early battery anomalies.

[0005] (2) Currently, vehicle battery warning can only be achieved based on the vehicle's data by monitoring whether voltage, temperature and other information exceed the safety threshold specified in the battery product manual. It is impossible to make a detailed prediction based on operating conditions such as ambient temperature and driving behavior;

[0006] (3) When the fault information is reported, the voltage, temperature and other information have exceeded the safety threshold, and the vehicle is facing a major safety hazard, and it is impossible to issue an early warning in the early stage of battery abnormality. Summary of the invention

[0007] In view of the above-mentioned shortcomings of the prior art, an object of the present invention is to provide a training method, system and detection method of an abnormality recognition model, so as to solve the problem in the prior art that early abnormalities of vehicle batteries cannot be detected.

[0008] A first aspect of the present invention provides a training method for an abnormality recognition model, comprising: based on a big data platform, obtaining the operating data of each battery cell of each vehicle at preset intervals, wherein the operating data includes voltage data and operating condition characteristics;

[0009] Based on the voltage data, a mean voltage standard deviation of each battery cell is calculated, and a maximum value of the mean voltage standard deviation of each battery cell is selected to form a first data set;

[0010] Based on the operating condition characteristics, extracting a plurality of vehicle characteristic data;

[0011] Performing correlation analysis on each of the vehicle feature data and the first data set, and selecting vehicle feature data with correlation greater than a preset threshold as a first training set;

[0012] Based on the first training set, the anomaly recognition model is iteratively trained under the constraint of the loss function to obtain a trained anomaly recognition model.

[0013] In one embodiment of the present invention, the step of calculating the mean voltage standard deviation of each battery cell per day based on the voltage data includes:

[0014] Obtaining the de-averaged voltage of each battery cell at a preset time interval according to the voltage data, wherein the de-averaged voltage = the voltage value of the battery cell at a preset time interval minus the voltage average of each battery cell of the vehicle at a preset time interval;

[0015] Based on the de-mean voltage, the standard deviation of the de-mean voltage of each battery cell per day is calculated using a standard deviation formula.

[0016] In one embodiment of the present invention, the operating condition characteristics include: any one or a combination of the driving intensity, operating temperature, battery aging degree, and operating power average position.

[0017] In one embodiment of the present invention, the step of performing correlation analysis on each of the vehicle characteristic data and the first data set includes:

[0018] The Pearson correlation coefficient is used to evaluate the correlation between each of the vehicle characteristic data and the first data set.

[0019] In one embodiment of the present invention, the method further includes: filtering the first data set and the first training set, and selecting data meeting the filtering condition as the second data set and the second training set;

[0020] Based on the second training set, iteratively train the anomaly recognition model under the constraint of the loss function to obtain a trained anomaly recognition model;

[0021] Among them, the filtering conditions include: the total SOC value of the vehicle battery pack is greater than 50%, the number of daily operation data of the vehicle is greater than 1,000, and the total current absolute value of the battery cell is greater than 2 amperes.

[0022] In one embodiment of the present invention, the step of iteratively training the anomaly recognition model based on the second training set under the constraint of the loss function to obtain the trained anomaly recognition model includes:

[0023] Step 1: input the second training set into the anomaly recognition model, calculate the predicted value, use the loss function to calculate the loss value between the predicted value and the standard value, and modify the network parameters of the original prediction model according to the loss value, wherein the standard value comes from the second data set;

[0024] Step 2: Repeat step 1 until the preset number of iterations is completed or the loss value is less than the preset threshold, and select the model with the smallest loss as the trained anomaly recognition model.

[0025] A second aspect of the present invention further provides a training system for an abnormality recognition model, comprising:

[0026] A sample acquisition module is used to acquire the operating data of each battery cell of each vehicle at preset intervals based on the big data platform, wherein the operating data includes voltage data and operating condition characteristics;

[0027] A data processing module, used to calculate the mean voltage standard deviation of each battery cell based on the voltage data, select the maximum value of the mean voltage standard deviation of each battery cell to form a first data set; and also used to extract multiple vehicle characteristic data based on the operating condition characteristics;

[0028] A selection module, configured to perform a correlation analysis on each of the vehicle feature data and the first data set, and select the vehicle feature data with a correlation greater than a preset threshold as a first training set;

[0029] The model training module is used to iteratively train the anomaly recognition model based on the first training set under the constraint of the loss function to obtain a trained anomaly recognition model.

[0030] The third aspect of the present invention also provides a detection method, comprising: obtaining the current operating data of the vehicle, inputting it into a trained abnormal recognition model, obtaining a recognition result, and determining whether the battery of the vehicle is abnormal based on the recognition result; wherein the operating data includes the voltage data and operating condition characteristics of the vehicle; the trained abnormal recognition model is obtained using any one of the training methods for training an abnormal recognition model of the first aspect of the present invention.

[0031] The fourth aspect of the present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method steps described in the training method of an abnormality recognition model of the first aspect of the present invention or the method steps of a detection method of the third aspect of the present invention are implemented.

[0032] The fifth aspect of the present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the method steps described in the training method of an abnormality recognition model in the first aspect of the present invention or the method steps of a detection method in the third aspect of the present invention are implemented.

[0033] As described above, the training method, system and detection method of an abnormality recognition model of the present invention have the following beneficial effects:

[0034] The present invention is based on a big data platform to obtain the operating data of each battery cell of each vehicle, wherein the operating data includes voltage data and operating condition characteristics; a first data set is obtained based on the voltage data, multiple vehicle characteristic data are obtained based on the operating condition characteristics, and each vehicle characteristic data with a high correlation with the first data set is screened out as a first training set, and the first training set is used to train an abnormality recognition model. The present invention obtains training data based on a big data platform, and selects operating condition characteristics with a high correlation with voltage data as input features, which effectively improves the recognition accuracy of the abnormality recognition model obtained by training, can realize the refined prediction of early abnormalities of vehicle batteries, and eliminate safety hazards of vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 Shown is a schematic flow chart of the training method in the first embodiment of the present invention.

[0036] Figure 2 Shown is a structural block diagram of a training system in a second embodiment of the present invention.

[0037] Figure 3 FIG. 1 is a schematic diagram of a computer device according to a third embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0039] It should be noted that the illustrations provided in the following embodiments are only used to schematically illustrate the basic concept of the present invention. The figures only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0040] See also Figure 1 The first embodiment of the present invention relates to a training method for an abnormality recognition model, which specifically includes:

[0041] Step 101: Based on the big data platform, the operating data of each battery cell of each vehicle is obtained at preset intervals.

[0042] Specifically, with the rapid development of Internet of Vehicles, 5G, and big data technology, a large amount of vehicle data is transmitted in real time through 4G / 5G and stored in the national big data platform, which makes it possible to upgrade the operating carrier of the vehicle battery early abnormality detection algorithm from a single-chip microcomputer to a cloud server cluster. The vehicles involved in this embodiment all have communication functions and can transmit operating data to the big data platform in real time, wherein the operating data includes the voltage data and operating condition characteristics of each vehicle. Specifically, the operating condition characteristics include any one or a combination of the driving intensity, operating temperature, battery aging, and average position of the operating condition power. It should be understood that the big data platform cleans the operating data uploaded by each vehicle, filters out invalid or abnormal data, and classifies and saves the retained operating data according to the preset strategy. It should be noted that the above-mentioned operating condition characteristics are only examples. In actual applications, in order to improve the accuracy, more operating condition characteristics can also be selected as feature data for model training.

[0043] Through the big data platform, users can obtain the operating data of each battery cell of each vehicle at preset intervals to form a data set as training data for the anomaly recognition model; the operating data includes the voltage data, operating condition characteristics, total SOC value of each vehicle's battery pack, the amount of daily operating data of each vehicle, the power and current data of each battery cell of each vehicle, etc.

[0044] Step 102: Acquire a first data set and a plurality of vehicle characteristic data based on the operating data.

[0045] Specifically, the first data set is obtained based on the voltage data of each vehicle, wherein the steps of obtaining include:

[0046] At preset time intervals, the de-averaged voltage of each battery cell of each vehicle is obtained based on the voltage data, wherein the de-averaged voltage = the voltage value of the battery cell at the preset time interval - the voltage average of each battery cell of the vehicle at the preset time interval; for example, if the preset time is 10s, then for each battery cell, a de-averaged voltage can be obtained every 10s.

[0047] Based on the de-mean voltage, the standard deviation of the de-mean voltage of each battery cell per day is calculated using the standard deviation formula; it should be understood that the standard deviation S is also called the standard deviation, and its expression is:

[0048]

[0049] n is the number of de-averaged voltages obtained by a battery cell every day, x i is the ith de-averaged voltage, is the arithmetic mean of n de-averaged voltages.

[0050] The maximum value of the mean-valued voltage standard deviation of each battery cell is selected to form a first data set. If the number of battery cells is m, the dimension of the first data set is m*1.

[0051] To continue explaining, each vehicle feature data is extracted based on the operating condition features. The user can select the features required for training from the operating condition features for extraction as needed. For example, if n vehicle feature data are extracted from the above operating condition features, the dimension of the extracted vehicle feature data is m*n.

[0052] Step 103 : performing correlation analysis on each vehicle feature data and the first data set, and selecting vehicle feature data with correlation greater than a preset threshold as the first training set.

[0053] Specifically, this embodiment uses the Pearson correlation coefficient to evaluate the correlation between each vehicle feature data and the first data set. The correlation index is used to determine whether two feature data are correlated. This embodiment uses the Pearson correlation coefficient to evaluate the correlation between two feature data. Generally, the closer the correlation index is to 0, the weaker the correlation; the closer the correlation index is to 1 or -1, the stronger the linear correlation. In this embodiment, the preset threshold value of the correlation is set to 0.25. When the correlation index of two feature data is greater than 0.25, the feature data with a lower information content index is eliminated. It should be understood that in the field of natural sciences, the Pearson correlation coefficient is widely used to measure the degree of correlation between two variables. The Pearson correlation coefficient between two variables is defined as the quotient of the covariance and standard deviation between the two variables:

[0054]

[0055] Among them, X is one feature data, Y is another feature data; ρ X ,Y is the Pearson correlation coefficient between the two feature data; E() is the expected value.

[0056] By adopting the above scheme, vehicle feature data with a high correlation with the prediction target can be selected as input features to train the anomaly recognition model, thereby improving the recognition accuracy of the anomaly recognition model. For example, k vehicle feature data with a high correlation with the first data set are selected from n vehicle feature data as the first training set. At this time, the dimension of the first training set is m*k.

[0057] Step 104: Based on the first training set, iteratively train the anomaly recognition model under the constraint of the loss function to obtain a trained anomaly recognition model.

[0058] Specifically, the steps of training the anomaly recognition model include:

[0059] Step 1: Input the first training set into the anomaly recognition model, calculate the predicted value, use the loss function to calculate the loss value of the predicted value and the standard value, and correct the network parameters of the original prediction model according to the loss value, wherein the standard value comes from the first data set.

[0060] The abnormal identification model in this embodiment is trained on the basis of a fully connected neural network. It should be understood that a fully connected neural network is a commonly used prediction model in the prior art, which is mainly composed of an input layer, two hidden layers and an output layer connected in sequence. The distribution of neurons in each layer is 4, 32, 32, 1, and the activation function of each hidden layer adopts the ReLU function. The fully connected neural network uses rmsprop as an optimizer. Furthermore, the loss function adopts the mean square error (MSE), which refers to the average value of the square of the distance between the model prediction value f(x) and the true value y of the sample. The formula is as follows:

[0061]

[0062] Step 2: Repeat step 1 until the preset number of iterations is completed or the loss value is less than the preset threshold, and select the model with the smallest loss as the trained anomaly recognition model.

[0063] Furthermore, it also includes step 3, randomly dividing a part of the above-mentioned first training set according to a preset ratio as a first test set, inputting the first test set into the trained anomaly recognition to verify its reliability, and if the prediction accuracy is greater than a preset threshold, the trained prediction model is used as the final anomaly recognition model, otherwise, repeating steps 1 to 2 to retrain the anomaly recognition model. When modeling, the training set is usually used as the expected distribution and the test set is used as the actual distribution.

[0064] It should be understood that for different vehicle models, the focus during training is different, and the selected operating condition characteristics are also different; in addition, for the same vehicle model, during the training process, the selected operating condition characteristics can be adjusted in real time according to the accuracy of the model, and the operating condition characteristics with poor effects can be eliminated and new operating condition characteristics can be added.

[0065] Furthermore, in order to improve the recognition accuracy of the abnormal recognition model, the first data set and the first training set can also be filtered, and the data that meets the filtering conditions can be selected as the second data set and the second training set, wherein the filtering conditions include: the total SOC value of the vehicle battery pack is greater than 50%, the number of daily operation data of the vehicle is greater than 1000, and the absolute value of the total current of the battery cell is greater than 2 amperes. It should be noted that the operation data also includes the current data of each battery cell, the SOC value during operation, and the number of valid operation data uploaded by the vehicle. In a feasible embodiment, assuming that r battery cells are filtered, the dimension of the second data set is r*1, and the dimension of the second training set is r*k. The filtered second training set is input into the original abnormal recognition model to train the abnormal recognition model. The training steps are similar to those of the first training set. Only the standard value at this time comes from the second data set. In order to save space, it will not be repeated here.

[0066] Furthermore, the present embodiment also relates to a detection method, the steps of which include: obtaining current operating data of a vehicle, the operating data including voltage data and operating condition characteristics of the vehicle, inputting the operating data into the trained abnormality recognition model obtained in steps 101-104, obtaining a recognition result, and determining whether the battery of the vehicle is abnormal based on the recognition result.

[0067] It can be seen that this embodiment is based on a big data platform to obtain the operating data of each battery cell of each vehicle, wherein the operating data includes voltage data and operating condition characteristics; a first data set is obtained based on the voltage data, multiple vehicle characteristic data are obtained based on the operating condition characteristics, and each vehicle characteristic data with a high correlation with the first data set is screened out as a first training set, and the first training set is used to train an abnormality recognition model. The present invention obtains training data based on a big data platform, and selects operating condition characteristics with a high correlation with voltage data as input features, which effectively improves the recognition accuracy of the abnormality recognition model obtained by training, and can achieve refined prediction of early abnormalities of vehicle batteries, eliminating safety hazards of vehicles.

[0068] See also Figure 2 The second embodiment of the present invention relates to a training system for an abnormality recognition model, which corresponds to the training method for an abnormality recognition model disclosed in the first embodiment, and specifically includes:

[0069] The sample acquisition module is used to obtain the operating data of each battery cell of each vehicle at preset intervals based on the big data platform, where the operating data includes voltage data and operating condition characteristics.

[0070] Specifically, with the rapid development of Internet of Vehicles, 5G, and big data technology, a large amount of vehicle data is transmitted in real time through 4G / 5G and stored in the national big data platform, which makes it possible to upgrade the operating carrier of the vehicle battery early abnormality detection algorithm from a single-chip microcomputer to a cloud server cluster. The vehicles involved in this embodiment all have communication functions and can transmit operating data to the big data platform in real time, wherein the operating data includes the voltage data and operating condition characteristics of each vehicle. Specifically, the operating condition characteristics include any one or a combination of the driving intensity, operating temperature, battery aging, and average position of the operating condition power. It should be understood that the big data platform cleans the operating data uploaded by each vehicle, filters out invalid or abnormal data, and classifies and saves the retained operating data according to the preset strategy. It should be noted that the above-mentioned operating condition characteristics are only examples. In actual applications, in order to improve the accuracy, more operating condition characteristics can also be selected as feature data for model training.

[0071] Through the big data platform, users can obtain the operating data of each battery cell of each vehicle at preset intervals to form a data set as training data for the anomaly recognition model; the operating data includes the voltage data, operating condition characteristics, total SOC value of each vehicle's battery pack, the amount of daily operating data of each vehicle, the power and current data of each battery cell of each vehicle, etc.

[0072] The data processing module is used to calculate the mean voltage standard deviation of each battery cell based on the voltage data, select the maximum value of the mean voltage standard deviation of each battery cell to form a first data set; and is also used to extract multiple vehicle characteristic data based on operating condition characteristics.

[0073] Specifically, the first data set is obtained based on the voltage data of each vehicle, wherein the steps of obtaining include:

[0074] At preset time intervals, the de-averaged voltage of each battery cell of each vehicle is obtained based on the voltage data, wherein the de-averaged voltage = the voltage value of the battery cell at the preset time interval - the voltage average of each battery cell of the vehicle at the preset time interval; for example, if the preset time is 10s, then for each battery cell, a de-averaged voltage can be obtained every 10s.

[0075] Based on the de-mean voltage, the standard deviation of the de-mean voltage of each battery cell per day is calculated using the standard deviation formula; it should be understood that the standard deviation S is also called the standard deviation, and its expression is:

[0076]

[0077] n is the number of de-averaged voltages obtained by a battery cell every day, x i is the ith de-averaged voltage, is the arithmetic mean of n de-averaged voltages.

[0078] The maximum value of the mean-valued voltage standard deviation of each battery cell is selected to form a first data set. If the number of battery cells is m, the dimension of the first data set is m*1.

[0079] To continue explaining, each vehicle feature data is extracted based on the operating condition features. The user can select the features required for training from the operating condition features for extraction as needed. For example, if n vehicle feature data are extracted from the above operating condition features, the dimension of the extracted vehicle feature data is m*n.

[0080] The selection module is used to perform correlation analysis on each vehicle feature data and the first data set, and select vehicle feature data with correlation greater than a preset threshold as the first training set.

[0081] Specifically, this embodiment uses the Pearson correlation coefficient to evaluate the correlation between each vehicle feature data and the first data set. The correlation index is used to determine whether two feature data are correlated. This embodiment uses the Pearson correlation coefficient to evaluate the correlation between two feature data. Generally, the closer the correlation index is to 0, the weaker the correlation; the closer the correlation index is to 1 or -1, the stronger the linear correlation. In this embodiment, the preset threshold value of the correlation is set to 0.25. When the correlation index of two feature data is greater than 0.25, the feature data with a lower information content index is eliminated. It should be understood that in the field of natural sciences, the Pearson correlation coefficient is widely used to measure the degree of correlation between two variables. The Pearson correlation coefficient between two variables is defined as the quotient of the covariance and standard deviation between the two variables:

[0082]

[0083] Among them, X is one feature data, Y is another feature data; ρ X ,Y is the Pearson correlation coefficient between the two feature data; E() is the expected value.

[0084] By adopting the above scheme, vehicle feature data with a high correlation with the prediction target can be selected as input features to train the anomaly recognition model, thereby improving the recognition accuracy of the anomaly recognition model. For example, k vehicle feature data with a high correlation with the first data set are selected from n vehicle feature data as the first training set. At this time, the dimension of the first training set is m*k.

[0085] The model training module is used to iteratively train the anomaly recognition model based on the first training set under the constraint of the loss function to obtain a trained anomaly recognition model.

[0086] Specifically, the steps of training the anomaly recognition model include:

[0087] Step 1: Input the first training set into the anomaly recognition model, calculate the predicted value, use the loss function to calculate the loss value of the predicted value and the standard value, and correct the network parameters of the original prediction model according to the loss value, wherein the standard value comes from the first data set.

[0088] The abnormal identification model in this embodiment is trained on the basis of a fully connected neural network. It should be understood that a fully connected neural network is a commonly used prediction model in the prior art, which is mainly composed of an input layer, two hidden layers and an output layer connected in sequence. The distribution of neurons in each layer is 4, 32, 32, 1, and the activation function of each hidden layer adopts the ReLU function. The fully connected neural network uses rmsprop as an optimizer. Furthermore, the loss function adopts the mean square error (MSE), which refers to the average value of the square of the distance between the model prediction value f(x) and the true value y of the sample. The formula is as follows:

[0089]

[0090] Step 2: Repeat step 1 until the preset number of iterations is completed or the loss value is less than the preset threshold, and select the model with the smallest loss as the trained anomaly recognition model.

[0091] Furthermore, it also includes step 3, randomly dividing a part of the above-mentioned first training set according to a preset ratio as a first test set, inputting the first test set into the trained anomaly recognition to verify its reliability, and if the prediction accuracy is greater than a preset threshold, the trained prediction model is used as the final anomaly recognition model, otherwise, repeating steps 1 to 2 to retrain the anomaly recognition model. When modeling, the training set is usually used as the expected distribution and the test set is used as the actual distribution.

[0092] It should be understood that for different vehicle models, the focus during training is different, and the selected operating condition characteristics are also different; in addition, for the same vehicle model, during the training process, the selected operating condition characteristics can be adjusted in real time according to the accuracy of the model, and the operating condition characteristics with poor effects can be eliminated and new operating condition characteristics can be added.

[0093] Furthermore, in order to improve the recognition accuracy of the abnormal recognition model, the selection module can also filter the first data set and the first training set, and select data that meets the filtering conditions as the second data set and the second training set, wherein the filtering conditions include: the total SOC value of the vehicle battery pack is greater than 50%, the number of daily operating data of the vehicle is greater than 1,000, and the absolute value of the total current of the battery cell is greater than 2 amperes. Assuming that r battery cells are filtered out, the dimension of the second data set is r*1, and the dimension of the second training set is r*k. The filtered second training set is input into the original abnormal recognition model, and the abnormal recognition model is trained. The training steps are similar to those of the first training set, except that the standard values ​​at this time come from the second data set. To save space, they will not be repeated here.

[0094] It can be seen that this embodiment is based on a big data platform to obtain the operating data of each battery cell of each vehicle, wherein the operating data includes voltage data and operating condition characteristics; a first data set is obtained based on the voltage data, multiple vehicle characteristic data are obtained based on the operating condition characteristics, and each vehicle characteristic data with a high correlation with the first data set is screened out as a first training set, and the first training set is used to train an abnormality recognition model. The present invention obtains training data based on a big data platform, and selects operating condition characteristics with a high correlation with voltage data as input features, which effectively improves the recognition accuracy of the abnormality recognition model obtained by training, and can achieve refined prediction of early abnormalities of vehicle batteries, eliminating safety hazards of vehicles.

[0095] See also Figure 3 The third embodiment of the present invention relates to a computer device, comprising a memory 301, a processor 302, and a computer program stored in the memory 301 and executable on the processor 302. When the processor 302 executes the computer program, the following steps are implemented:

[0096] Based on the big data platform, the operating data of each battery cell of each vehicle is obtained at preset intervals, where the operating data includes voltage data and operating condition characteristics;

[0097] Based on the voltage data, the mean voltage standard deviation of each battery cell is calculated, and the maximum value of the mean voltage standard deviation of each battery cell is selected to form a first data set;

[0098] Based on the working condition characteristics, multiple vehicle characteristic data are extracted;

[0099] Performing correlation analysis on each vehicle feature data and the first data set, and selecting vehicle feature data with correlation greater than a preset threshold as the first training set;

[0100] Based on the first training set, the anomaly recognition model is iteratively trained under the constraint of the loss function to obtain a trained anomaly recognition model.

[0101] Or implement the following steps:

[0102] The current operating data of the vehicle is obtained and input into the trained abnormality recognition model to obtain a recognition result, and according to the recognition result, it is determined whether the battery of the vehicle is abnormal; wherein the operating data includes the voltage data and operating condition characteristics of the vehicle.

[0103] The memory 301 and the processor 302 are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors 302 and the memory 301 together. The bus can also connect various other circuits such as peripheral devices 303, regulators 304, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor 302 is transmitted on a wireless medium through an antenna, and further, the antenna also receives data and transmits the data to the processor 302.

[0104] The processor 302 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management and other control functions. The memory 301 can be used to store data used by the processor 302 when performing operations.

[0105] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0106] A fourth embodiment of the present invention relates to a storage medium having a computer program stored thereon, which, when executed by a processor, implements the following steps:

[0107] Based on the big data platform, the operating data of each battery cell of each vehicle is obtained at preset intervals, where the operating data includes voltage data and operating condition characteristics;

[0108] Based on the voltage data, the mean voltage standard deviation of each battery cell is calculated, and the maximum value of the mean voltage standard deviation of each battery cell is selected to form a first data set;

[0109] Based on the working condition characteristics, multiple vehicle characteristic data are extracted;

[0110] Performing correlation analysis on each vehicle feature data and the first data set, and selecting vehicle feature data with correlation greater than a preset threshold as the first training set;

[0111] Based on the first training set, the anomaly recognition model is iteratively trained under the constraint of the loss function to obtain a trained anomaly recognition model.

[0112] Or implement the following steps:

[0113] The current operating data of the vehicle is obtained and input into the trained abnormality recognition model to obtain a recognition result, and according to the recognition result, it is determined whether the battery of the vehicle is abnormal; wherein the operating data includes the voltage data and operating condition characteristics of the vehicle.

[0114] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0115] In summary, the training method, system and detection method of an abnormal recognition model of the present invention obtains the operating data of each battery cell of each vehicle based on a big data platform, wherein the operating data includes voltage data and operating condition characteristics; obtains a first data set based on the voltage data, obtains multiple vehicle characteristic data based on the operating condition characteristics, and screens out each vehicle characteristic data with a high correlation with the first data set as a first training set, and uses the first training set to train to obtain an abnormal recognition model. The present invention obtains training data based on a big data platform, and selects operating condition characteristics with a high correlation with voltage data as input features, which effectively improves the recognition accuracy of the trained abnormal recognition model, can realize the refined prediction of early abnormalities of vehicle batteries, and eliminates the safety hazards of vehicles. Therefore, the present invention effectively overcomes the various shortcomings in the prior art and has a high industrial utilization value.

[0116] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A training method for an anomaly recognition model, characterized in that: include: Based on the big data platform, the operating data of each battery cell of each vehicle is obtained at preset intervals, wherein the operating data includes voltage data and operating condition characteristics; Based on the voltage data, a mean voltage standard deviation of each battery cell is calculated, and a maximum value of the mean voltage standard deviation of each battery cell is selected to form a first data set; Based on the operating condition characteristics, extracting a plurality of vehicle characteristic data; Performing correlation analysis on each of the vehicle feature data and the first data set, and selecting vehicle feature data with correlation greater than a preset threshold as a first training set; Based on the first training set, iteratively train the anomaly recognition model under the constraint of the loss function to obtain a trained anomaly recognition model; The step of obtaining the trained abnormality recognition model includes: filtering the first data set and the first training set, and selecting data meeting the filtering conditions as the second data set and the second training set; and Based on the second training set, iteratively train the anomaly recognition model under the constraint of the loss function to obtain a trained anomaly recognition model; Among them, the filtering conditions include: the total SOC value of the vehicle battery pack is greater than 50%, the number of daily operation data of the vehicle is greater than 1,000, and the total current absolute value of the battery cell is greater than 2 amperes.

2. The training method according to claim 1, characterized in that: The step of calculating the mean voltage standard deviation of each battery cell every day based on the voltage data comprises: Obtaining the de-averaged voltage of each battery cell at a preset time interval according to the voltage data, wherein the de-averaged voltage = the voltage value of the battery cell at a preset time interval minus the voltage average of each battery cell of the vehicle at a preset time interval; Based on the de-mean voltage, the standard deviation of the de-mean voltage of each battery cell per day is calculated using a standard deviation formula.

3. The training method according to claim 1, characterized in that: The operating condition characteristics include: any one or a combination of the driving intensity, operating temperature, battery aging degree, and average operating power position.

4. The training method according to claim 3, characterized in that: The step of performing correlation analysis on each of the vehicle characteristic data and the first data set comprises: The Pearson correlation coefficient is used to evaluate the correlation between each of the vehicle characteristic data and the first data set.

5. The training method according to claim 1, characterized in that: The step of iteratively training the anomaly recognition model based on the second training set under the constraint of the loss function to obtain the trained anomaly recognition model comprises: Step 1: input the second training set into the anomaly recognition model, calculate the predicted value, use the loss function to calculate the loss value between the predicted value and the standard value, and modify the network parameters of the original prediction model according to the loss value, wherein the standard value comes from the second data set; Step 2: Repeat step 1 until the preset number of iterations is completed or the loss value is less than the preset threshold, and select the model with the smallest loss as the trained anomaly recognition model.

6. A training system for an anomaly recognition model, characterized in that: include: A sample acquisition module is used to acquire the operating data of each battery cell of each vehicle at preset intervals based on the big data platform, wherein the operating data includes voltage data and operating condition characteristics; A data processing module, used to calculate the mean voltage standard deviation of each battery cell based on the voltage data, select the maximum value of the mean voltage standard deviation of each battery cell to form a first data set; and also used to extract multiple vehicle characteristic data based on the operating condition characteristics; A selection module, configured to perform a correlation analysis on each of the vehicle feature data and the first data set, and select the vehicle feature data with a correlation greater than a preset threshold as a first training set; A model training module, used for iteratively training the anomaly recognition model based on the first training set under the constraint of a loss function to obtain a trained anomaly recognition model, wherein the model training module is used for filtering the first data set and the first training set, selecting data that meets the filtering condition as a second data set and a second training set, and iteratively training the anomaly recognition model based on the second training set under the constraint of a loss function to obtain a trained anomaly recognition model; Among them, the filtering conditions include: the total SOC value of the vehicle battery pack is greater than 50%, the number of daily operation data of the vehicle is greater than 1,000, and the total current absolute value of the battery cell is greater than 2 amperes.

7. A detection method, characterized in that: include: The current operating data of the vehicle is obtained and input into a trained abnormality recognition model to obtain a recognition result, and whether the battery of the vehicle is abnormal is determined based on the recognition result; wherein the operating data includes voltage data and operating condition characteristics of the vehicle, and the trained abnormality recognition model is obtained using the training method described in any one of claims 1 to 5.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the training method described in any one of claims 1 to 5 or the steps of the detection method described in claim 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the training method described in any one of claims 1 to 5 or the steps of the detection method described in claim 7 are implemented.

Images