Modeling method and system for user charging behavior portrait

Abstract

The invention provides a modeling method and system for a user charging behavior portrait. Wherein the charging time point, the driving mileage and the battery attenuation data of a user are collected, a regional charging demand thermodynamic diagram is generated in combination with a driving track, and the electric energy consumption and the battery attenuation data are fused to form energy consumption characteristics. In an underground parking lot and other satellite signal missing areas, a differential positioning base station and vehicle-mounted inertial navigation are used for compensating positioning, and continuous space-time trajectory coordinates are generated. Energy consumption characteristics and track coordinates are analyzed through deep reinforcement learning, the correlation among the charging time period, the pile distribution density and the power grid load is identified, and a charging demand space-time weight coefficient is generated; and constructing a user charging behavior feature vector based on the coefficient and power grid data, performing similarity matching classification, and finally establishing a user charging behavior portrait model. According to the technical scheme provided by the invention, accurate prediction of the charging demand and space-time collaborative optimization of the power grid load are realized.

Description

Technical Field

[0001] The present application relates to the field of modeling technology, and in particular to a modeling method and system for user charging behavior portraits. Background Art

[0002] Cross-regional charging service scenarios need to solve the problem of the spatiotemporal dynamic coupling of user charging period preferences and vehicle movement trajectories. The method is required to be able to: integrate multi-source heterogeneous data (such as cross-city highway charging records, commuting trajectories within the city, and holiday travel chains) to explore the spatiotemporal migration patterns of user charging behavior; model the implicit association between charging demand and traffic network status (such as congestion index, length of stay in service areas), and predict the trigger threshold of users' cross-regional charging decisions; dynamically correct user portrait labels to support the optimization of charging pile layout and the generation of time-sharing pricing strategies.

[0003] The current mainstream solution adopts a joint modeling framework of spatiotemporal trajectory embedding and behavioral pattern decoupling: the temporal dependency characteristics of vehicle movement trajectories are extracted through gated recurrent units, and a spatiotemporal graph of charging events is simultaneously constructed, and the correlation of cross-regional charging stations is captured using a graph attention network; a behavioral decoupling module is designed to decompose user charging behaviors into basic needs (such as commuting charging) and random disturbances (such as emergency detour charging), and to simulate charging decision deviations in different scenarios; the dynamic weight parameters of user portraits are output through a regional charging hotspot prediction model.

[0004] However, the spatiotemporal trajectory embedding model is not capable of representing sparse sampling points of long-distance cross-city trajectories (such as continuous driving on the highway for more than 6 hours), resulting in biased prediction of charging demand; the behavior decoupling module relies on predefined behavior category labels and cannot adaptively identify new mixed modes (such as charging time splits caused by temporary work during travel). In addition, the sample generation quality of the adversarial generative network is poor in low-probability events such as extreme weather (such as heavy rain causing a surge in charging demand), resulting in a sharp drop in the credibility of the portrait label. Summary of the invention

[0005] The present application provides a modeling method and system for user charging behavior portraits, which are used to solve the problem in the prior art that it is impossible to dynamically associate user charging behavior with regional power grid load, charging pile distribution and vehicle spatiotemporal trajectory for collaborative analysis in a scenario where satellite signals are missing.

[0006] In a first aspect, the present application provides a method for modeling a user charging behavior profile, including: Collecting the user's charging time point, driving mileage and battery capacity decay data, generating a regional charging demand heat map by analyzing the spatial distribution of the charging time point, the driving mileage and the acquired vehicle driving trajectory, and integrating the power consumption data obtained through the vehicle interface with the battery capacity decay data to generate energy consumption characteristics; Deploy differential positioning base stations within the coverage area corresponding to the heat map of charging demand in the said area. When the vehicle enters an underground parking lot where satellite signals are missing, trigger the in-vehicle inertial navigation function to perform real-time compensation on the positioning coordinates recorded by the differential positioning base stations, and generate spatio-temporal trajectory coordinates that are continuously connected to the positioning coordinates. Jointly analyze the energy consumption characteristics and spatio-temporal trajectory coordinates to identify the correlation between the vehicle charging period, the distribution density of charging piles, and the peak and valley periods of the regional power grid load, and convert the correlation into a spatio-temporal weight coefficient of charging demand through a deep reinforcement learning model. Based on the spatio-temporal weight coefficient of charging demand and the obtained data on the peak and valley periods of the power grid load, construct a user charging behavior feature vector, and perform similarity matching on the user charging behavior feature vector to classify the user charging behavior categories. According to the user charging behavior categories, combine and correlate the spatio-temporal weight coefficient of charging demand, the charging time point, and the spatio-temporal trajectory coordinates to establish a user charging behavior portrait model.

[0007] Optionally, jointly analyze the energy consumption characteristics and spatio-temporal trajectory coordinates to identify the correlation between the vehicle charging period, the distribution density of charging piles, and the peak and valley periods of the regional power grid load, and convert the correlation into a spatio-temporal weight coefficient of charging demand through a deep reinforcement learning model, including: Perform time window matching on the power consumption rate per unit time in the energy consumption characteristics and the position movement rate in the spatio-temporal trajectory coordinates to generate a matching result reflecting the corresponding relationship between the vehicle's position change and the power consumption rate during the charging period. According to the vehicle position change in the matching result, extract the regional coverage characteristics of the distribution density of charging piles during the corresponding charging period, and associate them with the start time points of the peak and valley periods of the power grid load to generate dynamic correlation parameters. Analyze the change in the distribution density of charging piles in the dynamic correlation parameters to obtain the coupling relationship describing the charging period and the distribution density of charging piles, and synchronously analyze the fluctuation of the peak and valley periods of the power grid load to obtain the fluctuation characteristics of the peak and valley periods of the power grid load. The coupling relationship represents the charging period selection tendency, and the fluctuation characteristics represent the power grid load response ability. Perform interaction analysis on the charging period selection tendency and the power grid load response ability to generate correlation relationship parameters; through the training of a deep reinforcement learning model, convert the correlation relationship parameters into a spatio-temporal weight coefficient of charging demand.

[0008] Optionally, analyze the change in the distribution density of charging piles among the dynamic association parameters to obtain the coupling relationship between the charging time period and the distribution density of charging piles, and synchronously analyze the fluctuation of the peak and valley periods of the grid load to obtain the fluctuation characteristics of the peak and valley periods of the grid load. The coupling relationship characterizes the charging time period selection tendency, and the fluctuation characteristics characterize the grid load response ability, including: Segment the data of the change in the distribution density of charging piles among the dynamic association parameters by charging time period, and count the difference in the density of the distribution density of charging piles within each charging time period to generate coupling relationship data describing the change in the charging time period and the distribution density of charging piles; Synchronously extract the load value change data of the peak and valley periods of the grid load among the dynamic association parameters, and calculate the change amplitude of the load value within each peak and valley period of the grid load to generate fluctuation characteristic data of the peak and valley periods of the grid load; The difference in the density of the distribution density of charging piles in the coupling relationship data characterizes the charging pile selection tendency of users during the charging time period; The change amplitude of the load value in the fluctuation characteristic data characterizes the charging demand response ability of the grid during the peak and valley periods.

[0009] Optionally, generate a regional charging demand heat map by analyzing the charging time point, the driving mileage, and the spatial distribution of the obtained vehicle driving trajectories. At the same time, fuse the power consumption data obtained through the vehicle interface with the battery capacity attenuation data to generate energy consumption characteristics, including: Obtain the recording time of the charging time point and the geographical coordinates of the vehicle driving trajectory, map each charging time point to the grid of the preset geographical coordinates, and count the number of occurrences of the charging time point in each grid to generate a charging grid; According to the driving mileage, calculate the total driving mileage of the vehicle within the coverage area of each charging grid, and superimpose the total driving mileage on the number of occurrences of the charging time point in the corresponding grid. Generate a charging demand density grid through the superimposed value; Based on the superimposed value of each grid in the charging demand density grid, construct a regional charging demand heat map; Obtain the power consumption data per unit time of the vehicle before and after the charging time point through the vehicle interface, and perform a superimposed calculation on the power consumption data and the capacity loss ratio in the battery capacity attenuation data to obtain an energy consumption parameter; Calculate the charging demand density of each grid in the regional charging demand heat map, and perform an associated mapping on the charging demand density and the energy consumption characteristic parameter to output the energy consumption characteristic.

[0010] Optionally, deploy differential positioning base stations within the coverage area corresponding to the heat map of regional charging demand. When the vehicle enters an underground parking lot where satellite signals are missing, trigger the in-vehicle inertial navigation function to perform real-time compensation on the positioning coordinates recorded by the differential positioning base stations, and generate spatio-temporal trajectory coordinates that are continuously connected to the positioning coordinates, including: Within the coverage area of the heat map of regional charging demand, deploy differential positioning base stations according to the position boundaries of the charging demand dense areas in the heat map of regional charging demand, and determine the signal coverage range of each differential positioning base station; When the vehicle enters an underground parking lot where satellite signals are missing, use the in-vehicle signal detection device to identify whether the vehicle's position has deviated from the signal coverage range, mark the area that has deviated from the signal coverage range as a signal missing area, and trigger the in-vehicle inertial navigation function to start; Through the in-vehicle inertial navigation function, collect the moving direction and speed data of the vehicle in the signal missing area in real time, and combine with the positioning coordinates recorded by the differential positioning base stations to calculate the relative displacement of the vehicle in the signal missing area; Superimpose the relative displacement with the positioning coordinates recorded by the differential positioning base stations to generate the compensated positioning coordinates of the vehicle in the signal missing area, and connect the compensated positioning coordinates with the positioning coordinates recorded by the differential positioning base stations within the signal coverage range in chronological order to obtain a sequence of positioning coordinates; According to the sequence of positioning coordinates, extract the complete movement trajectory of the vehicle from the signal coverage area to the signal missing area, and generate spatio-temporal trajectory coordinates that are continuously connected to the positioning coordinates recorded by the differential positioning base stations.

[0011] Optionally, based on the spatio-temporal weight coefficient of charging demand and the obtained power grid load peak-valley period data, construct a user charging behavior feature vector, and perform similarity matching on the user charging behavior feature vector to classify user charging behavior categories, including: Extract the numerical parameters representing the charging time period selection tendency in the spatio-temporal weight coefficient of charging demand, and combine with the peak-valley time nodes recorded in the power grid load peak-valley period data to correspondingly combine the numerical parameters with the peak-valley time nodes to form a basic parameter set of user charging behavior; According to the corresponding relationship between the charging time period selection tendency and the peak-valley time nodes in the basic parameter set, perform quantitative scoring on each user's charging time period selection tendency to generate user charging behavior description data; Extract the charging tendency quantization value in the user charging behavior description data, and perform standardization processing on the charging tendency quantization value to obtain a user charging behavior feature vector; Calculate the difference in the charging tendency quantization values between the user charging behavior feature vectors, set the similarity determination criteria between different user behaviors based on the difference in the charging tendency quantization values, and group and match users through the similarity determination criteria; Based on the grouping and matching results, group users with a difference in the charging tendency quantization value less than a set threshold into the same charging behavior category to output the user charging behavior category division result.

[0012] Optionally, according to the user charging behavior category, combine and associate the charging demand spatio-temporal weight coefficient, the user charging time point, and the spatio-temporal trajectory coordinates to establish a user charging behavior portrait model, including: Extract the charging period preference parameters of each user charging behavior category from the user charging behavior category, and the charging period preference parameters are determined by the historical distribution law of the charging time point; Extract the charging location distribution parameters corresponding to the user charging behavior category according to the boundary of the dense area of the spatio-temporal trajectory coordinates; Perform superposition calculation on the charging demand spatio-temporal weight coefficient and the charging time point, and generate a weight distribution parameter reflecting the user's choice of charging at different periods according to the superposition result; Perform position association mapping on the weight distribution parameter and the charging location distribution parameter to establish a dynamic association relationship between the charging period selection tendency and the charging location distribution; Adjust the combination ratio of the charging period preference parameters and the charging location distribution parameters in the user charging behavior category according to the matching degree between the charging period selection tendency and the charging location distribution in the dynamic association relationship; Integrate the charging period preference parameters, the charging location distribution parameters, and the weight distribution parameters based on the adjusted combination ratio to construct the user charging behavior portrait model.

[0013] In a second aspect, the present application provides a modeling system for a user charging behavior portrait, including: A collection module for collecting the user's charging time point, driving mileage, and battery capacity attenuation data, generating a regional charging demand heat map by analyzing the charging time point, the driving mileage, and the spatial distribution of the obtained vehicle driving trajectory, and at the same time fusing the power consumption data obtained through the vehicle interface with the battery capacity attenuation data to generate an energy consumption feature; A deployment module for deploying differential positioning base stations within the coverage range corresponding to the regional charging demand heat map, and triggering the vehicle-mounted inertial navigation function to perform real-time compensation on the positioning coordinates recorded by the differential positioning base stations when the vehicle enters an underground parking lot with missing satellite signals, generating spatio-temporal trajectory coordinates that are continuously connected to the positioning coordinates; An analysis module, configured to jointly analyze the energy consumption characteristics and spatio-temporal trajectory coordinates to identify the correlation between the vehicle charging period, the distribution density of charging piles, and the peak and valley periods of the regional power grid load, and convert the correlation into a spatio-temporal weight coefficient of charging demand through a deep reinforcement learning model; A construction module, configured to construct a user charging behavior feature vector based on the spatio-temporal weight coefficient of charging demand and the obtained data of the peak and valley periods of the power grid load, and perform similarity matching on the user charging behavior feature vector to classify the user charging behavior categories; An establishment module, configured to combine and associate the spatio-temporal weight coefficient of charging demand, the charging time point, and the spatio-temporal trajectory coordinates according to the user charging behavior categories to establish a user charging behavior portrait model.

[0014] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for modeling a user charging behavior portrait as described in the first aspect above.

[0015] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it implements a method for modeling a user charging behavior portrait as described in the first aspect.

[0016] This application collects the charging time points, driving mileage, and battery capacity attenuation data of users, generates a regional charging demand heat map by analyzing the spatial distribution of the charging time points, the driving mileage, and the obtained vehicle driving trajectory, and at the same time fuses the power consumption data obtained through the vehicle interface with the battery capacity attenuation data to generate energy consumption characteristics, capable of fusing multi-dimensional user behavior and vehicle status data to construct a charging demand space prediction model, and realizing the dynamic correlation analysis of the battery health status and the energy usage pattern; by deploying differential positioning base stations within the coverage area corresponding to the regional charging demand heat map, when the vehicle enters an underground parking lot where satellite signals are missing, the on-vehicle inertial navigation function is triggered to compensate the positioning coordinates recorded by the differential positioning base stations in real time, generating space-time trajectory coordinates that are continuously connected to the positioning coordinates, capable of maintaining the space-time continuity of vehicle positioning data in complex indoor and outdoor scenarios and eliminating the interference of signal blind spots on charging behavior tracking; by jointly analyzing the energy consumption characteristics and the space-time trajectory coordinates to identify the correlation between the vehicle charging period, the distribution density of charging piles, and the peak and valley periods of the regional power grid load, and converting the correlation into a charging demand space-time weight coefficient through a deep reinforcement learning model, capable of quantifying the dynamic coupling effect of charging demand, power grid load, and infrastructure distribution, and establishing an extensible energy supply and demand optimization decision-making framework; by constructing a user charging behavior feature vector based on the charging demand space-time weight coefficient and the obtained peak and valley period data of the power grid load, and performing similarity matching on the user charging behavior feature vector to classify the user charging behavior categories, capable of realizing the refined classification of user charging patterns through cluster analysis and providing data support for differential service strategies; by combining and correlating the charging demand space-time weight coefficient, the charging time point, and the space-time trajectory coordinates according to the user charging behavior categories to establish a user charging behavior portrait model, capable of fusing space-time attributes and behavior characteristics to construct a multi-dimensional user portrait and improving the accuracy of intelligent charging service recommendations and power grid load predictions.

[0017] Furthermore, by matching the electricity consumption rate with the position movement rate through a time window, a real-time correlation mapping of the vehicle's dynamic behavior and energy consumption during the charging period is achieved, accurately capturing the microscopic fluctuation law of the spatio-temporal distribution of charging demand; combined with the regional coverage characteristics of the charging pile distribution density and the generation of dynamic correlation parameters during the peak and valley periods of the power grid load, a multi-dimensional coupling analysis framework for the layout of charging infrastructure, user behavior preferences, and power grid carrying capacity is established; by exploring the coupling relationship between the charging period selection tendency and the power grid load response ability, the potential impact path of user charging behavior on the power grid load is revealed, forming a two-way interaction mechanism model that takes into account user habits and power grid stability; using a deep reinforcement learning model to transform the interaction parameters into spatio-temporal weight coefficients, breaking through the limitations of traditional static weight allocation methods, and realizing the adaptive collaborative optimization of charging demand prediction and power grid dispatching strategies; finally, a dynamic decision-making model integrating vehicle movement characteristics, charging pile distribution density, and power grid load fluctuations is constructed, providing a two-way adjustment mechanism for the smart grid that takes into account the elasticity of the user demand side and the stability of the power grid supply side, significantly improving the regional charging resource scheduling efficiency and the power grid load peak-valley balancing ability.

[0018] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. Brief Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 Shows a flowchart of a method for modeling a user charging behavior portrait provided by the present application; Figure 2 Shows a schematic structural diagram of a system for modeling a user charging behavior portrait provided by the present application; Figure 3 Shows a schematic structural diagram of a computing device provided by the present application. Detailed Embodiments

[0021] In order to enable those skilled in the art to better understand the solution of the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application.

[0022] In some processes described in the specification, claims, and the above-mentioned drawings of the present application, a plurality of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear herein or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are different types.

[0023] Researchers found that existing charging behavior analysis solutions rely on single charging data or static power grid load forecasting, making it difficult to dynamically integrate user behavior, battery degradation, and underground scene positioning compensation, resulting in large deviations in charging demand forecasting, insufficient trajectory continuity, and low grid coordination efficiency. Based on this, a user charging behavior portrait modeling and dynamic optimization method is provided, which can achieve precise coordination and adaptation of charging demand and grid load through multi-source data spatio-temporal fusion and deep reinforcement learning modeling. The technical solution of the present application is applicable to scenarios such as intelligent charging scheduling, underground parking lot navigation optimization, and regional power grid load balancing.

[0024] The entire R & D process embodies a closed-loop optimization mechanism of multi-modal data-driven and user behavior dynamic modeling, aiming to overcome the defects of data dimension fragmentation, spatio-temporal trajectory compensation lag, and user behavior modeling staticization in existing solutions. Through the joint analysis of the charging demand heat map and underground positioning compensation, it breaks through the dependence of traditional charging prediction on a single ground scenario; based on the dynamic conversion of the charging demand weight by the deep reinforcement learning model, it solves the complexity of the correlation modeling between the peak and valley periods of the power grid load and user behavior; by combining the classification of user behavior categories and the combined association of the portrait model, it eliminates the adaptation deviation between the charging demand prediction and the real scenario; finally, through the closed-loop iterative optimization mechanism, it realizes the two-way improvement of the accuracy of the user behavior portrait and the efficiency of the power grid scheduling. This method forms a full-link dynamic coordination from data fusion to portrait modeling, significantly enhancing the intelligent management ability of charging resources in complex scenarios.

[0025] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0026] Figure 1 The flowchart of a modeling method for a user charging behavior portrait is provided for the embodiments of the present application, asFigure 1 As shown, the method includes: 101. Collect the charging time points, driving mileage, and battery capacity attenuation data of the user. Generate a regional charging demand heat map by analyzing the spatial distribution of the charging time points, the driving mileage, and the obtained vehicle driving trajectory. At the same time, fuse the power consumption data obtained through the vehicle interface with the battery capacity attenuation data to generate an energy consumption feature; In this step, the charging time point refers to the specific time information when the user charges the vehicle. The driving mileage refers to the distance data traveled by the vehicle within a specific time period. The battery capacity attenuation data refers to the quantitative index of the decline of battery performance over time. The vehicle driving trajectory refers to the sequence of position coordinates recorded during the vehicle's driving. The spatial distribution refers to the distribution characteristics of charging demand in geographical locations. The regional charging demand heat map refers to a visualization model reflecting the charging demand intensity in different regions. The power consumption data refers to the power usage record during the vehicle's driving. The energy consumption feature refers to a comprehensive index reflecting the vehicle's energy usage pattern.

[0027] In the embodiment of the present application, first, collect the charging time points (such as daily charging periods), driving mileage (such as single-day driving kilometers), and battery capacity attenuation data (such as remaining capacity percentage) of the user through an in-vehicle terminal, and at the same time combine the geographical coordinates of the vehicle driving trajectory obtained by the GPS module. Secondly, perform a spatial overlay analysis on the charging time points and the driving trajectory, and use a heat map generation algorithm (such as kernel density estimation) to statistically analyze the charging frequency and driving mileage distribution in different regions to generate a regional charging demand heat map reflecting the charging demand density. Then, extract the power consumption data (such as power consumption per 100 kilometers) from the vehicle interface, and integrate it with the battery capacity attenuation data through a weighted fusion algorithm (such as principal component analysis) to generate an energy consumption feature representing the vehicle's energy consumption characteristics after eliminating noise. Finally, store the heat map and the energy consumption feature in association for providing input for subsequent spatio-temporal analysis.

[0028] In a new energy vehicle big data platform in Shenzhen, the system collects the charging time points, driving mileage, and battery capacity attenuation data of 100,000 electric vehicles every day. Through analysis, it is found that the charging time points of vehicles in the Futian CBD area are concentrated from 8 to 10 pm on weekdays, and the monthly driving mileage of vehicles in this area is generally 40% higher than that of suburban users. Combining with the spatial distribution of vehicle driving trajectories, the system identifies that there is a small charging peak at noon on weekdays in the Science Park area, which highly coincides with the vehicle flow during the dining hours in the surrounding commercial areas. At the same time, after fusing the power consumption data obtained through the vehicle interface with the battery capacity attenuation data, the platform finds that when vehicles with a vehicle age of more than three years are fast-charged in high-temperature weather, the battery efficiency degradation rate is twice as fast as that of new vehicles. Based on the above data, the system generates a regional charging demand heat map covering the whole city, where the red-highlighted area shows that the charging demand during the evening peak in Nanshan District exceeds 1.5 times the supply capacity of the pile group, and marks the abnormal energy consumption zone in the aging vehicle aggregation area.

[0029] 102. Deploy differential positioning base stations within the coverage area corresponding to the regional charging demand heat map. When a vehicle enters an underground parking lot where satellite signals are missing, trigger the vehicle-mounted inertial navigation function to perform real-time compensation on the positioning coordinates recorded by the differential positioning base stations, and generate space-time trajectory coordinates that are continuously connected to the positioning coordinates. In this step, the differential positioning base station refers to a ground reference station used to improve positioning accuracy. The absence of satellite signals refers to the regional state where satellite positioning signals cannot be received. The vehicle-mounted inertial navigation function refers to the technology of independently calculating the position using vehicle sensors. The positioning coordinates refer to the longitude and latitude data of the vehicle's position. Real-time compensation refers to the supplementary calculation of the position when the signal is missing. The space-time trajectory coordinates refer to the continuous position record including the time and space dimensions.

[0030] In the embodiment of the present application, first, deploy differential positioning base stations in the high-frequency charging areas identified by the regional charging demand heat map to achieve centimeter-level positioning calibration by enhancing satellite signals. Secondly, when the vehicle drives into an underground parking lot where satellite signals are interrupted, the vehicle-mounted system automatically activates the inertial navigation function, calculates the vehicle position in real time based on the gyroscope and accelerometer data, and dynamically compensates the positioning coordinates of the differential base stations. Then, combine the compensated coordinates with the time stamps when the vehicle enters and exits the parking lot to generate continuous and unbroken space-time trajectory coordinates. Finally, transmit the corrected trajectory coordinates to the cloud database in real time to update the vehicle position information.

[0031] In response to the charging demand blind spots in the underground parking lots in the Luohu Commercial District in the regional charging demand heat map, the Municipal Transport Commission has deployed differential positioning base stations in six large parking lots such as MixC and KKMALL. When a vehicle enters the area where satellite signals are interrupted in the basement, the in-vehicle inertial navigation is immediately activated. By fusing data on wheel speed and steering wheel angle, it compensates the entrance coordinates recorded by the differential positioning base station in real time. When a Tesla Model Y owner enters the basement from Shennan East Road, the system continuously generates spatio-temporal trajectory coordinates with sub-meter accuracy, completely recording the path of the vehicle moving from Area C on the B2 floor to the charging pile in Area D. The compensated trajectory data shows that the average distance that vehicles circle around to find charging piles between 12:00 and 2:00 noon in this basement reaches 300 meters, revealing the efficiency loss problem caused by unclear charging pile signs. This data is synchronously fed back to the heat map abnormal area analysis module.

[0032] 103. Jointly analyze the energy consumption characteristics and spatio-temporal trajectory coordinates to identify the correlation between the vehicle charging period, the distribution density of charging piles, and the peak and valley periods of the regional power grid load, and convert the correlation into a spatio-temporal weight coefficient of charging demand through a deep reinforcement learning model; In this step, joint analysis refers to a processing method for cross-verifying multi-source data. The distribution density of charging piles refers to the number of charging facilities in a unit area. The peak and valley periods of the regional power grid load refer to the time periods when the load of the power system changes between high and low. The correlation refers to the degree of mutual influence between different factors. The deep reinforcement learning model refers to a machine learning method that optimizes decisions through a reward mechanism. The spatio-temporal weight coefficient of charging demand refers to a parameter that quantifies the charging demand in the spatio-temporal dimension.

[0033] In the embodiment of this application, first, align the energy consumption characteristics generated in step 101 and the spatio-temporal trajectory coordinates in step 102 according to the time series, and extract the correlation data between energy consumption and location of the vehicle within a specific period and geographical range. Secondly, analyze the dependence relationship between the charging period, the distribution density of charging piles, and the peak and valley periods of the regional power grid load through association rule mining technology, and identify the conflict or cooperation mode between the charging demand and the power grid load. Then, construct a deep reinforcement learning model, input the correlation into the model for policy training, and output a spatio-temporal weight coefficient of charging demand that quantifies the charging demand priority under different spatio-temporal conditions. Finally, bind the weight coefficient to the power grid load data to form a dynamic scheduling policy library.

[0034] The platform jointly analyzes the energy consumption characteristics and spatio-temporal trajectory coordinates, and discovers the superposition effect of vehicle charging demand and the peak value of the airport grid load during the peak flight period at the Bao'an Airport charging station. Through 100,000 training iterations, the deep reinforcement learning model identifies the positive correlation between the utilization rate of fast chargers at the airport at 7:00 am on weekdays and the grid transmission loss, and converts it into the spatio-temporal weight coefficient of charging demand. The model output shows that the spatio-temporal weight coefficient of charging demand in the Longgang residential area reaches 0.78 at 9:00 pm, while the grid load in this area at the same time period is at 60% of the daily peak level, revealing the problem of resource allocation that community charging piles do not effectively utilize the low valley electricity price of the grid. This coefficient is marked as a key parameter for dispatching optimization.

[0035] 104. Based on the spatio-temporal weight coefficient of charging demand and the obtained data of the peak and valley periods of the grid load, construct a user charging behavior feature vector, and perform similarity matching on the user charging behavior feature vector to divide the user charging behavior categories; In this step, the data of the peak and valley periods of the grid load refers to the time characteristics of the load fluctuation of the power system. The user charging behavior feature vector refers to a multi-dimensional feature combination that describes charging habits. Similarity matching refers to an analysis method for calculating the degree of proximity between feature vectors. The user charging behavior category refers to the user group divided according to charging habits.

[0036] In the embodiment of the present application, first, integrate the spatio-temporal weight coefficient of charging demand in step 103 and the data of the peak and valley periods of the grid load obtained externally, extract the charging time, location and energy consumption characteristics according to the user ID, and construct a multi-dimensional user charging behavior feature vector. Secondly, use the clustering algorithm to perform similarity matching on the feature vectors, and divide different user charging behavior categories according to the charging mode rules (such as charging at fixed times, random charging) and the grid load response ability (such as valley period preference). Finally, associate the classification label with the user portrait and store it in the user behavior database to provide a basis for personalized charging strategies.

[0037] Based on the spatio-temporal weight coefficient of charging demand and the peak and valley data of the grid load, the system constructs a user charging behavior feature vector including dimensions such as charging frequency, period preference, and grid response sensitivity. The data of a fleet of operating vehicles shows that there are significant differences in indicators such as the charging proportion at 3:00 am and the utilization rate of the grid valley period in its charging behavior feature vector compared with the private car group. Through cosine similarity matching, the platform divides users into three major categories: "grid response type", "fixed habit type", and "random demand type". Among them, the charging proportion of "grid response type" users during the electricity price valley period reaches 85%, and this type of user is identified as the best target group for demand-side management.

[0038] 105. Combine and correlate the charging demand spatio-temporal weight coefficient, the charging time point, and the spatio-temporal trajectory coordinates according to the user charging behavior category to establish a user charging behavior portrait model.

[0039] In this step, combination and correlation refer to the processing method of integrating and correlating multiple features. The user charging behavior portrait model refers to the data model describing the user charging characteristics.

[0040] In the embodiment of the present application, first, based on the user charging behavior category divided in step 104, extract the charging time, spatial location, and energy consumption weight characteristics of typical users. Secondly, perform multi-dimensional combination of the charging time point in step 101, the charging demand spatio-temporal weight coefficient in step 103, and the spatio-temporal trajectory coordinates in step 102 according to the user ID to construct a feature matrix reflecting the user charging habit. Then, input it into the machine learning model to train the user charging behavior portrait model and output the user charging behavior label (such as "charging in the residential area at night on weekdays"). Finally, deploy the portrait model to the charging scheduling system to generate personalized charging suggestions in real time, forming a data-driven closed-loop optimization process.

[0041] Combined with the user charging behavior category, the platform establishes a refined user charging behavior portrait model. For the group of office workers in the science and technology park among the "fixed habit type" users, the model combines and correlates their charging time point at 8 pm, the daily driving mileage of 15 kilometers, and the commuting trajectory coordinates from Futian to Nanshan, and outputs the portrait label of "charging during the cross-district commuting peak". Based on this portrait, the charging operator adds a photovoltaic-storage-charging integrated station at the Tongle Checkpoint on the Guangzhou-Shenzhen Expressway to guide users to complete charging on their way home from work. After three months of implementation, the charging queue time at the core urban stations for this group on weekday evenings is shortened by 12 minutes, and the cross-district charging ratio is increased to 37%, verifying the two-way optimization value of the portrait model for charging facility planning and user behavior guidance.

[0042] To sum up, steps 101 to 105 achieve multi-dimensional data fusion modeling and accurate portrait construction of user charging behavior. By integrating multi-source data such as charging time points, driving mileage, and battery capacity attenuation, the system constructs a heat map reflecting the regional charging demand distribution, and forms an energy consumption dynamic model by integrating the power consumption characteristics. The combination of differential positioning base station deployment and inertial navigation compensation technology breaks through the positioning continuity problem in the scenario of satellite signal loss, ensuring the complete acquisition of spatio-temporal trajectory coordinates. The intelligent analysis of the correlation relationship between charging time periods, pile group density, and grid load by the deep reinforcement learning model realizes the accurate construction and classification of user behavior feature vectors. The finally established user charging behavior portrait model dynamically correlates the spatio-temporal weight coefficient with the charging preference, providing a multi-dimensional user behavior cognitive framework for power grid scheduling and charging service optimization.

[0043] In order to integrate the correlation between the vehicle's electric energy consumption rate and its spatio-temporal trajectory, establish a collaborative analysis framework for charging demand and grid load, dynamically analyze the coupling law between vehicle position changes and charging pile distribution density during the charging period based on the time window matching mechanism, quantify the constraint relationship between the charging period selection tendency and the grid response ability in combination with the peak-valley fluctuation characteristics of the grid load, use deep reinforcement learning to convert multi-dimensional dynamic correlation parameters into spatio-temporal weight coefficients, realize the intelligent prediction of the spatio-temporal distribution of charging demand and resource adaptation, thereby optimizing the charging resource scheduling strategy, alleviating the peak-valley load pressure of the grid, improving the charging efficiency in high-density areas and the operation stability of the grid, and finally constructing a global optimization decision-making model that takes into account user charging preferences and the dynamic carrying capacity of the grid.

[0044] In some embodiments, as described in step 103, the energy consumption characteristics and spatio-temporal trajectory coordinates are jointly analyzed to identify the correlation between the vehicle charging period, the charging pile distribution density, and the peak-valley periods of the regional grid load, and the deep reinforcement learning model is used to convert the correlation into the spatio-temporal weight coefficient of the charging demand, including: 201. Perform time window matching on the power consumption rate per unit time in the energy consumption characteristics and the position movement rate in the spatio-temporal trajectory coordinates to generate a matching result reflecting the corresponding relationship between the vehicle's position change and the power consumption rate during the charging period; In step 201, the power consumption rate per unit time refers to the power consumption speed of the vehicle within a time unit. The position movement rate refers to the displacement change amount of the vehicle per unit time. Time window matching refers to aligning different data streams within a specific time period. The matching result refers to the analysis conclusion reflecting the relationship between power consumption and position change.

[0045] In the embodiments of the present application, first, the power consumption rate per unit time (such as the power consumption per hour) is extracted from the energy consumption characteristics obtained in step 101, and at the same time, the position movement rate (such as the distance the vehicle moves per minute) is extracted from the spatio-temporal trajectory coordinates in step 102. Secondly, the time series data of the two are segmented and matched according to a fixed window (such as 15 minutes) through the time window alignment algorithm to generate a matching result reflecting the corresponding relationship between the vehicle's position change and the power consumption rate during the same period. Then, the matching results are classified and stored according to the user ID, and the data differences between the charging period and the non-charging period are marked to provide a structured input for subsequent analysis.

[0046] 202. According to the vehicle position change in the matching result, extract the regional coverage characteristics of the charging pile distribution density during the corresponding charging period and associate them with the start time points of the peak-valley periods of the grid load to generate dynamic correlation parameters; In step 202, the area coverage feature refers to the distribution characteristics of charging piles in space. The dynamic correlation parameter refers to the variable reflecting the relationship between the charging facilities and the grid load. The starting time point refers to the critical time mark of the grid load change.

[0047] In the embodiment of the present application, first, based on the matching result of step 201, a set of geographical coordinates corresponding to the vehicle position change during the charging period is extracted, and the distribution density of coordinate points is statistically calculated through the kernel density estimation algorithm to generate an area coverage feature reflecting the charging pile coverage density (such as the number of charging piles per square kilometer in a certain area). Secondly, the peak-valley period data of the grid load provided by the power grid company is obtained (such as the peak period is 18:00 - 22:00), the area coverage feature is associated with the starting time point of the peak-valley period (such as 18:00) on the time axis, and a dynamic correlation parameter is generated through the weighted fusion algorithm (such as the correlation strength between the charging pile utilization rate during the peak period and the grid load). Finally, the parameters are stored according to the area-period dimension for subsequent coupling relationship analysis.

[0048] 203. Analyze the change in the charging pile distribution density in the dynamic correlation parameter to obtain the coupling relationship between the charging period and the charging pile distribution density, and synchronously analyze the fluctuation condition of the grid load peak-valley period to obtain the fluctuation characteristics of the grid load peak-valley period. The coupling relationship characterizes the charging period selection tendency, and the fluctuation characteristics characterize the grid load response ability; In step 203, the coupling relationship refers to the degree of mutual influence between the charging period and the charging pile density. The fluctuation characteristics refer to the regular characteristics of the grid load changing with time. The charging period selection tendency refers to the time characteristics of user preference for charging. The grid load response ability refers to the ability index of the power system to respond to load changes.

[0049] In the embodiment of the present application, first, perform time series decomposition on the dynamic correlation parameter of step 202, extract the law of the charging pile distribution density changing with the period (such as the density drops by 30% during the evening peak), and describe the coupling relationship between the charging period and the charging pile distribution density through a statistical model (such as the more concentrated the charging period, the greater the pressure on the charging pile density). Secondly, synchronously analyze the power fluctuation data of the grid load peak-valley period (such as the peak load is 50% higher than the valley value), and generate the fluctuation characteristics of the grid load peak-valley period through the trend fitting algorithm (such as the load change rate and the duration). Then, mark the coupling relationship as the charging period selection tendency (such as users prefer to avoid charging during the peak period), and mark the fluctuation characteristics as the grid load response ability (such as more charging demands can be carried during the valley period), and the two respectively characterize user behavior and grid state.

[0050] 204. Perform an interaction analysis on the charging period selection tendency and the grid load response ability to generate a correlation relationship parameter; In step 204, interaction analysis refers to an analysis method for studying the mutual influence of multiple factors. The correlation relationship parameter refers to a numerical index that quantifies the intensity of the interaction between factors.

[0051] In the embodiment of the present application, first, the charging period selection tendency in step 203 (such as the user's tendency to charge during the valley period) and the grid load response ability (such as sufficient load margin during the valley period) are subjected to interaction analysis, and the degree of cooperation or conflict between the two is calculated through the Pearson correlation coefficient (such as a positive correlation indicating that the user behavior matches the grid state). Secondly, a linear regression model is constructed according to the correlation coefficient to quantify the alleviating or aggravating effect of the charging period selection on the grid load, and a correlation relationship parameter that comprehensively reflects the user-grid interaction intensity is generated (such as a cooperation coefficient of 0.8 indicating a high degree of matching). Finally, the parameter is bound to the region-period dimension and input into the policy optimization module.

[0052] 205. Through the training of the deep reinforcement learning model, the correlation relationship parameter is converted into a spatio-temporal weight coefficient of the charging demand.

[0053] In step 205, training refers to the process of optimizing model parameters through data. Conversion refers to the operation of converting the analysis result into a computable parameter. The spatio-temporal weight coefficient of the charging demand refers to a quantified value that reflects the charging demand in the spatio-temporal dimension.

[0054] In the embodiment of the present application, first, a deep reinforcement learning model (such as DQN) is constructed, and the correlation relationship parameter in step 204 is used as the environmental state input, and the charging demand scheduling policy is used as the action space. Secondly, the model policy is optimized through multiple rounds of training iterations (such as the Q-learning update rule), so that the model learns to dynamically adjust the spatio-temporal allocation weight of the charging demand under a specific charging period selection tendency and grid load response ability. Then, the policy output by the model is converted into a quantifiable spatio-temporal weight coefficient of the charging demand (such as a weight of 0.6 for the business district during the evening peak), which represents the charging priority under different spatio-temporal conditions. Finally, the coefficient is synchronized to the charging scheduling system to guide the allocation of charging pile resources and the regulation of the grid load in real time, forming a closed-loop control link.

[0055] The following is a specific example: In the intelligent charging management system for the business district, the charging behavior portrait of users accurately guides load regulation. The system matches the charging rate of electric network-hailing vehicles around commercial buildings with their pick-up and driving trajectories in real time (step 201), and it is found that the charging rate fluctuates violently when vehicles move frequently and short distances in the core business district during the evening rush hour. By analyzing the distribution of charging pile groups in the charging hotspots (step 202), it is identified that the charging piles in the underground parking lots of office buildings show super-dense usage characteristics from 18:00 to 20:00. Synchronously correlating with the grid load curve, it is found that this period coincides with the evening rush hour of the regional substation. In-depth analysis shows that the user's charging time selection is highly concentrated in the vulnerable grid periods (step 203), forming a dangerous coupling between the charging demand pulse and the power supply capacity depression. The system quantifies this spatio-temporal mismatch relationship into an interaction parameter (step 204), and the deep reinforcement learning model generates a dynamic weight coefficient based on this (step 205), and intelligently recommends the strategy of "charging 1 hour in advance + collaborative use of charging piles across buildings" for high-frequency users. When a large-scale exhibition is held in a business district, the system, based on the updated weight parameters, guides thirty network-hailing vehicles to divert to the idle charging piles in adjacent hotels, successfully reducing the peak value of the charging load in the core area. At the same time, five abnormal users (the charging time and pick-up trajectory continuously deviate) are identified through the portrait model, triggering the charging behavior review mechanism. This system realizes the multi-level linkage from the microscopic user portrait to the macroscopic power grid dispatching, reducing the frequency of overloading warnings for charging piles in the business district.

[0056] In summary, steps 201 to 205 achieve the dynamic correlation modeling of the spatio-temporal characteristics of charging demand and the intelligent weight conversion. By dynamically correlating the power consumption rate and the position movement rate through the time window matching technology, the system reveals the deep coupling relationship between the charging time selection and the distribution density of charging piles. The dynamic correlation parameter generation mechanism effectively quantifies the influence degree of the peak-valley fluctuation of the grid load on the charging behavior, breaking through the limitations of traditional static correlation analysis. The deep reinforcement learning model transforms the multi-dimensional correlation parameters into spatio-temporal weight coefficients through interaction analysis, forming an interpretable quantification method for charging demand. This technical solution realizes the intelligent mapping from physical data to behavior characteristics, providing a high-precision parametric model for dynamic charging demand prediction.

[0057] In some embodiments, as described in step 203, analyzing the change in the distribution density of charging piles in the dynamic correlation parameters to obtain the coupling relationship between the charging time and the distribution density of charging piles, and synchronously analyzing the fluctuation of the grid load peak-valley period to obtain the fluctuation characteristics of the grid load peak-valley period. The coupling relationship represents the charging time selection tendency, and the fluctuation characteristics represent the grid load response ability, including: 301. Segment the data of the charging pile distribution density change in the dynamic association parameters by charging time periods, and statistically analyze the difference in the density of the charging pile distribution within each charging time period, so as to generate data describing the coupling relationship between the charging time period and the change in the charging pile distribution density; In step 301, the segmentation of the charging time period refers to classifying the charging behavior according to time intervals. The difference in density refers to the comparison result of the charging pile utilization rate in different time periods. The coupling relationship data refers to the quantitative index reflecting the association between the charging time period and the charging pile density.

[0058] In the embodiment of the present application, first, extract the data of the change in the charging pile distribution density from the dynamic association parameters generated in step 202, divide the time window according to the charging time periods (such as morning peak, evening trough), and divide the whole day into several continuous time periods. Secondly, use the kernel density estimation algorithm for the charging pile distribution density data within each time period to statistically analyze the difference in density in different regions (such as the density in the commercial area is higher than that in the residential area). Then, by comparing the density change amplitude between adjacent time periods (such as the density in the evening peak drops by 20% compared with that during the day), generate a time series association table describing the coupling relationship between the charging time period and the change in the charging pile distribution density, that is, the coupling relationship data. Finally, store the data according to the region-time period dimension to provide input for analyzing the user selection tendency.

[0059] 302. Synchronously extract the data of the change in the load value during the peak and trough periods of the power grid load in the dynamic association parameters, and calculate the change amplitude of the load value within each peak and trough period of the power grid load, so as to generate the fluctuation characteristic data of the peak and trough periods of the power grid load; In step 302, the data of the change in the load value refers to the recorded information of the power grid load fluctuation. The change amplitude refers to the quantitative value of the load difference between the peak and trough periods. The fluctuation characteristic data refers to the data set describing the power grid load change law.

[0060] In the embodiment of the present application, first, synchronously extract the data of the change in the load value during the peak and trough periods of the power grid load from the dynamic association parameters, and divide the time window according to the peak and trough periods (such as the peak period 18:00-22:00). Secondly, calculate the fluctuation trajectory of the load value through the moving average algorithm within each window, and statistically analyze its change amplitude (such as the peak value increases by 50% compared with the trough value). Then, identify the steep increase or slow decrease mode of the load change through the trend analysis algorithm (such as the load in the evening peak climbs rapidly), and generate the fluctuation characteristic data describing the power grid load fluctuation law. Finally, bind the data with the peak and trough period labels for evaluating the power grid response ability.

[0061] 303. The difference in the density of the charging pile distribution in the coupling relationship data characterizes the charging pile selection tendency of users during the charging time period; In step 303, the charging pile selection tendency refers to the behavioral characteristics of users preferring to use specific charging piles. The difference in density refers to the time distribution characteristics of the charging pile utilization rate.

[0062] In the embodiments of the present application, first, an intensity difference index within different charging time periods is extracted from the coupling relationship data in step 301 (for example, the occupancy rate of charging piles in the business district during the evening peak is 80%). Secondly, the preference rules of users for selecting charging piles at different times are identified through a behavior pattern analysis algorithm (such as decision tree classification) (for example, users avoid crowded areas during peak hours). Then, the intensity difference is mapped to the probability distribution of users' selection of charging piles (for example, the probability of users selecting low-density areas during peak hours increases), and the specific manifestations of the charging time period selection tendency are clarified. Finally, the tendency data is associated with the user portrait to support subsequent policy optimization.

[0063] 304. The change amplitude of the load value in the fluctuation feature data characterizes the charging demand response ability of the power grid during peak and valley periods.

[0064] In step 304, the charging demand response ability refers to the ability index of the power grid to cope with changes in charging load. The change amplitude refers to the intensity quantization value of the power grid load fluctuation.

[0065] In the embodiments of the present application, first, the change amplitude of the load value is extracted from the fluctuation feature data in step 302 (for example, the load fluctuation range during the valley period is ±10%). Secondly, the response upper limit of the power grid to the charging demand at different times is quantified through a power grid carrying capacity evaluation model (such as elastic coefficient calculation) (for example, the valley period can additionally carry 30% of the charging load). Then, the change amplitude is associated with the response upper limit to generate a level label characterizing the power grid load response ability (such as high response, medium response, low response). Finally, the response ability data is synchronized to the dispatching system to guide the charging demand shunt decision.

[0066] The following is a specific example: In the intelligent charging management system of tourist attractions, the user charging behavior portrait system accurately balances the supply and demand relationship. When the scenic area implements the time-sharing charging guidance strategy, the system divides the usage data of the charging pile group in the lakeside charging area into three time periods: morning, noon, and evening (step 301). It is found that the usage density of charging piles during the lunch period of holidays is 2.3 times the peak value of weekday evenings, and generates coupling relationship data showing that "the charging hotspots during lunch time gather towards the viewing platform". The load fluctuation curve of the scenic area substation is analyzed simultaneously (step 302), and the load change amplitude during the 14:00-16:00 period during holidays is captured. The larger than that on weekdays, the formation of fluctuation characteristic data characterizing the pressure risk of the power grid. The system identifies the behavior preference of tourists to "stop and charge" through the coupling relationship data (step 303), while the fluctuation characteristic data reveals that the response margin of the power grid is insufficient during the high load period at noon (step 304). Based on this, the intelligent scheduling model automatically activates the diversion strategy during the lunch period of holidays: 30% of the charging demand is guided to the charging piles at the edge of the parking lot through APP push, and the scenic area energy storage system is linked to improve the response capability of the local power grid. When monitoring a surge in tourists on a certain weekend caused the charging piles around the lake to be overloaded, the system dynamically adjusted the coupling relationship weight and temporarily enabled the access rights of the pile group in the backup charging area, successfully avoiding the main substation tripping accident. The portrait model is continuously optimized, and a new module for identifying night charging behavior characteristics in camping areas has been added to increase the utilization rate of charging piles during non-peak hours, forming a closed-loop management mechanism from deconstructing spatiotemporal characteristics to enhancing grid resilience.

[0067] In summary, steps 301 to 304 achieve a refined analysis of the distribution of charging piles and the characteristics of grid load fluctuations. By segmenting the time periods of changes in charging pile density, the system constructs a quantitative evaluation system for the tendency to select charging time periods, accurately describing the time preference characteristics of user charging behavior. The dynamic calculation model of the fluctuation amplitude of grid load during peak and valley periods reveals the boundary of the regional grid's ability to respond to fluctuations in charging demand. The two-dimensional feature extraction of charging pile selection tendency and grid response capability forms a dynamic balance analysis framework for charging demand and power supply capacity. This technical solution deeply binds user behavior preferences with grid operation status, providing a two-way collaborative decision-making basis for the optimal configuration of charging resources.

[0068] In some embodiments, as described in step 101, a regional charging demand heat map is generated by analyzing the spatial distribution of the charging time point, the driving mileage and the obtained vehicle driving trajectory, and the power consumption data obtained through the vehicle interface is integrated with the battery capacity decay data to generate energy consumption characteristics, including: 401. Obtain the recording time of the charging time point and the geographic coordinates of the vehicle driving trajectory, map each charging time point to a preset grid of geographic coordinates, and count the number of occurrences of the charging time point in each grid to generate a charging grid; In step 401, the recorded time refers to the specific timestamp when the charging behavior occurs. The geographical coordinates refer to the spatial positioning data of the vehicle's location. The grid refers to the regular unit that divides the geographical area. The charging grid refers to the spatial unit that contains the charging behavior statistics. The occurrence times refer to the frequency of the charging behavior that occurs within the grid.

[0069] In the embodiment of the present application, first, obtain the recorded time of the user's charging time point and the geographical coordinates of the vehicle's driving trajectory from the in-vehicle terminal, and divide the geographical space into uniform grids according to a fixed size (such as 1 square kilometer). Secondly, map the geographical coordinates corresponding to each charging time point to the grid to which it belongs, and count the occurrence times of the charging time points within each grid (such as 50 charging events are recorded in a certain grid). Then, mark the charging event frequency for each grid through the spatial database to generate a charging grid with the grid as the unit and the frequency as the attribute. Finally, store the charging grid data as a raster layer to provide a basis for subsequent density analysis.

[0070] 402. Calculate the total driving mileage of the vehicle within the coverage area of each charging grid according to the driving mileage, and superimpose the total driving mileage on the occurrence times of the charging time points of the corresponding grid, and generate a charging demand density grid through the superimposed value; In step 402, the total driving mileage refers to the total distance traveled by the vehicle within the grid area. Superimposing refers to the cumulative calculation of different data within the grid. The charging demand density grid refers to a spatial model that reflects the charging demand intensity of the region.

[0071] In the embodiment of the present application, first, based on the charging grid generated in step 401, extract the geographical boundary range of each grid, and screen out the driving mileage records that occur within the coverage area of the grid from the vehicle driving data. Secondly, accumulate the driving mileage within each grid to calculate the total driving mileage (such as the vehicle has traveled a total of 500 kilometers in a certain grid). Then, perform weighted superposition of the total driving mileage and the occurrence times of the charging time points according to the grid (such as the weight of the number of times is 70% and the weight of the mileage is 30%) to generate a charging demand density grid that reflects the charging demand intensity. Finally, normalize the superimposed value and store it as a density matrix.

[0072] 403. Construct a regional charging demand heat map based on the superimposed values of each grid in the charging demand density grid; In step 403, the superimposed value refers to the comprehensive quantization value of the charging behavior within the grid. The regional charging demand heat map refers to a map that visually displays the distribution of charging demand.

[0073] In the embodiments of the present application, first, the superimposed value of each grid is extracted from the charging demand density grid in step 402 (for example, the value of a certain grid is 85), and the value is mapped to a color gradient (for example, red represents high density) through a heat map generation algorithm (such as kernel density estimation). Secondly, the color gradient is combined with the geographical grid coordinates to construct a visual regional charging demand heat map, intuitively showing the charging demand distribution in different regions. Then, the density values of the uncovered areas are filled through a spatial interpolation algorithm (such as filling with the mean value of adjacent grids) to ensure the continuity and integrity of the heat map. Finally, the heat map data is synchronized to the scheduling system to support dynamic charging resource allocation.

[0074] 404. Obtain the power consumption data of the vehicle per unit time before and after the charging time point through the vehicle interface, and perform a superimposed calculation on the power consumption data and the capacity loss ratio in the battery capacity attenuation data to obtain an energy consumption parameter. In step 404, the unit time refers to a set fixed time interval. The capacity loss ratio refers to the percentage of battery performance attenuation. The energy consumption parameter refers to an index comprehensively reflecting the energy usage characteristics of the vehicle.

[0075] In the embodiments of the present application, first, the power consumption data per unit time before and after the charging time point is collected in real time through the vehicle interface (such as the average power consumption in the hour before charging), and at the same time, the capacity loss ratio in the historical battery capacity attenuation data is obtained from the battery management system (such as the remaining battery capacity is 80% of the initial value). Secondly, the power consumption data is smoothed according to a time window (such as using a moving average to eliminate fluctuations), and the processed power consumption value and the capacity loss ratio are weighted and superimposed (such as the power consumption accounts for 60% and the capacity attenuation accounts for 40%) to generate an energy consumption parameter comprehensively reflecting the vehicle's energy consumption status. Finally, the parameter is associated and stored according to the vehicle ID for energy characteristic analysis.

[0076] 405. Calculate the charging demand density of each grid in the regional charging demand heat map, and perform an associated mapping on the charging demand density and the energy consumption characteristic parameter to output the energy consumption characteristics.

[0077] In step 405, the charging demand density refers to the intensity value of the charging demand per unit area. The associated mapping refers to an operation of establishing a corresponding relationship between different characteristics. The energy consumption characteristic refers to a comprehensive index reflecting the vehicle's energy usage pattern.

[0078] In the embodiments of the present application, first, the charging demand density of each grid is extracted from the regional charging demand heat map in step 403 (for example, the density value of a certain grid is 0.8), and at the same time, the energy consumption parameters of the vehicles in the corresponding grid are obtained from step 404 (for example, the average energy consumption parameter is 75). Secondly, the charging demand density and the energy consumption parameters are matched by grid through a feature association algorithm (such as principal component analysis) to construct an energy consumption feature vector reflecting the "charging demand - energy consumption level" association relationship. Then, the feature vector is standardized (for example, normalized to the range of 0 - 1) to form input data for model training. Finally, the feature vector is injected into the charging scheduling model to optimize the spatio-temporal allocation efficiency of the charging strategy.

[0079] The following is a specific example: In the community shared electric vehicle charging scheduling system, the user charging behavior portrait system precisely optimizes resource allocation. When the system accesses the operation data of fifty shared cars in the community, first, the charging coordinates of the vehicles in the residential area and the commercial street are mapped to a charging grid with a precision of 200 meters (step 401), and it is identified that there is a charging hot spot in the grid west of the community kindergarten at 7 pm. The algorithm accumulates the driving mileage of the vehicles in this grid for round trips to the shopping center (step 402), and superimposes it on the charging frequency to generate a charging demand density grid showing the characteristics of "high-frequency short-distance supplementary charging". Based on the constructed three-dimensional heat map of the community (step 403), the demand overload area of the commercial street charging piles during the evening peak period is clearly presented. By synchronously analyzing the energy consumption curves of the vehicles before and after charging (step 404), it is found that there is a sudden increase in energy consumption within two hours after charging for vehicles with a decrease in battery health, and energy consumption parameters including the impact of battery attenuation are generated. The system spatially associates the peak charging density area in the heat map with the distribution of high-energy consumption vehicles (step 405) and outputs the energy consumption characteristics of "fast charging piles for high-decay vehicles concentrated in the commercial area". Accordingly, the operation and maintenance team installs slow charging piles in the kindergarten grid to divert vehicles with poor health, and configures a dynamic power regulation module for the fast charging piles in the commercial street. When it is monitored that the charging density shifts to the community stadium on weekends, the system automatically triggers the charging pile mode switch, successfully balancing the charging demands of new and old battery vehicles, and forming a closed-loop management mechanism from geographical feature analysis to facility dynamic adjustment.

[0080] In summary, steps 401 to 405 achieve grid-based precise modeling of the charging demand heat map and energy consumption characteristics. Through the grid mapping technology of charging time points and driving mileage, the system constructs a spatial distribution model reflecting the charging demand density in the region. The superposition calculation of power consumption data and battery attenuation parameters innovatively integrates short-term energy consumption dynamics and long-term battery loss characteristics. The correlation mapping mechanism between the heat map grid and energy consumption characteristics breaks through the limitations of traditional single-dimensional data analysis and forms a composite characteristic model of spatial distribution and energy consumption. This technical solution realizes cross-domain characteristic fusion from the geographical space to the energy dimension and provides a high-precision spatial energy consumption benchmark for charging behavior analysis.

[0081] In some embodiments, as described in step 102, differential positioning base stations are deployed within the coverage area corresponding to the regional charging demand heat map. When the vehicle enters an underground parking lot where satellite signals are missing, the on-vehicle inertial navigation function is triggered to perform real-time compensation on the positioning coordinates recorded by the differential positioning base stations, generating spatio-temporal trajectory coordinates that are continuously connected to the positioning coordinates, including: 501. Within the coverage area of the regional charging demand heat map, deploy differential positioning base stations according to the position boundaries of the charging demand intensive areas in the regional charging demand heat map, and determine the signal coverage range of each differential positioning base station; In step 501, the charging demand intensive area refers to the geographical range with high demand intensity in the charging demand heat map. The position boundary refers to the spatial boundary of the charging demand intensive area. The signal coverage range refers to the geographical area where the differential positioning base station works effectively.

[0082] In the embodiments of the present application, first, based on the regional charging demand heat map generated in step 403, identify the boundary coordinates (such as the longitude and latitude of polygon vertices) of the charging demand intensive areas in the map. Secondly, install differential positioning base stations within the boundary range according to a preset density (such as deploying 2 base stations per square kilometer) to ensure that the signal coverage ranges of the base stations overlap each other. Then, determine the signal coverage range of each base station (such as a radius of 500 meters) through signal strength testing, and mark the boundary of the coverage area in the geographic information system. Finally, synchronize the base station coordinates and coverage range data to the vehicle navigation system to provide a benchmark for signal detection.

[0083] 502. When the vehicle enters an underground parking lot where satellite signals are missing, use the on-vehicle signal detection device to identify whether the vehicle's position has deviated from the signal coverage range, mark the area where the deviation from the signal coverage range occurs as a signal missing area, and trigger the start of the on-vehicle inertial navigation function; In step 502, the on-vehicle signal detection device refers to a vehicle device that monitors the status of positioning signals. The signal missing area refers to the physical space where positioning signals cannot be received. Starting refers to the operation instruction to activate the inertial navigation function.

[0084] In the embodiments of the present application, first, when the vehicle enters an underground parking lot where satellite signals are missing, the in-vehicle signal detection device continuously monitors the signal strength of the surrounding differential positioning base stations. Secondly, if the signal strength continuously drops below the set threshold (e.g., below -90 dBm) and the duration exceeds the tolerance (e.g., 5 seconds), it is determined that the vehicle's position has moved out of the signal coverage area, and this area is marked as a signal missing area. Then, the in-vehicle control module is triggered to activate the inertial navigation function and switch to the data acquisition mode of the gyroscope and accelerometer. Finally, a signal loss event log is sent to the cloud, recording the trigger time and location.

[0085] 503. Real-time collect the moving direction and speed data of the vehicle in the signal missing area through the in-vehicle inertial navigation function, and combine the positioning coordinates recorded by the differential positioning base station to calculate the relative displacement of the vehicle in the signal missing area; In step 503, the moving direction data refers to the real-time record of the vehicle's driving direction. The speed data refers to the measured value of the vehicle's instantaneous speed. The relative displacement refers to the moving distance and direction of the vehicle in the signal missing area.

[0086] In the embodiments of the present application, first, the moving direction (e.g., heading angle) and speed data (e.g., meters per second) of the vehicle in the signal missing area are real-time collected through the in-vehicle inertial navigation function. Secondly, the positioning coordinates (e.g., longitude and latitude) recorded by the last differential positioning base station before the vehicle enters the signal missing area are extracted from the base station database in step 501. Then, the relative displacement of the vehicle (e.g., moving 50 meters eastward) is calculated based on the inertial navigation data, and the dead reckoning method is used to convert the displacement into a position increment relative to the base station coordinates. Finally, the increment data is temporarily stored in the local cache waiting for coordinate compensation.

[0087] 504. Superimpose the relative displacement with the positioning coordinates recorded by the differential positioning base station to generate the compensated positioning coordinates of the vehicle in the signal missing area, and connect the compensated positioning coordinates with the positioning coordinates recorded by the differential positioning base station within the signal coverage area in chronological order to obtain a positioning coordinate sequence; In step 504, the compensated positioning coordinates refer to the position data calculated by inertial navigation. The chronological order refers to the sequential relationship arranged in time sequence. The positioning coordinate sequence refers to a dataset of continuously recorded vehicle positions.

[0088] In the embodiment of the present application, first, the relative displacement calculated in step 503 is superimposed on the positioning coordinates recorded by the differential positioning base station before entering the signal missing area (such as the initial coordinate longitude + 0.001 degrees). Secondly, the relative displacement is converted into absolute geographical coordinates through a coordinate conversion algorithm to generate compensated positioning coordinates (such as the updated longitude and latitude). Then, the compensated coordinates are connected to the original positioning coordinates within the coverage of the base station in chronological order to form a positioning coordinate sequence without breakpoints. Finally, the timing error is eliminated through time synchronization verification to ensure the continuity and smoothness of the coordinate sequence.

[0089] 505. Extract the complete movement trajectory of the vehicle from the signal coverage area to the signal missing area according to the positioning coordinate sequence, and generate a spatio-temporal trajectory coordinate that is continuously connected to the positioning coordinates recorded by the differential positioning base station.

[0090] In step 505, the complete movement trajectory refers to the driving path including the signal coverage and missing areas. Continuously connected means the seamless connection of the trajectories generated by different positioning methods. The spatio-temporal trajectory coordinate refers to the position record including the time and space dimensions.

[0091] In the embodiment of the present application, first, all coordinate points of the vehicle entering the signal missing area from the signal coverage area are extracted from the positioning coordinate sequence in step 504. Secondly, the coordinate gap at the moment of signal switching is filled through a trajectory interpolation algorithm (such as linearly interpolating to supplement the missing 0.5 seconds of data). Then, the interpolated coordinate sequence is smoothly connected to the trajectory segment recorded by the base station to generate a complete spatio-temporal trajectory coordinate (such as a longitude and latitude sequence including timestamps). Finally, the trajectory coordinates are transmitted back to the cloud and the vehicle position database is updated to complete the closed-loop positioning correction.

[0092] The following is a specific example: In the charging navigation system of the airport integrated transportation hub, the positioning compensation system accurately reconstructs the underground charging trajectory. When the system detects that the ground charging area of ​​the terminal presents an "L-shaped" thermal distribution (step 501), eight groups of differential positioning base stations are deployed along the charging-intensive channel from area A to area D, and their signal coverage accurately matches the online car-hailing connection route. When a new energy taxi enters the charging area on the B2 floor, the on-board device instantly recognizes the loss of satellite signals (step 502), and immediately activates the six-axis gyroscope and wheel speed sensor to start inertial navigation. During the vehicle's circuitous movement between charging piles (step 503), the system integrates the 23-degree turning data of the inertial navigation with the last positioning coordinates of the base station in real time, and accumulates and generates a displacement vector relative to the underground charging area. When the vehicle completes charging and drives towards the E12 exit (step 504), the algorithm seamlessly connects the compensated positioning coordinates of the underground path with the ground base station signal to form a three-dimensional trajectory sequence including the ramp climbing angle. Based on the complete spatiotemporal trajectory coordinates (step 505), the system identifies the abnormal behavior characteristics of the vehicle's "detour service channel after charging", triggering the charging efficiency review mechanism. During a rainstorm, the system successfully restored the actual length of time five vehicles stayed in the underground charging area by compensating the coordinate sequence, corrected the billing error caused by signal loss, and provided three-dimensional trajectory data support for the optimized layout of hub charging piles, forming a closed-loop management chain from positioning hardware deployment to behavior portrait generation.

[0093] In summary, steps 501 to 505 achieve high-precision positioning compensation and space-time trajectory reconstruction in complex scenarios. Through the dynamic coordination of differential positioning base station deployment and inertial navigation, the system constructs a positioning compensation mechanism for areas where satellite signals are missing. The relative displacement calculation and coordinate superposition algorithm effectively solve the positioning interruption problem in scenes such as underground parking lots, ensuring the continuity and integrity of the space-time trajectory. The intelligent trigger mechanism of signal coverage detection and inertial navigation realizes seamless switching of positioning modes. This technical solution improves the physical space positioning accuracy to the sub-meter level, providing highly reliable trajectory data support for the space-time analysis of charging behavior.

[0094] In some embodiments, as described in step 104, based on the spatiotemporal weight coefficient of the charging demand and the acquired peak and valley period data of the power grid load, a user charging behavior feature vector is constructed, and similarity matching is performed on the user charging behavior feature vector to classify the user charging behavior category, including: 601. Extracting the numerical parameter representing the charging period selection tendency from the charging demand spatiotemporal weight coefficient, combining the numerical parameter with the peak and valley time nodes recorded in the peak and valley period data of the power grid load, and correspondingly combining the numerical parameter with the peak and valley time nodes to form a basic parameter set of the user's charging behavior; In step 601, the numerical parameter refers to the digital feature that quantifies the charging period selection tendency. The peak-valley time node refers to the critical moment when the power grid load changes between high and low. The basic parameter set refers to the core feature combination that describes the user's charging behavior.

[0095] In the embodiment of the present application, first, a numerical parameter (such as the weight value of a certain period is 0.7) that characterizes the user's charging period selection tendency is extracted from the charging demand spatio-temporal weight coefficient generated in step 205. At the same time, the peak-valley time nodes (such as the peak period 18:00-20:00) in the power grid load peak-valley period data provided by the power grid company are obtained. Secondly, the numerical parameter is aligned with the time axis of the peak-valley time node according to the user ID, and the two are correspondingly combined through the time window matching algorithm (such as user A has a weight of 0.7 during the evening peak period). Then, the period weight of each user is bound to the peak-valley node to form a basic parameter set including the user, the period, the weight, and the peak-valley label. Finally, the set is stored in the user behavior database to provide input for quantitative analysis.

[0096] 602. According to the corresponding relationship between the charging period selection tendency and the peak-valley time node in the basic parameter set, a quantitative score is given to each charging period selection tendency to generate user charging behavior description data; In step 602, the quantitative scoring refers to the process of converting behavioral features into numerical evaluations. The user charging behavior description data refers to the quantitative record that reflects the charging habits.

[0097] In the embodiment of the present application, first, based on the basic parameter set in step 601, the numerical values of the user's charging period selection tendency under different peak-valley time nodes are extracted (such as user A has a weight of 0.9 during the valley period). Secondly, a quantitative scoring rule is designed (such as the higher the weight of the peak period selection, the lower the score), and the tendency numerical values of each user are standardized and scored (such as user A has a weight of 0.3 during the peak period and gets 30 points). Then, the scoring results are associated with the peak-valley period labels to generate user charging behavior description data that describes the user's charging behavior pattern (such as user A "valley period preference score 85"). Finally, the data is classified and stored according to the user ID to support subsequent feature extraction.

[0098] 603. Extract the charging tendency quantization value in the user charging behavior description data, and perform a standardization process on the charging tendency quantization value to obtain a user charging behavior feature vector; In step 603, the charging tendency quantization value refers to the numerical representation of the user's charging time preference. The standardization process refers to the operation of converting data into a unified dimension. The user charging behavior feature vector refers to the multi-dimensional feature combination that describes the charging habits.

[0099] In the embodiments of the present application, first, the charging tendency quantization values of all users are extracted from the user charging behavior description data in step 602 (for example, user A gets 85 points and user B gets 60 points). Secondly, the quantization values are converted into a unified dimension (such as the range of 0 - 1) through a standardization algorithm (such as Z-Score) to eliminate the magnitude differences between users. Then, the standardized values are constructed into a multi-dimensional vector according to the user ID to form a user charging behavior feature vector (such as user A's vector [0.8, 0.2]). Finally, the feature vector is stored in the model training library for similarity matching.

[0100] 604. Calculate the difference in the charging tendency quantization values between the user charging behavior feature vectors, set the similarity determination criteria between different user behaviors based on the difference in the charging tendency quantization values, and group and match users through the similarity determination criteria; In step 604, the difference in the charging tendency quantization value refers to the degree of difference in the charging preferences between users. The similarity determination criterion refers to the criterion for judging the proximity of user behaviors. Grouping and matching refers to the process of dividing user groups according to similarity.

[0101] In the embodiments of the present application, first, the charging tendency quantization values of all users are extracted from the user charging behavior feature vectors in step 603, the Euclidean distance or cosine similarity between every two users is calculated to obtain the difference in the charging tendency quantization values (for example, the difference between user A and B is 0.3). Secondly, the similarity determination criteria are set according to the difference distribution (for example, if the difference is less than 0.2, they are regarded as the same category), and users are grouped and matched through a clustering algorithm (such as hierarchical clustering). Then, the matching results are marked as preliminary grouping labels (for example, group 1 includes users A, C, and D). Finally, the grouped data is temporarily stored in the cache waiting for threshold verification.

[0102] 605. Based on the grouping and matching results, users with a difference in the charging tendency quantization value less than the set threshold are classified into the same charging behavior category to output the classification result of the user charging behavior category.

[0103] In step 605, the set threshold refers to the maximum allowable difference value for judging similar behaviors. The charging behavior category refers to the grouping of users with similar charging habits. The classification result refers to the final classification conclusion of the user group.

[0104] In the embodiments of the present application, first, based on the grouping and matching results in step 604, the difference in the charging tendency quantization values of each group of users is extracted and compared with the preset threshold (such as the difference upper limit of 0.25). Secondly, users with a difference less than the threshold are classified into the same category (for example, the difference in group 1 is 0.15 and remains as category 1). Then, the groups exceeding the threshold are re-divided (for example, the difference in group 2 is 0.3 and is split into category 2 and 3). Finally, the classification result of the user charging behavior category including the user ID and the category label is output and synchronized to the charging scheduling system to implement personalized strategy distribution.

[0105] The following is a specific example: In the electric vehicle charging management system of a science and technology park, the user behavior clustering system accurately identifies the characteristics of charging patterns. When dynamic time-of-use electricity prices are introduced in the park, the system extracts the weight coefficients of the charging time periods of employees' vehicles (step 601). Combining the two peak periods of 12:00 - 14:00 at noon and 18:00 - 20:00 in the evening recorded by the power grid, users who frequently choose to charge at noon are marked as the "off-peak charging tendency group". A quantitative score is given to the charging time period selections of 300 users (step 602), generating behavioral description data showing the "noon charging preference index". Through standardization processing (step 603), the index is converted into a comparable feature vector, and it is found that the noon charging characteristic values of employees in the R & D center are generally higher than those of users in the administrative building. The algorithm calculates the difference degree between the feature vectors (step 604), and sets a similarity threshold to divide users into four categories: high-frequency noon charging type, evening peak-dependent type, random charging type, and ultra-low valley preference type. Based on the threshold matching results (step 605), the system identifies that there are charging pile usage conflicts among 20 high-frequency noon users and automatically triggers the "noon charging reservation + diversion guidance" strategy. During a certain power grid upgrade period, the system preferentially pushes flexible charging suggestions to random charging type users according to the classification results, successfully transferring 18% of the noon demand to off-peak periods. This portrait model is continuously optimized. A new cross-park user behavior comparison module is added, and it is found that the evening charging tendency of users in the adjacent biomedical park is significantly higher than that in the science and technology park, providing decision-making support for regional power grid coordinated scheduling and forming a closed-loop optimization mechanism from individual feature extraction to group strategy formulation.

[0106] In summary, steps 601 to 605 achieve the extraction of quantitative characteristics and intelligent classification of user charging behaviors. Through the parametric mapping of the charging time period selection tendency and the peak-valley periods of the power grid, the system constructs a standardized description system for user behavior characteristics. The combination of the quantitative scoring mechanism and the standardization processing algorithm breaks through the comparability barrier of behavior data in different dimensions. The dynamic grouping and matching technology based on the similarity of feature vectors realizes the accurate clustering of user groups and the recognition of behavior patterns. This technical solution converts a large amount of user behavior data into an operable classification label system, establishing a scientific grouping basis for personalized charging service recommendations.

[0107] In some embodiments, as described in step 105, according to the user charging behavior category, combining the spatio-temporal weight coefficient of the charging demand, the user charging time point, and the spatio-temporal trajectory coordinates to establish a user charging behavior portrait model, including: 701. Extract the charging time period preference parameters of each user charging behavior category from the user charging behavior category, and the charging time period preference parameters are determined by the historical distribution law of the charging time points; In step 701, the charging period preference parameter refers to a quantitative index that reflects the user's charging time pattern. The historical distribution pattern refers to the statistical characteristics of the charging time points that occurred in the past.

[0108] In the embodiments of the present application, first, extract the user list of each category from the user charging behavior categories output in step 605, and obtain the historical charging time point records thereof. Secondly, analyze the concentration pattern of the user charging periods within each category through a time distribution statistical method (for example, 80% of the charging of a certain category of users occurs between 18:00 and 20:00), and calculate the proportion of the occurrence frequency of charging events in each period. Then, convert the frequency proportion into a standardized parameter (such as the late peak period parameter 0.8), and generate a charging period preference parameter that characterizes the overall habit of the category. Finally, bind the parameter to the category label and store it in the user portrait database.

[0109] 702. Extract the charging location distribution parameter corresponding to the user charging behavior category according to the boundary of the dense area of the spatio-temporal trajectory coordinates; In step 702, the dense area boundary refers to the geographical range that frequently appears in the spatio-temporal trajectory coordinates. The charging location distribution parameter refers to the spatial characteristics of the user's selected charging location.

[0110] In the embodiments of the present application, first, based on the spatio-temporal trajectory coordinates generated in step 505, extract the set of trajectory points corresponding to the user charging behavior category, and identify the dense area boundary through a kernel density estimation algorithm (for example, the concentration of coordinate points in a certain area exceeds the threshold). Secondly, count the occurrence times and distribution range of charging events in each dense area (for example, the proportion of charging events within a certain boundary is 60%), and generate a charging location distribution parameter that reflects the position aggregation intensity (such as the distribution parameter of area A is 0.7). Then, store the parameter according to the category - area dimension for subsequent correlation analysis.

[0111] 703. Perform superposition calculation on the charging demand spatio-temporal weight coefficient and the charging time point, and generate a weight distribution parameter that reflects the user's weight of choosing to charge at different times according to the superposition result; In step 703, the superposition calculation refers to the operation process of weighted fusion of different parameters. The weight distribution parameter refers to a quantitative index that reflects the importance of period selection.

[0112] In the embodiments of the present application, first, extract the weight values of the user at different times from the charging demand spatio-temporal weight coefficient in step 205 (for example, the weight of user A in the morning peak is 0.6), and align it with the charging time point records in step 701 according to the time window. Secondly, fuse the weight value and the charging time point frequency in the same period through a weighted superposition algorithm (for example, the weight 0.6 and the frequency 10 times are superposed to 6.0), and generate a weight distribution parameter that characterizes the user's period selection priority. Then, store the parameter according to the user - period dimension to support dynamic correlation mapping.

[0113] 704. Perform a position - associated mapping on the weight distribution parameter and the charging location distribution parameter to establish a dynamic association relationship between the charging period selection tendency and the charging location distribution; In step 704, the position - associated mapping refers to establishing the corresponding relationship between period selection and spatial distribution. The dynamic association relationship refers to the interaction mode among parameters that change over time.

[0114] In the embodiment of the present application, first, match the weight distribution parameter in step 703 (such as the weight of user A during the evening peak is 0.8) with the charging location distribution parameter in step 702 (such as the distribution parameter of area A is 0.7) according to geographical regions and periods. Secondly, calculate the coordination degree of the weight and the position parameter through a spatial association algorithm (such as the matching degree between the weight during the evening peak in area A and the position parameter is 0.75), and establish a dynamic association relationship (such as "users tend to choose area A during the evening peak period"). Then, store the association relationship in the relational database to support the adjustment of the combination ratio.

[0115] 705. Adjust the combination ratio of the charging period preference parameter and the charging location distribution parameter in the user charging behavior category according to the matching degree between the charging period selection tendency and the charging location distribution in the dynamic association relationship; In step 705, the matching degree refers to the strength of the consistency between period selection and position distribution. The combination ratio refers to the contribution weights of different parameters in the model.

[0116] In the embodiment of the present application, first, analyze the matching degree between the charging period selection tendency and the charging location distribution in the dynamic association relationship of step 704 (such as the matching degree of a certain category is 80%). Secondly, if the matching degree is lower than the preset threshold (such as 70%), then reduce the weight of the charging period preference parameter in this category proportionally (such as from 0.8 to 0.6), and at the same time increase the weight of the charging location distribution parameter (such as from 0.7 to 0.9). Then, perform iterative adjustment until the matching degree meets the standard to form an optimized parameter combination ratio. Finally, update the adjustment result to the user portrait database.

[0117] 706. Integrate the charging period preference parameter, the charging location distribution parameter, and the weight distribution parameter based on the adjusted combination ratio to construct the user charging behavior portrait model.

[0118] In step 706, the user charging behavior portrait model refers to a data model that comprehensively describes the charging characteristics of users. Integration refers to the process of systematically integrating multi - source parameters.

[0119] In the embodiments of the present application, first, the adjusted charging period preference parameters, charging location distribution parameters in step 705, and the weight distribution parameters in step 703 are integrated to construct a multi-dimensional feature matrix according to the user ID (such as period parameter 0.6, location parameter 0.9, weight 0.8). Secondly, the multi-dimensional matrix is reduced to an interpretable label system through a feature fusion algorithm (such as principal component analysis) to train a user charging behavior portrait model (such as a decision tree model). Then, the portrait labels output by the model (such as "peak evening area A preference type") are deployed to the charging recommendation system to complete the closed-loop modeling process.

[0120] The following is a specific example: In the electric vehicle charging management system in the university town, the user behavior portrait system accurately depicts the charging patterns of teachers and students. When the system analyzes the charging records of the faculty group, it extracts the preference parameters during the double-peak periods of 07:00 - 08:30 in the morning and 12:00 - 13:00 at noon (step 701), and simultaneously identifies the geographical boundaries where the charging hotspots are densely distributed in the parking lot of the teachers' apartment and the charging area of the experimental building (step 702). The algorithm superimposes the spatio-temporal weight coefficient of charging in the teaching area at noon on the time axis (step 703) to generate distribution parameters showing "a sharp increase in the charging weight of the experimental building at noon". Through correlation analysis, it is found that 90% of the high-weight charging behaviors at noon occur in the fast charging pile cluster on the west side of the experimental building (step 704). However, when some charging piles are under repair during the winter vacation, the system detects that the charging location during this period spreads to the liberal arts building area, and dynamically adjusts the proportion of the original area's location distribution parameters (step 705). Finally, the optimized period preference, location distribution, and dynamic weight parameters are integrated (step 706) to construct five types of fine portraits such as "faculty experimental building noon fast charging type" and "cross-campus mobile charging type". After the start of the spring semester, the system accurately predicts the peak usage period of the newly established charging area in the art building based on the updated model, coordinates the grid expansion two weeks in advance, and identifies through portrait differences that 10% of the newly recruited teachers have abnormal charging location offsets, triggering the vehicle usage specification training mechanism. The model continuously integrates the parameter changes during special periods such as winter and summer vacations to form a closed-loop optimization system from basic feature extraction to dynamic portrait calibration.

[0121] In summary, steps 701 to 706 achieve the dynamic association of multi-dimensional charging behavior characteristics and the optimization of the portrait model. Through the correlation mapping of the charging period preference parameters and the location distribution parameters, the system constructs an analysis framework for the spatio-temporal selection mode of user charging behavior. The dynamic superposition mechanism of the weight distribution parameters reveals the spatio-temporal priority law in the user's charging decision-making. The combined ratio adjustment algorithm optimizes the weight of the feature parameters according to the actual matching degree, ensuring the adaptive evolution ability of the portrait model. This technical solution realizes the intelligent sublimation from discrete behavior characteristics to a three-dimensional portrait model, providing a high-precision user behavior deduction platform for charging demand prediction and grid coordinated scheduling.

[0122] Figure 2 The following is a schematic structural diagram of a modeling system for a user's charging behavior portrait provided by an embodiment of the present application. As Figure 2 shown, the system includes: A collection module 21, configured to collect the user's charging time point, driving mileage, and battery capacity attenuation data, generate a regional charging demand heat map by analyzing the spatial distribution of the charging time point, the driving mileage, and the obtained vehicle driving trajectory, and at the same time fuse the power consumption data obtained through the vehicle interface with the battery capacity attenuation data to generate an energy consumption feature; A deployment module 22, configured to deploy differential positioning base stations within the coverage range corresponding to the regional charging demand heat map, and when the vehicle enters an underground parking lot with missing satellite signals, trigger the on-vehicle inertial navigation function to perform real-time compensation on the positioning coordinates recorded by the differential positioning base stations to generate spatio-temporal trajectory coordinates that are continuously connected to the positioning coordinates; An analysis module 23, configured to perform joint analysis on the energy consumption feature and the spatio-temporal trajectory coordinates to identify the correlation between the vehicle charging period, the distribution density of charging piles, and the peak-valley period of the regional power grid load, and convert the correlation into a charging demand spatio-temporal weight coefficient through a deep reinforcement learning model; A construction module 24, configured to construct a user charging behavior feature vector based on the charging demand spatio-temporal weight coefficient and the obtained power grid load peak-valley period data, and perform similarity matching on the user charging behavior feature vector to classify user charging behavior categories; An establishment module 25, configured to combine and associate the charging demand spatio-temporal weight coefficient, the charging time point, and the spatio-temporal trajectory coordinates according to the user charging behavior category to establish a user charging behavior portrait model.

[0123] Figure 2 The described modeling system for a user's charging behavior portrait can execute Figure 1 the modeling method for a user's charging behavior portrait described in the embodiment shown. The implementation principle and technical effects will not be elaborated. For the modeling system for a user's charging behavior portrait in the above embodiment, the specific manners in which each module and unit perform operations have been described in detail in the embodiment related to the method, and will not be elaborated here.

[0124] In a possible design, Figure 2 the modeling system for a user's charging behavior portrait in the embodiment shown can be implemented as a computing device. As Figure 3 shown, the computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, where the one or more computer instructions are called and executed by the processing component 32.

[0125] The processing component 32 is used for the above Figure 1 A method for modeling a user charging behavior portrait in the above-described embodiment.

[0126] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0127] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0128] Of course, the computing device may also necessarily include other components, such as input / output interfaces, display components, communication components, etc.

[0129] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0130] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0131] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above processing component, storage component, etc. may be basic server resources leased or purchased from a cloud computing platform.

[0132] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above Figure 1 A method for modeling a user charging behavior portrait in the above-described embodiment.

[0133] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0134] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative work.

[0135] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0136] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A modeling method for user charging behavior portrait, characterized in that: include: Collecting the user's charging time point, driving mileage and battery capacity decay data, generating a regional charging demand heat map by analyzing the spatial distribution of the charging time point, the driving mileage and the acquired vehicle driving trajectory, and integrating the power consumption data obtained through the vehicle interface with the battery capacity decay data to generate energy consumption characteristics; Deploy differential positioning base stations within the coverage area corresponding to the regional charging demand heat map. When a vehicle enters an underground parking lot where satellite signals are missing, the vehicle-mounted inertial navigation function is triggered to perform real-time compensation for the positioning coordinates recorded by the differential positioning base station, generating space-time trajectory coordinates that are continuously connected with the positioning coordinates. The energy consumption characteristics are jointly analyzed with the spatiotemporal trajectory coordinates to identify the correlation between the vehicle charging period and the distribution density of charging piles and the peak and valley periods of regional power grid load, and the correlation is converted into a spatiotemporal weight coefficient of charging demand through a deep reinforcement learning model; Based on the spatiotemporal weight coefficient of the charging demand and the acquired peak and valley period data of the power grid load, construct a user charging behavior feature vector, and perform similarity matching on the user charging behavior feature vector to classify the user charging behavior category; According to the user charging behavior category, the charging demand spatiotemporal weight coefficient, the charging time point and the spatiotemporal trajectory coordinates are combined and associated to establish a user charging behavior portrait model.

2. The method according to claim 1, characterized in that The energy consumption characteristics are jointly analyzed with the spatiotemporal trajectory coordinates to identify the correlation between the vehicle charging period and the distribution density of charging piles and the peak and valley periods of regional power grid load, and the correlation is converted into the spatiotemporal weight coefficient of charging demand through a deep reinforcement learning model, including: Performing time window matching of the unit time electric energy consumption rate in the energy consumption feature and the position movement rate in the space-time trajectory coordinates to generate a matching result reflecting the corresponding relationship between the position change of the vehicle during the charging period and the electric energy consumption rate; According to the vehicle position change in the matching result, extract the regional coverage characteristics of the charging pile distribution density in the corresponding charging period, and associate them with the starting time point of the peak and valley period of the power grid load to generate dynamic association parameters; Analyze the change of charging pile distribution density in the dynamic correlation parameters to obtain the coupling relationship describing the charging period and the charging pile distribution density, and simultaneously analyze the fluctuation of the power grid load during the peak and valley periods to obtain the fluctuation characteristics of the power grid load during the peak and valley periods, wherein the coupling relationship represents the tendency of charging period selection, and the fluctuation characteristics represent the load response capability of the power grid; Performing an interaction analysis on the charging period selection tendency and the grid load response capability to generate correlation relationship parameters; Through the training of the deep reinforcement learning model, the association relationship parameters are converted into the spatiotemporal weight coefficients of the charging demand.

3. The method according to claim 2, characterized in that Analyze the change of charging pile distribution density in the dynamic correlation parameter to obtain the coupling relationship describing the charging period and the charging pile distribution density, and simultaneously analyze the fluctuation of the peak and valley period of the power grid load to obtain the fluctuation characteristics of the peak and valley period of the power grid load. The coupling relationship represents the tendency of charging period selection, and the fluctuation characteristics represent the response capability of the power grid load, including: The data of the change of the distribution density of charging piles in the dynamic correlation parameter is divided into segments according to the charging period, and the density difference of the distribution density of charging piles in each charging period is counted to generate coupling relationship data describing the change of the charging period and the distribution density of charging piles; Synchronously extracting the load value change data of the power grid load peak and valley periods in the dynamic associated parameters, and calculating the change amplitude of the load value in each power grid load peak and valley period to generate the fluctuation characteristic data of the power grid load peak and valley period; The density difference of the charging pile distribution density in the coupling relationship data represents the user's charging pile selection tendency during the charging period; The variation range of the load value in the fluctuation characteristic data represents the charging demand response capability of the power grid during peak and valley periods.

4. The method according to claim 1, characterized in that A regional charging demand heat map is generated by analyzing the spatial distribution of the charging time point, the driving mileage and the acquired vehicle driving trajectory, and the power consumption data acquired through the vehicle interface is integrated with the battery capacity decay data to generate energy consumption characteristics, including: Obtaining the recorded time of the charging time point and the geographic coordinates of the vehicle's driving trajectory, mapping each charging time point to a preset grid of geographic coordinates, and counting the number of occurrences of the charging time point in each grid to generate a charging grid; According to the driving mileage, the total driving mileage of the vehicle in the coverage area of ​​each charging grid is calculated, and the total driving mileage is superimposed on the number of occurrences of the charging time point of the corresponding grid, and the charging demand density grid is generated through the superimposed values; Constructing a regional charging demand heat map based on the superimposed values ​​of each grid in the charging demand density grid; Obtaining the electric energy consumption data per unit time of the vehicle before and after the charging time point through the vehicle interface, and superimposing and calculating the electric energy consumption data with the capacity loss ratio in the battery capacity decay data to obtain the energy consumption parameter; The charging demand density of each grid in the regional charging demand heat map is calculated, and the charging demand density is associated and mapped with the energy consumption characteristic parameter to output the energy consumption characteristic.

5. The method according to claim 1, characterized in that Deploy differential positioning base stations within the coverage area corresponding to the regional charging demand heat map. When a vehicle enters an underground parking lot where satellite signals are missing, trigger the vehicle-mounted inertial navigation function to perform real-time compensation on the positioning coordinates recorded by the differential positioning base station, and generate space-time trajectory coordinates that are continuously connected with the positioning coordinates, including: Within the coverage of the regional charging demand heat map, differential positioning base stations are deployed according to the location boundaries of the charging demand-intensive areas in the regional charging demand heat map, and the signal coverage range of each differential positioning base station is determined; When a vehicle enters an underground parking lot where satellite signals are missing, the vehicle-mounted signal detection device identifies whether the vehicle position is out of the signal coverage range, marks the area out of the signal coverage range as a signal missing area, and triggers the vehicle-mounted inertial navigation function to start; The vehicle's moving direction and speed data in the signal loss area are collected in real time through the vehicle's inertial navigation function, and the relative displacement of the vehicle in the signal loss area is calculated in combination with the positioning coordinates recorded by the differential positioning base station; The relative displacement is superimposed with the positioning coordinates recorded by the differential positioning base station to generate the compensated positioning coordinates of the vehicle in the signal missing area, and the compensated positioning coordinates are connected with the positioning coordinates recorded by the differential positioning base station within the signal coverage range in chronological order to obtain a positioning coordinate sequence; According to the positioning coordinate sequence, the complete moving trajectory of the vehicle from the signal coverage area to the signal loss area is extracted to generate the spatiotemporal trajectory coordinates that are continuously connected with the positioning coordinates recorded by the differential positioning base station.

6. The method according to claim 1, characterized in that Based on the spatiotemporal weight coefficient of the charging demand and the acquired peak and valley period data of the power grid load, a user charging behavior feature vector is constructed, and similarity matching is performed on the user charging behavior feature vector to classify the user charging behavior category, including: Extracting the numerical parameter representing the charging period selection tendency from the charging demand spatiotemporal weight coefficient, combining the numerical parameter with the peak and valley time nodes recorded in the peak and valley period data of the power grid load, and correspondingly combining the numerical parameter with the peak and valley time nodes to form a basic parameter set of the user's charging behavior; According to the corresponding relationship between the charging period selection tendency and the peak and valley time nodes in the basic parameter set, each charging period selection tendency is quantitatively scored to generate user charging behavior description data; Extracting a charging tendency quantified value from the user charging behavior description data, and performing standardization processing on the charging tendency quantified value to obtain a user charging behavior feature vector; Calculating the charging tendency quantization value difference between the charging behavior feature vectors of the users, setting a similarity determination standard between different user behaviors based on the charging tendency quantization value difference, and grouping and matching the users according to the similarity determination standard; Based on the group matching results, users whose charging tendency quantization value differences are less than a set threshold are classified into the same charging behavior category to output the user charging behavior category classification results.

7. The method according to claim 1, characterized in that According to the user charging behavior category, the charging demand spatiotemporal weight coefficient, the user charging time point and the spatiotemporal trajectory coordinates are combined and associated to establish a user charging behavior portrait model, including: Extracting a charging period preference parameter of each user charging behavior category from the user charging behavior category, wherein the charging period preference parameter is determined by a historical distribution law of charging time points; Extracting charging location distribution parameters corresponding to the user charging behavior category according to the dense area boundary of the spatiotemporal trajectory coordinates; The charging demand spatiotemporal weight coefficient is superimposed on the charging time point, and a weight distribution parameter reflecting the user's choice of charging at different time periods is generated according to the superposition result; Performing position association mapping on the weight distribution parameter and the charging location distribution parameter to establish a dynamic association relationship between the charging time period selection tendency and the charging location distribution; According to the matching degree between the charging time period selection tendency and the charging location distribution in the dynamic association relationship, adjusting the combination ratio of the charging time period preference parameter and the charging location distribution parameter in the user charging behavior category; The charging period preference parameter, charging location distribution parameter and weight distribution parameter are integrated based on the adjusted combination ratio to construct the user charging behavior portrait model.

8. A modeling system for user charging behavior portrait, characterized in that: include: A collection module is used to collect the user's charging time point, driving mileage and battery capacity decay data, generate a regional charging demand heat map by analyzing the spatial distribution of the charging time point, the driving mileage and the acquired vehicle driving trajectory, and integrate the power consumption data obtained through the vehicle interface with the battery capacity decay data to generate energy consumption characteristics; A deployment module is used to deploy differential positioning base stations within the coverage area corresponding to the regional charging demand heat map. When a vehicle enters an underground parking lot where satellite signals are missing, the vehicle-mounted inertial navigation function is triggered to perform real-time compensation for the positioning coordinates recorded by the differential positioning base station, and generate space-time trajectory coordinates that are continuously connected with the positioning coordinates. An analysis module is used to jointly analyze the energy consumption characteristics and the spatiotemporal trajectory coordinates to identify the correlation between the vehicle charging period and the charging pile distribution density and the peak and valley periods of the regional power grid load, and convert the correlation into a spatiotemporal weight coefficient of the charging demand through a deep reinforcement learning model; A construction module, used to construct a user charging behavior feature vector based on the charging demand spatiotemporal weight coefficient and the acquired power grid load peak and valley time period data, and perform similarity matching on the user charging behavior feature vector to classify the user charging behavior category; A module is established, which is used to combine and associate the charging demand spatiotemporal weight coefficient, the charging time point and the spatiotemporal trajectory coordinates according to the user charging behavior category, so as to establish a user charging behavior portrait model.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a modeling method for a user charging behavior portrait as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a modeling method for a user charging behavior portrait as described in any one of claims 1 to 7 is implemented.

Images