Model training method, charging behavior prediction method, electronic device, and storage medium

Abstract

The application provides a model training method and device, a charging behavior prediction method, an electronic device and a storage medium. The model training method comprises: acquiring at least one historical time sequence data sequence of a vehicle; performing data sampling on the historical time sequence data sequence according to a preset sampling duration to obtain a plurality of samples; the samples correspond to a charging behavior occurrence probability, and the closer a sample is to a target sample in the plurality of samples, the greater the charging behavior occurrence probability corresponding to the sample is, the target sample being a sample corresponding to an occurrence time of a historical charging behavior; performing feature coding on the plurality of samples to obtain coded data; inputting the coded data into a recurrent neural network to perform model training, and obtaining a trained charging behavior prediction model; wherein the charging behavior prediction model is used to predict whether the vehicle is about to perform a charging behavior. The method can improve the accuracy of predicting the charging intention of the vehicle, and to a certain extent, is conducive to improving the charging experience of the vehicle owner.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vehicles, and more particularly, to a model training method and device, a charging behavior prediction method, an electronic device and a storage medium in the field of vehicles. BACKGROUND

[0002] Battery thermal management is a main function of an electric vehicle battery system, mainly to keep the battery pack always working in a suitable temperature range, so as to maintain the best working state of the battery pack. Especially when the electric vehicle is charging, the battery usually needs to be preheated to reach the best working temperature before charging. Only in this way can the best power output and input of the vehicle, the maximum available energy, and the longest cycle life be achieved. In order to improve the charging experience of the vehicle owner, the charging intention of the vehicle is predicted in the related art, but there is a problem of low prediction accuracy. SUMMARY

[0003] The present application provides a model training method, a charging behavior prediction method, an electronic device and a storage medium, which can improve the accuracy of predicting the charging intention of the vehicle, and to a certain extent, help to improve the charging experience of the vehicle owner.

[0004] In a first aspect, a model training method is provided, which includes: obtaining at least one historical time series data sequence of a vehicle; wherein the historical time series data sequence includes a plurality of preset type data corresponding to different time points, and the different time points include a time point of occurrence of a historical charging behavior; performing data sampling on the historical time series data sequence according to a preset sampling duration to obtain a plurality of samples; wherein the sample corresponds to a charging behavior occurrence probability, and the closer the sample is to a target sample in the plurality of samples, the greater the charging behavior occurrence probability corresponding to the sample is, and the target sample is a sample corresponding to the time point of occurrence of the historical charging behavior; performing feature encoding on the plurality of samples to obtain feature encoded data; inputting the feature encoded data into a recurrent neural network to perform model training, and obtaining a trained charging behavior prediction model; wherein the charging behavior prediction model is used to predict whether the vehicle will soon have a charging behavior.

[0005] In the technical solution, the at least one historical time sequence data sequence of the vehicle is obtained, the historical time sequence data sequence is sampled according to a preset sampling time length to obtain a plurality of samples, the plurality of samples are feature-encoded to obtain feature-encoded data, the feature-encoded data is input into a recurrent neural network for model training, and a trained charging behavior prediction model is obtained. The charging behavior prediction model can be trained to predict the probability of charging behavior. The historical time sequence data sequence includes a plurality of preset type data corresponding to different time points, and the different time points include a historical charging behavior occurrence time point, that is, the historical time sequence data sequence includes data corresponding to the historical charging behavior occurrence time point. Therefore, the historical time sequence data sequence can reflect the historical charging data of the vehicle, providing reasonable and accurate sample references for model training. The plurality of samples obtained by sampling the historical time sequence data sequence each correspond to a charging behavior occurrence probability. Considering that the closer a sample to the target sample corresponding to the historical charging behavior occurrence time point, the greater the possibility of charging behavior, the charging behavior occurrence probability of the sample closer to the target sample in the plurality of samples is set to be greater, which can reasonably and accurately reflect the possibility of charging behavior at the time point corresponding to the plurality of samples. Therefore, the charging behavior prediction model trained based on the samples can accurately and reasonably predict the charging behavior occurrence probability based on the current time sequence data sequence of the vehicle, that is, accurately and reasonably predict the charging intention of the vehicle, which is beneficial to improving the charging experience of the vehicle owner to a certain extent. For example, when the charging behavior occurrence probability is predicted to be high, that is, the charging intention of the vehicle is predicted, relevant measures can be taken in advance to improve the charging experience of the vehicle owner. The relevant measures include, for example, preheating the battery management system of the vehicle to make the battery enter a better working state when the user charges, thereby reducing the charging waiting time and improving the charging experience of the electric vehicle owner.

[0006] In combination with the first aspect, in some possible implementation manners, the historical charging behavior occurrence time point is a termination time point corresponding to the historical time sequence data sequence.

[0007] In combination with the first aspect, in some possible implementation manners, the charging behavior occurrence probability corresponding to the sample is determined based on a preset probability decay value and a sample distance between the sample and the target sample.

[0008] In combination with the first aspect, in some possible implementation manners, the at least one historical time sequence data sequence of the vehicle is obtained by collecting historical vehicle data and performing trip division on the historical vehicle data according to a preset trip interval time length to obtain a plurality of historical time sequence data sequences of the vehicle. Each historical time sequence data sequence corresponds to the trip interval time length.

[0009] With reference to the first aspect, in some possible implementation manners, the dividing, according to the preset trip interval duration, the historical vehicle data into the plurality of historical time-series data sequences of the vehicle comprises: determining a plurality of historical charging behavior occurrence time points according to the historical vehicle data; and taking each of the historical charging behavior occurrence time points as a starting point of trip division, obtaining historical vehicle data within the trip interval duration to obtain the plurality of historical time-series data sequences.

[0010] The second aspect provides a charging behavior prediction method, which comprises: obtaining a current time-series data sequence of a vehicle; wherein the current time-series data sequence comprises a plurality of preset type data corresponding to different time points respectively, a terminal time point corresponding to the current time-series data sequence is a current time point, and a starting time point corresponding to the current time-series data sequence is a target time point before the current time point; performing data sampling on the current time-series data sequence according to a preset sampling duration to obtain a plurality of sampling data; performing feature encoding on the plurality of sampling data to obtain feature-encoded sampling data; inputting the feature-encoded sampling data into a charging behavior prediction model to output a predicted charging behavior occurrence probability; wherein the charging behavior prediction model is the trained charging behavior prediction model obtained in the first aspect.

[0011] With reference to the second aspect, in some possible implementation manners, after the output of the predicted charging behavior occurrence probability, the method further comprises: when the charging behavior occurrence probability is greater than a preset probability, controlling a battery management system of the vehicle to perform battery preheating.

[0012] The third aspect provides a model training apparatus, which comprises: an obtaining module configured to obtain at least one historical time-series data sequence of a vehicle; wherein the historical time-series data sequence comprises a plurality of preset type data corresponding to different time points respectively, and the different time points comprise a historical charging behavior occurrence time point; a sampling module configured to perform data sampling on the historical time-series data sequence according to a preset sampling duration to obtain a plurality of samples; wherein the samples correspond to charging behavior occurrence probabilities, and the closer a sample is to a target sample in the plurality of samples, the greater the charging behavior occurrence probability corresponding to the sample is, and the target sample is a sample corresponding to the historical charging behavior occurrence time point; a feature encoding module configured to perform feature encoding on the plurality of samples to obtain feature-encoded data; and a training module configured to input the feature-encoded data into a recurrent neural network to perform model training to obtain a trained charging behavior prediction model; wherein the charging behavior prediction model is configured to predict whether a charging behavior is about to occur.

[0013] In some possible implementation manners, the occurrence time of the historical charging behavior is a terminal time corresponding to the historical time sequence data.

[0014] In some possible implementation manners, the occurrence probability of the charging behavior corresponding to the sample is determined based on a preset probability attenuation value and a sample distance between the sample and the target sample.

[0015] In some possible implementation manners, the obtaining module is specifically configured to: collect historical vehicle data; and perform trip division on the historical vehicle data according to a preset trip interval duration, to obtain the historical time sequence data sequences of the vehicle; and a duration corresponding to each historical time sequence data sequence is the trip interval duration.

[0016] In some possible implementation manners, the obtaining module is specifically configured to: determine occurrence times of a plurality of historical charging behaviors according to the historical vehicle data; and obtain the historical time sequence data sequences by taking each occurrence time of the historical charging behaviors as a starting point of trip division and obtaining historical vehicle data within the trip interval duration.

[0017] A fourth aspect provides an electronic device including a memory and a processor. The memory is configured to store executable program code, and the processor is configured to invoke and run the executable program code from the memory, so that the electronic device performs the method in the first aspect, any possible implementation manner of the first aspect, the second aspect, or any possible implementation manner of the second aspect.

[0018] A fifth aspect provides a computer program product, which includes computer program code. When the computer program code is run on a computer, the computer program code causes the computer to perform the method in the first aspect, any possible implementation manner of the first aspect, the second aspect, or any possible implementation manner of the second aspect.

[0019] A sixth aspect provides a computer-readable storage medium, which stores computer program code. When the computer program code is run on a computer, the computer program code causes the computer to perform the method in the first aspect, any possible implementation manner of the first aspect, the second aspect, or any possible implementation manner of the second aspect. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a flowchart of a model training method provided by an embodiment of the present application;

[0021] Figure 2is a schematic diagram of a plurality of samples obtained by sampling provided by an embodiment of the present application;

[0022] Figure 3 is a schematic diagram of a recurrent neural network provided by an embodiment of the present application;

[0023] Figure 4 is a schematic flow chart of a prediction method of charging behavior provided by an embodiment of the present application;

[0024] Figure 5 is a structural schematic diagram of a model training device provided by an embodiment of the present application;

[0025] Figure 6 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0026] The technical solutions in the present application will be described in detail below with reference to the drawings. In the description of the embodiments of the present application, unless otherwise specified, " / " represents the meaning of or, for example, A / B can represent A or B: "and / or" in the text only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which means that there are three cases of A alone, A and B, and B alone. In addition, in the description of the embodiments of the present application, "multiple" means two or more than two.

[0027] Hereinafter, the terms "first" and "second" are only for descriptive purposes, and cannot be understood as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more features.

[0028] New energy vehicles refer to vehicles using unconventional vehicle fuels (gasoline, diesel, etc.) as power sources. As the main power source of new energy vehicles, the importance of power batteries for new energy vehicles is self-evident. In the actual use of vehicles, the use conditions of batteries are complex and changeable. In order to improve the cruising range, the vehicle needs to arrange as many battery cells as possible in a certain space, so the space of the battery pack on the vehicle is very limited. A large amount of heat is generated by the battery during vehicle operation and accumulates in the relatively narrow space over time. Due to the dense stacking of battery cells in the battery pack, heat dissipation in the middle area is relatively more difficult to some extent, which exacerbates the temperature inconsistency between battery cells, which reduces the charging and discharging efficiency of the battery and affects the power of the battery. In severe cases, it can also cause thermal runaway, affecting the safety and service life of the system.

[0029] The temperature of a power battery has a significant impact on its performance, lifespan, and safety. At low temperatures, lithium-ion batteries experience increased internal resistance and reduced capacity; in extreme cases, the electrolyte may freeze, preventing the battery from discharging, thus easily leading to a decrease in the power output performance and driving range of electric vehicles. When charging new energy vehicles in low-temperature conditions, the Battery Management System (BMS) typically heats the battery to a suitable temperature before charging. Improper handling can lead to instantaneous overcharging, causing internal short circuits, and potentially resulting in smoke, fire, or even explosion. The safety issues of low-temperature charging of electric vehicle battery systems have significantly hindered the promotion of electric vehicles in cold regions.

[0030] Battery thermal management is a primary function of electric vehicle battery systems, ensuring the battery pack operates within a suitable temperature range to maintain optimal performance. This is especially crucial during charging, where the battery needs to be preheated to its optimal operating temperature before charging begins. This ensures the vehicle achieves optimal power output and input, maximum usable energy, and the longest cycle life. While some technologies predict the vehicle's charging intentions to improve the charging experience, these methods suffer from low prediction accuracy.

[0031] This specification provides a model training method, apparatus, method for predicting charging behavior, electronic device, and storage medium, which can solve the aforementioned problems existing in related technologies. Specifically, this specification first introduces embodiments of the model training method.

[0032] in, Figure 1 This is a flowchart illustrating a model training method provided in one embodiment of this specification. For example, the execution entity of the model training method provided in this embodiment is an electronic device with model training capabilities, such as a server, terminal, or cloud. For example, as... Figure 1 As shown, the model training method includes:

[0033] Step 101: Obtain at least one historical time-series data sequence of the vehicle. The historical time-series data sequence includes multiple preset data types corresponding to different times, and these different times include the times when historical charging activities occurred.

[0034] Step 102: Based on the preset sampling duration, sample the aforementioned historical time-series data sequence to obtain multiple samples. Each sample corresponds to a probability of charging behavior occurring, and the closer the sample is to the target sample, the higher the probability of charging behavior occurring. The target sample is the sample corresponding to the time when the aforementioned historical charging behavior occurred.

[0035] Step 103: feature encoding is performed on the plurality of samples to obtain feature encoded data.

[0036] Step 104: the feature encoded data is input into a recurrent neural network for model training to obtain a trained charging behavior prediction model. The charging behavior prediction model is used to predict whether a charging behavior will occur.

[0037] In Figure 1 The technical means provided by the model training method of the embodiment shown in the embodiment includes obtaining at least one historical time series data sequence of the vehicle, sampling the historical time series data sequence according to a preset sampling duration to obtain a plurality of samples, performing feature encoding on the plurality of samples to obtain feature encoded data, inputting the feature encoded data into a recurrent neural network for model training, and obtaining a trained charging behavior prediction model. The charging behavior prediction model can be trained to predict the probability of a charging behavior. Since the historical time series data sequence used to train the model includes a plurality of preset type data corresponding to different time points, the historical time series data sequence includes data corresponding to the occurrence time of the historical charging behavior, so that the historical time series data sequence can reflect the historical charging data of the vehicle to provide reasonable and accurate sample reference for training the model. The plurality of samples obtained by sampling the historical time series data sequence each correspond to a charging behavior occurrence probability. Considering that the closer the sample to the target sample corresponding to the occurrence time of the historical charging behavior, the greater the possibility of a charging behavior, the charging behavior occurrence probability of the sample closer to the target sample in the plurality of samples is set to be greater, which can reasonably and accurately reflect the possibility of a charging behavior at the time corresponding to the plurality of samples. Therefore, the charging behavior prediction model trained based on these samples can accurately and reasonably predict the probability of a charging behavior based on the current time series data sequence of the vehicle, that is, it can accurately and reasonably predict the charging intention of the vehicle, which is beneficial to improving the charging experience of the vehicle owner to a certain extent. For example, when the probability of a charging behavior is predicted to be high, that is, the vehicle is predicted to have a charging intention, relevant measures can be taken in advance to improve the charging experience of the vehicle owner. The relevant measures include, for example, preheating the battery control system of the vehicle to make the battery enter a better working state when the user charges, thereby reducing the charging waiting time and improving the charging experience of the electric vehicle owner.

[0038] The specific implementation of each step in the embodiment shown in Figure 1 is described in detail as follows:

[0039] In step 101, the historical time sequence data sequence of the vehicle includes a plurality of preset type data corresponding to respective different time points, and intervals between adjacent two time points can be the same or different. For example, the electronic device can take a certain time point in the past as a starting point, and obtain preset type data corresponding to N time points in the past, where intervals between adjacent two time points are the same. The different time points include a time point at which a historical charging behavior occurs, that is, the historical time sequence data sequence of the vehicle includes a plurality of preset type data corresponding to the time point at which the historical charging behavior occurs and a plurality of preset type data corresponding to time points near the time point at which the historical charging behavior occurs. The historical charging behavior can be understood as a charging behavior that has occurred in the past. The plurality of preset type data of the vehicle can be selected according to actual needs, and can include but is not limited to a vehicle identification number, a driving mileage, a battery state of charge (SOC), a battery health status, a fast / slow charging flag, a vehicle global positioning system (GPS) or a vehicle position, and the like.

[0040] In an exemplary embodiment, the number of historical time sequence data sequences obtained by the electronic device can be several, and the time length corresponding to each historical time sequence data sequence can be the same or different, which is not specifically limited in the present embodiment.

[0041] In an exemplary embodiment, the implementation of step 101 can include collecting historical vehicle data, dividing the historical vehicle data according to a preset travel interval time length to obtain a plurality of historical time sequence data sequences of the vehicle, and each historical time sequence data sequence corresponds to the travel interval time length.

[0042] For example, the electronic device can collect historical vehicle data through the Internet of Vehicles, and the historical vehicle data can include historical data of a plurality of types of vehicles, which can include but is not limited to a time point, a vehicle identification number, a driving mileage, a battery SOC, a time point at which a charging behavior occurs, a battery health status, a fast / slow charging flag, a vehicle GPS, and the like. The preset travel interval time length can be set according to actual needs, and can be valued within 20 minutes to 40 minutes, but the present embodiment is not specifically limited thereto. Assuming that the preset travel interval time length is 30 minutes, the historical vehicle data can be divided according to a time range covered by the historical vehicle data and a time point at which a historical charging behavior occurs, to obtain a plurality of historical time sequence data sequences with a time length of 30 minutes, and a plurality of time points corresponding to each historical time sequence data sequence include a time point at which a historical charging behavior occurs.

[0043] In an example embodiment, the occurrence time of the historical charging behavior can be the end time corresponding to the historical time series data sequence. For example, define the occurrence time of the historical charging behavior as Tc, and the preset travel interval duration as Q, the historical vehicle data within Tc-Q to Tc can be used as the charging travel data, i.e., as a historical time series data sequence.

[0044] In an example embodiment, the occurrence time of the historical charging behavior can also be the start time corresponding to the historical time series data sequence. For example, the historical vehicle data within Tc to Tc+Q can be used as the charging travel data, i.e., as a historical time series data sequence.

[0045] In an example embodiment, the implementation of the travel division of the historical vehicle data according to the preset travel interval duration to obtain the historical time series data sequence of the vehicle can include: determining the occurrence time of a plurality of historical charging behaviors according to the historical vehicle data; and taking each occurrence time of the historical charging behavior as the start point of the travel division, and obtaining the historical vehicle data within the travel interval duration to obtain a plurality of historical time series data sequences.

[0046] It can be understood that the vehicle can have a plurality of charging behaviors in the past history time, and therefore, in the embodiment, the occurrence time of the plurality of historical charging behaviors can be determined based on the historical vehicle data. Then, the occurrence time of each historical charging behavior can be taken as the start time Tc of the divided historical time series data sequence, and the historical vehicle data within the travel interval duration Q can be obtained. For example, define the occurrence time of the historical charging behavior as Tc, and the preset travel interval duration as Q, the historical vehicle data within Tc-Q to Tc can be used as the charging travel data, i.e., as a historical time series data sequence. Therefore, based on a plurality of Tc, a plurality of charging travel data, i.e., a plurality of historical time series data sequences, can be obtained.

[0047] In step 102, the electronic device can sample the historical time series data sequence according to a preset sampling duration to obtain a plurality of samples; each sample can correspond to a charging behavior occurrence probability, and the sample corresponding to the charging behavior occurrence probability in the plurality of samples is closer to the target sample, and the target sample is the sample corresponding to the occurrence time of the historical charging behavior in the plurality of samples.

[0048] The preset sampling duration can be set according to actual needs, for example, can be in 2 minutes to 6 minutes, however, the embodiment does not make specific limitation. Exemplarily, the preset sampling duration can include several different durations, so that several samples with different durations can be obtained. Optionally, the preset sampling duration can be a duration t, so that several samples with the same duration t can be obtained. Assuming that the duration corresponding to the historical time series data sequence is the travel interval duration, denoted as Q, and the preset sampling duration is a duration t, then for a historical time series data sequence, Q / t samples can be obtained at most.

[0049] Exemplarily, the electronic device can take the end time corresponding to the historical time series data sequence as the sampling starting point and extract t time data along the historical time direction, and each t time data extracted is taken as a sample. The electronic device can also take the start time corresponding to the historical time series data sequence as the sampling starting point and extract t time data along the future time direction, and each t time data extracted is taken as a sample.

[0050] Exemplarily, the target sample can include the historical time series data sequence of the occurrence time of the historical charging behavior. Alternatively, it can be understood that the target sample can include the historical vehicle data at the time of the charging behavior, that is, the target sample can represent the feature data of the vehicle at the time of the charging behavior. The charging behavior occurrence probability corresponding to the target sample can be 100%, the closer the other samples to the target sample, the greater the corresponding charging behavior occurrence probability, and the farther the distance from the target sample, the smaller the corresponding charging behavior occurrence probability.

[0051] In the exemplary embodiment, the charging behavior occurrence probability corresponding to the sample is determined based on a preset probability decay value and the sample distance between the sample and the target sample. The preset probability decay value can be set according to actual needs, for example, can be between 5% and 20%, however, the embodiment does not make specific limitation.

[0052] Exemplarily, referring to Figure 2, the electronic device takes the occurrence time Tc of the historical charging behavior as the sampling starting point, and sequentially extracts data within t time along the historical time direction (the arrow direction to the left in the figure) to obtain sample 0, sample 1,..., and sample n. The time corresponding to sample 0 includes the occurrence time Tc of the historical charging behavior, so sample 0 is the target sample, and thus the charging behavior occurrence probability corresponding to sample 0 is set to 100%. The charging behavior occurrence probability corresponding to each sample can be expressed as (100-n*s)%. n is the interval number of samples from Tc, that is, the sample distance between the sample and the target sample, and n ∈ (0, Q / t). s is a preset probability attenuation value, for example, s can be 10. The sample distance between sample 1 and sample 0 is 1, and the sample distance between sample 2 and sample 0 is 2, so the charging behavior occurrence probability corresponding to sample 1 can be (100-1*10)% = 90%, the charging behavior occurrence probability corresponding to sample 2 can be (100-1*20)% = 80%, and so on. The charging behavior occurrence probability corresponding to all samples can be obtained in this way.

[0053] In step 103, the electronic device can respectively perform feature encoding on the plurality of samples to obtain feature-encoded data. The feature encoding can be understood as: respectively performing numerical feature statistics on the data in a single sample, and the numerical features include but are not limited to mean, maximum value, minimum value, variance, etc.

[0054] In step 104, the electronic device can input the feature-encoded data obtained in step 103 into a recurrent neural network for iterative learning until the network converges, to obtain a trained charging behavior prediction model. The charging behavior prediction model is used to predict whether the vehicle will soon perform a charging behavior.

[0055] In an exemplary embodiment, the recurrent neural network can include a neural network layer, a fully connected layer, and a softmax layer. The neural network layer is used to perform dimension reduction processing and feature integration on the feature-encoded data to obtain a first feature vector. The fully connected layer is used to perform dimension reduction processing on the first feature to obtain a second feature vector, which can adapt to the input of the softmax layer. The softmax layer is used to normalize the second feature vector into a probability distribution vector. That is, the softmax layer can be used as the last layer of the recurrent neural network, which is used for the output of the multi-classification problem.

[0056] The neural network layer can be a Long Short Term Memory (LSTM) recurrent neural network layer, a Recurrent Neural Network (RNN) layer, a Gate Recurrent Unit (GRU) layer, or a transformer layer.

[0057] In an exemplary embodiment, a schematic diagram of the recurrent neural network can refer to Figure 3 , which comprises an attention module 301, a neural network layer 302, a fully connected layer 303, and a softmax layer 304. The attention module 301 is used to process the feature-encoded data, output a target feature vector, and achieve the effect of strengthening important features and weakening unimportant features. The important features are features associated with the prediction of the probability of charging behavior, i.e., features that need to be focused on when predicting the probability of charging behavior. The neural network layer 302 is used to reduce the dimension of the target feature vector and integrate the features to obtain a first feature vector. The fully connected layer is used to reduce the dimension of the first feature to obtain a second feature vector. The softmax layer is used to normalize the second feature vector into a probability distribution vector.

[0058] The attention module 301 can be a soft attention module, a hard attention module, a static attention module, a dynamic attention module, or a self-attention module.

[0059] In an exemplary embodiment, refer to Figure 3 , Figure 3 In the above embodiment, the frame number, the mileage, the battery SOC, the time point, the battery health status, the fast / slow charging flag, and the vehicle GPS sample data are input into the feature encoding layer, and the feature encoding layer outputs the feature-encoded data, so that the feature-encoded data are input into the recurrent neural network.

[0060] In this embodiment, the recurrent neural network is used to perform time series modeling on the historical vehicle data to obtain a charging behavior prediction model. The charging behavior prediction model can accurately predict whether the vehicle will soon perform a charging behavior.

[0061] The following describes an embodiment of the charging behavior prediction method of the embodiment of the present specification. The execution subject of the charging behavior prediction method is an electronic device, which can be a device with a charging behavior prediction function, such as a server, a vehicle terminal, or a cloud terminal. When the execution subject of the charging behavior prediction method is a server or a cloud terminal, the server or the cloud terminal can send the prediction result to the vehicle terminal after obtaining the prediction result, so that the vehicle terminal can know whether the vehicle will soon perform a charging behavior.

[0062] Figure 4 is a schematic flowchart of a charging behavior prediction method provided by an embodiment of the present application.

[0063] For example, Figure 4As shown, the method may include:

[0064] Step 401: Obtain the current time-series data sequence of the vehicle. The current time-series data sequence includes multiple preset data types corresponding to different times. The end time of the current time-series data sequence is the current time, and the start time of the current time-series data sequence is the target time preceding the current time.

[0065] Step 402: Based on the preset sampling duration, sample the current time series data sequence to obtain multiple sampled data.

[0066] Step 403: Perform feature encoding on the above multiple sampled data to obtain the feature-encoded sampled data.

[0067] Step 404: Input the sampled data after the above feature encoding into the charging behavior prediction model, and output the predicted probability of charging behavior.

[0068] In this embodiment, the charging behavior prediction model can accurately and reasonably predict the probability of charging behavior based on the vehicle's current time-series data sequence. This means it can accurately and reasonably predict the vehicle's charging intention, which to some extent improves the charging experience for car owners. For example, when a high probability of charging behavior is predicted, i.e., the vehicle is predicted to have a charging intention, relevant measures can be taken in advance to improve the charging experience for car owners. These measures could include preheating the battery by controlling the vehicle's battery management system, ensuring that the battery is already in its optimal working state when the user starts charging, reducing charging waiting time, and thus improving the charging experience for electric vehicle owners.

[0069] The following are Figure 4 The specific implementation methods of each step in the illustrated embodiments are described in detail below:

[0070] In step 401, the vehicle's current time-series data sequence includes multiple preset data types corresponding to different times. The interval between two adjacent times can be the same or different. The ending time corresponding to the current time-series data sequence is the current time, and the starting time corresponding to the current time-series data sequence is the target time before the current time. For example, the electronic device can use the current time as the starting point and acquire multiple preset data types corresponding to N times, where the interval between two adjacent times can be the same. The multiple preset data types may include, but are not limited to: VIN, mileage, battery SOC, battery health status, fast / slow charging indicators, vehicle GPS, etc.

[0071] For example, the interval length between the start time and the end time corresponding to the current time series data sequence can be the preset trip interval length Q mentioned in the above embodiments. For example, if the end time corresponding to the current time series data sequence is the current time Tin, the start time corresponding to the current time series data sequence is Tin-Q. That is, the current time series data sequence is the vehicle data between Tin-Q and Tin.

[0072] In step 402, the current time series data sequence can be sampled according to the preset sampling time t to obtain a plurality of sampling data. The sampling time t in this step is the same as the sampling time t mentioned in step 102 of the above embodiments, and details are not repeated here.

[0073] For example, the electronic device can take the current time Tin as the starting point, t as the sampling time, and continuously take Q / t sampling data, each of which corresponds to a time length of t.

[0074] In step 403, the electronic device can encode the features of the above plurality of sampling data to obtain the feature-encoded sampling data. Feature encoding can be understood as: respectively performing numerical feature statistics on the data in a single sampling data, and the numerical features include but are not limited to mean, maximum, minimum, variance, etc. The feature encoding method in this step is the same as the feature encoding method mentioned in step 103 of the above embodiments.

[0075] In step 404, the electronic device can input the above feature-encoded sampling data into the charging behavior prediction model to output the predicted charging behavior occurrence probability. The charging behavior prediction model is the trained charging behavior prediction model in any of the above embodiments. The charging behavior occurrence probability output in this step can be the probability of the vehicle possibly having a charging behavior within the sampling time t after the current time.

[0076] For example, if the charging behavior occurrence probability output in step 404 is greater than or equal to a preset probability, it can be determined that the prediction result is that a charging behavior will occur, that is, the vehicle is more likely to have a charging behavior within the sampling time t after the current time. If the charging behavior occurrence probability output in step 404 is less than the preset probability, it can be determined that the prediction result is that the vehicle is not likely to have a charging behavior within the sampling time t after the current time. The preset probability can be set according to actual needs, such as being between 80% and 95%, but this embodiment does not make specific limitations.

[0077] In the example embodiment, after step 404, the prediction method of the charging behavior can further include: when the probability of the charging behavior is greater than a preset probability, controlling the battery management system of the vehicle to preheat the battery. For example, when the probability of the charging behavior is greater than 90%, the controller in the vehicle can control the battery management system of the vehicle to preheat the battery, so that when the charging behavior actually occurs, the battery has been heated to a suitable working temperature. Assuming that the sampling duration t is 5 minutes, the current time is 9 o'clock, and the probability of the charging behavior in the next 5 minutes is predicted to be 90% at 9 o'clock, the battery management system of the vehicle can be controlled to preheat the battery, so that when the vehicle owner starts charging the vehicle at 9:05, the battery of the vehicle has been heated to a suitable working temperature, instead of waiting until 9:05 to control the battery management system of the vehicle to preheat the battery.

[0078] In this embodiment, considering that battery preheating needs to consume time, the charging intention of the vehicle is predicted based on the trained charging behavior prediction model, i.e., whether the vehicle will soon have a charging behavior, and when it is predicted that the vehicle will soon have a charging behavior, the battery can be preheated in advance, the waiting time for charging is shortened, and the charging experience of the vehicle owner is improved.

[0079] In summary, in the embodiment of the present application, the charging behavior prediction model is trained based on historical vehicle data and a recurrent neural network, so that whether the vehicle owner will charge the vehicle, i.e., whether the charging behavior will soon occur, is predicted in advance by the trained model. Based on the prediction result, the battery management system is controlled to preheat the battery, so that the battery can enter the working state in advance and the charging time is shortened.

[0080] The model training device of the embodiment of the present application will be described below, Figure 5 is a structural schematic diagram of a model training device provided by the embodiment of the present application.

[0081] As shown in the example, Figure 5 the device includes:

[0082] The acquisition module 501 is configured to acquire at least one historical time series data sequence of a vehicle; wherein the historical time series data sequence includes a plurality of preset type data corresponding to different time points, and the different time points include a historical charging behavior occurrence time point.

[0083] The sampling module 502 is configured to sample the historical time series data sequence according to a preset sampling duration to obtain a plurality of samples; wherein each sample corresponds to a charging behavior occurrence probability, and the closer the sample is to a target sample in the plurality of samples, the greater the charging behavior occurrence probability corresponding to the sample is, and the target sample is a sample corresponding to the historical charging behavior occurrence time point.

[0084] The feature encoding module 503 is configured to perform feature encoding on the plurality of samples to obtain feature-encoded data.

[0085] The training module 504 is configured to input the feature-encoded data into a recurrent neural network to perform model training, and obtain a trained charging behavior prediction model; the charging behavior prediction model is configured to predict whether a charging behavior will occur.

[0086] In a possible implementation, the occurrence time of the historical charging behavior corresponds to a terminal time of the historical time sequence data sequence.

[0087] In a possible implementation, the charging behavior occurrence probability corresponding to the sample is determined based on a preset probability decay value and a sample distance between the sample and the target sample.

[0088] In a possible implementation, the obtaining module is specifically configured to: collect historical vehicle data; and perform trip division on the historical vehicle data according to a preset trip interval duration, to obtain a plurality of historical time sequence data sequences of the vehicle; a duration corresponding to each historical time sequence data sequence is the trip interval duration.

[0089] In a possible implementation, the obtaining module is specifically configured to: determine occurrence times of a plurality of historical charging behaviors according to the historical vehicle data; and obtain a plurality of historical time sequence data sequences by taking each occurrence time of the historical charging behavior as a starting point of trip division and obtaining historical vehicle data within the trip interval duration.

[0090] The following describes an embodiment of an electronic device of an embodiment of the present specification, Figure 6 is a structural schematic diagram of an electronic device provided by an embodiment of the present application.

[0091] For example, as shown in Figure 6 The electronic device 600 includes a memory 601 and a processor 602, where the memory 601 stores executable program code, and the processor 602 is configured to invoke and execute the executable program code to perform a model training method or a charging behavior prediction method.

[0092] The present embodiment can divide the electronic device into functional modules according to the above method examples, for example, each functional module can be obtained, or two or more functions can be integrated into one processing module, and the integrated module can be implemented in the form of hardware. It should be noted that the division of the modules in the present embodiment is illustrative, and is only a logical functional division. In actual implementation, another division mode can be used.

[0093] In the case of adopting the respective functional modules corresponding to the respective functions, the electronic device can include an acquisition module, a sampling module, a feature encoding module, a training module, and the like. It should be noted that all related content of the respective steps involved in the above method embodiments can be referred to the function description of the corresponding functional modules, and will not be repeated here.

[0094] The electronic device provided in the embodiment is used to execute the above model training method or the charging behavior prediction method, and thus the same effects as the above implementation method can be achieved.

[0095] In the case of adopting the integrated unit, the electronic device can include a processing module and a storage module. The processing module can be used to control and manage the actions of the electronic device. The storage module can be used to support the electronic device to execute program codes and data, and the like.

[0096] The processing module can be a processor or a controller, which can realize or execute various exemplary logical blocks, modules and circuits described in combination with the disclosure of the present application. The processor can also be a combination of computing functions, such as one or more microprocessor combinations, a combination of digital signal processing (DSP) and microprocessor, and the like. The storage module can be a memory.

[0097] The embodiment also provides a computer readable storage medium, which stores computer program codes. When the computer program codes run on a computer, the computer is caused to execute the above related method steps to realize the model training method or the charging behavior prediction method in the above embodiment.

[0098] The embodiment also provides a computer program product. When the computer program product runs on a computer, the computer is caused to execute the above related steps to realize the model training method or the charging behavior prediction method in the above embodiment.

[0099] In addition, the electronic device provided by the embodiment of the present application can be a chip, an assembly or a module. The electronic device can include a connected processor and a memory. The memory is used to store instructions. When the electronic device runs, the processor can invoke and execute the instructions to enable the chip to execute the model training method or the charging behavior prediction method in the above embodiment.

[0100] The electronic device, the computer readable storage medium, the computer program product or the chip provided by the embodiment of the present application are all used to execute the corresponding method provided above, and thus the beneficial effects that can be achieved can be referred to the beneficial effects of the corresponding method provided above, which will not be repeated here.

[0101] Through the description of the above embodiments, those skilled in the art can understand that, for the convenience and brevity of description, only the division of the above functional modules is exemplified, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

[0102] In the embodiments provided in the present application, it should be understood that the disclosed device and method can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the modules or units is only a logical function division, and in actual implementation, there can be another division manner, for example, a plurality of units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed ones can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0103] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A model training method, characterized in that, The method comprises: obtaining at least one historical time series data sequence of a vehicle; wherein the historical time series data sequence comprises a plurality of preset type data and time data corresponding to different time points, the different time points include a historical charging behavior occurrence time point, and the plurality of preset type data comprises a vehicle frame number, a driving mileage, a battery state of charge, a battery health status, a fast / slow charging flag, and a vehicle position; According to a preset sampling time length, the historical time series data sequence is sampled to obtain a plurality of samples; wherein, the sample corresponds to a charging behavior occurrence probability, and the closer the sample is to the target sample in time, the greater the charging behavior occurrence probability corresponding to the sample is, the target sample is a sample corresponding to a time of occurrence of the historical charging behavior, the time of occurrence of the historical charging behavior is Tc, Tc is taken as a sampling starting point, the charging probability is 100%, and a Q time period is divided into n samples in a forward direction by a sampling time length t, and the charging behavior probability of each sample is expressed as (100-n s)%, n is a sample interval number of Tc, that is, a sample time distance between the sample and the target sample, n ∈ (0, Q / t), and s is a preset probability attenuation value. feature encoding is performed on the plurality of samples to obtain feature encoded data; the feature encoded data is input into a recurrent neural network for model training to obtain a trained charging behavior prediction model; wherein the charging behavior prediction model is used to predict a probability of a charging behavior occurring within a sampling duration.

2. The method of claim 1, wherein, The historical charging behavior occurrence time point is a termination time point corresponding to the historical time series data sequence.

3. The method of claim 1, wherein, The charging behavior occurrence probability corresponding to the sample is determined based on a preset probability decay value and a time distance between the sample and the target sample.

4. The method according to any one of claims 1 to 3, characterized in that, The obtaining of the at least one historical time series data sequence of the vehicle comprises: collecting historical vehicle data; dividing a trip of the historical vehicle data according to a preset trip interval duration to obtain a plurality of historical time series data sequences of the vehicle; wherein a duration corresponding to each historical time series data sequence is the trip interval duration.

5. The method of claim 4, wherein, The dividing of the trip of the historical vehicle data according to the preset trip interval duration to obtain the plurality of historical time series data sequences of the vehicle comprises: determining a plurality of historical charging behavior occurrence time points from the historical vehicle data; respectively taking each historical charging behavior occurrence time point as a starting point of trip division, and obtaining historical vehicle data within the trip interval duration to obtain a plurality of historical time series data sequences.

6. A method of predicting charging behavior, characterized by, The method comprises: obtaining a current time series data sequence of a vehicle; wherein the current time series data sequence comprises a plurality of preset type data and time data corresponding to different time points, a termination time point corresponding to the current time series data sequence is a current time point, a starting time point corresponding to the current time series data sequence is a target time point before the current time point, and the plurality of preset type data comprises a vehicle frame number, a driving mileage, a battery state of charge, a battery health status, a fast / slow charging flag, and a vehicle position; sampling data according to a preset sampling duration to obtain a plurality of sampling data; feature encoding is performed on the plurality of sampling data to obtain feature encoded sampling data; the feature encoded sampling data is input into a charging behavior prediction model to output a predicted charging behavior occurrence probability; wherein the charging behavior prediction model is a trained charging behavior prediction model obtained according to any one of claims 1 to 5.

7. The method of claim 6, wherein, After the output of the predicted charging behavior occurrence probability, the method further comprises: when the charging behavior occurrence probability is greater than a preset probability, controlling a battery management system of the vehicle to perform battery preheating.

8. A model training apparatus characterized by comprising: The device comprises: An acquisition module is configured to acquire at least one historical time series data sequence of a vehicle, wherein the historical time series data sequence comprises a plurality of preset type data and time data corresponding to different time points, the different time points include a time point of a historical charging behavior, and the plurality of preset type data includes a vehicle frame number, a driving mileage, a battery state of charge, a battery state of health, a fast / slow charging flag, and a vehicle position. The sampling module is configured to sample the historical time series data sequence according to a preset sampling duration to obtain a plurality of samples; wherein the samples correspond to charging behavior occurrence probabilities, and the closer the sample is to a target sample in time, the greater the charging behavior occurrence probability corresponding to the sample is, the target sample is a sample corresponding to a time of occurrence of the historical charging behavior, the time of occurrence of the historical charging behavior is Tc, Tc is taken as a sampling starting point, a charging probability is 100%, a Q time period is divided into n samples in a forward direction by a sampling duration t, and a charging behavior probability of each sample is represented as (100-n s)%, n is a number of intervals between the sample and Tc, that is, a sample time distance between the sample and the target sample, n ∈ (0, Q / t), and s is a preset probability attenuation value. A feature encoding module is configured to perform feature encoding on the plurality of samples to obtain feature-encoded data. A training module is configured to input the feature-encoded data into a recurrent neural network to perform model training, and obtain a trained charging behavior prediction model, wherein the charging behavior prediction model is configured to predict a probability of a charging behavior of the vehicle occurring within a sampling duration.

9. An electronic device, comprising: The electronic device comprises: a memory configured to store executable program code; a processor configured to call and run the executable program code from the memory, so that the electronic device performs the method according to any one of claims 1 to 5, or performs the method according to any one of claims 6 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, when the computer program is executed, the method according to any one of claims 1 to 5 is realized, or the method according to any one of claims 6 to 7 is realized.

Images