A charging load data analysis method and system based on actual dispatching services

Abstract

This invention proposes a charging load data analysis method and system based on actual dispatching operations, belonging to the field of charging load data analysis technology. First, cluster analysis is performed on several charging load data points within a preset time interval to obtain cluster centers corresponding to each charging group category. For each identified charging group category, appropriate data correction rules are selected; for residential users, data correction mainly refers to community relationship networks, and for taxi users, data correction mainly refers to driving trajectory information. Second, a first association graph model is generated from the corrected first complete data set to obtain a first group feature vector and a first equipment risk label. Finally, a first charging load analysis report containing fixed and dynamic fill items is generated. The technical solution of this invention can adaptively correct and analyze data according to the characteristics of different group categories, and can realize the dynamic generation of analysis reports.

Description

Technical Field

[0001] This invention belongs to the field of charging load data analysis technology, and particularly relates to a charging load data analysis method and system based on actual scheduling services. Background Technology

[0002] With the rapid growth of electric vehicle (EV) ownership, the impact of charging load on the power grid is becoming increasingly significant. Current technologies for charging load analysis largely rely on the statistical characteristics of individual users or sites, lacking in-depth analysis of the differences among user groups. This results in insufficient prediction accuracy and coarse-grained grid dispatching strategies. Furthermore, traditional methods do not fully consider the correlation between group behavior and equipment capacity, making it difficult to achieve refined load management. Summary of the Invention

[0003] The purpose of this invention is to provide a charging load data analysis method and system based on actual scheduling services, so as to solve the technical problem of insufficient accuracy in charging load analysis in the prior art.

[0004] This invention proposes a method for analyzing charging load data based on actual scheduling services, comprising the following steps:

[0005] S1: Perform a first regularization process on the first charging load data to obtain a first complete data set;

[0006] The first rule-based processing refers to correcting abnormal data in the first charging load data; the first charging load data includes multiple first charging load sub-data; each first charging load sub-data refers to the charging load data of a specific user in a specific area.

[0007] S2: Generate a first association graph model based on the first integrity data set, and determine at least one first group feature vector and at least one first device risk label based on the first association graph model;

[0008] S3: For each feature vector of the first group, select a first visualization template, and generate at least one first charging load analysis report based on at least one first device risk label and the first visualization template.

[0009] Preferably, step S1 includes the following sub-steps:

[0010] S11: Input each of the first charging load sub-data into the first charging load group classification model one by one to obtain multiple first charging load group types;

[0011] S12: Determine a first data correction rule based on the first charging load group type, and correct the corresponding first charging load sub-data according to the first data correction rule to obtain multiple second charging load sub-data.

[0012] S13: Determine the first integrity data set based on multiple second charging load sub-data.

[0013] Preferably, the first data correction rule includes:

[0014] S1211: Perform community multidimensional feature encoding on the first resident users corresponding to all the first charging load sub-data to obtain a first community graph neural network;

[0015] S1212: Calculate the second resident user with the highest similarity to each of the first resident users based on the first community graph neural network;

[0016] S1213: Determine the first data anomaly point in the first charging load sub-data, and determine the first data anomaly time point based on the first data anomaly point;

[0017] S1214: Determine the first replacement data value based on the first abnormal data time point, and use the first replacement data value to correct the first charging load sub-data to determine the second charging load sub-data;

[0018] The first replacement data value is obtained from the first charging load sub-data corresponding to the second residential user.

[0019] Preferably, the first data correction rule includes:

[0020] S1221: Determine the first trajectory information based on the first charging load sub-data, and obtain multiple first charging stations based on the first trajectory information;

[0021] S1222: Identify the first data anomaly in the first charging load sub-data, and repair the first data anomaly based on the historical charging load data of multiple first charging stations to determine the second charging load sub-data.

[0022] Preferably, in step S2, generating the first association graph model based on the first integrity data set includes the following steps:

[0023] S21: Divide the first complete data set into at least one first complete data subset, and determine at least one first group feature extraction rule for at least one first complete data subset;

[0024] S22: For at least one of the first complete data subsets, generate the first association graph model using the first group feature extraction rules.

[0025] Preferably, step S3 includes the following sub-steps:

[0026] S31: Based on each feature vector of the first group, determine the corresponding first group label, and select the first visualization template according to the first group label;

[0027] S32: Determine at least one first fill item based on the first visualization template, the first device risk label, and the first group feature vector;

[0028] S33: Generate at least one first charging load analysis report based on at least one of the first fill-in items and the first visualization template.

[0029] Preferably, the first community graph neural network includes multiple nodes and multiple edges connecting different nodes, wherein each node represents a resident user, and each edge represents the relationship between different nodes. Preferably, the first fill item includes fixed fill items and dynamic fill items.

[0030] Preferably, in step S32, the dynamic filling item is obtained after analyzing the first device risk label and the first group feature vector.

[0031] The present invention also proposes a charging load data analysis system based on actual scheduling services, which is used to implement the above-mentioned charging load data analysis method based on actual scheduling services.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0033] First, cluster analysis is performed on several charging load data points within a preset time interval to obtain cluster centers corresponding to each charging group category. For each identified charging group category, appropriate data correction rules are selected: for residential users, data correction is mainly based on community relationship networks; for taxi users, data correction is mainly based on driving trajectory information. Second, a first association graph model is generated from the corrected first complete data set to obtain a first group feature vector and a first equipment risk label. Finally, a first charging load analysis report containing fixed and dynamic fill items is generated. The technical solution of this invention can adaptively correct and analyze data according to the characteristics of different group categories, enabling dynamic generation of analysis reports. Attached Figure Description

[0034] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0035] Figure 1 This is an execution flowchart of a charging load data analysis method based on actual scheduling services in this invention;

[0036] Figure 2 This is a schematic diagram of the first association graph model. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Example

[0039] This embodiment proposes a charging load data analysis method based on actual scheduling services. The specific process is as follows: Figure 1 As shown, it includes the following steps:

[0040] S1: Perform a first regularization process on the first charging load data to obtain a first complete data set.

[0041] Different types of charging users typically have specific behavioral characteristics, and the data that can be used to characterize the charging load usually includes: historically stored station-line-transformer-customer topology, power grid operation data, metering data, marketing data, battery BMS data, GPS trajectory data, etc.

[0042] The combinations of highly sensitive data differ depending on the type of charging user. Common charging user types typically include taxi drivers, residents, and other commercial facilities. Taking taxi drivers as an example, the charging load data that is highly sensitive to them includes not only common grid operation data and metering data, but also GPS trajectory data. This is because the average daily mileage of taxi drivers is significantly influenced by GPS trajectory and weather data. For residential charging users, their driving trajectories are usually more fixed, so the data that is most sensitive to them is concentrated in battery management system (BMS) data and charging time data.

[0043] To facilitate the effective extraction of characteristic data for different types of charging users in subsequent steps, the first charging load data needs to undergo a first regularization process in this step. This first charging load data includes multiple first charging load sub-data sets.

[0044] Each first charging load sub-data refers to the charging load data of a specific user within a specific area, which may correspond to different user types. The first rule-making process refers to the data correction process for data anomalies and missing data that occur in the first charging load data.

[0045] Step S1 includes the following sub-steps:

[0046] S11: Input each first charging load sub-data into the first charging load group classification model one by one to obtain multiple first charging load group types.

[0047] Each first charging load sub-data includes at least driving trajectory data and metering data. The driving trajectory data refers to a user's driving trajectory within a preset time interval, and the metering data refers to a user's charging habits within the preset time interval.

[0048] In the first charging load group classification model, clustering is performed based on historical big data to obtain cluster centers corresponding to various user groups.

[0049] Specifically, multiple first historical charging load data points located within the same area as the first charging load sub-data point can be selected as sample data. Each first historical charging load data point includes a historical user's driving trajectory data and charging habit data within a preset time interval.

[0050] The driving trajectory data can include multiple first driving sub-trajectories traveled by a historical user within a preset time interval. The repetition rate of these multiple first driving sub-trajectories is calculated and used as the driving trajectory data. This is because the main charging load groups include ordinary residential users and taxi users. For ordinary residential users, their daily travel trajectories have a high probability of overlap, while for taxi users, their daily travel trajectories have a certain degree of randomness, thus the probability of high overlap is low. In summary, driving trajectory data used to characterize trajectory overlap rate can be used as one of the factors for generating cluster centers.

[0051] The charging habit data includes historical user preference charging time data within a preset time range. This is because residential users typically choose to slow charge after get off work, i.e., at night, which is both convenient for commuting and better for the battery. However, charging only at night usually cannot meet the power needs of taxi users, and taxi users also have higher requirements for charging efficiency. Therefore, taxi users usually choose fast charging mode, and the charging time is more random.

[0052] Therefore, for each first historical charging load data point, corresponding driving trajectory data and charging habit data can be analyzed and used as the two main dimensions to construct the first historical charging load vector. After clustering multiple first historical charging load vectors, at least one first user group cluster center can be obtained, where each first user group cluster center corresponds to a specific user group and characterizes the charging characteristics of that user group.

[0053] Preferably, the dimension of the first historical charging load vector can be set according to specific needs, and may also include data on the degree of influence of weather and battery BMS data.

[0054] Preferably, each first user group cluster center can correspond to either resident users or taxi users. When there is a need for more refined classification, multiple first user group cluster centers can be further subdivided within resident users. This two-level classification method allows for better targeting and accuracy in subsequent data analysis results and adjustment strategies.

[0055] For example, the first user group cluster center may include the following vector label: [route repetition rate 85%, habitual charging time interval 23:00-7:00]. When the similarity between the first charging load sub-data and the vector label of the first user group cluster center exceeds a preset value, it can be determined that the user group corresponding to the first charging load sub-data is a residential user, and the corresponding confidence level can be determined based on the calculated similarity value.

[0056] S12: Determine the first data correction rule based on the first charging load group type, and correct the corresponding first charging load sub-data according to the first data correction rule to obtain multiple second charging load sub-data.

[0057] As mentioned in step S11, different user groups have specific user characteristics. For example, the resident group has a high trajectory repetition rate and a relatively fixed charging time, while the taxi group has a low trajectory repetition rate and a relatively random charging time.

[0058] Different data correction rules are needed to correct the corresponding first charging load sub-data for different user groups with different user characteristics.

[0059] The situations requiring correction mainly include two categories: missing data values ​​and abnormal data values. Missing data values ​​are mainly identified by checking whether the charging load data values ​​and the time series satisfy a one-to-one correspondence; if not, there is missing data. Abnormal data values ​​are mainly identified by calculating whether the data value fluctuations exceed preset values; if they exceed preset values, it is considered an abnormal data value situation.

[0060] The following section introduces the corresponding first data correction rules for the two main user groups (resident users and taxi users):

[0061] 1. The data correction rules for residential users are as follows:

[0062] S1211: Perform community multidimensional feature encoding on the first resident users corresponding to all first charging load sub-data to obtain the first community graph neural network.

[0063] The first community's graph neural network includes multiple nodes and multiple edges connecting different nodes. Each node represents a resident user, and each edge connecting different nodes represents the relationship between those different nodes.

[0064] 1) Node feature construction:

[0065] A feature vector is generated for each resident user. An example feature vector is as follows:

[0066] f i =[Family size; Household age; Number of air conditioners; Electric vehicle power; Historical average load]∈R 5

[0067] The above are just examples of feature vector dimensions. These dimensions can be adjusted as needed to characterize the electricity consumption habits of residential users, thereby facilitating the subsequent correction of abnormal charging load values.

[0068] 2) Edge weight definition:

[0069] If Resident User 1 and Resident User 2 belong to the same community physical unit (such as a building), the edge weight can be assigned a value of 1.0; if they belong to adjacent community units, the weight can be assigned a value of 0.5. The above assignment rules can also be adjusted according to specific needs.

[0070] S1212: Calculate the second resident user with the highest similarity to each first resident user based on the graph neural network of the first community.

[0071] The calculation process mainly considers two factors: first, the similarity of the feature vectors of each first resident user with multiple neighboring first resident users; and second, the edge weights of each first resident user with multiple neighboring first resident users. Combining these two factors determines the similarity of each first resident user with other neighboring resident users, and the resident user with the highest similarity is selected as the second resident user corresponding to the first resident user.

[0072] S1213: Identify the first data anomaly point in the first charging load sub-data, and determine the first data anomaly time point based on the first data anomaly point.

[0073] In this step, it is necessary to determine the time point at which the data anomaly occurred for each data anomaly point in the first charging load sub-data.

[0074] S1214: Determine the first replacement data value based on the first abnormal data time point, and use the first replacement data value to correct the first charging load sub-data to determine the second charging load sub-data.

[0075] In this step, it is necessary to find the corresponding first replacement data value in the charging load sub-data of the second residential user based on the first data anomaly time point, so as to correct the first charging load sub-data and determine the second charging load sub-data.

[0076] If the first residential user and the second residential user do not have charging load data at the same time on the same day, the charging load data at the same time on different natural days can be selected as the first replacement data value.

[0077] By following the above steps, a second resident with similar electricity consumption habits to the first resident can be found through the community network, and the charging load data of the second resident can be used to correct the charging load data of the first resident.

[0078] 2. The data recovery rules for taxi users are as follows:

[0079] For taxi users, their charging load data is primarily determined by the charging characteristics of their frequently used charging stations. Therefore, their frequently used charging stations can be identified based on their driving trajectory information, and then the data can be corrected by using charging load data from other charging stations at the same stations.

[0080] S1221: Determine the first trajectory information based on the first charging load sub-data, and obtain multiple first charging stations based on the first trajectory information.

[0081] The first trajectory information refers to the trajectory information that a taxi user will frequently travel through in the first charging load sub-data corresponding to the taxi user.

[0082] By comparing the first trajectory information with the map information, the charging stations along the first trajectory information can be identified, and these charging stations can be identified as the first charging stations.

[0083] S1222: Identify the first data anomaly in the first charging load sub-data, and repair the first data anomaly based on the historical charging load data of multiple first charging stations to determine the second charging load sub-data.

[0084] In this step, based on the first data anomaly points presented in the first charging load sub-data, the first charging load sub-data is repaired using historical charging load data from multiple first charging stations at the same absolute or relative time to obtain the second charging load sub-data.

[0085] S13: Determine the first integrity data set based on multiple second charging load sub-data.

[0086] The first complete data set contains multiple second charging load sub-data items. Each second charging load sub-data item includes a charging user's various charging data, a first group label, and a first confidence level.

[0087] The first group label includes at least resident users and taxi users, and the first confidence level refers to the probability of correct classification.

[0088] S2: Generate a first association graph model based on the first integrity data set, and determine at least one first group feature vector and at least one first device risk label based on the first association graph model.

[0089] In step S1, different data correction strategies were adopted to correct each first charging load sub-data according to the characteristics of different charging load groups, laying a good foundation for the subsequent data processing.

[0090] In this step, the first complete dataset needs to be fully analyzed and mined to obtain feature vectors and device risk labels for different user groups, thereby providing data support for subsequent adaptive data display.

[0091] Step S2 includes the following sub-steps:

[0092] S21: Divide the first integrity data set into at least one first integrity data subset, and determine at least one first group feature extraction rule for the at least one first integrity data subset.

[0093] The first complete data set contains first charging load sub-data corresponding to different types of user groups, such as residential users and taxi users.

[0094] In this step, the first integrity data set can be divided into at least one first integrity data subset based on the user type label contained in each first charging load sub-data. Each first integrity data subset corresponds to a user group category.

[0095] Since different user groups correspond to different charging load characteristics, it is necessary to select corresponding group feature extraction rules based on the characteristics of each user group. The following explains the group feature extraction rules for residential users and taxi users:

[0096] 1. Rules for extracting group characteristics of residential users:

[0097] The LSTM model is used to superimpose periodic decomposition, and the valley filling ratio R is output. valley and regularity index S reg .

[0098] Among them, Gu Chong accounted for R valley Specifically, this involves analyzing residential users' charging data within a preset time interval to obtain the ratio of charging time during off-peak hours to total charging time; the regularity index S reg It represents the probability that a residential user will perform charging behavior according to a specified pattern in each preset time period.

[0099] For example, a residential user's off-peak charging rate is 85%, and the regularity index is 0.92 (indicating that the charging behavior is highly regular).

[0100] 2. Rules for extracting group characteristics of taxi users:

[0101] Since taxi users typically choose fast charging mode, pulse intensity and fast charging frequency are important analytical data.

[0102] For taxi users, charging load data for a preset time interval is extracted from the first charging load sub-data, and then the pulse intensity I is extracted. p and fast charging frequency F c The calculation formula is as follows:

[0103]

[0104] Where V(t) is the pulse voltage at a specified time point, and t1 and t2 are the start and end time points, respectively.

[0105] S22: For at least one first complete data subset, generate a first association graph model using the first group feature extraction rules.

[0106] In this step, the correlation between user groups and power supply equipment is quantified to dynamically assess equipment overload risk, providing data support for subsequent strategy generation.

[0107] like Figure 2 As shown, the first association graph model includes multiple nodes and edges. The definitions of nodes and edges are as follows:

[0108] 1. Node:

[0109] User group node g: generated from the group classification results of step S1 (e.g., taxis, residents, logistics fleets, commercial facilities).

[0110] Power supply equipment node d: includes power grid equipment such as transformers and distribution lines, and its attributes include capacity C. d (e.g., transformer rated capacity 600kW).

[0111] 2. Edge weight R g→d :

[0112] The overload risk of user group g to device d is represented by the following formula:

[0113]

[0114] in:

[0115] I g The group influence factor (calculated by weighting the features extracted in step S21, such as taxi I) g =1.2, Resident I g =0.8);

[0116] P g,max The maximum charging power for user group g (from the feature extraction results of step S21, such as taxi group P) g,max =500kW);

[0117] C d The capacity of device d (obtained from power grid topology data, such as transformer capacity C) d =600kW).

[0118] Specifically, for each subset of data in the first complete data set, the corresponding first group feature extraction rule is used to extract group features, and a first association graph model is generated based on the extracted group features.

[0119] S23: Generate at least one first group feature vector and at least one first device risk label based on the first association graph model.

[0120] The risk label for each device can be determined through S22, and this risk label is obtained by checking R. g→dThe determination is made through numerical judgment. An example judgment rule is as follows:

[0121] R g→d >1: High risk (insufficient equipment capacity);

[0122] 0.8 < R g→d ≤1: Medium risk (close to capacity limit);

[0123] R g→d ≤0.8: Low risk (sufficient capacity).

[0124] The first group feature vector mainly includes the charging load feature data of the specified group extracted according to the first group feature extraction rules.

[0125] S3: For each first group feature vector, select the first visualization template and generate at least one first charging load analysis report.

[0126] Different charging load groups typically require different types of data to be displayed. Therefore, in this step, for at least one first group feature vector and at least one first device risk label output in step S2, it is necessary to select a first visualization template corresponding to each first group feature vector, and generate at least one first charging load analysis report for the displayed content, so that subsequent steps can generate adjustment strategies based on the content of the first charging load analysis report.

[0127] S31: Based on each first group feature vector, determine the corresponding first group label, and select the first visualization template according to the first group label.

[0128] This step involves dynamic matching of visual templates.

[0129] 1. Template Library Construction: Multiple visual templates are pre-designed, each corresponding to a specific group tag. For example:

[0130] Taxi template: Includes real-time heat map (displaying charging hotspot areas), dynamic line graph (showing peak and off-peak load changes), and GPS trajectory heat map;

[0131] Resident template: Includes bar chart (comparing charging demand at different time periods) and pie chart (the percentage of charging types).

[0132] In each type of template, each type of display diagram includes at least one filler item. The first charging load analysis report can be obtained by filling at least one filler item with data.

[0133] 2. Template matching logic:

[0134] Template matching algorithms (such as tag-based keyword matching or semantic similarity calculation) are used to associate group tags with templates in the template library.

[0135] For example, if the tag is "taxi", the system will automatically call a template containing a GPS trajectory heatmap.

[0136] S32: Determine at least one first filler item based on the first visualization template, the first device risk label, and the first group feature vector.

[0137] The first visualization template includes some fixed fill items; however, the first group feature vector may contain dynamic fill items that pose a significant risk. Therefore, a combination of fixed and dynamic fill items can be used as the first fill item.

[0138] The dynamic adaptation rules for the first fill item are as follows:

[0139] The content to be filled is dynamically selected based on the range of feature values. For example, a dynamically filled item could be:

[0140] If the "overload risk" is greater than 1, add a red warning icon and the text "high risk" to the template;

[0141] If the "valley charging ratio" is less than 30%, a strategy suggestion of "increasing the nighttime charging ratio" will be generated.

[0142] It can also integrate multi-dimensional data:

[0143] By combining power grid topology data (such as equipment capacity) and user behavior data, composite fill-in items are generated. For example:

[0144] Transformer capacity information is overlaid on the heat map, and "Equipment overload risk: high / medium / low" is marked.

[0145] Preferably, the dynamically filled items can be determined according to a pre-set mapping relationship.

[0146] S33: Generate at least one first charging load analysis report based on at least one first fill item and the first visualization template.

[0147] In this step, for each group type, multiple fixed fill items are selected from the first fill items and filled into the first visualization template, and multiple dynamic fill items are selected from the first fill items and added to the first visualization template.

[0148] The present invention also proposes a charging load data analysis system based on actual scheduling services, for executing the above-mentioned charging load data analysis method based on actual scheduling services.

[0149] This invention performs cluster analysis on several charging load data points within a preset time interval to obtain cluster centers corresponding to each charging group category. For each identified charging group category, appropriate data correction rules are selected: for residential users, data correction is primarily based on community relationship networks, and for taxi users, it is primarily based on driving trajectory information. Next, a first association graph model is generated from the corrected first complete data set to obtain a first group feature vector and a first equipment risk label. Finally, a first charging load analysis report containing fixed and dynamic fill items is generated. This invention as a whole can adaptively correct and analyze data according to the characteristics of different group categories, enabling dynamic generation of analysis reports.

[0150] The above description is only a preferred embodiment of the present invention. Therefore, all equivalent changes or modifications made to the structure, features and principles described in the scope of the present invention are included in the scope of the present invention.

Claims

1. A method for analyzing charging load data based on actual dispatching services, characterized in that, Includes the following steps: S1: Perform a first regularization process on the first charging load data to obtain a first complete data set; The first rule-based processing refers to correcting abnormal data appearing in the first charging load data; the first charging load data includes multiple first charging load sub-data; each first charging load sub-data refers to the charging load data of a specific user in a specific area; S2: Generate a first association graph model based on the first integrity data set, and determine at least one first group feature vector and at least one first device risk label based on the first association graph model; S3: For each feature vector of the first group, select a first visualization template, and generate at least one first charging load analysis report based on at least one first device risk label and the first visualization template; Step S1 includes the following sub-steps: S11: Input each of the first charging load sub-data into the first charging load group classification model one by one to obtain multiple first charging load group types; S12: Determine a first data correction rule based on the first charging load group type, and correct the corresponding first charging load sub-data according to the first data correction rule to obtain multiple second charging load sub-data. S13: Determine the first integrity data set based on multiple second charging load sub-data; For residential users, the first data correction rule includes: S1211: Perform community multidimensional feature encoding on the first resident users corresponding to all the first charging load sub-data to obtain a first community graph neural network; S1212: Calculate the second resident user with the highest similarity to each of the first resident users based on the first community graph neural network; S1213: Determine the first data anomaly point in the first charging load sub-data, and determine the first data anomaly time point based on the first data anomaly point; S1214: Determine the first replacement data value based on the first abnormal data time point, and use the first replacement data value to correct the first charging load sub-data to determine the second charging load sub-data; The first replacement data value is obtained from the first charging load sub-data corresponding to the second residential user; Specifically, for taxi users, the first data correction rule includes: S1221: Determine the first trajectory information based on the first charging load sub-data, and obtain multiple first charging stations based on the first trajectory information; S1222: Identify the first data anomaly in the first charging load sub-data, and repair the first data anomaly based on the historical charging load data of multiple first charging stations to determine the second charging load sub-data.

2. The charging load data analysis method based on actual scheduling services according to claim 1, characterized in that, In step S2, generating the first association graph model based on the first complete data set includes the following steps: S21: Divide the first complete data set into at least one first complete data subset, and determine at least one first group feature extraction rule for at least one first complete data subset; S22: For at least one of the first complete data subsets, generate the first association graph model using the first group feature extraction rules.

3. The charging load data analysis method based on actual scheduling services according to claim 2, characterized in that, Step S3 includes the following sub-steps: S31: Based on each feature vector of the first group, determine the corresponding first group label, and select the first visualization template according to the first group label; S32: Determine at least one first fill item based on the first visualization template, the first device risk label, and the first group feature vector; S33: Generate at least one first charging load analysis report based on at least one of the first fill-in items and the first visualization template.

4. The charging load data analysis method based on actual scheduling services according to claim 3, characterized in that, The first community graph neural network includes multiple nodes and multiple edges connecting different nodes, where each node represents a resident user and each edge represents the relationship between different nodes.

5. The charging load data analysis method based on actual scheduling services according to claim 4, characterized in that, The first fill item includes fixed fill items and dynamic fill items.

6. The charging load data analysis method based on actual scheduling services according to claim 5, characterized in that, In step S32, the dynamic filling item is obtained after analyzing the first device risk label and the first group feature vector.

7. A charging load data analysis system based on actual dispatching services, characterized in that, This method is used to implement the charging load data analysis method based on actual scheduling services as described in any one of claims 1-6.

Images