Cross-domain electric vehicle charging station load forecasting method and system based on transfer learning

Abstract

The present invention discloses a cross-domain electric vehicle charging station load prediction method and system based on transfer learning, comprising: obtaining source domain building energy consumption data and target domain charging station energy consumption data, and performing trend extraction respectively to obtain source domain trend prediction data and target domain trend prediction data; obtaining source domain residual data and target domain residual data respectively according to the source domain trend prediction data and the target domain trend prediction data, performing similarity analysis on the source domain residual data and the target domain residual data to obtain source domain residual data to be migrated; obtaining a cross-domain transfer learning network after model pre-training and model fine-tuning based on the source domain residual data to be migrated, predicting the target domain residual data according to the cross-domain transfer learning network to obtain target domain residual prediction data, and superimposing the target domain residual prediction data with the target domain trend prediction data to obtain the target domain charging station load prediction result. The problem of less data, poor prediction accuracy and negative migration of electric vehicle charging stations is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle charging station load prediction, and in particular to a cross-domain electric vehicle charging station load prediction method and system based on transfer learning. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Regarding the load data of electric vehicle charging stations, on the one hand, the newly built charging stations have a short time of use, so the data volume is small; on the other hand, although some charging stations have been built and put into operation for a period of time, the data credibility is low due to poor post-operation and maintenance. Therefore, due to the current problem of small amount of electric vehicle charging load data and low data credibility, the load forecasting data has certain limitations and the forecasting accuracy is low.

[0004] At present, building energy consumption prediction is relatively mature, with rich data sets, and deep network models based on building energy consumption prediction have good performance; however, the premise of good model performance is a large amount of historical data. When the sample data is small, the model performance is poor. For electric vehicle charging station load data, the main problems are small data volume and low data credibility. Directly using deep network models for prediction has poor results.

[0005] Transfer learning, as an important branch of machine learning, focuses on transferring learned knowledge to new problems. However, if the knowledge of the building domain is directly transferred to the electric vehicle charging station, and the building energy consumption data is directly used to predict the electric vehicle charging station, it involves cross-domain problems. The energy consumption patterns of the two are different, and the energy consumption data is not similar enough to meet the similarity analysis of transfer learning, so negative transfer problems will occur after cross-domain migration. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a cross-domain electric vehicle charging station load prediction method and system based on transfer learning, which extracts trends from source domain and target domain data respectively to obtain residual data of the source domain and target domain, pre-trains a neural network based on the source domain residual data, and migrates the data to the target domain residual data after fine-tuning, so as to solve the problems of small amount of electric vehicle charging station data, poor prediction accuracy and cross-domain negative transfer.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] In a first aspect, the present invention provides a cross-domain electric vehicle charging station load prediction method based on transfer learning, comprising:

[0009] Obtain source domain building energy consumption data and target domain charging station energy consumption data, and perform trend extraction respectively to obtain source domain trend prediction data and target domain trend prediction data;

[0010] According to the source domain trend prediction data and the target domain trend prediction data, source domain residual data and target domain residual data are obtained respectively, and similarity analysis is performed on the source domain residual data and the target domain residual data to obtain the source domain residual data to be migrated;

[0011] A cross-domain transfer learning network is obtained after model pre-training and model fine-tuning based on the source domain residual data to be migrated. The target domain residual data is predicted based on the cross-domain transfer learning network to obtain the target domain residual prediction data. The target domain residual prediction data is superimposed with the target domain trend prediction data to obtain the target domain charging station load prediction result.

[0012] As an optional implementation, the source domain building energy consumption data is subtracted from the source domain trend prediction data to obtain the source domain residual data, and the target domain charging station energy consumption data is subtracted from the target domain trend prediction data to obtain the target domain residual data.

[0013] As an optional implementation, the DDTW method is used to calculate the similarity between the source domain residual data and the target domain residual data, and the source domain residual data to be migrated is obtained by setting a similarity judgment threshold.

[0014] As an optional implementation method, the source domain residual data to be migrated is used to pre-train the LSTM neural network, and the trained LSTM neural network is fine-tuned to obtain a cross-domain transfer learning network.

[0015] As an optional implementation method, the model fine-tuning process includes source domain training and target domain training; specifically, the LSTM neural network includes seven LSTM layers and one fully connected layer. During source domain training, the last three LSTM layers are locked, and the first four LSTM layers are trained according to the source domain residual data to be migrated, and the parameters and weights of each LSTM layer are saved; during target domain training, the last three LSTM layers are unlocked, and the last three LSTM layers are trained according to the target domain residual data, and the first four LSTM layers are trained with the target domain residual data according to the parameters and weights learned from the source domain training, and the fully connected layer is updated at the same time.

[0016] As an optional implementation, data preprocessing is performed on the source domain building energy consumption data and the target domain charging station energy consumption data before trend extraction, including data normalization processing using a maximum-minimum standardization method.

[0017] As an optional implementation, support vector machines are used to extract trends from the preprocessed source domain building energy consumption data and target domain charging station energy consumption data.

[0018] In a second aspect, the present invention provides a cross-domain electric vehicle charging station load forecasting system based on transfer learning, comprising:

[0019] The cross-domain electric vehicle charging station load forecasting system based on transfer learning is characterized by including:

[0020] A trend extraction module is configured to obtain source domain building energy consumption data and target domain charging station energy consumption data, and perform trend extraction respectively to obtain source domain trend prediction data and target domain trend prediction data;

[0021] The residual extraction module is configured to obtain source domain residual data and target domain residual data according to the source domain trend prediction data and the target domain trend prediction data, respectively, and perform similarity analysis on the source domain residual data and the target domain residual data to obtain the source domain residual data to be migrated;

[0022] The cross-domain migration module is configured to obtain a cross-domain transfer learning network based on the source domain residual data to be migrated after model pre-training and model fine-tuning, predict the target domain residual data based on the cross-domain transfer learning network, obtain the target domain residual prediction data, and superimpose the target domain residual prediction data with the target domain trend prediction data to obtain the target domain charging station load prediction result.

[0023] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in the first aspect is performed.

[0024] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The present invention proposes a cross-domain electric vehicle charging station load prediction method and system based on transfer learning, which extracts trends from source domain and target domain data respectively to obtain residual data of the source domain and target domain, pre-trains a neural network based on the source domain residual data, and migrates the data to the target domain residual data after fine-tuning, thereby solving the problems of small amount of data, poor prediction accuracy and cross-domain negative migration of electric vehicle charging stations.

[0027] The present invention proposes a cross-domain electric vehicle charging station load prediction method and system based on transfer learning. Building energy consumption data is selected as the source domain data set in view of the advantages of rich building energy consumption data sets and high reliability, while the target domain electric vehicle charging station data sets are small and of low credibility, which solves the current problems of insufficient charging station data sets and low data quality; the migration direction is positioned to the residual data after trend prediction, and the LSTM neural network is trained according to the source domain residual data after trend prediction, and cross-domain migration is performed on the target domain residual data to avoid negative migration.

[0028] The present invention proposes a cross-domain electric vehicle charging station load prediction method and system based on transfer learning. In the model fine-tuning strategy, the last few layers are locked during source domain training and unlocked during target domain training. The last few layers are adapted to the target domain data to achieve a more accurate prediction. The fine-tuned network has better prediction accuracy and lower training burden.

[0029] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0031] Figure 1 A flow chart of a cross-domain electric vehicle charging station load prediction method based on transfer learning provided in Example 1 of the present invention;

[0032] Figure 2 A block diagram of a cross-domain electric vehicle charging station load prediction method based on transfer learning provided in Example 1 of the present invention;

[0033] Figure 3 A schematic diagram of similarity analysis provided in Example 1 of the present invention;

[0034] Figure 4 Schematic diagram of the model fine-tuning strategy provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0035] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0036] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0037] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0038] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0039] Example 1

[0040] This embodiment provides a cross-domain electric vehicle charging station load prediction method based on transfer learning. Figure 1-Figure 2 As shown, including:

[0041] Obtain source domain building energy consumption data and target domain charging station energy consumption data, and perform trend extraction respectively to obtain source domain trend prediction data and target domain trend prediction data;

[0042] According to the source domain trend prediction data and the target domain trend prediction data, the source domain residual data and the target domain residual data are obtained respectively;

[0043] Perform similarity analysis on the source domain residual data and the target domain residual data to obtain the source domain residual data to be migrated;

[0044] A cross-domain transfer learning network is obtained after model pre-training and model fine-tuning based on the source domain residual data to be migrated. The target domain residual data is predicted based on the cross-domain transfer learning network to obtain the target domain residual prediction data. The target domain residual prediction data is superimposed with the target domain trend prediction data to obtain the target domain charging station load prediction result.

[0045] In this embodiment, the source domain building energy consumption data and the target domain charging station energy consumption data are divided into 70% training set and 30% test set respectively; the training set and the test set are reconstructed using the sliding time window length r; in this embodiment, r is taken as 30 as an example, that is, the sliding time window cuts out the first 30 data as the model input, and the 31st data input is used as the model output to form a time series window; the sliding time window rolls back one cycle each time to form a new time series window, until this window rolls to the last time point, and the reconstructed data is used to train the prediction model; the specific formulas are as follows:

[0046]

[0047]

[0048] Among them, T is the source domain, G is the target domain, t is the t-th data output, r is the length of the sliding time window, is the source domain output, Input for the source domain; is the target domain output, Enter for the target domain.

[0049] In this embodiment, when acquiring the energy consumption data of the source domain building and the target domain charging station, the similarity between the cycles of the building and the charging station is taken into consideration, and the data of the daily cycle and the weekly cycle are used as the input for model training.

[0050] In the following trend extraction and residual prediction processes, a sliding time window L=30 is used to reconstruct data.

[0051] In this embodiment, data preprocessing is performed on the source domain building energy consumption data and the target domain charging station energy consumption data;

[0052] Specifically, the maximum-minimum standardization method is used to normalize the data; the maximum-minimum standardization is to perform a linear transformation on the original data, and minA and minB are the maximum and minimum values ​​of the data respectively. The original value x of the data is mapped to the value x′ in the interval [0,1] through the maximum-minimum standardization, as shown in formula (3):

[0053]

[0054] In this embodiment, trend extraction is performed on the preprocessed source domain building energy consumption data and target domain charging station energy consumption data respectively; specifically, support vector machine (SVM) is used for trend extraction;

[0055] Support vector machine (SVM) was proposed for binary classification problems, but SVR samples ultimately have only one category. The optimal hyperplane it seeks is not the one that separates two or more categories of sample points as closely as SVM does, but the one that minimizes the total deviation of all sample points from the hyperplane.

[0056] In this embodiment, the source domain trend prediction data and the target domain trend prediction data are obtained through trend extraction, and the source domain residual data and the target domain residual data are obtained therefrom;

[0057] Specifically, the source domain residual data is obtained by subtracting the source domain trend prediction data from the source domain building energy consumption data, and the target domain residual data is obtained by subtracting the target domain trend prediction data from the target domain charging station energy consumption data. The obtained source domain residual data and target domain residual data are the objects of true cross-domain migration; as shown in formula (4)-formula (5):

[0058]

[0059]

[0060] Among them, Y t T is the source domain building energy consumption data, is the source domain trend prediction data, is the source domain residual data; Y t G is the target domain charging station energy consumption data, For the target domain trend prediction data, is the target domain residual data.

[0061] In this embodiment, since the use of building energy consumption data to predict the load of electric vehicle charging stations involves cross-domain issues, the building domain and the charging station energy consumption data are not similar enough. When the knowledge learned in the source domain has a negative effect on the learning in the target domain, "negative transfer" will occur. Therefore, in order to solve the problem of negative transfer, this embodiment proposes a method of trend extraction and error transfer, which converts the transferred object into a residual, does not involve domain issues, and avoids the "negative transfer" caused by cross-domain.

[0062] In this embodiment, the source domain residual data and the target domain residual data are subjected to similarity analysis, and the source domain residual data to be migrated is obtained by setting a similarity judgment threshold;

[0063] like Figure 3 As shown in FIG. 1 , specifically, DDTW is used to calculate the similarity between the source domain residual data and the target domain residual data. The output of DDTW is the distance between the two. That is, the smaller the output result, the higher the similarity, and vice versa.

[0064] Set a similarity judgment threshold. If the similarity judgment threshold is met, transfer learning is performed. If it is not met, the source domain data is replaced until the similarity judgment is met.

[0065] DDTW uses the first-order derivative of the estimated time series data to solve the continuity and monotonicity conditions of the warping path in DTW (dynamic time warping) by considering the "shape", which causes various distortions in the time series data alignment process. At this time, the elements in the "distance matrix" are no longer the distance between two points, but the square of the difference of the first-order derivative of the time series data at the two points, as shown in the following formula:

[0066]

[0067]

[0068] D x [c,q]=[Dx [c],D x [q]] (8)

[0069] Among them, D x [q] is the source domain distance matrix, D x [c] represents the target domain distance matrix; D x [c, q] represents the distance cost matrix;

[0070] Find the Euclidean distance between each element of the distance cost matrix The sum is the DDTW output result.

[0071] In this embodiment, a certain source domain residual data is randomly distributed into n parts according to a random distributor, each part of the data is subjected to a DDTW operation with the target domain residual data, and finally the n DDTW results are averaged to obtain the final result.

[0072] In this embodiment, after the source domain residual data to be migrated is preprocessed by maximum-minimum normalization, the LSTM neural network is pre-trained with the source domain residual data to be migrated, and a cross-domain transfer learning network is obtained after model fine-tuning. The target domain residual data is predicted according to the cross-domain transfer learning network to obtain target domain residual prediction data, and the target domain residual prediction data is superimposed with the target domain trend prediction data to obtain the final prediction result.

[0073] like Figure 4 As shown in the figure, the model fine-tuning process includes: the LSTM neural network adopts a network structure of seven LSTM layers and one fully connected layer, and MSE is used as the loss function to evaluate the network performance. The number of units in each layer is 200, Dropout (0.5) is added to each LSTM layer, and the Adam algorithm is used to optimize the network training;

[0074] The training includes source domain training and target domain training. For source domain training, the last three layers of LSTM are locked for source domain training, and the first four layers of LSTM are pre-trained according to the source domain data, and the parameters and weights of each layer of LSTM are saved. For target domain training, the last three layers of LSTM are unlocked, and the first four layers are trained according to the parameters and weights learned from the source domain for target domain data. The parameters and weights of the last three layers are not changed to adapt to the target domain training data to achieve more accurate prediction. At the same time, the fully connected layer is updated, and a small learning rate (0.0001) is set to prevent overfitting.

[0075] In the model fine-tuning strategy of this embodiment, the last few layers are locked during source domain training and unlocked during target domain training. The last few layers are adapted to the target domain data to achieve more accurate prediction. After fine-tuning, the network has better prediction accuracy and lower training burden.

[0076] In this embodiment, the target domain residual data is predicted according to the cross-domain transfer learning network to obtain the target domain residual prediction data, and the target domain residual prediction data is superimposed on the target domain trend prediction data to obtain the target domain final prediction data:

[0077]

[0078] in, is the target domain residual prediction data, Final prediction data for the target domain.

[0079] This embodiment selects building energy consumption data as the source domain data set because the building energy consumption data set is rich and reliable, while the target domain electric vehicle charging station data set is small and has low credibility, which solves the current problems of insufficient charging station data sets and low data quality. This embodiment locates the migration direction to the residual data after trend prediction, trains the LSTM neural network based on the source domain residual data after trend prediction, and performs cross-domain migration on the target domain residual data to avoid negative migration.

[0080] Example 2

[0081] This embodiment provides a cross-domain electric vehicle charging station load prediction system based on transfer learning, including:

[0082] A trend extraction module is configured to obtain source domain building energy consumption data and target domain charging station energy consumption data, and perform trend extraction respectively to obtain source domain trend prediction data and target domain trend prediction data;

[0083] The residual extraction module is configured to obtain source domain residual data and target domain residual data according to the source domain trend prediction data and the target domain trend prediction data, respectively, and perform similarity analysis on the source domain residual data and the target domain residual data to obtain the source domain residual data to be migrated;

[0084] The cross-domain migration module is configured to obtain a cross-domain transfer learning network based on the source domain residual data to be migrated after model pre-training and model fine-tuning, predict the target domain residual data based on the cross-domain transfer learning network, obtain the target domain residual prediction data, and superimpose the target domain residual prediction data with the target domain trend prediction data to obtain the target domain charging station load prediction result.

[0085] It should be noted that the above modules correspond to the steps described in Example 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer executable instructions.

[0086] In further embodiments, there is also provided:

[0087] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, it will not be described in detail here.

[0088] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0089] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0090] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in Example 1 is completed.

[0091] The method in Example 1 can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it is not described in detail here.

[0092] Those skilled in the art will appreciate that the units, i.e., algorithm steps, of the various examples described in conjunction with this embodiment can be implemented in electronic hardware or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0093] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A cross-domain electric vehicle charging station load forecasting method based on transfer learning, characterized in that: include: Obtain source domain building energy consumption data and target domain charging station energy consumption data, and perform trend extraction respectively to obtain source domain trend prediction data and target domain trend prediction data; According to the source domain trend prediction data and the target domain trend prediction data, source domain residual data and target domain residual data are obtained respectively, and similarity analysis is performed on the source domain residual data and the target domain residual data to obtain the source domain residual data to be migrated; A cross-domain transfer learning network is obtained after model pre-training and model fine-tuning based on the source domain residual data to be transferred. Specifically, the LSTM neural network is pre-trained based on the source domain residual data to be transferred, and the cross-domain transfer learning network is obtained after model fine-tuning of the trained LSTM neural network. The model fine-tuning process includes source domain training and target domain training; specifically, the LSTM neural network includes seven LSTM layers and one fully connected layer. During source domain training, the last three LSTM layers are locked, the first four LSTM layers are trained according to the source domain residual data to be migrated, and the parameters and weights of each LSTM layer are saved; during target domain training, the last three LSTM layers are unlocked, and the last three LSTM layers are trained according to the target domain residual data, the first four LSTM layers are trained according to the target domain residual data according to the parameters and weights learned from the source domain training, and the fully connected layer is updated at the same time; The target domain residual data is predicted according to the cross-domain transfer learning network to obtain the target domain residual prediction data. The target domain residual prediction data is superimposed with the target domain trend prediction data to obtain the target domain charging station load prediction result.

2. The cross-domain electric vehicle charging station load forecasting method based on transfer learning as claimed in claim 1 is characterized in that: The source domain residual data is obtained by subtracting the source domain trend prediction data from the source domain building energy consumption data, and the target domain residual data is obtained by subtracting the target domain trend prediction data from the target domain charging station energy consumption data.

3. The cross-domain electric vehicle charging station load forecasting method based on transfer learning as claimed in claim 1, characterized in that: The DDTW method is used to calculate the similarity between the source domain residual data and the target domain residual data. By setting the similarity judgment threshold, the source domain residual data to be migrated is obtained.

4. The cross-domain electric vehicle charging station load forecasting method based on transfer learning as claimed in claim 1, characterized in that: Before trend extraction of source domain building energy consumption data and target domain charging station energy consumption data, data preprocessing is performed, including data normalization using the maximum-minimum standardization method.

5. The cross-domain electric vehicle charging station load forecasting method based on transfer learning as claimed in claim 4 is characterized in that: Support vector machine is used to extract trends from the preprocessed source domain building energy consumption data and target domain charging station energy consumption data.

6. A cross-domain electric vehicle charging station load forecasting system based on transfer learning, characterized in that: include: A trend extraction module is configured to obtain source domain building energy consumption data and target domain charging station energy consumption data, and perform trend extraction respectively to obtain source domain trend prediction data and target domain trend prediction data; The residual extraction module is configured to obtain source domain residual data and target domain residual data according to the source domain trend prediction data and the target domain trend prediction data, respectively, and perform similarity analysis on the source domain residual data and the target domain residual data to obtain the source domain residual data to be migrated; The cross-domain migration module is configured to obtain a cross-domain migration learning network after model pre-training and model fine-tuning according to the source domain residual data to be migrated. Specifically, the LSTM neural network is pre-trained with the source domain residual data to be migrated, and the cross-domain migration learning network is obtained after model fine-tuning of the trained LSTM neural network; The model fine-tuning process includes source domain training and target domain training; specifically, the LSTM neural network includes seven LSTM layers and one fully connected layer. During source domain training, the last three LSTM layers are locked, the first four LSTM layers are trained according to the source domain residual data to be migrated, and the parameters and weights of each LSTM layer are saved; during target domain training, the last three LSTM layers are unlocked, and the last three LSTM layers are trained according to the target domain residual data, the first four LSTM layers are trained according to the target domain residual data according to the parameters and weights learned from the source domain training, and the fully connected layer is updated at the same time; The target domain residual data is predicted according to the cross-domain transfer learning network to obtain the target domain residual prediction data. The target domain residual prediction data is superimposed with the target domain trend prediction data to obtain the target domain charging station load prediction result.

7. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 5 is completed.

8. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 5.

Images