Method and system for judging charging mode based on convolutional self-coding extraction features

Abstract

The invention relates to the technical field of batteries, and discloses a method and system for judging a charging mode based on convolutional self-encoding feature extraction, and the method comprises the steps: collecting a plurality of historical charging fragments of a vehicle, the historical charging fragments comprising a plurality of groups of feature data changing with the SOC; performing format conversion, data filling and standardization on each group of feature data in each historical charging segment to obtain first charging data; inputting each group of first charging data in each historical charging segment into a convolutional auto-encoder for feature extraction, and respectively outputting extracted data corresponding to each group of first charging data; averaging each group of first charging data to obtain second charging data; splicing the second charging data into the corresponding feature data to obtain third charging data; and dividing different charging modes in the third charging data by adopting a density clustering method. The method has the advantage that accurate charging mode identification is realized by combining the convolution auto-encoder and the density clustering method.

Description

Technical Field

[0001] The present invention relates to the field of battery technology, and in particular to a method and system for judging a charging mode based on convolutional autoencoding feature extraction. Background Art

[0002] With the rapid popularization of new energy vehicles, the coverage of charging infrastructure and the quality of charging services have gradually become important aspects of user experience and industry development. At present, the research on the charging mode of new energy vehicles is mainly concentrated from the perspective of charging pile operators, focusing on the utilization efficiency of charging piles, site distribution optimization and load management. However, from the user's perspective, that is, the research on the charging mode of new energy vehicles themselves is relatively rare.

[0003] Existing research on charging modes usually relies on infrastructure data provided by charging piles, such as charging time, power curves, and load conditions. The acquisition of such data is more dependent on the intelligence level of charging piles, and is mainly concentrated in cities with dense charging facilities. However, at this stage, most charging piles are not equipped with smart devices, which makes it difficult for users to accurately understand in real time how long it will take for their vehicles to charge at a specific charging pile.

[0004] In addition, the data provided by charging piles is mostly centered on infrastructure operations, lacking a comprehensive description of vehicle charging behavior and patterns. This information gap not only limits users' effective planning of charging behavior, but also poses challenges to regional energy management and optimization of charging services.

[0005] Therefore, starting from the perspective of new energy vehicles themselves, taking user behavior and vehicle charging characteristics as entry points, exploring the division and analysis of multi-dimensional charging modes has become a difficult problem that needs to be solved urgently. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a method and system for judging the charging mode based on convolutional autoencoder feature extraction. By combining the convolutional autoencoder and the density clustering method, accurate charging mode recognition is achieved, and users are provided with support such as charging time prediction and pattern recognition. It also provides a scientific basis for optimizing the layout of charging piles, charging behavior management and regional energy scheduling.

[0007] In a first aspect, the present invention provides a method for determining a charging mode based on feature extraction using a convolutional autoencoder, comprising the following steps:

[0008] S1. Collect multiple historical charging segments of the vehicle, wherein the historical charging segments include multiple sets of characteristic data that change with SOC, and the characteristic data include charging current, charging voltage, ambient temperature and time;

[0009] S2, performing format conversion on each set of characteristic data in each historical charging segment, performing data filling on each set of characteristic data after format conversion, and standardizing each set of characteristic data after data filling to obtain first charging data;

[0010] S3, inputting each group of first charging data in each historical charging segment into a convolutional autoencoder for feature extraction, and outputting the extracted data corresponding to each group of first charging data respectively;

[0011] S4, taking the average of each group of first charging data to obtain second charging data; splicing the second charging data to the extracted data of the corresponding first charging data to obtain third charging data;

[0012] S5. Use a density clustering method to divide different charging modes in the third charging data.

[0013] Further, in S2, the format conversion includes:

[0014] S21, setting the first value, the second value and the fixed value to be positive constants less than 1, and the second value minus the first value equals the fixed value;

[0015] S22, in the process of the SOC changing from the first value to the second value, taking the average of the corresponding charging current, charging voltage and ambient temperature respectively, and calculating the total time required;

[0016] S23, the first value is added to the fixed value, and S22 is repeated until the SOC reaches a maximum value, thereby obtaining four sets of sequences of charging current, charging voltage, ambient temperature and time as the SOC increases.

[0017] Further, in S2, each set of feature data after format conversion is respectively filled with data, including:

[0018] The low-rank matrix filling method is used to unify the dimensions of the four groups of sequences of each historical charging segment.

[0019] Further, in S2, the steps of normalizing the groups of characteristic data after data filling to obtain first charging data include:

[0020] The four groups of sequences with unified dimensions in each historical charging segment are scaled to the interval [0,1].

[0021] Further, the convolutional autoencoder comprises:

[0022] Three parallel feature extraction paths and two fully connected layers that aggregate the three feature extraction paths;

[0023] Among them, each feature extraction path passes through the convolution layer, batch normalization layer and activation function in sequence; in the convolution layer, the convolution kernel sizes of the three feature extraction paths are 5×5, 7×7 and 9×9 respectively; each fully connected layer is followed by an activation function.

[0024] In a second aspect, the present invention provides a system for determining a charging mode based on feature extraction using a convolutional autoencoder, comprising:

[0025] Collection module: collects multiple historical charging segments of the vehicle, wherein the historical charging segments include multiple sets of characteristic data that change with SOC, and the characteristic data include charging current, charging voltage, ambient temperature and time;

[0026] Feature extraction module: connected to the acquisition module, used to convert the formats of each group of feature data in each historical charging segment, perform data filling on each group of feature data after format conversion, and standardize each group of feature data after data filling to obtain first charging data; input each group of first charging data in each historical charging segment into a convolutional autoencoder for feature extraction, and output the extracted data corresponding to each group of first charging data;

[0027] Clustering module: connected to the feature extraction module, used to take the average of each group of first charging data to obtain second charging data; splice the second charging data into the extracted data of the corresponding first charging data to obtain third charging data; and use density clustering method to divide different charging modes in the third charging data.

[0028] In a third aspect, the present invention provides an electronic device, the electronic device comprising:

[0029] Processor and memory;

[0030] The processor is used to execute the steps of the method described in the first aspect by calling the program or instruction stored in the memory.

[0031] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a program or instruction, wherein the program or instruction enables a computer to execute the steps of the method described in the first aspect.

[0032] The embodiments of the present invention have the following technical effects:

[0033] (1) The present application comprehensively considers the relationship between charging current, charging voltage, ambient temperature, time and charging performance, and performs format conversion, data filling and other processing on the acquired historical charging fragments, which greatly reduces the inconsistency of the data and ensures the integrity and accuracy of the data; by extracting the deep features of each group of feature data by the convolutional autoencoder and classifying them by density clustering method, it can realize efficient and accurate charging mode recognition when the charging data is complex and diverse, and improve the robustness and applicability of the analysis; wherein, the convolutional autoencoder increases the convolution path and adds the fully connected layer, so that the feature extraction is more comprehensive and more accurate;

[0034] (2) This application breaks away from the limitation of traditional reliance on charging pile infrastructure data by analyzing the charging data characteristics of new energy vehicles themselves, realizes the classification and analysis of charging modes from the perspective of user vehicles, and makes up for the shortcomings of existing research in user-side charging mode identification;

[0035] (3) The charging mode analysis results generated by this application can provide users with a basis for charging time prediction and charging behavior recommendations, helping users to better plan their charging behavior, reduce charging anxiety, and improve the user experience of new energy vehicles;

[0036] (4) The charging mode analysis results generated by this application can provide a scientific basis for charging pile operators, support charging pile site optimization and charging station load management; at the same time, by mining the charging behavior characteristics of regional users, it can provide more accurate demand forecasts for power system dispatching;

[0037] (5) This application does not rely on specific intelligent charging pile equipment, and can achieve mode classification only through vehicle charging data, which has wide applicability; it can run efficiently on multiple hardware platforms such as edge computing devices and cloud servers, facilitating large-scale promotion and deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 It is a flow chart of a method for judging a charging mode based on extracting features by convolutional autoencoding provided by an embodiment of the present invention;

[0040] Figure 2 is a structural diagram of a convolutional autoencoder provided by an embodiment of the present invention;

[0041] Figure 3 is a characteristic distribution diagram of the experimental group provided in Example 2 of the present invention;

[0042] Figure 4 is a characteristic distribution diagram of the control group provided in Example 2 of the present invention;

[0043] Figure 5 is a characteristic distribution diagram of the first vehicle provided in Example 3 of the present invention;

[0044] Figure 6 This is a characteristic distribution diagram of the second vehicle provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0046] First, the following instructions are given:

[0047] (1) New energy vehicle charging mode refers to the charging behavior and characteristics of the vehicle under different charging conditions, involving the dynamic changes of charging current, voltage, temperature, SOC and other parameters. These modes are affected by multiple factors, including battery characteristics, charging pile capacity, user behavior and external environmental conditions.

[0048] During the charging process, different charging modes will show significant differences. For example, the slow charging mode is characterized by low current and is usually used in home charging scenarios; the fast charging mode is characterized by higher current and power and is mostly used in public charging facilities; and the supercharging mode, that is, the super fast charging mode, aims at extremely high power and rapid energy replenishment, and has high technical requirements for batteries and charging facilities. In these modes, there are obvious mode differences in the characteristics of battery temperature rise, SOC change rate, current and voltage fluctuations, etc.

[0049] For example, in constant current mode, the current remains constant and the voltage gradually increases with SOC; in constant voltage mode, the voltage remains constant and the current gradually decreases. The temperature will affect the chemical reaction rate and internal resistance of the battery: at low temperatures, the internal resistance increases, limiting the rate of increase of current and voltage, and at high temperatures, the current needs to be limited to avoid thermal runaway. These characteristic parameters jointly determine the choice of charging mode and the timing of switching to ensure charging efficiency and safety.

[0050] Therefore, voltage, current and temperature can be used as key characteristic parameters to determine the charging mode.

[0051] (2) It should be noted that the charging process of a vehicle can be divided into different charging modes according to different charging conditions, and the charging process under the same charging mode has similar charging curves. When using a multi-layer perceptron to simulate the charging process under a unified charging mode, the correlation coefficient is very high.

[0052] Figure 1 is a flow chart of a method for determining a charging mode based on extracting features by convolutional autoencoder according to an embodiment of the present invention. Figure 1 , specifically including:

[0053] S1. Collect multiple historical charging segments of the vehicle, wherein the historical charging segments include multiple sets of characteristic data that change with SOC, and the characteristic data include charging current, charging voltage, ambient temperature and time;

[0054] The historical charging segment includes a complete charging process in the historical charging data, that is, the charging time is continuous and uninterrupted, rather than the process of SOC from 1% to 100%. The charging voltage, charging current, etc. can be a curve or a constant value.

[0055] S2, performing format conversion on each set of characteristic data in each historical charging segment, performing data filling on each set of characteristic data after format conversion, and standardizing each set of characteristic data after data filling to obtain first charging data;

[0056] Specifically, the format conversion method is:

[0057] S21, setting the first value, the second value and the fixed value to be positive constants less than 1, and the second value minus the first value equals the fixed value;

[0058] S22, in the process of the SOC changing from the first value to the second value, taking the average of the corresponding charging current, charging voltage and ambient temperature respectively, and calculating the total time required;

[0059] S23, the first value is added to the fixed value, and S22 is repeated until the SOC reaches a maximum value, thereby obtaining four sets of sequences of charging current, charging voltage, ambient temperature and time as the SOC increases.

[0060] Assuming the first value is 20% and the second value is 21%, the charging current, charging voltage and ambient temperature are averaged during this process, and the corresponding time is the sum of the time of this stage; if the first value is increased to 21%, the same operation is performed in the process of SOC from 21% to 22%, and the accumulation is performed in sequence.

[0061] The low-rank matrix filling method is used to unify the dimensions of the four groups of sequences of each historical charging segment. The charging current, charging voltage, ambient temperature and time during the SOC from 0 to 100% are filled. Because the charging duration of the acquired charging segments is inconsistent, it is difficult to unify the data dimensions. In the subsequent density clustering algorithm, a unified data dimension is required, so the low-rank matrix filling method is used here to ensure the unification of the data dimension and reduce the possibility of large data filling deviation affecting subsequent calculations.

[0062] The four groups of sequences with unified dimensions in each historical charging segment are scaled to the interval of [0, 1] to eliminate the influence of different feature dimensions and improve the performance and stability of subsequent algorithms.

[0063] S3, inputting each group of first charging data in each historical charging segment into a convolutional autoencoder for feature extraction, and outputting the extracted data corresponding to each group of first charging data respectively;

[0064] See also Figure 2 The convolutional autoencoder includes three parallel feature extraction paths and two fully connected layers that aggregate the three feature extraction paths; each feature extraction path passes through the convolution layer, batch normalization layer and activation function in sequence; in the convolution layer, the convolution kernel sizes of the three feature extraction paths are 5×5, 7×7 and 9×9 respectively; each fully connected layer is followed by an activation function.

[0065] For high-frequency fluctuations, such as the instantaneous changes caused by noise and uncontrollable factors during the charging process, smaller convolution kernels can capture these details more accurately, but 3×3 convolution kernels cannot capture effective features in data such as current and voltage, and insufficient feature extraction will affect model performance. The 5×5 convolution kernel is moderate and can smooth short-term fluctuations, filter noise, and effectively extract local features in the data, ensuring that the model has better feature representation capabilities on time series data, and is suitable for capturing local features in a short period of time, such as small fluctuations in current and voltage within a small time range.

[0066] The 7×7 convolution kernel is suitable for capturing feature extraction of medium time spans, which can smooth local noise and capture medium-scale curve change trends. In the charging curve, medium-length fluctuations may reflect the gradual changes in current and voltage due to battery characteristics or environmental conditions.

[0067] The 9×9 convolution kernel is suitable for extracting long-term trends and global features, such as the gradual change of charging voltage in the constant voltage stage, or the overall change curve of SOC over time. It can filter high-frequency noise in a short period of time and focus more on the low-frequency characteristics of the curve.

[0068] The above design ensures that the model can adapt to local fluctuations in the charging process while grasping the laws of global changes.

[0069] After feature extraction, batch normalization layer and activation function are used to enhance training stability and nonlinear expression ability. Finally, local features extracted by convolutional layers of different paths are aggregated and compressed to form a global representation through a fully connected layer, and deep feature representation is output, that is, extracted data. The activation function of this application uses relu function.

[0070] S4, taking the average of each group of first charging data to obtain second charging data; splicing the second charging data to the extracted data of the corresponding first charging data to obtain third charging data;

[0071] The role of splicing is to fully integrate the temporal features and statistical features to ensure the comprehensiveness and accuracy of feature expression, thereby providing more effective data input for subsequent charging mode division and improving the classification performance and result stability of the density clustering method.

[0072] S5. Use a density clustering method to divide different charging modes in the third charging data.

[0073] The density clustering method can cluster high-density areas in the feature space based on the density characteristics of the data distribution without pre-specifying the number of clusters. The density clustering method has advantages in dealing with irregularly shaped clusters and can identify abnormal data points, i.e., noise, making it suitable for situations where the charging data is unevenly distributed and has outliers, ultimately achieving accurate division of charging modes. After the division, it is also necessary to manually define the types of each divided charging mode.

[0074] In summary, this application divides the charging modes from the perspective of new energy vehicles by combining convolutional autoencoders with density clustering.

[0075] After the classification, the vehicle charging status can be accurately predicted based on the classified data and types:

[0076] 1. Prediction of charging time: Based on the charging mode, it is possible to infer the approximate time required for the vehicle to be fully charged from different SOC states, such as 10%-20%, under a certain charging mode, including the shortest time and the longest time. This prediction can help car owners better plan charging time and provide a basis for charging efficiency evaluation for vehicle operations.

[0077] 2. Charging segment classification and full-process inference: Through methods such as decision trees, we can determine the charging mode to which a part of the charging segment belongs, and infer the duration of the complete charging process based on the mode.

[0078] Based on the above technical solution, the present application comprehensively considers the relationship between charging current, charging voltage, ambient temperature, time and charging performance, and performs format conversion, data filling and other processing on the acquired historical charging fragments, which greatly reduces the inconsistency of the data and ensures the integrity and accuracy of the data; by using the convolutional autoencoder to extract the deep features of each group of feature data and classify them through density clustering, it can achieve efficient and accurate charging mode recognition when the charging data is complex and diverse, thereby improving the robustness and applicability of the analysis; among them, the convolutional autoencoder increases the convolution path and adds a fully connected layer to make the feature extraction more comprehensive and accurate.

[0079] Example 2

[0080] Based on the content of Example 1, this application experimentally verifies the importance of convolutional autoencoders in feature extraction, and compares and analyzes the distribution of the first charging data on the coordinate axis under the conditions of “using convolutional autoencoders” and “not using convolutional autoencoders”.

[0081] 1) Experimental design

[0082] Experimental group: The first charging data of a certain vehicle was processed by a convolutional autoencoder to extract deep features, and cluster analysis was performed using a density clustering method;

[0083] Control group: Density clustering analysis was performed directly using the first charge data without being processed by the convolutional autoencoder.

[0084] The principal component analysis method is used to reduce the high-dimensional features to two dimensions to show the distribution of samples on the two-dimensional coordinate axis. One point represents a charging segment, and points with the same color and shape belong to the same charging mode.

[0085] 2) Experimental results

[0086] The distribution of experimental group characteristics can be found in Figure 3 : When the convolutional autoencoder extracts features, the sample points are clearly distributed in the two-dimensional space, the characteristics of different charging modes form clearly separated clusters, there are significant boundaries between the categories, and the clustering effect is good.

[0087] For the distribution of control group characteristics, see Figure 4 : When the first charging data is directly used for clustering, the sample points are distributed chaotically, the categories of each charging mode overlap significantly, and some samples are difficult to distinguish, resulting in unstable clustering results.

[0088] 3) The convolutional autoencoder can extract more representative and stable features, improve the separation of sample points in the feature space, and thus significantly enhance the classification performance of the density clustering method. The control group without the convolutional autoencoder showed the shortcomings of more overlapping distributions and poor clustering effects, proving the necessity and effectiveness of the convolutional autoencoder for feature extraction.

[0089] Example 3

[0090] Based on the foregoing embodiments, the present application provides a specific embodiment for analyzing the charging data of new energy vehicles in a certain region in a certain year and month.

[0091] 1) Experimental data

[0092] Data source: collected from new energy vehicle charging records in the area, covering the complete charging process data of 20 vehicles.

[0093] Data content: Each record contains time, charging current, charging voltage, ambient temperature, SOC and other characteristics, covering different charging behaviors and scenarios.

[0094] 2) Experimental methods

[0095] After data preprocessing and deep features extracted by convolutional autoencoder, density clustering method was used for cluster analysis;

[0096] Different from other clustering methods, density clustering does not preset the number of clustering categories and can dynamically identify the number and structure of clusters based on data distribution.

[0097] 3) Experimental results

[0098] Sample distribution diagram: The high-dimensional deep features are reduced by principal component analysis to visualize the sample distribution. Take two cars as an example. Figure 5 and Figure 6 :

[0099] The same color and shape represent the same cluster category. The density clustering method adaptively divides the pattern according to the data distribution. The sample points form several clusters in the two-dimensional feature space, and the boundaries between categories are clear and densely distributed. Figure 5 The display divides the charging segments into two categories, Figure 6 The display can divide the charging segment into 4 different modes.

[0100] Based on all the data and the processing results of the density clustering method, a total of 5 charging modes are divided, and the characteristics of each mode are as follows:

[0101] Mode 1: low current and long time charging;

[0102] Mode 2: High current fast charging, SOC increases rapidly;

[0103] Mode 3: Supercharge mode, with high current and voltage and significant temperature rise;

[0104] Mode 4: intermittent charging, SOC fluctuations are large;

[0105] Mode 5: Abnormal charging behavior, mainly noise or irregular data points, that is, charging fragments with irregular data and no reference value.

[0106] 4) Experimental conclusion

[0107] Dynamic clustering capability: The density clustering method can automatically adjust the number of categories according to the density distribution of the data, without the need to preset the number of clusters, and flexibly adapt to the structure of different data sets;

[0108] The clustering effect is good: the sample points are clearly distributed in the feature space, and there is a clear separation between the categories, which can effectively capture the characteristics of different charging behaviors;

[0109] Outlier identification: Density clustering can identify low-density areas as noise points, which helps to discover abnormal charging behavior.

[0110] This embodiment verifies that the convolutional autoencoder combined with the density clustering method can flexibly and accurately divide the charging mode without limiting the number of categories, providing strong technical support for the charging behavior analysis of new energy vehicles.

[0111] Example 4

[0112] Based on the above embodiments, the present application provides a system for determining a charging mode based on feature extraction using convolutional autoencoder, including:

[0113] Collection module: collects multiple historical charging segments of the vehicle, wherein the historical charging segments include multiple sets of characteristic data that change with SOC, and the characteristic data include charging current, charging voltage, ambient temperature and time;

[0114] Feature extraction module: connected to the acquisition module, used to convert the formats of each group of feature data in each historical charging segment, perform data filling on each group of feature data after format conversion, and standardize each group of feature data after data filling to obtain first charging data; input each group of first charging data in each historical charging segment into a convolutional autoencoder for feature extraction, and output the extracted data corresponding to each group of first charging data;

[0115] Clustering module: connected to the feature extraction module, used to take the average of each group of first charging data to obtain second charging data; splice the second charging data into the extracted data of the corresponding first charging data to obtain third charging data; and use density clustering method to divide different charging modes in the third charging data.

[0116] For other contents, please refer to Example 1, which will not be described in detail here.

[0117] Example 5

[0118] Based on the aforementioned embodiments, the present application provides an electronic device, comprising: a processor and a memory; the processor is used to execute the steps of the method in Embodiment 1 by calling a program or instruction stored in the memory.

[0119] Example 6

[0120] Based on the aforementioned embodiments, the present application provides a computer-readable storage medium, which stores a program or instruction, and the program or instruction enables a computer to execute the steps of the method in Embodiment 1.

[0121] It should be noted that the terms used in the present invention are only for describing specific embodiments, rather than limiting the scope of the present application. As shown in the present specification, unless the context clearly indicates an exception, the words "one", "a", "a kind of" and / or "the" do not specifically refer to the singular, but may also include the plural. The terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of more restrictions, the elements defined by the sentence "include one..." do not exclude the presence of other identical elements in the process, method or device including the elements.

[0122] It should also be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", etc. should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for judging charging mode based on convolutional autoencoder feature extraction, characterized in that: The steps include: S1. Collect multiple historical charging segments of the vehicle, wherein the historical charging segments include multiple sets of characteristic data that change with SOC, and the characteristic data include charging current, charging voltage, ambient temperature and time; S2, performing format conversion on each set of characteristic data in each historical charging segment, performing data filling on each set of characteristic data after format conversion, and standardizing each set of characteristic data after data filling to obtain first charging data; S3, inputting each group of first charging data in each historical charging segment into a convolutional autoencoder for feature extraction, and outputting the extracted data corresponding to each group of first charging data respectively; S4, taking the average of each group of first charging data to obtain second charging data; splicing the second charging data to the extracted data of the corresponding first charging data to obtain third charging data; S5. Use a density clustering method to divide different charging modes in the third charging data.

2. According to claim 1, a method for judging charging mode based on convolutional autoencoder feature extraction is characterized in that: In S2, the format conversion includes: S21, setting the first value, the second value and the fixed value to be positive constants less than 1, and the second value minus the first value equals the fixed value; S22, in the process of the SOC changing from the first value to the second value, taking the average of the corresponding charging current, charging voltage and ambient temperature respectively, and calculating the total time required; S23, the first value is added to the fixed value, and S22 is repeated until the SOC reaches a maximum value, thereby obtaining four sets of sequences of charging current, charging voltage, ambient temperature and time as the SOC increases.

3. The method for judging charging mode based on convolutional autoencoder feature extraction according to claim 2 is characterized in that: In S2, each set of feature data after format conversion is filled with data respectively, including: The low-rank matrix filling method is used to unify the dimensions of the four groups of sequences of each historical charging segment.

4. The method for judging charging mode based on convolutional autoencoder feature extraction according to claim 3 is characterized in that: In S2, the step of standardizing each group of characteristic data after data filling to obtain first charging data includes: The four groups of sequences with unified dimensions in each historical charging segment are scaled to the interval [0,1] respectively.

5. The method for judging charging mode based on convolutional autoencoder feature extraction according to claim 1, characterized in that: The convolutional autoencoder comprises: Three parallel feature extraction paths and two fully connected layers that aggregate the three feature extraction paths; Among them, each feature extraction path passes through the convolution layer, batch normalization layer and activation function in sequence; in the convolution layer, the convolution kernel sizes of the three feature extraction paths are 5×5, 7×7 and 9×9 respectively; each fully connected layer is followed by an activation function.

6. A system for judging charging mode based on convolutional autoencoder feature extraction, used to execute the method for judging charging mode based on convolutional autoencoder feature extraction as described in any one of claims 1 to 5, characterized in that: Includes the following modules: Collection module: collects multiple historical charging segments of the vehicle, wherein the historical charging segments include multiple sets of characteristic data that change with SOC, and the characteristic data include charging current, charging voltage, ambient temperature and time; Feature extraction module: connected to the acquisition module, used to convert the formats of each group of feature data in each historical charging segment, perform data filling on each group of feature data after format conversion, and standardize each group of feature data after data filling to obtain first charging data; input each group of first charging data in each historical charging segment into a convolutional autoencoder for feature extraction, and output the extracted data corresponding to each group of first charging data; Clustering module: connected to the feature extraction module, used to average each group of first charging data to obtain second charging data; The second charging data are respectively spliced ​​into the extracted data of the corresponding first charging data to obtain third charging data; and different charging modes in the third charging data are divided using a density clustering method.

7. An electronic device, characterized in that: The electronic device comprises: Processor and memory; The processor is used to execute the steps of the method according to any one of claims 1 to 5 by calling the program or instruction stored in the memory.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program or instruction, and the program or instruction enables a computer to execute the steps of the method according to any one of claims 1 to 5.

Images