Method and apparatus for training model, method and apparatus for estimating remaining charging time, and electronic device

Abstract

The present application relates to the technical field of charging estimation. Provided are a method and apparatus for training a model, a method and apparatus for estimating a remaining charging time, and an electronic device. The method for training a model comprises: collecting historical charging data of a vehicle; dividing the overall charging process of a battery into a plurality of charging paths, performing feature extraction on the basis of the historical charging data and according to the charging paths obtained after division, and performing sorting to obtain charging sample data of each charging path; and using the charging sample data of each charging path to train a preset time estimation model, so as to obtain a trained remaining-charging-time estimation model. In the present application, by means of the superposition of a plurality of charging paths into which an overall charging process is divided, the charging time of each charging path can be dynamically calculated, thereby improving the estimation accuracy of the remaining charging time.

Description

Model training method, remaining charging time estimation method, device and electronic equipment

[0001] Cross-reference to Related Applications

[0002] This application claims priority to Chinese Patent Application No. CN202410719226.0, filed on June 5, 2024, entitled “Model training method, remaining charging time estimation method, device and electronic equipment”, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD

[0003] The present application belongs to the technical field of charging estimation, in particular to a model training method, a remaining charging time estimation method, a device and an electronic equipment. BACKGROUND

[0004] With the rapid development of electric vehicle technology, as one of the core components, the performance of the battery management system directly affects the use efficiency and user experience of electric vehicles. The estimation of the remaining charging time (RCT) is an important function in the battery management system, which can help users to reasonably arrange the charging plan.

[0005] Currently, the estimation of the remaining charging time mainly relies on static battery models and simple linear or nonlinear regression methods. These methods are often based on standard charging curves under laboratory conditions, and the remaining charging time is estimated by measuring the instantaneous voltage, current and temperature of the battery and other parameters. However, due to the influence of battery aging, environmental temperature changes and other factors, the accuracy of existing estimation based on fixed models and parameters is not high. Therefore, there is an urgent need for a solution that can improve the accuracy of the remaining charging time estimation.

[0006] SUMMARY

[0007] The present application provides a model training method, a remaining charging time estimation method, a device and an electronic equipment to improve the accuracy of the remaining charging time estimation.

[0008] In a first aspect, the embodiments of the present application provide a model training method, comprising:

[0009] Collecting historical charging data of a vehicle;

[0010] Dividing the overall charging process of the battery into multiple charging paths, and performing feature extraction based on the historical charging data according to the divided charging paths to obtain charging sample data of each charging path;

[0011] Training a preset time estimation model using the charging sample data of each charging path to obtain a trained remaining charging time estimation model;

[0012] The charging sample data includes sample feature data and a corresponding charging result label in a current charging path, and the charging result label includes charging time and charging temperature rise.

[0013] In the embodiments of the present application, by dividing the overall charging process into multiple charging paths, the charging time of each charging path can be dynamically calculated, and the estimation accuracy of the remaining charging time is improved.

[0014] In some possible embodiments, the sample feature data includes at least one of a charging path, a vehicle model, a charging type, an ambient temperature, a battery initial temperature, a battery health value, and a current matching degree.

[0015] The current matching degree is a ratio of an actual battery charging current to a battery charging request current in the corresponding charging path.

[0016] In the embodiments of the present application, one or more feature data of the charging path, the vehicle model, the charging type, the ambient temperature, the battery initial temperature, the battery health value, and the current matching degree are extracted, so that the estimation accuracy of the remaining charging time can be further improved.

[0017] In some possible embodiments, after the preset time estimation model is trained by using the charging sample data of each charging path to obtain the trained remaining charging time estimation model, the method further includes:

[0018] The historical charging data of the vehicle is continuously collected, and updated charging sample data is obtained based on the newly collected historical charging data;

[0019] The remaining charging time estimation model is iteratively trained by using the updated charging sample data according to a preset period.

[0020] In the embodiments of the present application, by continuously collecting charging data of various vehicles, updating sample data according to a set period, and iteratively training the estimation model, the estimation accuracy of the remaining charging time can be further improved.

[0021] In a second aspect, the embodiments of the present application provide a remaining charging time estimation method, including:

[0022] In response to a charging time estimation request of a target vehicle, charging feature data is obtained based on the charging time estimation request;

[0023] A to-be-estimated total charging path corresponding to the charging time estimation request is determined.

[0024] for each charging path corresponding to the total charging path to be estimated, based on the charging feature data, a pre-trained remaining charging time estimation model is used to sequentially obtain the charging time of each charging path;

[0025] Based on the obtained charging time, the remaining charging time estimation result corresponding to the charging time estimation request is obtained by statistics;

[0026] The pre-trained remaining charging time estimation model is trained by any one of the model training methods; and the data items of the charging feature data and the data items of the sample feature data correspond to each other.

[0027] In the embodiments of the present application, the charging time of each charging path is obtained by using the pre-trained model to statistically obtain the overall remaining charging time, thereby improving the estimation accuracy of the remaining charging time.

[0028] In some possible embodiments, the charging feature data includes a battery initial temperature;

[0029] The pre-trained remaining charging time estimation model is used to sequentially obtain the charging time of each charging path based on the charging feature data, including:

[0030] For each charging path corresponding to the total charging path to be estimated, based on the charging feature data, a pre-trained remaining charging time estimation model is used to sequentially obtain the charging temperature rise of each charging path;

[0031] The battery initial temperature of each charging path except the initial charging path is calculated based on the battery initial temperature and the charging temperature rise of the previous charging path; and the initial charging path is the first charging path corresponding to the total charging path to be estimated.

[0032] In the embodiments of the present application, the battery initial temperature of each charging path is determined according to the charging temperature rise of the previous charging path, so as to dynamically calculate the charging time of each charging path, thereby further improving the estimation accuracy of the remaining charging time.

[0033] In some possible embodiments, the determination of the total charging path to be estimated corresponding to the charging time estimation request includes:

[0034] Based on the charging time estimation request, a starting power and a cutoff power are determined, a charging path corresponding to the starting power is determined as a starting charging path, a charging path corresponding to the cutoff power is determined as an ending charging path, and the starting charging path, the ending charging path and an intermediate charging path are determined as the total charging path to be estimated; wherein the intermediate charging path is a charging path between the starting charging path and the ending charging path.

[0035] statistically determine the remaining charging time estimation result corresponding to the charging time estimation request based on the obtained charging time, including:

[0036] determine a first charging time corresponding to the starting charging path based on the starting electric quantity, and determine a second charging time corresponding to the ending charging path based on the ending electric quantity;

[0037] statistically determine the remaining charging time estimation result corresponding to the charging time estimation request based on the first charging time, the second charging time, and the charging time corresponding to the intermediate charging path.

[0038] In the embodiments of the present application, the accurate charging time is determined according to the proportion of the charging path where the starting electric quantity and the ending electric quantity are located, so as to further improve the estimation accuracy of the remaining charging time.

[0039] In some possible embodiments, the response to the charging time estimation request of the target vehicle and the acquisition of the charging feature data based on the charging time estimation request include:

[0040] In response to the charging time estimation request of the target vehicle, the current stable time of the target vehicle after the charging time estimation request is determined, and the charging feature data of the target vehicle at the current stable time is acquired;

[0041] The current stable time is the end time of a current stable window, and the current stable window is a charging current in a preset time window that is in a preset fluctuation range.

[0042] In the embodiments of the present application, when responding to the charging time estimation request, the estimation is performed according to the charging feature data at the current stable time by judging whether the current is stable, so as to further improve the estimation accuracy of the remaining charging time.

[0043] In some possible embodiments, after the statistical determination of the charging time based on the obtained charging time to obtain the remaining charging time estimation result corresponding to the charging time estimation request, the method further includes:

[0044] The current matching degree of the target vehicle during charging is acquired in real time; wherein the current matching degree is a ratio of an actual charging average current in the preset time window to a charging request average current;

[0045] When it is judged that the fluctuation of the current matching degree exceeds a preset matching degree fluctuation range, the charging feature data of the target vehicle at the current fluctuation time is re-acquired, and the remaining charging time estimation result is updated based on the charging feature data.

[0046] In the embodiments of the present application, by reacquiring the charging feature data and updating the estimation result when the current matching degree fluctuates greatly, the estimation accuracy of the remaining charging time is further improved.

[0047] In a third aspect, the embodiments of the present application provide a model training device, comprising:

[0048] a data collection module, configured to collect historical charging data of a vehicle;

[0049] a feature extraction module, configured to divide the overall charging process of the battery into a plurality of charging paths, and perform feature extraction based on the historical charging data according to the divided charging paths, to obtain charging sample data of each charging path;

[0050] a model training module, configured to train a preset time estimation model using the charging sample data of each charging path, to obtain a trained remaining charging time estimation model;

[0051] wherein the charging sample data comprises sample feature data and a corresponding charging result label under a current charging path, and the charging result label comprises charging time and charging temperature rise.

[0052] In a fourth aspect, the embodiments of the present application provide a remaining charging time estimation device, comprising:

[0053] a data acquisition module, configured to acquire charging feature data based on a charging time estimation request of a target vehicle in response to the charging time estimation request;

[0054] a path determination module, configured to determine a total charging path to be estimated corresponding to the charging time estimation request;

[0055] a time estimation module, configured to, for each charging path corresponding to the total charging path to be estimated, acquire charging time of each charging path based on the charging feature data and using a pre-trained remaining charging time estimation model;

[0056] a time statistics module, configured to perform statistics based on the acquired charging time, to obtain a remaining charging time estimation result corresponding to the charging time estimation request;

[0057] wherein the pre-trained remaining charging time estimation model is trained by any of the model training methods; and the data items of the charging feature data and the data items of the sample feature data correspond to each other.

[0058] In a fifth aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the method of any of the embodiments when running the program.

[0059] In a sixth aspect, a computer readable storage medium is provided, which stores a computer program, and the computer program, when run by a processor, implements the method of any of the embodiments.

[0060] In a seventh aspect, a vehicle is provided, which includes the electronic device of the fifth aspect.

[0061] In an eighth aspect, a computer program product is provided, which includes a computer program, and the computer program, when run by a processor, implements the method of any of the embodiments of the first aspect.

[0062] Other features and advantages of the present application will be further described in the following description with reference to the drawings, and part of the features and advantages will become apparent from the description, or will be learned by practice of the embodiments of the present application. The purposes and other advantages of the present application can be achieved and obtained by the structure particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments of the present application. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as limiting the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0064] FIG. 1 is a flow diagram of a model training method provided by an embodiment of the present application;

[0065] FIG. 2 is a flow diagram of a remaining charging time estimation method provided by an embodiment of the present application;

[0066] FIG. 3 is a structural diagram of a model training device provided by an embodiment of the present application;

[0067] FIG. 4 is a structural diagram of a remaining charging time estimation device provided by an embodiment of the present application;

[0068] FIG. 5 is a structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0069] The technical solutions in the embodiments of the present application will be clearly and completely described in connection with the drawings in the embodiments of the present application.

[0070] It should be noted that the term "multiple" in the embodiments of the present application refers to two or more, and therefore, "multiple" can also be understood as "at least two" in the embodiments of the present application. The association relationship of the associated objects is described, which means that there can be three relationships, for example, A and / or B, which can represent the following three cases: A exists alone, A and B exist together, and B exists alone. In addition, the character " / ", if not specially stated, generally represents an "or" relationship between the associated objects before and after it.

[0071] It should be noted that the existing remaining charging time estimation scheme mainly has the following problems: 1. Most of the existing methods are based on fixed models and parameters, which are difficult to adapt to the influence of battery aging, battery temperature change, environmental temperature change and other factors, resulting in low estimation accuracy; 2. The battery charging process will cause the battery temperature to rise, and the charging performance under different battery temperature changes and environmental temperature changes is different, and the existing technology lacks dynamic adaptability to these changes, resulting in low estimation accuracy; 3. The existing technology fails to fully utilize the advantages of cloud computing and big data technology, lacks effective analysis of a large amount of historical charging data, and also leads to low estimation accuracy of the remaining charging time.

[0072] In view of the problems existing in the prior art, the embodiments of the present application provide a model training method and a remaining charging time estimation method, thereby effectively improving the estimation accuracy.

[0073] As shown in FIG. 1, the embodiments of the present application provide a model training method, which can include the following steps:

[0074] S101, collecting historical charging data of a vehicle;

[0075] Specifically, the collection of historical charging data can be realized based on high-frequency new energy vehicle burying point data. The collected historical charging data can be classified and stored and analyzed according to vehicle models, charging types and the like. In order to ensure the accuracy of parameter calculation, the data acquisition frequency can be configured to be greater than 1hz.

[0076] S102, dividing the overall charging process of the battery into multiple charging paths, and extracting features based on the historical charging data according to the divided charging paths to arrange charging sample data of each charging path. The charging sample data includes sample feature data under the current charging path and corresponding charging result labels, and the charging result labels include charging time and charging temperature rise.

[0077] Specifically, the overall charging process refers to the process of battery power from 0% to 100%, which can be divided into multiple charging paths according to the needs, for example, 0%-1%, 1%-2%, 2%-3%, …, 99%-100%, a total of 100 charging paths. In addition to average division, a specific division method can also be used, for example, 0%-10%, 10%-50%, 50%-100%, a total of three charging paths. The specific charging path division method can be set according to specific needs.

[0078] For the obtained historical charging data, feature extraction is performed according to each charging path obtained by division to obtain charging sample data of each charging path. The charging sample data includes sample feature data and corresponding charging result labels of each charging path. The sample feature data refers to initial feature data that can be directly obtained or calculated, and the charging result label refers to the actual charging time and charging temperature rise corresponding to each charging path.

[0079] S103, training the preset time estimation model using the charging sample data of each charging path to obtain a trained remaining charging time estimation model;

[0080] Specifically, the charging sample data of each charging path is input into the preset time estimation model, and the time estimation model is trained to enable the time estimation model to learn the corresponding relationship between the sample feature data and the corresponding charging result label, and finally obtain a trained remaining charging time estimation model. Subsequently, by collecting or calculating the feature data of the battery, the corresponding remaining charging time estimation result can be output by the remaining charging time estimation model.

[0081] In some possible embodiments, further, the sample feature data includes at least one of the charging path, the vehicle model, the charging type, the ambient temperature, the battery initial temperature, the battery health value, and the current matching degree.

[0082] The current matching degree is the ratio of the actual battery charging current to the battery charging request current under the corresponding charging path.

[0083] It can be understood that the more data types and quantities included in the sample feature data, the better the effect of the trained model, and the higher the accuracy of the remaining charging time estimation. The sample feature data is extracted and calculated from the historical charging data, for example, the current matching degree is the ratio of the actual battery charging current to the battery charging request current under the corresponding charging path.

[0084] It can be understood that the charging path, the vehicle model and the charging type can be directly used as the type of the sorted data, that is, the feature data is sorted and stored according to the charging path, the vehicle model and the charging type respectively, and is used to train the corresponding pair of residual charging time estimation models. Therefore, the charging path, the vehicle model and the charging type do not need to be included in the range of model training as sample feature data.

[0085] It should be noted that the charging path, the vehicle model and the charging type can also be used as sample feature data for training the model. Specifically, the data of the charging path, the vehicle model and the charging type can be indexed, vector encoded and subjected to other preprocessing operations to meet the needs of model training.

[0086] In some possible embodiments, after step 103, the method can further include:

[0087] Continuously collecting historical charging data of the vehicle, and sorting updated charging sample data based on the newly collected historical charging data;

[0088] Iteratively training the residual charging time estimation model using the updated charging sample data according to a preset period.

[0089] It should be noted that as time goes by, the charging data of each vehicle will change. In order to adapt to this state change, the charging sample data can be updated and sorted at a certain period, and the residual charging time estimation model can be retrained using the updated data for iterative training of the model, so as to continuously improve the estimation accuracy of the model.

[0090] In a second aspect, the embodiments of the present application provide a residual charging time estimation method, which can include the following steps:

[0091] S201, in response to a charging time estimation request of a target vehicle, and based on the charging time estimation request, charging feature data is obtained;

[0092] It should be noted that the embodiments of the present application can realize the estimation of the residual charging time in a vehicle-cloud cooperative manner, that is, in a vehicle-end request and cloud-end response mode, which saves resource consumption while ensuring efficient execution of the prediction task and real-time updating of the model. Through centralized processing of complex data analysis and prediction tasks in the cloud, and using real-time data requests from the vehicle end, efficient use of resources and fast response are realized, and the operation efficiency and prediction performance of the entire system are optimized.

[0093] Specifically, when the target vehicle needs to obtain the residual charging time, the charging time estimation request can be triggered and the related data can be uploaded, and the cloud responds to the charging time estimation request of the target vehicle and obtains the related charging feature data.

[0094] S202, determine the total charging path to be estimated corresponding to the charging time estimation request.

[0095] It can be understood that according to the charging time estimation request, the total charging path that the target vehicle wants to estimate can be known, that is, the total path from the current power (SOC, state of charge) to the cut-off power. For example, the current power of the battery is 20 (the unit can be percentage point), and the user preset cut-off power is 80, so the total charging path to be estimated corresponding to the charging time estimation request is 20-80, which contains a total of 60 charging paths.

[0096] S203, for each charging path corresponding to the total charging path to be estimated, based on the charging feature data, the pre-trained remaining charging time estimation model is used to obtain the charging time of each charging path in turn. Wherein, the pre-trained remaining charging time estimation model is obtained by training the above any model training method; the data items of the charging feature data correspond to each other with the data items of the sample feature data.

[0097] It should be noted that based on the obtained charging feature data, the charging time of each charging path can be obtained based on the remaining charging time estimation model respectively.

[0098] It should be noted that the data items of the charging feature data correspond to each other with the data items of the sample feature data, which means that the number and type of the obtained charging feature data are consistent with the number and type of the sample feature data when training the model.

[0099] S204, based on the obtained charging time, statistics are performed to obtain the remaining charging time estimation result corresponding to the charging time estimation request.

[0100] Specifically, by superimposing the charging time of each charging path calculated, the final remaining charging time estimation result is obtained.

[0101] In some possible embodiments, the charging feature data includes the initial temperature of the battery;

[0102] In step S203, based on the charging feature data, the pre-trained remaining charging time estimation model is used to obtain the charging time of each charging path in turn, which can include:

[0103] For each charging path corresponding to the total charging path to be estimated, based on the charging feature data, the pre-trained remaining charging time estimation model is used to obtain the charging temperature rise of each charging path in turn;

[0104] Wherein, the initial temperature of the battery of each charging path except the initial charging path is calculated based on the initial temperature of the battery of the last charging path and the charging temperature rise; the initial charging path is the first charging path corresponding to the total charging path to be estimated.

[0105] It should be noted that since the initial temperature of the battery of each charging path is affected by the temperature rise of the previous charging path, the charging characteristic data needs to be obtained in sequence according to the order of the charging paths, and the charging time estimation needs to be performed in sequence.

[0106] Specifically, for the first charging path (initial charging path) corresponding to the total charging path to be estimated, the corresponding charging characteristic data can be directly obtained, which can include seven characteristic data such as charging path, vehicle model, charging type, ambient temperature, initial battery temperature, battery health value, and current matching degree. For each charging path in the total charging path to be estimated except the initial charging path, the initial temperature of the battery is calculated according to the initial temperature of the battery of the previous charging path and the charging temperature rise.

[0107] In some possible embodiments, step S202 of determining the total charging path to be estimated corresponding to the charging time estimation request comprises:

[0108] determining the initial charge and the cut-off charge based on the charging time estimation request, determining the charging path corresponding to the initial charge as the initial charging path, determining the charging path corresponding to the cut-off charge as the end charging path, and determining the initial charging path, the end charging path, and the intermediate charging path as the total charging path to be estimated; wherein the intermediate charging path is the charging path between the initial charging path and the end charging path;

[0109] Step S204, based on the obtained charging time, statistics are obtained, and the remaining charging time estimation result corresponding to the charging time estimation request can include:

[0110] determining the first charging time corresponding to the initial charging path based on the initial charge, and determining the second charging time corresponding to the end charging path based on the cut-off charge;

[0111] based on the first charging time, the second charging time, and the charging time corresponding to the intermediate charging path, statistics are obtained, and the remaining charging time estimation result corresponding to the charging time estimation request is obtained.

[0112] It should be noted that in order to improve the estimation efficiency, the total charging path to be estimated can be counted according to an integer, for example, the unit of the charging path is 1, and the accuracy of the current charge (initial charge) and the cut-off charge corresponding to the charging time estimation request is 0.1. Then, the total charging path to be estimated can be determined by rounding or rounding up. For example, the current charge is 19.4, and the cut-off charge is 80.6. Then, the total charging path to be estimated can be 19-80, 19-81, or 20-81, and the specific approximate acquisition method can be set according to the demand.

[0113] In the embodiments of the present application, in order to improve the accuracy of the estimation of the remaining charging time, for the starting charging path and the ending charging path, the actual proportion is calculated according to the starting electric quantity and the cut-off electric quantity respectively to obtain the actual charging time. Specifically, for the starting charging path, if the current electric quantity is 19.4, the starting charging path is the 20th charging path, and after the charging time of the 20th charging path is calculated, the charging time is included in the final time estimation result according to the ratio of (20-19.4) / 1. Similarly, for the cut-off electric quantity of 80.6, the charging time of the 81st charging path is included in the final time estimation result according to the ratio of (80.6-80) / 1.

[0114] In some possible embodiments, step S201, in response to the charging time estimation request of the target vehicle, and based on the charging time estimation request, the charging feature data can include:

[0115] In response to the charging time estimation request of the target vehicle, the current stable time of the target vehicle after the charging time estimation request is determined, and the charging feature data of the target vehicle at the current stable time is obtained;

[0116] The current stable time is the end time of the current stable window, and the current stable window is that the charging current in the preset time window is in the preset fluctuation range.

[0117] It should be noted that after the target vehicle initiates the charging time estimation request, the time when the charging current enters the stable state can be determined first, and the charging feature data at this time is obtained for the estimation of the remaining charging time, so as to further improve the estimation accuracy of the remaining charging time.

[0118] Specifically, a time window can be set, if the charging current of each frame in the time window is in the preset fluctuation range, it is considered that the charging enters the stable stage, and the end time of the time window is used as the current stable time, so that the charging feature data obtained based on the current stable time is used for the estimation of the remaining charging time.

[0119] In some possible embodiments, after step S204, the following steps can also be included:

[0120] The current matching degree of the target vehicle during charging is obtained in real time; wherein the current matching degree is the ratio of the actual charging average current to the charging request average current in the preset time window;

[0121] When it is judged that the fluctuation of the current matching degree exceeds the preset matching degree fluctuation range, the charging feature data of the target vehicle at the current fluctuation time is re-obtained, and the remaining charging time estimation result is updated based on the charging feature data.

[0122] It should be noted that after the feedback of the remaining charging time result, if it is judged that the fluctuation of the current matching degree is too large, the charging characteristic data needs to be re-acquired and the remaining charging time estimation result needs to be re-determined.

[0123] Specifically, the current matching degree of the target vehicle during charging is calculated in real time; wherein the current matching degree is the ratio of the actual charging average current to the charging request average current in a preset time window; by judging the current matching degree of each time window compared with the current matching degree of the last time window, if it exceeds a certain fluctuation range, the end time of the current time window is taken as the current fluctuation time, the charging characteristic data at the current fluctuation time is acquired to re-estimate the remaining charging time.

[0124] The model training method and the remaining charging time estimation method of the application will be introduced below in combination with a specific scenario.

[0125] The embodiments of the application are mainly realized by the following steps:

[0126] Step one: acquire vehicle historical charging data and perform data preprocessing.

[0127] (1) Data acquisition is realized based on high-frequency new energy vehicle embedded data, and exemplary required data signals can include: vehicle frame number, current time, vehicle model code, real SOC, user preset upper limit SOC for charging, battery health value, battery current, battery request current, temperature sampling list, battery charging state, longitude, latitude, etc.

[0128] (2) The collected data is preprocessed, which can include: 1, for the same row of vehicle frame number and current time, de-duplication processing is performed; 2, data exceeding the preset range (such as-40 to 160) in the temperature sampling data is removed; 3, data with battery current sampling value exceeding the preset range (such as-2000-2000) is removed.

[0129] Step two, feature extraction of each charging path.

[0130] A cloud data table is constructed for each charging path, for example, the entire charging process is divided into 0-1, 1-2, 2-3, …, 99-100, a total of 100 charging paths. According to the historical charging data, the charging time, battery temperature rise, and related feature data affecting the result of each charging path are extracted.

[0131] Taking a charging event of a certain vehicle as an example, the charging path SOC i represents the real SOC from i-1 to i, such as representing the process from 19 to 20, wherein i is an integer from 1 to 100.

[0132] The accuracy of the real SOC is generally 0.1, in order to improve the uniformity of the charging path, the integer precision charging path can be extracted as the subsequent path standard. For example, if the real SOC is charged from 19.4 to 79.4, the characteristics of 59 charging paths of 20-21, 21-22, 23-24, …, 78-79 are extracted.

[0133] For any charging path SOC i , record the time when the real SOC just changes to i-1 as t i-1 , and the time when it is charged to i as t i .

[0134] (1) Calculate the initial temperature of the battery at t i-1 :

[0135] Where N is the number of battery temperature samples, t_list ti-1k is the temperature of the kth temperature sample at t i-1 .

[0136] (2) Calculate the ambient temperature e_temp i under the SOC i charging path:

[0137] Where N is the total number of frames from t i-1 to t i , and e_temp k is the ambient temperature at the kth time.

[0138] (3) Calculate the charging current matching degree under the SOC i charging path:

[0139] In actual charging scenarios, due to the complexity of charging, there are scenarios where the same pile is charged by multiple vehicles at the same time, and the power supply of the charging pile is dispersed, resulting in the actual charging current of the battery, which cannot meet the demand current of the battery. In order to mark this situation, the feature of the matching degree of the actual charging current and the demand current of the battery is introduced to identify the matching state of the actual charging current and the demand current under each charging path. There is:

[0140] Where: b_current k is the actual charging current at k, n_current k is the charging request current at k, and N is the total number of frames from t i-1 to t i .

[0141] (4) Record the battery health value soh i-1 at t i :

[0142] (5) Record the charging type c_type at time t i-1 i ;

[0143] (6) Record the vehicle type v_type i ;

[0144] (7) Calculate the charging time c_time under each charging path i : c_time i = t i -t i-1 (s)

[0145] (8) Calculate the charging temperature b_temp_dif under each charging path i b_temp_dif i = b_temp ti -b_temp ti-1 (℃)

[0146] Where: b_temp ti is the battery temperature at time t i , and b_temp ti-1 is the battery temperature at time t i-1 .

[0147] The following table is an example of the charging sample data collected and extracted:

[0148] Step three, quality cleaning of feature data.

[0149] For the data quality of the collected data, in order to ensure the accuracy of feature calculation, the following conditions need to be met:

[0150] (1) The switching point position of each charging path cannot have data loss, that is, at the time t i-1 , t i when SOC becomes an integer bit, the data needs to be continuous and cannot have data loss, to ensure that the switching bit data is reliable.

[0151] (2) Ensure the credibility of the cumulative calculation method feature, here introduces the data loss rate to evaluate the credibility of the feature, such as the data collection frequency is x hertz, there is: s_count i = (t i -t i-1 ) × x

[0152] Where s_count i is the number of data frames that should have under the condition of no data loss. ​

[0153] Thus there are:

[0154] wherein r_count i is the actual number of data frames under the corresponding charging path.

[0155] extract L_rate i More than 95% of all data, forming a feature data set, so as to improve the data accuracy.

[0156] According to the vehicle frame number vin, the cloud performs vehicle model association to obtain the vehicle model characteristics v_type i .

[0157] Step four, training of the remaining charging time estimation model.

[0158] According to the historical charging data, the feature extraction in step two is performed on all historical charging events of all vehicles to obtain charging sample data used for subsequent training. The training of the model can be completed based on a preset BP neural network algorithm, for example, with the charging path, vehicle model, charging type, ambient temperature, battery initial temperature, battery health, current matching degree, charging time, battery temperature rise, and the like as input.

[0159] When predicting the battery temperature rise and charging time of different paths based on the BP neural network, the effective data in step three is selected to form an original data set S base , wherein the structure form is {(X1,Y1,Z1)(X2,Y2,Z2),…,(X n ,Y n ,Z n )}, and any one sample is denoted as (X i ,Y i ,Z i ), X i is a feature, and the form is (X i1 ,X i2 ,X i3 ,X i4 ,X i5 ,X i6 ,X i7 ), and the seven features are charging path, vehicle model, charging type, ambient temperature, battery initial temperature, battery health, and current matching degree. Correspondingly, Y i ,Z i is a sample label (charging result label), Y i represents the charging time, and Z i represents the battery temperature rise. The BP neural network model training process is as follows:

[0160] (1) Data preprocessing:

[0161] SOC of charging path i For the charging interval, an integer sequence is used to represent the charging interval, such as i represents the SOC i , that is, the process of charging the battery SOC from i-1 to i. An embedding layer is added to the model, which maps the index of each charging path (such as 1 representing the path of 0-1) to a fixed-size dense vector. The vector is learned through training and can capture the potential relationship between paths. The input dimension is set to 100 and the output dimension is set to 20.

[0162] The category features such as vehicle model (vehicle characteristics) and charging type are preprocessed into numerical data that the model can understand. Since there is no order relationship between categories, One-hot encoding can be used for feature processing. Taking the charging type as an example, first, identify all the charging types, which are 3 in this embodiment: slow charging, fast charging, and super fast charging. Create a vector of length 3 for each charging type. For example, "slow charging" can be encoded as [1, 0, 0], "fast charging" as [0, 1, 0], and "super fast charging" as [0, 0, 1]. Similarly, create a One-hot vector for the vehicle model according to the number of vehicle models.

[0163] For the environmental temperature, battery initial temperature, battery health value, and current matching degree features, to solve the problem of inconsistent value ranges of different features in the sample and improve the convergence speed of subsequent data algorithms, the features can be normalized. In this embodiment, Z-score standardization is used for normalization. For any feature X iJ (J = 1, 2, 3, …, n), find its mathematical expectation as std(X iJ ), further, let the converted feature be X iJ _new, and the calculation method of the conversion process is as follows:

[0164] (2) Data set division:

[0165] Randomly divide the original data set, with 80% as the training set and 20% as the test set.

[0166] (3) Establish a neural network:

[0167] An exemplary 5-layer deep neural network can be designed, including an input layer, 3 hidden layers, and an output layer, to ensure sufficient model complexity to capture the nonlinear characteristics of the charging process. Each layer can be configured with 10 neurons, evenly distributed among the hidden layers, to form stable information processing channels. The input layer receives the feature vector after data preprocessing, and the output layer corresponds to the two prediction targets of charging time and battery temperature rise, so the output layer is set to 2 neurons.

[0168] (4) Activation function:

[0169] To effectively capture and process nonlinear relationships while avoiding the problem of gradient disappearance, ReLU can be used as the activation function in all hidden layers of the model. Its advantages include simple calculation, fast convergence speed, and keeping the gradient unsaturated in the positive interval. The output layer can use a linear activation function to directly output the predicted value, which is suitable for regression problems.

[0170] (5) Loss function:

[0171] Mean squared error (MSE) can be used as the loss function. Let the result after the kth training in the data set be Sresult_k = {(X1, Y 1_k ,Z 1_k ),(X2, Y 2_k ,Z 2_k ),…,(X n ,Y n_k ,Z n_k )}. Since there are two predicted values, the mean squared error of the two predicted values is added. In addition, due to the different importance in the prediction task, a weight factor is introduced to ensure that the charging time has a higher weight in training.

[0172] The error can be calculated as follows:

[0173] where n is the total number of samples, Y i is the actual value of the i-th sample, Y i_k and Z i_k are the predicted values of the charging time and temperature rise after the kth training, respectively. ω1 and ω2 are the weights of the charging time prediction and battery temperature rise prediction, which can be set to 0.7 and 0.3, respectively. By adjusting the weights, the direction of model optimization is controlled to ensure the best overall performance.

[0174] (6) Optimization function

[0175] The steepest descent method can be selected, which is based on the idea of iterating in the direction of the maximum gradient. The basic formula is as follows:

[0176] In the formula, ω represents the weight of any neuron, θ represents the threshold of any neuron, η is the learning rate, the initial learning rate η is set to 0.01, and the learning rate is gradually reduced as the training progresses. Δω k is the neuron weight adjustment amount after the kth training, and Δθ k is the neuron threshold adjustment amount after the kth training. By taking the partial derivative of the weight of each neuron, etc., the change amount of the current weight and threshold is obtained according to the set learning rate, and the parameter modification is completed, and then retraining is performed.

[0177] (7) Training end condition:

[0178] The maximum number of training epochs is set to 3000, and the loss function threshold is set to 0.03. The initial threshold and weight of each layer of neurons are set to the default value. When the model reaches the maximum number of training times or the loss function value is lower than the set threshold, it means that the model training is completed.

[0179] (8) Model verification:

[0180] The verification set is input into the trained model, and when the verification set error is less than the set threshold, it is considered that the model training effect is good, otherwise the parameters need to be adjusted and the model needs to be retrained, and the cycle is repeated until the model precision meets the standard.

[0181] (9) Data update and model iterative training:

[0182] As time goes on, the system continuously collects new charging data from each vehicle. The new data is integrated into the existing data set to form an updated data set, and the model is evaluated and iteratively trained at a predetermined period (e.g., monthly) to continuously improve the prediction accuracy.

[0183] Step five, vehicle request phase:

[0184] The trained model is deployed to the cloud server, and the cloud receives requests from vehicles to calculate the remaining charging time and provides a prediction service for the remaining charging time. Through the API interface, users can view the remaining charging time prediction results and charging suggestions in real time on the vehicle system or mobile application, so as to more reasonably plan the charging time and travel. The specific method is as follows:

[0185] Vehicle request response scenario:

[0186] (1) Initial charging, stable charging current stage;

[0187] (2) When the current matching degree changes.

[0188] In order to circumvent a complex charging scene such as overcharging, when the charging current and the demand current fluctuate obviously, it indicates that the charging scene changes, at this time, the cloud request is re-performed, and the remaining charging time is corrected.

[0189] Response scenario one:

[0190] 1. First, calculate the charging current stable value, and exemplarily take 10s window to perform charging current sliding window. When the data in the window meets the following conditions, it is determined that the current stable period is entered, that is:

[0191] Where b_current i is the charging current of the i-th second in the window.

[0192] Therefore, we have:

[0193] Where w_s current is the charging stable current in the window period.

[0194] Get the average demand current of the battery in the current stable period window:

[0195] Where w_n current is the window average demand current, and n_current i is the charging request current of the i-th second.

[0196] Therefore, we have:

[0197] Where m_rate is the current matching degree at the initial charging stage.

[0198] 2. The time of the first request sent and responded by the cloud is the end time t i of the current stable window.

[0199] 3. The time stamp, vehicle frame number, current SOC of the battery, vehicle model, charging type, environment temperature, battery temperature, battery health value, current matching degree, charging stable current, user set charging cutoff SOC, and latitude and longitude information at t i moment are sent to the cloud, and the cloud estimation response is waited.

[0200] Response scenario two:

[0201] In actual charging scenarios, there are often multiple vehicles in a pile, which may cause the initial charging current to not meet the demand current. As other vehicles in the same pile leave, the output current of the charging pile can meet the requested current. There are also cases where there is originally one vehicle per pile, but as the charging pile is used by other vehicles, the output current of the charging pile cannot meet the demand current. Therefore, when the charging matching degree changes significantly, the vehicle end re-issues a request to estimate the remaining charging time (or the server re-acquires feature data to update the estimation result), so as to obtain an accurate estimated time again.

[0202] At the beginning of charging, the vehicle end records the charging current matching degree at time t i The vehicle end continues to calculate the charging current matching degree. If the real-time charging matching degree exceeds the fluctuation range of the last recorded charging matching degree, the relevant feature parameters are repeatedly calculated and sent to the cloud end to update the estimation result of the remaining charging time, and the new charging current matching degree is recorded for subsequent comparison.

[0203] Determination of battery matching degree change: take 10s as the window, calculate the current matching degree in the window, and if the following conditions are met, consider it as the condition for triggering the re-estimation of the remaining charging time:

[0204] Or,

[0205] m_rate f is the current matching degree at the last request.

[0206] If the condition for triggering the re-estimation of the remaining charging time is met, the timestamp, vehicle frame number, real SOC of the battery, vehicle model, charging type, environmental temperature, battery temperature, battery health value, current matching degree, stable charging current, user-set charging cutoff SOC, latitude and longitude information, etc. at time t i are sent to the cloud end, and the cloud end estimation response is waited for again.

[0207] Step six, cloud prediction stage:

[0208] After the cloud end receives the request, it predicts the remaining charging time and returns the result of each response to the vehicle end and the user's mobile phone end. The specific method is as follows:

[0209] For example: the real SOC (current capacity) of the battery received by the cloud end is k, and the accuracy of k is 0.1. The user's preset charging cutoff SOC is m, and the accuracy of m is also 0.1.

[0210] kf can be obtained by rounding k up, mf can be obtained by rounding m up, the charging time and the charging temperature rise of each sub-path from kf to mf are calculated to count the remaining charging time (for example, k = 19.4, m = 80.6, kf = 20, mf = 81, and 62 charging paths need to be calculated at this time).

[0211] (1) For the initial path, that is, SOC = k to SOC = kf, the SOC kf The estimated charging type c_type k , the environmental temperature e_temp k , the battery temperature b_temp k , the battery health value soh k , the current supply-demand matching degree m_rate k , and the vehicle model are sent to the cloud model to obtain the charging time c_time kf and the battery temperature rise b_temp_dif kf of the path. kf

[0212] The charging time of SOC = k to SOC = kf is: c_time k = (kf-k) x c_time kf

[0213] The battery temperature rise of SOC = k to SOC = kf is: b_temp_dif k = (kf-k) x b_temp_dif kf

[0214] (2) For any path (intermediate charging path) of SOC kf to SOC mf-1 , since it involves iteration of battery temperature rise and battery initial temperature, the charging time and the battery temperature rise of the path need to be calculated in sequence:

[0215] Take the path of SOC kf+i for example (the range of i is 1 to mf-kf-1):

[0216] The battery charging path is SOC kf+i , and the vehicle type, the charging type, and the battery health value remain unchanged.

[0217] For the environmental temperature, the weather API is used for calculation according to the latitude and longitude information.

[0218] For the battery initial temperature b_temp kf+i , there is:

[0219] ​For current demand matching value m_rate kf+i , according to the cloud pre-stored current charging data table, the table lookup obtains the battery demand current r_current kf+i in this charging path, and has:

[0220] c_current k is the initial uploaded stable charging current, m_rate kf+i If greater than 1, process it as 1.

[0221] Substitute the above 7 characteristic parameters into the model to obtain the SOC kf+i charging time c_time kf+i of the charging path, and the battery temperature rise b_temp_dif kf+1 .

[0222] Since the initial temperature of the battery is affected by the estimation of the battery temperature rise of the previous path, the estimation of each charging path must be performed in sequence.

[0223] (3) For the end path, the SOC is charged from mf-1 to m, and the calculation method of all characteristic parameters is consistent with that of the intermediate path. After substituting into the model, the SOC mf charging time c_time mf of the path can be obtained.

[0224] Then the charging time from SOC mf-1 to SOC m is: c_time m =(m-mf+1)×c_time mf

[0225] Finally:

[0226] After the calculation is completed, it is returned to the vehicle end according to the vehicle frame number, and the charging time is updated and reminded on the mobile phone end.

[0227] It should be noted that the above only describes the scenario that the battery is charged immediately after the current charging. If there is a heat management time (charging heating / cooling) before the battery is charged, the heat management time needs to be added. The cloud can train a heat management time estimation model according to a similar training method.

[0228] Please refer to FIG. 3, which shows the composition block diagram of the model training device provided by some embodiments of the present application. It should be understood that the model training device corresponds to the above-mentioned method embodiment of FIG. 1, and can perform each step involved in the above-mentioned method embodiment. The specific functions of the model training device can be referred to the description in the above, and to avoid repetition, the detailed description is appropriately omitted here.

[0229] The model training apparatus of FIG. 3 comprises at least one software function module which can be stored in the form of software or firmware in the memory or solidified in the model training apparatus, and the model training apparatus comprises:

[0230] The data collection module 310 is configured to collect historical charging data of the vehicle.

[0231] The feature extraction module 320 is configured to divide the overall charging process of the battery into a plurality of charging paths, and perform feature extraction based on the historical charging data according to the divided charging paths, so as to obtain charging sample data of each charging path.

[0232] The model training module 330 is configured to train the preset time estimation model by using the charging sample data of each charging path, so as to obtain a trained remaining charging time estimation model.

[0233] The charging sample data comprises sample feature data and a corresponding charging result label in the current charging path, and the charging result label comprises charging time and charging temperature rise.

[0234] Referring to FIG. 4, FIG. 4 shows a composition block diagram of a remaining charging time estimation apparatus according to some embodiments of the present application. It should be understood that the remaining charging time estimation apparatus corresponds to the method embodiment of FIG. 2 described above, and can perform each step involved in the method embodiment described above. The specific functions of the remaining charging time estimation apparatus can be referred to the description above, and the detailed description is appropriately omitted here to avoid repetition.

[0235] The remaining charging time estimation apparatus of FIG. 4 comprises at least one software function module which can be stored in the form of software or firmware in the memory or solidified in the remaining charging time estimation apparatus, and the remaining charging time estimation apparatus comprises:

[0236] The data acquisition module 410 is configured to acquire charging feature data based on a charging time estimation request of a target vehicle in response to the charging time estimation request.

[0237] The path determination module 420 is configured to determine a total charging path to be estimated corresponding to the charging time estimation request.

[0238] The time estimation module 430 is configured to, for each charging path corresponding to the total charging path to be estimated, acquire charging time of each charging path based on the charging feature data and by using the pre-trained remaining charging time estimation model in sequence.

[0239] The time statistics module 440 is configured to perform statistics based on the acquired charging time, so as to obtain a remaining charging time estimation result corresponding to the charging time estimation request.

[0240] The pre-trained residual charging time estimation model is trained by any model training method; and the data items of the charging feature data and the data items of the sample feature data correspond to each other.

[0241] It can be understood that the above-mentioned device embodiments correspond to the method embodiments of the present application. The device provided by the embodiments of the present application can implement the method provided by any one of the method embodiments of the present application.

[0242] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-mentioned device can refer to the corresponding process in the foregoing method, and will not be described in detail here.

[0243] As shown in FIG. 5, some embodiments of the present application provide an electronic device 500, which includes a memory 510, a processor 520, and a computer program stored in the memory 510 and executable on the processor 520, wherein the processor 520 reads the program from the memory 510 through the bus 530 and implements the method of any embodiment of the residual charging time estimation method as described above when executing the program.

[0244] The processor 520 can process digital signals and can include various computing structures. For example, a complex instruction set computer structure, a reduced instruction set computer structure, or a structure implementing a combination of multiple instruction sets. In some examples, the processor 520 can be a microprocessor.

[0245] The memory 510 can be used to store instructions executed by the processor 520 or data related to the execution process of the instructions. These instructions and / or data can include code for implementing some or all functions of one or more modules described in the embodiments of the present application. The processor 520 of the embodiments of the present disclosure can be used to execute the instructions in the memory 510 to implement the method shown in the above. The memory 510 includes a dynamic random access memory, a static random access memory, a flash memory, an optical memory, or other memories well known to those skilled in the art.

[0246] Some embodiments of the present application also provide a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to perform the method described in the method embodiments.

[0247] Some embodiments of the present application also provide a computer program product, which, when executed on a computer, causes the computer to perform the method described in the method embodiments.

[0248] It should be noted that each of the embodiments in the present specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same parts of each embodiment can be mutually referred to.

[0249] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can also be implemented in other ways. The apparatus embodiments described above are only schematic. For example, the flowcharts and block diagrams in the drawings show the possible implementation architectures, functions and operations of the apparatus, method and computer program product according to the embodiments of the present application. In this regard, each block in the flowcharts or block diagrams can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in a different order than that shown in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and the combination of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0250] In addition, the functional modules in each of the embodiments of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0251] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0252] The above merely provides an example of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, and thus, once an item is defined in one drawing, it need not be further defined and explained in subsequent drawings.

[0253] The above merely provides an example of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, and thus, once an item is defined in one drawing, it need not be further defined and explained in subsequent drawings.

[0254] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one from another entity or action without necessarily requiring or implying any actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without limitation, an element preceded by "comprises... a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus.

[0255] The above merely provides an example of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, and thus, once an item is defined in one drawing, it need not be further defined and explained in subsequent drawings.

Claims

1. A model training method, characterized in that, The method comprises: collecting historical charging data of a vehicle; dividing a whole charging process of a battery into multiple charging paths, and performing feature extraction on the historical charging data according to the divided charging paths to obtain charging sample data of each charging path; training a preset time estimation model by using the charging sample data of each charging path to obtain a trained remaining charging time estimation model; wherein the charging sample data comprises sample feature data and corresponding charging result labels in a current charging path, and the charging result labels comprise charging time and charging temperature rise.

2. The model training method of claim 1, wherein, The sample feature data comprises at least one of a charging path, a vehicle model, a charging type, an ambient temperature, a battery initial temperature, a battery health value, and a current matching degree. The current matching degree is a ratio of an actual charging current of the battery in the corresponding charging path to a battery charging request current.

3. The model training method of claim 1, wherein, After the training of the preset time estimation model by using the charging sample data of each charging path to obtain the trained remaining charging time estimation model, the method further comprises: continuously collecting historical charging data of the vehicle, and obtaining updated charging sample data based on the newly collected historical charging data; iteratively training the remaining charging time estimation model by using the updated charging sample data according to a preset period.

4. A method of estimating a remaining charge time, characterized by, The method comprises: responding to a charging time estimation request of a target vehicle, and obtaining charging feature data based on the charging time estimation request; determining a total charging path to be estimated corresponding to the charging time estimation request; for each charging path corresponding to the total charging path to be estimated, obtaining charging time of each charging path in sequence by using a pre-trained remaining charging time estimation model based on the charging feature data; statistically obtaining a remaining charging time estimation result corresponding to the charging time estimation request based on the obtained charging time; wherein the pre-trained remaining charging time estimation model is obtained by the model training method of any one of claims 1-3; and data items of the charging feature data correspond to data items of the sample feature data.

5. The method of claim 4, wherein The charging feature data comprises a battery initial temperature. The method of obtaining charging time of each charging path in sequence by using the pre-trained remaining charging time estimation model based on the charging feature data comprises: for each charging path corresponding to the total charging path to be estimated, obtaining charging temperature rise of each charging path in sequence by using the pre-trained remaining charging time estimation model based on the charging feature data; wherein the battery initial temperature of each charging path except an initial charging path is calculated based on the battery initial temperature and the charging temperature rise of a previous charging path; and the initial charging path is a first charging path corresponding to the total charging path to be estimated.

6. The method of claim 4, wherein The method of determining the total charging path to be estimated corresponding to the charging time estimation request comprises: determining a starting power and a cut-off power based on the charging time estimation request, and determining a first charging path corresponding to the starting power and a last charging path corresponding to the cut-off power as the total charging path to be estimated. The starting charging path is determined, the charging path corresponding to the cut-off electric quantity is determined as an ending charging path, and the starting charging path, the ending charging path and an intermediate charging path are determined as the total charging path to be estimated; wherein the intermediate charging path is a charging path between the starting charging path and the ending charging path; The obtained charging time is counted to obtain a remaining charging time estimation result corresponding to the charging time estimation request, including: The first charging time corresponding to the starting charging path is determined based on the starting electric quantity, and the second charging time corresponding to the ending charging path is determined based on the cut-off electric quantity; The first charging time, the second charging time and the charging time corresponding to the intermediate charging path are counted to obtain a remaining charging time estimation result corresponding to the charging time estimation request.

7. The remaining charge time estimation method according to claim 4, characterized by, The charging feature data is obtained based on the charging time estimation request in response to the charging time estimation request of the target vehicle, including: In response to the charging time estimation request of the target vehicle, a current stable time of the target vehicle after the charging time estimation request is determined, and charging feature data of the target vehicle at the current stable time is obtained; The current stable time is an end time of a current stable window, and the current stable window is a charging current in a preset time window.

8. The method of claim 7, wherein After the obtained charging time is counted to obtain a remaining charging time estimation result corresponding to the charging time estimation request, further including: The current matching degree of the target vehicle during charging is obtained in real time; wherein the current matching degree is a ratio of an actual charging average current to a charging request average current in the preset time window; When it is judged that the fluctuation of the current matching degree exceeds a preset matching degree fluctuation range, the charging feature data of the target vehicle at the current fluctuation time is re-obtained, and the remaining charging time estimation result is updated based on the charging feature data.

9. A model training apparatus, comprising: Including: The data collection module is used for collecting historical charging data of the vehicle; The feature extraction module is used for dividing the overall charging process of the battery into a plurality of charging paths, and extracting features based on the historical charging data according to the divided charging paths, and arranging to obtain charging sample data of each charging path; The model training module is used for training a preset time estimation model using charging sample data of each charging path to obtain a trained remaining charging time estimation model; The charging sample data includes sample feature data under the current charging path and a corresponding charging result label, and the charging result label includes charging time and charging temperature rise.

10. A remaining charge time estimation device characterized by comprising: Including: The data acquisition module is used for obtaining charging feature data based on the charging time estimation request in response to the charging time estimation request of the target vehicle; The path determination module is used for determining a total charging path to be estimated corresponding to the charging time estimation request; a time estimation module, configured to, for each charging path corresponding to the total charging path to be estimated, obtain charging time of each charging path in sequence based on the charging feature data and by using a pre-trained residual charging time estimation model; a time statistics module, configured to perform statistics based on the obtained charging time to obtain a residual charging time estimation result corresponding to the charging time estimation request; wherein the pre-trained residual charging time estimation model is obtained by training the model training method in any one of claims 1-3; and data items of the charging feature data and data items of the sample feature data correspond to each other. A computer readable storage medium, having stored thereon a computer program, wherein the computer program, when executed by a processor, causes the processor to carry out the method of any one of claims 1-8.

11. An electronic device, comprising: A computer readable storage medium, having stored thereon a computer program, wherein the computer program, when executed by a processor, causes the processor to carry out the method of any one of claims 1-8.

12. A computer-readable storage medium, characterized in that, An electronic device as claimed in claim 11.

13. A vehicle characterized by comprising: ​

Images