Trajectory prediction-based migration method and system for urban mobile charging pile

Abstract

The application discloses a kind of trajectory prediction-based urban mobile charging pile migration method.The method obtains the position information of user terminal dispersion, constructs trajectory embedding sequence after vectorization processing, obtains the trip trajectory table based on trajectory embedding sequence.The historical trajectory T of user in trip trajectory table is segmented into n segment trajectories, trajectory grouping is carried out with segment trajectory similarity, the address preference probability of user in different group is determined, the complete trajectory of user is predicted according to the address preference probability and trajectory confidence of user charging, the site expectation value of any user route subinterval is determined according to complete trajectory, and the highest expected place in trip trajectory table is selected as the recommended urban mobile charging pile site location.The application also discloses a kind of migration system for realizing trajectory prediction-based urban mobile charging pile migration method.

Description

Technical Field

[0001] This invention relates to data processing technology, and more particularly to a method and system for migrating urban mobile charging stations based on trajectory prediction. Background Technology

[0002] In existing technologies, the relocation and site selection methods for service facilities such as mobile charging piles mainly employ kernel density to determine demand points, and then select the coverage points with the highest density after screening. For example, the literature "A Site Selection Method for Taxi Service Stations Based on Service Demand" (Ye Zhen et al., China Academy of Transportation Sciences, 2019) proposes a method for determining the site selection of taxi service stations using kernel density estimation. Another example is CN202010351142.8, which discloses a charging pile site selection and deployment method and system, using geographic identification technology to determine the distribution of the surrounding environment, thereby determining the most efficient site selection. The above-mentioned techniques require a large sample size of location data as support during trajectory prediction, and the prediction granularity is limited to "location points" rather than "location trajectories." Therefore, existing technologies aim to identify vehicle trajectories as accurately as possible based on fragmented trajectories, relying on limited location information and focusing on the user's mobile terminal. Furthermore, existing technologies aim to provide a site selection method for urban mobile charging piles by using clustering and grouping methods to perform efficient trajectory prediction based on vehicle movement and preferences. Summary of the Invention

[0003] To address the aforementioned issues, this invention provides a method and system for migrating urban mobile charging stations based on trajectory prediction. By accessing limited location information from user vehicle terminals, the method uses the similarity of segment trajectories within the user vehicle as a grouping criterion, and combines user personal information to provide a confidence level condition for trajectory prediction. The method predicts the vehicle's travel trajectory through trajectory grouping and address preference probability, and uses overlapping locations in the travel trajectories as recommended sites for mobile charging stations.

[0004] The objective of this invention can be achieved through the following technical solutions:

[0005] A migration method for urban mobile charging stations based on trajectory prediction includes the following steps:

[0006] Step 1: Determine the addressing range of the urban mobile charging station, retrieve the geospatial data of the addressing range, and divide the addressing range into multiple sub-ranges based on the geospatial data;

[0007] Step 2: Retrieve location and time information from multiple user terminal devices, extract spatial and temporal feature vectors, and generate a trajectory embedding sequence;

[0008] Step 3: Map the trajectory embedding sequence to geospatial data to generate at least one user travel trajectory table. The travel trajectory table is stored in the user database and contains the user's time-related historical trajectories T.

[0009] Step 4: Divide the sub-interval into a first sub-interval with an access frequency greater than the baseline frequency and a second sub-interval with an access frequency less than or equal to the baseline frequency. The trajectory prediction unit divides the user's historical trajectory T into n segment trajectories, T={T i |i≤n}, the trajectory T of each segment i It intersects only with the unique first subinterval;

[0010] Step 5: Determine the trajectory function f(T) of any user based on multiple fragment trajectories of any user terminal device. i The similarity of trajectory functions of at least two users is compared by using a loss function. The trajectories of users with high similarity are grouped separately, and the address preference probability of users in each group is recorded.

[0011] Step 6: Call the user's attribute parameters and set the trajectory confidence K for any segment of the historical trajectory T based on the user's attribute parameters. Ti Predict the user's complete trajectory;

[0012] Step 7: Count the number of complete user trajectories within multiple sub-intervals and determine the expected location value for each sub-interval;

[0013] Step 8: Select the sub-interval with the largest expected location value as the target address of a city's mobile charging station. If the city's mobile charging stations have been allocated, proceed to step 9; otherwise, proceed to step 8.

[0014] Step 9: Based on the Euclidean distance between the target address and the selected sub-interval, adjust the expected location value of any unselected sub-interval, and return to Step 7;

[0015] Step 10: Migrate all city mobile charging stations to the corresponding sub-area based on the target address.

[0016] In this invention, in step 8, the unselected urban mobile charging pile with the smallest distance from the sub-interval is selected.

[0017] In this invention, the addressing interval is represented by transforming the discrete vector of the relocation location information of urban mobile charging piles into a low-dimensional vector. The addressing interval contains geospatial data of the relocation location of urban mobile charging piles.

[0018] In this invention, the trajectory embedding sequence is composed of spatial feature vectors and temporal feature vectors. The spatial feature vectors identify the user's geographical coordinates, which include three variables: longitude, latitude, and the administrative planning area to which the user belongs. The temporal feature vectors are the time values ​​of the user's check-in at each verification code location. The corresponding time values ​​and location information together constitute a time matrix.

[0019] In this invention, the time matrix is ​​obtained based on the reparameterization method of spatial feature vectors and time feature vectors. The location information includes hidden layer vectors in the time matrix. The hidden layer vectors provide location topology relationships for the extraction of spatial feature vectors and time feature vectors of other surrounding locations.

[0020] In this invention, the travel trajectory table is a collection of location information and time from the user terminal, and the trajectory cloud terminal is a cloud storage terminal device consisting of a shared data server and a wireless network communication protocol.

[0021] In this invention, the user's historical trajectory T is segmented into fragment trajectories based on the complexity of the moving shape, the frequency of user access, and the time threshold.

[0022] In this invention, the trajectory confidence level K is determined based on the frequency of occurrence of attribute parameters in the user's travel trajectory table. Tn .

[0023] In this invention, the attribute parameters include at least the user's home address and work address, and a trajectory confidence K is assigned to the user's corresponding trajectory based on the user's attribute parameters. Tn .

[0024] A migration system for urban mobile charging piles based on trajectory prediction is disclosed. This migration system is used in the aforementioned migration method for urban mobile charging piles based on trajectory prediction. The system includes a user database, a trajectory acquisition unit, a data processing unit, a data analysis unit, a trajectory prediction unit, and a command sending unit.

[0025] The user database stores attribute parameters for multiple users;

[0026] The trajectory acquisition unit retrieves the geospatial data of the addressing interval and generates n segment trajectories;

[0027] The data processing unit determines the trajectory function of any user based on multiple fragment trajectories of any user terminal device;

[0028] The data analysis unit counts the number of complete user trajectories within multiple sub-intervals to determine the expected location value for each sub-interval.

[0029] The trajectory prediction unit sets the trajectory confidence level based on the user's attribute parameters and predicts the user's complete trajectory.

[0030] The instruction sending unit determines the expected location value for each sub-interval and determines the target address of the urban mobile charging pile based on the expected location value.

[0031] The migration method and system for urban mobile charging piles based on trajectory prediction of the present invention have the following beneficial effects: by acquiring user location information, it conforms to actual application and has high data availability; by segmenting user historical trajectories into fragment trajectories and identifying them using fragment trajectories, the complex calculation process is reduced and the cost can be effectively controlled; by grouping trajectories according to user address preference probabilities and introducing user personal information as a confidence standard, the accuracy of trajectory prediction under limited sample capacity can be improved; the site selection of urban mobile charging piles after migration based on this method is more in line with the travel habits of vehicles in the area, thereby improving the radiation rate and coverage of urban mobile charging piles after migration. Attached Figure Description

[0032] Figure 1 This is a flowchart of the migration method for urban mobile charging piles based on trajectory prediction according to the present invention;

[0033] Figure 2 This is a schematic diagram of the dynamic topology constructed by trajectory embedding sequence in this invention;

[0034] Figure 3 This is a schematic diagram illustrating the prediction results of different preference points and their segment trajectories in this invention.

[0035] Figure 4 This is a block diagram of the migration system for urban mobile charging piles based on trajectory prediction according to the present invention. Detailed Implementation

[0036] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0037] In existing technologies, the site selection for urban mobile charging stations is generally based on the density of the city's permanent residents and the density of people gathering in public places. This site selection method has a significant subjective bias, and the radiation density and coverage of the completed urban mobile charging stations cannot be guaranteed. This invention obtains the travel trajectories of residents in a region through trajectory prediction, transforming the scattered and concentrated location data information obtained through user terminals into trajectory route maps, and completing the migration of mobile charging stations based on the trajectory distribution. This invention uses an iterative algorithm based on historical trajectories to select different sub-intervals as migration targets, enabling the periodic migration of mobile charging stations in different areas. Compared to existing fixed-site charging stations, this invention is better able to adapt to situations where user density changes.

[0038] Example 1

[0039] This embodiment details a method and system for relocating urban mobile charging stations based on trajectory prediction, used to recommend optimal relocation sites for existing urban mobile charging stations. (Refer to...) Figure 1 The method includes the following steps:

[0040] Step 1: Determine the addressing range of the urban mobile charging station. The trajectory acquisition unit retrieves the geospatial data of the addressing range and divides the addressing range into multiple sub-ranges based on the geospatial data.

[0041] The addressing interval is the range defined by the selected city's mobile charging stations. The addressing interval information is pre-set and input into the trajectory prediction unit for processing using the boundary latitude and longitude. The trajectory acquisition unit retrieves the geospatial data of the determined addressing interval through the service terminal platform and divides the addressing interval into multiple sub-intervals in the form of a grid. The grid density determines the accuracy of the final migration target location; a higher grid density results in higher accuracy. In this embodiment, the number of grids is 0.1% of the area of ​​the addressing interval in the geographic space.

[0042] Step 2: Obtain the GPRS location information of the user terminal, access the data terminal platform through the API interface, call the address and time information of the user terminal location, extract the spatial feature vector and time feature vector of the user location, and generate a trajectory embedding sequence.

[0043] Step 3: Map the trajectory embedding sequence to geospatial data to generate at least one user travel trajectory table. This travel trajectory table is stored in the user database and contains the user's time-related historical trajectories T. Since location information and the corresponding time points can be obtained through the data terminal platform, but the time and location for trajectory prediction are unknown, it is necessary to extract the temporal and spatial features of the user's location before predicting the user's trajectory. Figure 2 The extraction of the temporal and spatial feature vectors specifically includes the following three steps:

[0044] Step 31: Call the city's public area data, extract the access time and geographical location of the user's location, construct a time vector using a variational autoencoder, and construct the corresponding spatiotemporal features by connecting the temporal feature domain and spatial feature of the user and location.

[0045] Step 32: Map the user's time characteristics to the city's geospatial space, such as the check-in verification code location and the user's dynamic topology. Figure 2 As shown, if the user has location information recorded, then the two nodes are connected.

[0046] Step 33: Based on the above steps, the system platform generates at least one user travel trajectory table. The travel trajectory table includes the spatial characteristics of the check-in and verification location within the city, the dynamic topological structure of the check-in and verification location and the user, and the user's historical trajectory T within a specific time period. The travel trajectory table is stored in the user database.

[0047] In this embodiment, the historical trajectory generated by a user within a specific time period is T={T1,T2,…,T…} n}, define Q ij This indicates that user A was at a specific timestamp t. j The obtained location information, j∈[1,n], Q ij It includes the longitude, latitude, and administrative division of the check-in verification location. After a user successfully checks in on the system, a link is recorded.

[0048] Specifically, for multiple historical trajectories generated by unknown users, where there is no location information data on the user's end, or the location information data recorded on the user's end is too insufficient, resulting in multiple unlinked trajectories, a set T of unlinked trajectories is constructed. i ={T1,T2,…,T m}, establish a user set U={u1,u2,…,u n}, (m>n), establish a mapping relationship T between the set of unlinked trajectories and the set of users. i →U.

[0049] Step 4: Divide the sub-interval into a first sub-interval with an access frequency greater than the baseline frequency and a second sub-interval with an access frequency less than or equal to the baseline frequency. The trajectory prediction unit divides the user's historical trajectory T into n segment trajectories, T={T i |i≤n}, the trajectory T of each segment i It intersects only with the unique first subinterval.

[0050] The user's historical trajectory T is segmented into fragmented trajectories based on three methods: a mechanism for classifying movement shape complexity, a mechanism for classifying access frequency to check-in verification points, and a mechanism for classifying based on time thresholds. Different segmentation methods for historical trajectory T result in different numbers and attributes of fragmented trajectories. In this embodiment, the preferred segmentation mechanism uses a time threshold as the criterion for historical trajectory segmentation, with the time threshold set to 5 hours.

[0051] In this embodiment, the top 20% of locations by frequency within the urban area are designated as the second sub-interval, and locations with a frequency greater than the top 20% are designated as the first sub-interval. The second sub-interval constitutes one of the criteria for user trajectory grouping, while the first sub-interval has no reference value for trajectory grouping. (Refer to...) Figure 3In a sub-interval, the frequency accessed 10% or more below the base frequency is defined as a noise interval. If several noise intervals exist within multiple adjacent grid regions, the set of several noise points is considered as the second sub-interval. During trajectory prediction, the segment trajectory contained in the second sub-interval contains multiple sets of segment trajectories.

[0052] Step 5: The trajectory prediction unit determines the trajectory function f(T) of any user based on multiple segment trajectories of any user terminal device. i The method compares the similarity of trajectory functions of at least two users using a loss function. Trajectories of users with high similarity are grouped separately, and the address preference probability of users in each group is recorded. Within the time interval [a,b], the loss function is Loss(f(x1),f(x2),…,f(x...). n ))= a b [f(x1)-f(x2)-…-f(xn)] 2 dx;

[0053] Divide the historical trajectory T into multiple segment trajectories T n In a trajectory embedding sequence, fragmented trajectories constitute a time-dependent multivariable function. The similarity of user trajectory functions is determined by comparing the fragments. Since the trajectories are unordered, the trajectory function f(T) obtained by placing fragmented trajectories into the trajectory embedding sequence... n The loss function α can also be discrete. In this preferred embodiment, a loss function is used to compare the loss values ​​α of multiple trajectory functions, defined within the interval [a, b]. The trajectory function loss value α = Loss(f(x1), f(x2), ..., f(x...). n The degree of deviation of the function loss value α from 0 is used as the standard for trajectory function similarity. The greater the deviation of the loss value α from 0, the lower the similarity of the compared trajectory functions, and vice versa.

[0054] Step 6: The trajectory prediction unit calls the user's attribute parameters and sets the trajectory confidence level K for any segment of the historical trajectory T based on the user's attribute parameters. Ti According to the address preference probability and trajectory confidence K of users in each group Ti Predict the user's complete trajectory.

[0055] In this embodiment, the user terminal accesses the data terminal platform via an API interface to retrieve the user's personal information, including home address and work address. Attribute parameters of the user's personal information are calculated using a Long Short-Term Memory (LSTM) network and a Recurrent Gate Unit (GRU). The frequency of occurrence of these attribute parameters is compared with the frequency in the user's travel trajectory table to determine the trajectory confidence level K. TnThe coefficient value, which represents whether the user's location information in the second sub-interval is the user's residential or work address, is converted into a weighted coefficient in a binomial distribution. This weighted coefficient is then used as the trajectory confidence K. Tn The coefficient values ​​are determined. In the LSTM model, the activation sigmoid function is used to label the density attribute points of the trajectory fragments at the check-in location. In the tanh layer of the neural network, the output gate can define the state information contained in the user's check-in verification location, which includes the spatiotemporal feature parameters after rasterization. Based on this, a link between the trajectory and the user is constructed, saving the user's historical trajectory and the sub-intervals traversed. The softmax function is used to complete the mapping between the user and the complete predicted trajectory.

[0056] Step 7: Count the number of complete user trajectories within multiple sub-intervals to determine the expected location value for each sub-interval. Specifically, the trajectory prediction unit iterates through all travel trajectory tables. If a sub-interval in a travel trajectory table overlaps at least once across all travel trajectory tables, the frequency of that sub-interval is recorded. The overlap frequency of multiple sub-intervals in the travel trajectory tables is calculated, and an expected value is assigned to each sub-interval based on the overlap frequency. These expected values ​​are then sorted as recommended relocation locations for urban mobile charging stations.

[0057] Step 8: Select the sub-interval with the highest expected location value as the target address for a city's mobile charging station. If all city mobile charging stations have been allocated, proceed to Step 9; otherwise, proceed to Step 8. This embodiment can employ a data stack structure, sequentially arranging city mobile charging stations and selecting the unselected station with the smallest distance to the sub-interval. This method ensures that sub-intervals requiring mobile charging stations can select the nearest available station, reducing the travel distance of the charging stations.

[0058] The preferred regional location encoding method in this embodiment involves processing the city spatial map into a 5-bit raster and encoding the latitude and longitude data values ​​of the region in the two-dimensional space into corresponding strings. Different encoding levels represent different ranges. The longer the string, the closer the similar strings are.

[0059] Step 9: Based on the Euclidean distance between the target address and the selected sub-interval, adjust the expected location value of any unselected sub-interval, and return to Step 7.

[0060] In this embodiment, the number of intersections between the target address's surrounding sub-intervals and the predicted user trajectory is used as the frequency. An expected value is assigned to each sub-interval based on this frequency. Specifically, if the sample size for predicting user travel trajectories is small, or if the frequency of intersections between the sub-intervals and the predicted trajectory is low, the relocation and site selection of mobile charging stations in multiple cities cannot be completed in one go. Based on the Euclidean distance between the sub-intervals and the target address, target addresses that are as close as possible to the sub-intervals are selected for secondary allocation.

[0061] Step 10: Based on the target address, migrate all city mobile charging stations to the corresponding sub-area to complete the migration process of city mobile charging stations.

[0062] Reference Figure 4 A migration system for urban mobile charging piles based on trajectory prediction includes a trajectory acquisition unit, a trajectory prediction unit, a user database, and a user database. The user database contains all the information from the trajectory acquisition unit, the trajectory prediction unit, and the user's location information data through a terminal platform. The terminal platform uses the user's home address and workplace as the basis for the trajectory prediction unit to configure confidence for any trajectory.

[0063] Example 2

[0064] like Figure 4 This embodiment describes a migration system for urban mobile charging stations based on trajectory prediction, used to implement the aforementioned migration method for urban mobile charging stations based on trajectory prediction. The migration system includes a user database, a trajectory acquisition unit, a data processing unit, a data analysis unit, a trajectory prediction unit, and a command sending unit. The user database stores attribute parameters of multiple users; the trajectory acquisition unit retrieves geospatial data of the addressing interval and generates n segment trajectories; the data processing unit determines the trajectory function of any user based on the multiple segment trajectories of any user terminal device. The data analysis unit counts the number of complete user trajectories within multiple sub-intervals and determines the expected location value for each sub-interval. The trajectory prediction unit sets the trajectory confidence level based on the user's attribute parameters and predicts the user's complete trajectory. The command sending unit determines the expected location value for each sub-interval and determines the target address of the urban mobile charging station based on the expected location value.

[0065] Example 3

[0066] This embodiment further discloses the method for predicting the probability of user travel address preferences according to the present invention. This method uses a modified trajectory function and trajectory confidence level to determine the predicted probability of user travel address preferences, and based on this, determines the expected location value for the migration of urban mobile charging stations to the corresponding sub-intervals.

[0067] In this embodiment, since the user address preference probability and its grouping are not fixed and are relatively greatly affected by time, it is preferable to statistically analyze the address preference probability and use a time-aware decay function to assess the relative importance of the address preference probability over time. In this embodiment, for a user moving from any preference point X to another arbitrary preference point Y, the address preference probability model... Where t is the start time of the historical trajectory during the trip, t i With t j denoted as the time when the system terminal receives a successful connection after the user uploads location information at preference point X and preference point Y, respectively; M is the number of times the user accesses preference point Y from preference point X; N is the number of times the user accesses preference point X in the historical trajectory T; and ρ is the hyperparameter in the time-aware decay function.

[0068] Based on the address preference probability model within any group, the user's next unknown travel destination can only be locations experienced in their historical trajectory T. However, this does not exclude situations where location data is scarce or trajectory data is too sparse. In such cases, this model will prioritize trajectory prediction based on the trajectory grouping results. In this embodiment, to minimize this problem, the system can optimize the personal preference model by comparing multiple prediction results with actual results. The expected value γ from preference point X to preference point Y is continuously adjusted. The prediction model is as follows: Where ρ is a hyperparameter in the time-aware decay function, and γ=[γ0,γ1,γ2] T s is a matrix obtained by the system through observing the historical trajectories of each user using a personal preference model. The personal preference matrix constructed for the personal preference model.

[0069] To achieve better prediction results, this invention continuously optimizes the prediction model. By accessing the terminal platform, a large amount of historical user trajectory data can be obtained, providing a good data sample for model optimization. In this embodiment, three months of user trajectory data are selected as training samples, divided into weekday training samples and holiday training samples. After training, the model can predict trajectories based on the date parameters in the current time matrix, distinguishing between weekdays and holidays.

[0070] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A migration method for urban mobile charging stations based on trajectory prediction, characterized in that, Includes the following steps: Step 1: Determine the addressing range of the urban mobile charging station, retrieve the geospatial data of the addressing range, and divide the addressing range into multiple sub-ranges based on the geospatial data; Step 2: Retrieve location and time information from multiple user terminal devices, extract spatial and temporal feature vectors, and generate a trajectory embedding sequence; Step 3: Map the trajectory embedding sequence to geospatial data to generate at least one user travel trajectory table. The travel trajectory table is stored in the user database and contains the user's time-related historical trajectories T. Step 4: Divide the sub-interval into a first sub-interval with an access frequency greater than the baseline frequency and a second sub-interval with an access frequency less than or equal to the baseline frequency. The trajectory prediction unit divides the user's historical trajectory T into n segment trajectories, T={T i |i≤n}, the trajectory T of each segment i It intersects only with the unique first subinterval; Step 5: Determine the trajectory function f(T) of any user based on multiple fragment trajectories of any user terminal device. i The similarity of trajectory functions of at least two users is compared by using a loss function. The trajectories of users with high similarity are grouped separately, and the address preference probability of users in each group is recorded. Step 6: Call the user's attribute parameters and set the trajectory confidence K for any segment of the historical trajectory T based on the user's attribute parameters. Ti Predict the user's complete trajectory; Step 7: Count the number of complete user trajectories within multiple sub-intervals and determine the expected location value for each sub-interval; Step 8: Select the sub-interval with the largest expected location value as the target address of a city's mobile charging station. If the city's mobile charging stations have been allocated, proceed to step 9; otherwise, proceed to step 7. Step 9: Adjust the expected location value of any unselected sub-interval based on the Euclidean distance between the target address and the selected sub-interval; Step 10: Migrate all city mobile charging stations to the corresponding sub-area based on the target address.

2. The migration method for urban mobile charging piles based on trajectory prediction according to claim 1, characterized in that, In step 8, the unselected urban mobile charging pile with the smallest distance from the sub-interval is selected.

3. The migration method for urban mobile charging piles based on trajectory prediction according to claim 1, characterized in that, The addressing interval is represented by transforming the discrete vector of the relocation location information of urban mobile charging piles into a low-dimensional vector. The addressing interval contains the geospatial data of the relocation location of urban mobile charging piles.

4. The migration method for urban mobile charging piles based on trajectory prediction according to claim 1, characterized in that, The trajectory embedding sequence is composed of spatial feature vectors and temporal feature vectors. The spatial feature vectors identify the user's geographic coordinates, which include three variables: longitude, latitude, and the administrative region to which the user belongs. The temporal feature vectors are the time values ​​of the user's check-in at each verification code location. The corresponding time values ​​and location information together form a time matrix.

5. The migration method for urban mobile charging piles based on trajectory prediction according to claim 4, characterized in that, The time matrix is ​​obtained based on the reparameterization method of spatial feature vectors and temporal feature vectors. The location information includes the hidden layer vectors in the time matrix. The hidden layer vectors provide the location topology for the extraction of spatial feature vectors and temporal feature vectors of other surrounding locations.

6. The migration method for urban mobile charging piles based on trajectory prediction according to claim 1, characterized in that, The travel trajectory table is a collection of location information and time from the user terminal, while the trajectory cloud terminal is a cloud storage terminal device consisting of a shared data server and a wireless network communication protocol.

7. The migration method for urban mobile charging piles based on trajectory prediction according to claim 1, characterized in that, The user's historical trajectory T is segmented into fragment trajectories based on the complexity of the moving shape, the frequency of user access, and the time threshold.

8. The migration method for urban mobile charging piles based on trajectory prediction according to claim 1, characterized in that, The trajectory confidence level K is determined based on the frequency of occurrence of attribute parameters in the user's travel trajectory table. Tn .

9. The migration method for urban mobile charging piles based on trajectory prediction according to claim 1, characterized in that, The attribute parameters must include at least the user's home address and work address. Based on these attribute parameters, a trajectory confidence score K is assigned to the user's corresponding trajectory. Tn .

10. A migration system for implementing the migration method for urban mobile charging piles based on trajectory prediction according to claim 1, characterized in that, The migration system includes a user database, a trajectory acquisition unit, a data processing unit, a data analysis unit, a trajectory prediction unit, and a command sending unit. The user database stores attribute parameters for multiple users; The trajectory acquisition unit retrieves the geospatial data of the addressing interval and generates n segment trajectories; The data processing unit determines the trajectory function of any user based on multiple fragment trajectories of any user terminal device; The data analysis unit counts the number of complete user trajectories within multiple sub-intervals to determine the expected location value for each sub-interval. The trajectory prediction unit sets the trajectory confidence level based on the user's attribute parameters and predicts the user's complete trajectory. The instruction sending unit determines the expected location value for each sub-interval and determines the target address of the urban mobile charging pile based on the expected location value.

Images