Battery management method, device and readable storage medium

Abstract

An embodiment of the present application provides a battery management method, device, and readable storage medium. The method includes: obtaining battery repair information, vehicle status information, and battery power consumption habit information of a vehicle from a database built on a server according to a preset battery life impact factor; obtaining a battery fault prediction result according to a battery fault prediction model, battery repair information, vehicle status information, and battery power consumption habit information, where the battery fault prediction model includes a full amount feature analysis sub-model, a category feature analysis sub-model, and a fault prediction sub-model; and managing the battery of the vehicle according to the battery fault prediction result. By using the full amount feature analysis sub-model and the category feature analysis sub-model pre-trained through a machine learning algorithm, a battery fault prediction result with higher accuracy can be obtained, thereby improving the effectiveness of vehicle management.

Description

Technical Field

[0001] This application relates to the field of data mining technology, and particularly to a battery management method, apparatus, and readable storage medium. Background Art

[0002] With the increasing emphasis on environmental protection by people and the continuous development of new energy technologies, electric vehicles have been increasingly recognized by people. A battery is an important component of an electric vehicle, and effective management of the battery is of crucial significance for the normal operation of the electric vehicle.

[0003] In traditional methods, predictive maintenance is usually adopted for the maintenance management of the vehicle's battery, which is also a widely used management method currently. When using the "predictive maintenance" battery management method, usually, the prediction result of battery failure is obtained based on a large amount of historical battery maintenance information first, and then, it is determined whether the vehicle's battery needs to be repaired according to the prediction result of battery failure. The accuracy rate of the battery failure prediction result obtained by the above method is relatively low. Summary of the Invention

[0004] Embodiments of this application provide a battery management method, apparatus, and readable storage medium to improve the accuracy rate of the battery failure prediction result, and further improve the effectiveness of the vehicle's battery management.

[0005] In a first aspect, embodiments of this application provide a battery management method, including:

[0006] Obtain the battery maintenance information, vehicle status information, and battery power consumption habit information of the vehicle according to a preset battery life impact factor;

[0007] Obtain the battery failure prediction result according to the battery failure prediction model, the battery maintenance information, the vehicle status information, and the battery power consumption habit information; wherein, the battery failure prediction model includes a full - volume feature analysis sub - model, a category feature analysis sub - model, and a failure prediction sub - model;

[0008] Manage the vehicle's battery according to the battery failure prediction result.

[0009] In some possible designs, the obtaining the battery failure prediction result according to the battery failure prediction model, the battery maintenance information, the vehicle status information, and the battery power consumption habit information includes:

[0010] Obtain a full - volume feature analysis result according to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the full - volume feature analysis sub - model;

[0011] Obtain the category feature analysis result according to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the category feature analysis sub-model;

[0012] Obtain the battery fault prediction result according to the full feature analysis result, the category feature analysis result, and the fault prediction sub-model.

[0013] In some possible designs, the full feature analysis sub-model includes: a gradient boosting decision tree (GBDT) model and a factorization machine (FM) model;

[0014] The obtaining of the full feature analysis result according to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the full feature analysis sub-model includes:

[0015] Perform data processing on the battery maintenance information, the vehicle status information, and the battery power consumption habit information to obtain the input features corresponding to the full feature analysis sub-model, and the data processing includes at least one of data cleaning, single feature analysis, factor analysis, and non-numeric feature conversion;

[0016] Input the input features into multiple decision trees included in the GBDT model;

[0017] According to the FM model, perform feature combination on the leaf nodes of the multiple decision trees to obtain the full feature analysis result.

[0018] In some possible designs, the performing of feature combination on the leaf nodes of the multiple decision trees according to the FM model to obtain the full feature analysis result includes:

[0019] According to the FM model, perform feature combination on the leaf nodes of the multiple decision trees to obtain a combined feature vector;

[0020] According to a fully connected layer, perform feature dimension conversion on the combined feature vector to obtain the full feature analysis result, and the feature dimension of the full feature analysis result is the same as the feature dimension of the category feature analysis result.

[0021] In some possible designs, the category feature analysis sub-model is a deep neural networks (DNN) model, and the category feature analysis sub-model includes a connected embedding layer and multiple fully connected layers;

[0022] Obtaining a category feature analysis result according to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the category feature analysis sub-model, including:

[0023] Extracting target category features corresponding to the preset target category from the extracted battery maintenance information, the vehicle status information, and the battery power consumption habit information according to the preset target category;

[0024] Densifying the input target category features according to the embedding layer to obtain processed target category features;

[0025] Performing feature combination on the processed target category features according to the connected multiple fully-connected layers to obtain the category feature analysis result.

[0026] In some possible designs, fusing the full quantity feature analysis result and the category feature analysis result to obtain a fused feature vector, including:

[0027] Fusing the full quantity feature analysis result and the category feature analysis result according to the connected multiple fully-connected layers, the first weight corresponding to the full quantity feature analysis result, and the second weight corresponding to the category feature analysis result to obtain a fused feature vector.

[0028] In some possible designs, the first weight is equal to the second weight.

[0029] In some possible designs, obtaining the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle according to the preset battery life influencing factors, including:

[0030] Obtaining the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle from a database built on a server according to the preset battery life influencing factors;

[0031] Wherein, the database includes data related to the vehicle, and the data related to the vehicle includes the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle.

[0032] In a second aspect, an embodiment of the present application further provides a battery management device, including:

[0033] An obtaining module, configured to obtain the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle according to the preset battery life influencing factors;

[0034] A processing module, configured to obtain the battery fault prediction result according to the battery fault prediction model, the battery maintenance information, the vehicle status information, and the battery power consumption habit information; wherein, the battery fault prediction model includes a full-feature analysis sub-model, a category-feature analysis sub-model, and a fault prediction sub-model; and manage the vehicle's battery according to the battery fault prediction result.

[0035] In a third aspect, an embodiment of the present application further provides an electronic device, including: a memory and a processor, the memory is coupled to the processor;

[0036] The memory is used to store computer program instructions;

[0037] The processor executes the computer program instructions to execute the method according to any one of the first aspect.

[0038] In a fourth aspect, an embodiment of the present application further provides a readable storage medium, including: computer program instructions;

[0039] When the computer program instructions run on an electronic device, the electronic device is caused to execute the method according to any one of the first aspect.

[0040] In a fifth aspect, an embodiment of the present application further provides a program product, the program product includes a computer program, the computer program is stored in a storage medium, and at least one processor of the battery management device can read the computer program from the storage medium, and the at least one processor executes the computer program to cause the battery management device to execute the method according to any one of the first aspect.

[0041] The embodiments of the present application provide a battery management method, apparatus, and readable storage medium. The method includes: obtaining battery maintenance information, vehicle status information, and battery power consumption habit information of a vehicle according to preset battery life impact factors; obtaining a battery fault prediction result according to a battery fault prediction model, battery maintenance information, vehicle status information, and battery power consumption habit information, where the battery fault prediction model includes a full - feature analysis sub - model, a category - feature analysis sub - model, and a fault prediction sub - model; and managing the battery of the vehicle according to the battery fault prediction result. The embodiments of the present application use the full - feature analysis sub - model pre - trained by a machine learning algorithm to process the full features with sufficient feature combinations, use the category - feature analysis sub - model pre - trained by a machine learning algorithm to process the categorical features, and use the fault prediction sub - model to analyze the combined result of the output of the full - feature analysis sub - module and the output of the category - feature analysis sub - model, so as to obtain a battery fault prediction result with a relatively high accuracy, and further improve the effectiveness of vehicle management. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0043] Figure 1 It is a scenario diagram applicable to the battery management method provided by an embodiment of the present application;

[0044] Figure 2 It is a flowchart of the battery management method provided by an embodiment of the present application;

[0045] Figure 3 It is a schematic structural diagram of the battery fault prediction model provided by an embodiment of the present application;

[0046] Figure 4 It is a schematic structural diagram of the battery fault prediction model provided by another embodiment of the present application;

[0047] Figure 5 It is a flowchart of the battery management method provided by an embodiment of the present application;

[0048] Figure 6 It is an architecture diagram of the battery management method provided by an embodiment of the present application;

[0049] Figure 7 It is a schematic structural diagram of the battery management apparatus provided by an embodiment of the present application;

[0050] Figure 8 This is a schematic structural diagram of a battery management device provided by another embodiment of the present application. Detailed implementation manners

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0052] A battery is an important component of an electric vehicle, and the effective management of the battery is of crucial significance to the normal operation of the electric vehicle. In traditional methods, there are three ways to repair and manage the vehicle's battery: First, after-sales repair, that is, repair when there is a fault. This method belongs to unplanned maintenance and may cause unnecessary losses to the vehicle owner; Second, preventive maintenance, also known as scheduled maintenance, that is, regularly check and repair the vehicle's battery to eliminate possible and hidden fault risks. This method may lead to over-maintenance and improper maintenance, thus bringing many inconveniences to the vehicle management; Third, predictive maintenance, that is, analyze based on historical battery repair information to predict whether the vehicle's battery will fail, and then manage the vehicle's battery. This is also a widely used method at present.

[0053] In traditional methods, when using the "predictive maintenance" battery management method, usually, first obtain the prediction result of battery failure based on a large amount of historical battery repair information, and then determine whether it is necessary to repair the vehicle's battery according to the prediction result of battery failure. The accuracy rate of the battery failure prediction result obtained by the above method is relatively low, thereby affecting the effectiveness of the vehicle's battery management.

[0054] Based on this, the embodiments of the present application provide a battery management method to improve the accuracy of battery fault prediction results. First, the predictive maintenance for the battery in this solution can be defined as a regression and classification task. The regression task is to predict the remaining service life of the vehicle battery; the classification task is to predict whether the vehicle battery will fail within a certain period in the future. Therefore, the core concept of the battery management method provided by the embodiments of the present application is as follows: obtaining a large amount of vehicle status data and battery-related data, and extracting effective features therefrom; using the full-feature analysis sub-model pre-trained by a machine learning algorithm to process the full features with sufficient feature combinations, using the category feature analysis sub-model pre-trained by a machine learning algorithm to process the categorical features, and using the fault prediction sub-model to analyze the combined results of the output of the full-feature analysis sub-module and the output of the category feature analysis sub-model, so as to obtain a battery fault prediction result with a relatively high accuracy. The battery fault prediction result may include the remaining service life of the battery and / or the probability value of the battery failing in the future; based on the battery fault prediction result, the effectiveness of vehicle management can be further improved.

[0055] Figure 1 FIG. is a diagram of the applicable scenario of the battery management method provided by an embodiment of the present application. As Figure 1 shown, this scenario includes: at least one vehicle 101 and a battery management device 102.

[0056] Among them, the vehicle 101 includes: a battery 1010, a sensor assembly 1012, and a communication assembly 1014. The battery 1010 is used to provide electrical energy for each component of the vehicle 101. For example, the battery can provide electrical energy for the multimedia device on the vehicle, and the battery can provide electrical energy for the engine, etc.; the sensor assembly 1012 is used to collect vehicle status information. The sensor assembly 1012 can be a group of sensor assemblies, and each sensor assembly is used to collect different vehicle status information. The vehicle status information includes, for example, but is not limited to, the total vehicle mileage, the total vehicle fuel consumption, the total engine revolutions, the real-time vehicle speed, the longitude and latitude information, the engine speed, the real-time fuel consumption, the brake information, the throttle information, the atmospheric temperature, the coolant temperature, the urea tank temperature, etc.; the communication assembly 1014 can send the vehicle status data collected by the sensor assembly 1012 to the battery management device 102. The communication assembly 1014 is, for example, an in-vehicle T-BOX device, and the in-vehicle T-BOX device can perform data interaction with the battery management device 102 through a wireless network.

[0057] The battery management device 102 is configured to obtain the status information of the vehicle collected by the sensor assembly 1012 of the vehicle 101; and obtain the battery maintenance information and battery power consumption habit information of the vehicle 101; and based on the obtained vehicle status information, battery maintenance information, battery power consumption habit information, and a pre-trained battery fault prediction model, obtain a battery fault prediction result, and perform vehicle management according to the battery fault prediction result.

[0058] The battery management device 102 may be a server, and the server may be a group of servers, or multiple groups of servers, and may be one type or multiple types of servers; alternatively, the battery management device 102 may also be implemented by software and run on a cloud server cluster, and the embodiments of the present application do not limit this. Wherein, when the battery management device 102 runs on a cloud server cluster, a database may be built on the cloud server cluster, and the database is used to store data related to the battery of the vehicle 101. The data related to the battery may include, for example, battery maintenance data and battery power consumption habit information. The battery management device 102 may obtain the battery maintenance data and battery power consumption habit information of the vehicle 101 from the database.

[0059] It should be understood that the database may also be used to store other vehicle-related data, such as vehicle status information reported by the vehicle, vehicle scheduling information, vehicle maintenance information, and so on. It should be noted that the vehicle maintenance information includes but is not limited to the battery maintenance information of the vehicle. Exemplarily, it may also include maintenance information of other components of the vehicle except the battery, such as engine maintenance information and tire maintenance information.

[0060] In some cases, it may further include: a terminal device 103, and the terminal device 103 is the terminal device corresponding to the vehicle 101. If the battery management device 102 determines that the battery fault prediction result of the vehicle 101 meets a preset condition according to the battery fault prediction result, the battery management device 102 may send a prompt message to the terminal device 103, and the terminal device 103 displays the received prompt message, so that the user (such as the vehicle owner) and the driver can obtain the prompt message in time, and can check the status of the battery according to the prompt message, thereby ensuring the safe and reliable operation of the vehicle.

[0061] The battery management method provided by the embodiments of the present application will be introduced in detail through several specific embodiments below.

[0062] Figure 2 It is a flowchart of the battery management method provided by an embodiment of the present application. In the embodiments of the present application, the execution subject is taken as the battery management device for illustration. As Figure 2 shown, the method of this embodiment includes:

[0063] S101. Obtain the battery maintenance information, vehicle status information, and battery power consumption habit information of the vehicle according to the preset battery life impact factors.

[0064] First, the analysis of the battery life impact factors in this solution may include:

[0065] 1. Vehicle condition, that is, the condition of the vehicle. For example, the running time of the engine, the remaining battery power, and so on.

[0066] 2. Road condition, that is, the road condition. For example, dangerous sections, uphill and downhill, etc.

[0067] 3. Driving behavior, such as the number of engine starts, the mileage between two adjacent engine starts, entertainment time, braking behavior, etc.

[0068] 4. The user's vehicle usage habits (which can reflect the battery power consumption habits), such as the duration of using in-vehicle electronic devices after the vehicle is turned off, etc.

[0069] Based on the above factors, obtain the battery maintenance information, vehicle status information, and battery power consumption habit information of the vehicle from the relevant original data of the vehicle.

[0070] Among them, the battery maintenance information of the vehicle can be extracted from the vehicle's maintenance information, and the battery maintenance information may include: maintenance time, fault description, fault cause, fault handling solution, and so on.

[0071] A possible implementation method is that the battery management device can obtain the battery maintenance information from the database built in the cloud server cluster environment.

[0072] The vehicle status data may include vehicle static information, such as but not limited to the total mileage, total fuel consumption, total engine revolutions of the vehicle, etc. The vehicle status data may also include vehicle real-time status information, such as but not limited to real-time vehicle speed, longitude and latitude information, engine speed, real-time fuel consumption, braking information, throttle information, atmospheric temperature, coolant temperature, urea tank temperature, etc.

[0073] Among them, if the method shown in Figure 1 is adopted, the vehicle status information is collected through the vehicle's sensor component, and the sensor component can collect the vehicle status information periodically or at specific moments.

[0074] A possible implementation method is that the collected vehicle status information can be sent to the database built in the cloud server cluster environment through the vehicle's communication component, and the battery management device can obtain the vehicle status information from the database; another possible implementation method is that the vehicle's communication component sends the collected vehicle status information to the battery management device.

[0075] The battery power usage habit information is used to reflect the user's habit of using the vehicle's battery. The user's habit of using the vehicle's battery has a relatively serious impact on the service life of the battery. For example, some users use the electronic devices in the vehicle for a long time after the vehicle is turned off, or use electronic devices with a relatively large power for a long time, causing the battery to discharge and shortening the service life of the battery. Exemplarily, the battery power usage habit information includes, but is not limited to: the battery discharge duration after the vehicle is turned off, the battery discharge frequency after the vehicle is turned off, the duration and frequency of the battery discharge current being greater than a preset threshold, etc.

[0076] In a possible implementation, the battery management device obtains the battery power usage habit information from a database built in a cloud server cluster environment.

[0077] S102. Obtain a battery fault prediction result according to the battery fault prediction model, the battery repair information, the vehicle status information, and the battery power usage habit information; wherein, the battery fault prediction model includes a full - quantity feature analysis sub - model, a category feature analysis sub - model, and a fault prediction sub - model.

[0078] In a possible implementation, as shown in Figure 3 The battery fault prediction model includes a full - quantity feature analysis sub - model, a category feature analysis sub - model, and a fault prediction sub - model. The output terminals of the full - quantity feature analysis sub - model and the category feature analysis sub - model are both connected to the input terminal of the fault prediction sub - model. The full - quantity feature analysis sub - model and the category feature analysis sub - model respectively correspond to corresponding input features. Among them, the full - quantity feature analysis sub - model is used to output a full - quantity feature analysis result according to the corresponding input features; the category feature analysis sub - model is used to output a category feature analysis result according to the corresponding input features; the fault prediction sub - model is used to fuse the full - quantity feature analysis result and the category feature analysis result, and obtain a battery fault prediction result according to the fused feature vector, where the battery fault prediction result is used to indicate the possibility of the vehicle's battery having a fault.

[0079] In a possible implementation, the battery fault prediction result may include a probability value of the battery having a fault. Among them, the higher the probability value of the battery having a fault, the greater the possibility of the battery having a fault; the lower the probability value of the battery having a fault, the lower the possibility of the battery having a fault. By numericalizing the battery fault prediction result, the battery fault prediction result is more intuitive and is conducive to the user quickly determining whether the vehicle's battery will have a fault.

[0080] In some other possible implementation manners, the battery fault prediction result may include the remaining life of the battery. The longer the remaining life of the battery is, the lower the probability of the battery failing; the shorter the remaining life of the battery is, the higher the probability of the battery failing.

[0081] In some other possible implementation manners, the battery fault prediction result may include both the probability value of the battery failing and the remaining life of the battery.

[0082] In Figure 3 Based on the structure of the battery fault prediction model shown in Figure 4 As shown in a possible implementation manner, the full - scale feature analysis sub - model includes: a GBDT model and a factorization machine FM model. Among them, the output end of the GBDT model is connected to the input end of the FM model, and the output end of the FM model is connected to the input end of the fault prediction sub - model through a fully - connected layer. Since a large number of sparse features output by the GBDT model cannot be directly connected to the fully - connected layer, an FM model is set between the GBDT model and the fully - connected layer. The FM model processes the sparse features output by the GBDT model, and the fully - connected layer after the FM model recombines and processes the features output by the FM model to ensure that the feature dimension of the full - scale feature analysis result output by the full - scale feature analysis sub - model is consistent with the feature dimension of the category feature analysis result output by the category feature analysis sub - model.

[0083] Among them, the input features of the full - scale feature analysis sub - model can be obtained through data processing such as data pre - processing, data cleaning, single - feature analysis, factor analysis, and data conversion on the original data in the database. Among them, data pre - processing can be to conduct a preliminary analysis and screening of the original data to obtain data in three aspects: vehicle maintenance information, vehicle status information, and battery power consumption habits information. The purpose of data cleaning is to determine whether the data is available, which can specifically include the processing of outliers and missing values, etc. Single - feature analysis includes, for a certain feature, comparing the data of vehicles with battery failures and the data of vehicles without battery failures, analyzing the differences between the average value, median, maximum value, minimum value, etc., to determine whether there are differences. If the differences are not significant, it means that these data have no practical effect on battery fault prediction and can be deleted from the data set. Data conversion includes: converting non - numerical features into numerical features. For example, converting categorical features in strings, such as vehicle models and vehicle series, into numerical features to facilitate subsequent model learning.

[0084] Optionally, as Figure 4As shown, the category feature analysis sub-model can be a DNN model, and the DNN model includes: an embedding layer and multiple connected fully-connected layers. Among them, the embedding layer is used to densify the input target category features, and the multiple connected fully-connected layers are used to combine the processed target category features output by the embedding layer and output the category feature analysis result.

[0085] Specifically, the input features corresponding to the category feature analysis sub-model can be target category features with category properties extracted from the original data in the database, and the target category features can be represented in the form of target category feature vectors. In practical applications, the accuracy of the battery fault prediction result obtained by the category feature analysis sub-model based on the input target category feature vector with category properties is higher than that obtained by the category feature analysis sub-model based on the input full-scale feature vector. Since the GBDT model has already performed sufficient feature combination, therefore, inputting the full-scale feature vector to the category feature analysis sub-model has a relatively small impact on the accuracy of the battery fault prediction result; therefore, this solution starts from another dimension and inputs the target category feature vector with category properties to the category feature analysis sub-model to improve the accuracy of the battery fault prediction result.

[0086] In a possible implementation, the number of fully-connected layers included in the category feature analysis sub-model is 2. This solution ensures the accuracy of the result output by the category feature analysis sub-model by setting an appropriate number of fully-connected layers, while reducing the computational complexity and improving the battery fault prediction efficiency.

[0087] Optionally, as Figure 4 shown, the fault prediction sub-model includes: a feature fusion layer, multiple connected fully-connected layers, and an activation layer (sigmoid layer). Among them, the feature fusion layer (concat layer) is used to fuse the full-scale feature analysis result output by the fully-connected layer included in the full-scale feature analysis sub-model and the category feature analysis result output by the fully-connected layer included in the category feature analysis sub-model. The feature dimension of the full-scale feature analysis result output by the fully-connected layer included in the full-scale feature analysis sub-model is the same as the feature dimension of the category feature analysis result output by the fully-connected layer included in the category feature analysis sub-model; the feature fusion layer inputs the fused feature vector to the multiple connected fully-connected layers, and the multiple connected fully-connected layers perform feature combination on the fused feature vector. The multiple connected fully-connected layers output the feature vector after feature combination to the sigmoid layer, and the sigmoid layer outputs the battery fault prediction result according to the feature vector after feature combination, for example, the battery fault prediction probability value.

[0088] Optionally, the feature fusion layer may fuse the full - volume feature analysis result and the category feature analysis result according to the first weight corresponding to the full - volume feature analysis result and the second weight corresponding to the category feature analysis result to obtain a fused feature vector. By comprehensively considering the influence of the full - volume feature analysis result on the battery fault prediction result and the influence of the category feature analysis result on the battery fault prediction result, setting the first weight and the second weight is beneficial to improving the accuracy of the battery fault prediction result.

[0089] It should be understood that Figure 4 the multiple fully - connected layers shown in can be the same or have differences, and the specific parameters of each fully - connected layer are set according to the actual situation.

[0090] Optionally, the first weight is equal to the second weight. Of course, the first weight and the second weight can also be unequal and can be set according to actual needs. The embodiments of the present application are not limited to this.

[0091] S103. Manage the vehicle's battery according to the battery fault prediction result.

[0092] In a possible implementation, the battery management device can determine the probability that the vehicle's battery fails according to the battery fault prediction result; and can determine whether to send a prompt message according to the probability that the vehicle's battery fails and a preset condition, where the prompt message is used to prompt the user that the probability of the vehicle's battery failure is relatively high. In addition, the prompt message can also include information prompting the user to check whether the indicators of the battery are normal.

[0093] Optionally, the preset condition can be numerically converted into a specific value. If the probability that the battery fails is higher than the preset threshold, the battery management device sends a prompt message to the client. If the probability that the battery fails is less than or equal to the preset threshold, the battery management device can determine that the battery can operate normally.

[0094] Exemplarily, when the probability that the battery fails is higher than the preset threshold, the battery management device sends a prompt message to the client. The prompt message is, for example, "The probability of the vehicle's battery failure is relatively high. It is recommended to check whether the battery is normal."

[0095] In some possible cases, the battery management device can also print the prompt message in the log to remind the back - end staff.

[0096] In this embodiment, battery maintenance information, vehicle status information, and battery power consumption habit information of the vehicle are obtained; according to the battery fault prediction model, battery maintenance information, vehicle status information, and battery power consumption habit information, a battery fault prediction result is obtained, where the battery fault prediction model includes a full - feature analysis sub - model, a category - feature analysis sub - model, and a fault prediction sub - model; according to the battery fault prediction result, the battery of the vehicle is managed. In the embodiment of the present application, the full - feature analysis sub - model pre - trained by a machine learning algorithm is used to process the full - features with sufficient feature combinations, the category - feature analysis sub - model pre - trained by a machine learning algorithm is used to process the categorical features, and the fault prediction sub - model is used to analyze the combined result of the output of the full - feature analysis sub - module and the output of the category - feature analysis sub - model, so as to obtain a battery fault prediction result with a higher accuracy, thereby improving the effectiveness of vehicle management.

[0097] In the above Figure 2 In the shown embodiment, the battery fault prediction model is pre - trained by a machine learning algorithm. Figure 5 It is a flowchart of a battery management method provided by an embodiment of the present application, where Figure 5 how to obtain the battery fault prediction model is introduced in detail in the shown embodiment.

[0098] Specifically, it may include the following steps:

[0099] Step 1: Obtain the original vehicle maintenance information and vehicle status information from the database, and obtain an effective data set according to the original vehicle maintenance information and vehicle status information (that is, Figure 5 the data set shown in

[0100] Among them, the original vehicle maintenance information includes vehicle battery maintenance information and maintenance information of other components of the vehicle. Therefore, it is necessary to screen and analyze the original vehicle maintenance information to obtain the battery maintenance information.

[0101] The vehicle status information includes vehicle real - time status information and vehicle static information. Among them, the amount of data of the vehicle real - time status information is large and the degree of dispersion is high. Therefore, it is necessary to analyze the vehicle real - time status information to obtain effective data, such as aggregated category - feature data: the average vehicle speed, average mileage, fuel consumption per 100 kilometers corresponding to the vehicle in 1 year, 1 month, and 1 day respectively.

[0102] The processed vehicle real - time status information, battery maintenance information, vehicle static information, and vehicle usage habit information obtained from the database are merged according to the vehicle identifier (such as the vehicle ID) and time stamp (the time stamp can be, for example, the time stamp carried when the information is generated), so as to obtain an effective data set.

[0103] Step 2: Perform data processing on the data included in the valid data set, such as data preprocessing, data cleaning, single-feature analysis, factor analysis, and data transformation (i.e., non-numeric feature transformation).

[0104] The purpose of this step is to: based on the preset battery life impact factors, perform data processing on the valid data set and extract valid features. The extracted valid features may include: vehicle maintenance information, vehicle status information, and vehicle operation information (i.e., battery power consumption habits information), corresponding features in these three aspects.

[0105] Among them, the battery life impact factors may include: 1. Vehicle condition, that is, the condition of the vehicle. For example, the engine running duration, the remaining battery power, etc. 2. Road condition, that is, the road condition. For example, dangerous sections, uphill and downhill, etc. 3. Driving behavior, such as the number of engine starts, the mileage between two adjacent engine starts, entertainment time, braking behavior, etc. 4. The user's vehicle usage habits (which can reflect the battery power consumption habits), such as the duration of using in-vehicle electronic devices after the vehicle is turned off.

[0106] In this step, the purpose of data cleaning is to determine whether the data is available, which may specifically include the processing of outliers and missing values, etc. Single-feature analysis includes, for a certain feature, comparing the data of vehicles with battery failures and the data of vehicles without battery failures, analyzing the differences between the mean, median, maximum, minimum, etc., and determining whether there are differences. If the differences are not significant, it means that these data have no practical effect on battery failure prediction and can be deleted from the data set. Data transformation includes: converting non-numeric features into numeric features. For example, converting categorical features of strings, such as vehicle models and vehicle series, into numeric features to facilitate subsequent model learning.

[0107] Step 3: Train and test the model.

[0108] Among them, the data output in Step 2 can be divided into a training set and a test set. The data included in the training set is used for model training and parameter optimization, and the data included in the test set is used to verify whether the performance of the model meets the preset conditions.

[0109] In the process of training the model, by using different machine learning algorithms for model training and parameter optimization, continuously improve the prediction performance of the model, and continuously improve the model structure. In this solution, the improved model structure can be as shown in Figure 3 and Figure 4 as shown in, and a combination of GBDT model, FM model, and DNN model is adopted.

[0110] In practical applications, since there is a significant difference in the number of maintenance vehicles and normal vehicles, in order to ensure the prediction performance of the model obtained through training, in this solution, F1 and AUC are used as evaluation indicators for the model prediction performance.

[0111] Among them, F1, that is, the F1 score, is an indicator used in statistics to measure the accuracy of binary classification models. F1 takes into account both the accuracy and recall rate of the classification model. The larger the value of F1, the higher the prediction performance of the model; the smaller the value of F1, the lower the prediction performance of the model.

[0112] AUC (area under curve) is defined as the area enclosed by the ROC curve and the coordinate axes. In this solution, the larger the value of AUC, the higher the prediction performance of the model; the lower the value of AUC, the lower the prediction performance of the model.

[0113] In practical applications, when the battery fault prediction model uses the GBDT model alone, the value of AUC of the battery fault prediction model is 0.69; when the battery fault prediction model uses the DNN model alone, the value of AUC of the battery fault prediction model is 0.71; when the battery fault prediction model uses a combination of the GBDT model, FM model, and DNN model, the value of AUC of the battery fault prediction model is 0.75. Thus, it can be seen that in practical applications, using a combination of the GBDT model, FM model, and DNN model for the battery fault prediction model can obtain better prediction performance. Among them, for the specific structure of the model, reference can be made to Figure 3 and Figure 4 the description of the embodiments shown, which will not be elaborated here.

[0114] Step Four: Deploy the model.

[0115] After the battery fault prediction model is trained, package the battery fault prediction model into a battery fault prediction service program and deploy it on an online server. Use this battery fault prediction service to obtain the battery fault prediction results of online vehicles.

[0116] Figure 6 This is the architecture diagram of the battery management method provided by an embodiment of this application. As Figure 6 shown, it may include: a running environment, a database layer, a business data layer, a data processing layer, a model prediction and output layer.

[0117] Among them, the running environment can be a cloud server cluster.

[0118] The database layer is a database service built on the cloud server; the database layer can include one or more databases, and each database can be used to store different types of data.

[0119] Exemplarily, the database layer includes three databases, namely Database 1, Database 2, and Database 3. Among them, Database 1 is used to store the maintenance information of the vehicle, Database 2 is used to store the status information returned by the vehicle, and Database 3 is used to store the driving habits information of the user.

[0120] It should be noted that the database layer can also run on an independent server, or it can also run in an environment composed of a cloud server and an independent server. This application does not limit this. Figure 6 The embodiments shown are only examples and do not limit the operating environment.

[0121] The business data layer is the original data layer associated with battery prediction. Among them, the original data associated with battery prediction can include: vehicle static information, weather information, road network information data, and so on.

[0122] The data processing layer processes the original data to generate feature data that can be applied to the battery fault prediction model. Among them, the data processing layer can perform data preprocessing, data cleaning, single feature analysis and factor analysis, data conversion (i.e., non-numeric feature conversion), and other data processing on the original data.

[0123] In a possible implementation, the process of the data processing layer processing the original data can be an online automated feature extraction process. The feature extraction script can process the original data into feature data that meets the model input requirements, and the feature extraction script will store these processed feature data in the feature database.

[0124] Optionally, the feature extraction script can run periodically according to requirements. For example, the feature extraction script runs once a day; the feature extraction script can also run at a specific time. For example, the feature extraction script can run once at 8:00 am and 6:00 pm every day. This application embodiment does not limit this.

[0125] The model prediction and output layer inputs the feature data output by the data processing layer into the battery fault prediction model and outputs the battery fault prediction result. Among them, the battery fault prediction model can be obtained through Figure 5 the method shown.

[0126] Specifically, in this solution, after the battery fault prediction model is trained, it can be packaged into a battery fault prediction service program so as to deploy the battery fault prediction model to run on an online server. After the battery fault prediction service program is deployed on the online server, the battery fault prediction service program can obtain feature data from the database in a set manner, and then obtain the battery fault prediction results of all vehicles online. Among them, the battery fault prediction result is a value between 0 and 1, and the battery fault prediction results of all vehicles can be stored in the corresponding prediction result list.

[0127] The battery management script will screen out the vehicles whose battery fault probability exceeds the preset threshold from the prediction result list, and send a prompt message to the user corresponding to the vehicle (such as Figure 1 the terminal device corresponding to the vehicle shown in

[0128] Figure 7 FIG. is a schematic structural diagram of a battery management device provided by an embodiment of the present application. As Figure 7 shown, the battery management device 700 provided in this embodiment includes: an acquisition module 701 and a processing module 702.

[0129] The acquisition module 701 is configured to obtain the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle according to the preset battery life influencing factors;

[0130] The processing module 702 is configured to obtain the battery fault prediction result according to the battery fault prediction model, the battery maintenance information, the vehicle status information, and the battery power consumption habit information; wherein, the battery fault prediction model includes a full-scale feature analysis sub-model, a category feature analysis sub-model, and a fault prediction sub-model; and manage the battery of the vehicle according to the battery fault prediction result.

[0131] In some possible designs, the processing module 702 is specifically configured to:

[0132] Obtain a full-scale feature analysis result according to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the full-scale feature analysis sub-model; obtain a category feature analysis result according to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the category feature analysis sub-model; and obtain the battery fault prediction result according to the full-scale feature analysis result, the category feature analysis result, and the fault prediction sub-model.

[0133] In some possible designs, the full - scale feature analysis sub - model includes: a GBDT model and an FM model; a processing module 702, specifically configured to: perform data processing on the battery maintenance information, the vehicle status information, and the battery power consumption habit information to obtain input features corresponding to the full - scale feature analysis sub - model, where the data processing includes at least one of data cleaning, single - feature analysis, factor analysis, and non - numerical feature conversion; input the input features into multiple decision trees included in the GBDT model; and perform feature combination on leaf nodes of the multiple decision trees according to the FM model to obtain the full - scale feature analysis result.

[0134] In some possible designs, the processing module 702 is specifically configured to: perform feature combination on leaf nodes of the multiple decision trees according to the FM model to obtain a combined feature vector; and perform feature - dimension conversion on the combined feature vector according to a fully - connected layer to obtain the full - scale feature analysis result, where the feature dimension of the full - scale feature analysis result is the same as that of the category feature analysis result.

[0135] In some possible designs, the category feature analysis sub - model is a deep neural network DNN model, and the category feature analysis sub - model includes a connected embedding layer and multiple fully - connected layers;

[0136] The processing module 702 is specifically configured to extract target category features corresponding to a preset target category from the battery maintenance information, the vehicle status information, and the battery power consumption habit information according to the preset target category; perform densification processing on the input target category features according to the embedding layer to obtain processed target category features; and perform feature combination on the processed target category features according to the connected multiple fully - connected layers to obtain the category feature analysis result.

[0137] In some possible designs, the processing module 702 is specifically configured to fuse the full - scale feature analysis result and the category feature analysis result according to the connected multiple fully - connected layers, a first weight corresponding to the full - scale feature analysis result, and a second weight corresponding to the category feature analysis result to obtain a fused feature vector.

[0138] In some possible designs, the first weight is equal to the second weight.

[0139] In some possible designs, the obtaining module 701 is specifically configured to obtain the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle from a database built on a server according to a preset battery life impact factor; wherein, the database includes data related to the vehicle, and the data related to the vehicle includes the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle.

[0140] The battery management device provided by the embodiment of the present application can be used to execute the technical solutions of any of the above method embodiments, and its implementation principle and technical effects are similar. For reference, please refer to the description of the foregoing method embodiments, and details will not be repeated here.

[0141] Figure 8 FIG. is a schematic structural diagram of a battery management device provided by another embodiment of the present application. The battery management device 800 provided in this embodiment includes: a memory 801, a processor 802, and computer program instructions.

[0142] Among them, the computer program instructions are stored in the memory 801 and are configured to be executed by the processor 802. When the processor 802 executes the computer program instructions, the battery management device executes the technical solutions of any of the above method embodiments. The relevant descriptions can be understood by referring to the corresponding descriptions and effects of the steps in the method embodiments, and details will not be elaborated here.

[0143] Among them, in this embodiment, the memory 801 and the processor 802 are connected through a bus 803.

[0144] In some possible designs, the battery management device 800 further includes an interface 804 for communicating with other devices. For example, the battery management device 800 can communicate with an in-vehicle T-BOX device through the interface 804.

[0145] The embodiment of the present invention further provides a readable storage medium, on which a computer program is stored. When the computer program is executed by an electronic device, the electronic device executes the technical solutions of any of the above method embodiments.

[0146] The embodiment of the present invention further provides a program product, the program product includes a computer program, the computer program is stored in a readable storage medium, and at least one processor of the battery management device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to enable the battery management device to execute the technical solutions of any of the above method embodiments.

[0147] In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of devices or modules can be in electrical, mechanical or other forms.

[0148] The modules described as separate components may or may not be physically separated. The components displayed as modules may or may not be physical modules, that is, they can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0149] In addition, the functional modules in each embodiment of the present invention can be integrated in a processing module, or each module can exist physically alone, or two or more modules can be integrated in one module. The above integrated modules can be implemented in the form of hardware or in the form of a combination of hardware and software functional modules.

[0150] The program code for implementing the method of the embodiments of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to the processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing devices, so that when the program codes are executed by the processor or controller, the functions / operations specified in the flowchart and / or block diagram are implemented. The program codes can be executed entirely on the machine, partially on the machine, executed partially on the machine as an independent software package and partially on a remote machine, or executed entirely on a remote machine or server.

[0151] In the context of the present invention, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0152] In addition, although the operations are depicted in a particular order, this should be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed to achieve the desired result. In certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the foregoing discussion, these should not be construed as limiting the scope of the present disclosure. Certain features described in the context of separate embodiments can also be implemented in combination in a single implementation. Conversely, the various features described in the context of a single implementation can also be implemented separately or in any suitable sub-combination in multiple implementations.

[0153] Finally, it should be noted that although the subject matter has been described in language specific to structural features and / or methodological logical acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. On the contrary, the specific features and acts described above are merely example forms of implementing the claims; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A battery management method, characterized in that, Including: Obtain the battery maintenance information, vehicle status information, and battery power consumption habit information of the vehicle according to the preset battery life impact factors; Obtain the battery fault prediction result according to the battery fault prediction model, the battery maintenance information, the vehicle status information, and the battery power consumption habit information; wherein, the battery fault prediction model includes a full - quantity feature analysis sub - model, a category feature analysis sub - model, and a fault prediction sub - model; the full - quantity feature analysis sub - model includes a GBDT model and an FM model, the GBDT model is used to output sparse features, and the FM model is used to process the sparse features and output the full - quantity feature analysis result; the category feature analysis sub - model is a deep neural network DNN model, including an embedding layer and multiple fully - connected layers, the embedding layer is used to perform densification processing on the input target category features, and the connected multiple fully - connected layers are used to perform feature combination on the processed target category features output by the embedding layer and output the category feature analysis result; the fault prediction sub - model includes: a feature fusion layer, multiple connected fully - connected layers, and an activation layer, the feature fusion layer fuses the full - quantity feature analysis result and the category feature analysis result, the connected multiple fully - connected layers perform feature combination on the fused feature vector and output the feature - combined feature vector to the activation layer, and the activation layer outputs the battery fault prediction result according to the feature - combined feature vector; manage the battery of the vehicle according to the battery fault prediction result.

2. The method according to claim 1, wherein The step of obtaining the battery fault prediction result according to the battery fault prediction model, the battery maintenance information, the vehicle status information, and the battery power consumption habit information includes: Obtain the full - quantity feature analysis result according to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the full - quantity feature analysis sub - model; Obtain the category feature analysis result according to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the category feature analysis sub - model; obtain the battery fault prediction result according to the full - quantity feature analysis result, the category feature analysis result, and the fault prediction sub - model; The step of obtaining the battery fault prediction result according to the full - quantity feature analysis result, the category feature analysis result, and the fault prediction sub - model includes: Fuse the full - quantity feature analysis result and the category feature analysis result according to the fault prediction sub - model to obtain a fused feature vector; Obtain the battery fault prediction result according to the fused feature vector.

3. The method according to claim 2, wherein The step of obtaining the full - quantity feature analysis result according to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the full - quantity feature analysis sub - model includes: Perform data processing on the battery maintenance information, the vehicle status information, and the battery power consumption habit information to obtain the input features corresponding to the full-feature analysis sub-model. The data processing includes at least one of data cleaning, single-feature analysis, factor analysis, and non-numerical feature conversion; Input the input features into multiple decision trees included in the GBDT model; According to the FM model, perform feature combination on the leaf nodes of the multiple decision trees to obtain the full-feature analysis result.

4. The method according to claim 3, wherein The step of "According to the FM model, perform feature combination on the leaf nodes of the multiple decision trees to obtain the full-feature analysis result" includes: According to the FM model, perform feature combination on the leaf nodes of the multiple decision trees to obtain a combined feature vector; According to the fully connected layer, perform feature dimension conversion on the combined feature vector to obtain the full-feature analysis result. The feature dimension of the full-feature analysis result is the same as the feature dimension of the category feature analysis result.

5. The method according to claim 2, characterized in that, The category feature analysis sub-model is a deep neural network DNN model, and the category feature analysis sub-model includes a connected embedding layer and multiple fully connected layers; The step of "According to the battery maintenance information, the vehicle status information, the battery power consumption habit information, and the category feature analysis sub-model, obtain the category feature analysis result" includes: According to the preset target category, extract the target category features corresponding to the preset target category from the battery maintenance information, the vehicle status information, and the battery power consumption habit information; According to the embedding layer, perform densification processing on the input target category features to obtain processed target category features; According to the connected multiple fully connected layers, perform feature combination on the processed target category features to obtain the category feature analysis result.

6. The method according to any one of claims 2 to 5, characterized in that, The step of "Fuse the full-feature analysis result and the category feature analysis result to obtain a fused feature vector" includes: According to the connected multiple fully connected layers, the first weight corresponding to the full-feature analysis result, and the second weight corresponding to the category feature analysis result, fuse the full-feature analysis result and the category feature analysis result to obtain a fused feature vector.

7. The method according to any one of claims 1 to 5, characterized in that, The step of "According to the preset battery life impact factor, obtain the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle" includes: According to the preset battery life impact factor, obtain the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle from the database built on the server; Wherein, the database includes data related to the vehicle, and the data related to the vehicle includes the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle.

8. A battery management device, characterized in that, It includes: An acquisition module for obtaining the battery maintenance information, the vehicle status information, and the battery power consumption habit information of the vehicle according to the preset battery life impact factor; A processing module, configured to obtain the battery fault prediction result according to the battery fault prediction model, the battery maintenance information, the vehicle status information, and the battery power consumption habit information; wherein, the battery fault prediction model includes a full - scale feature analysis sub - model, a category feature analysis sub - model, and a fault prediction sub - model; the full - scale feature analysis sub - model includes a GBDT model and an FM model, the GBDT model is used to output sparse features, and the FM model is used to process the sparse features and output the full - scale feature analysis result; the category feature analysis sub - model is a deep neural network DNN model, including an embedding layer and multiple fully - connected layers, the embedding layer is used to perform densification processing on the input target category features, and the connected multiple fully - connected layers are used to perform feature combination on the processed target category features output by the embedding layer and output the category feature analysis result; the fault prediction sub - model includes: a feature fusion layer, multiple connected fully - connected layers, and an activation layer, the feature fusion layer fuses the full - scale feature analysis result and the category feature analysis result, the connected multiple fully - connected layers perform feature combination on the fused feature vector and output the feature - combined feature vector to the activation layer, and the activation layer outputs the battery fault prediction result according to the feature - combined feature vector, and outputs the battery fault prediction result through the activation layer; and manage the battery of the vehicle according to the battery fault prediction result.

9. An electronic device, characterized in that, Comprising: A memory and a processor, the memory is coupled with the processor; The memory is used to store computer program instructions; The processor executes the computer program instructions to execute the method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that, Comprising: Computer program instructions; When the computer program instructions run on an electronic device, the electronic device is enabled to execute the method according to any one of claims 1 to 7.

Images