User charging behavior multi-dimensional feature space-time quantification method considering cross-domain factors

Abstract

The invention discloses a user charging behavior multi-dimensional feature space-time quantification method considering cross-domain factors, which objectively quantifies the charging behavior of a user based on user travel rules, charging service level, power grid price level, traffic network data and meteorological factors of the location. According to the method, the influence of the charging behavior of the user on surrounding road traffic is put forward, the blank in the field is made up, the influence of the charging behavior on the area where the charging behavior is located is fully considered, a traditional single evaluation dimension is exceeded, basic capability construction of vehicle-road-network collaborative optimization can be well supported, and the method has good application prospects. The scientificity and precision of the charging and battery swap infrastructure planning are improved, the user behavior modeling theory is enriched, and cross-field technology fusion is promoted.

Description

TECHNICAL FIELD

[0001] The present application relates to the fields of traffic and energy integration and user behavior analysis, and in particular to a method for spatiotemporal quantification of multi-dimensional features of user charging and swapping behavior considering cross-field factors. BACKGROUND

[0002] With the continuous growth of the number of electric vehicles, their charging and swapping behavior has become a key link in the coupling regulation of the traffic system and the power system. Under the background of the "double carbon" goal and the construction of smart cities, electric vehicles, as mobile energy storage units, not only affect the spatiotemporal distribution of power grid load, but also are closely related to traffic efficiency, infrastructure layout, and environmental factors. In related technologies, through the coordinated operation of travel trajectory analysis, charging facility state monitoring, and power grid dispatching strategies, a preliminary correlation model of user behavior and system operation is constructed. Specifically, this technical system covers the entire process from user decision mechanism to charging and swapping behavior prediction, including vehicle state identification, charging service evaluation, traffic congestion response, price sensitivity analysis, and weather influence modeling, and preliminarily realizes multi-dimensional perception of user behavior and partial system coordination.

[0003] However, in existing user charging and swapping behavior modeling methods, single-dimensional data such as user preferences, charging convenience, or economic efficiency are directly used for analysis, and key cross-field factors such as "vehicle-pile-road-grid-weather" are not systematically integrated, which may result in insufficient behavior characterization dimensions and difficulty in accurately reflecting the user's real decision-making mechanism in complex environments. Specifically, existing technologies usually construct models based on static or local data, lack coordinated modeling of dynamic factors such as traffic congestion index, power grid time-of-use price, charging pile queuing time, and extreme weather probability, thereby affecting the spatiotemporal accuracy and system response capability of behavior prediction. Among them, although vehicle state and charging facility service level are considered by some studies, the coupling relationship between them and the traffic network, power grid dispatching, and weather conditions has not been effectively quantified, limiting the depth and breadth of traffic and energy system collaborative optimization. SUMMARY

[0004] The present application aims to at least partially solve one of the technical problems in the related art.

[0005] The present application proposes a method for spatiotemporal quantification of multi-dimensional features of user charging and swapping behavior considering cross-field factors.

[0006] Another object of the present application is to propose a device for spatiotemporal quantification of multi-dimensional features of user charging and swapping behavior considering cross-field factors.

[0007] To achieve the above-mentioned purposes, the present application proposes, in one aspect, a method for spatiotemporal quantification of multi-dimensional features of user charging and swapping behavior considering cross-field factors, comprising:

[0008] S1, a multi-dimensional behavior characteristic index system including vehicle state, charging facility service level, traffic operating condition, power grid price level and meteorological factor is constructed, and influence of each dimension on user charging and battery swapping behavior is quantitatively represented;

[0009] S2, based on the multi-dimensional behavior characteristic index system, user travel trajectory, charging pile operation data, traffic congestion index, time-of-use electricity price information and meteorological observation data are fused, and standardized processing and space-time alignment of multi-source heterogeneous data are performed;

[0010] S3, a multi-item logistic regression model and a nonlinear least square method are used to calibrate parameters and model probability of user selecting charging and battery swapping in different time periods and charging station selection probability, so that dynamic prediction of user behavior is realized;

[0011] S4, a cross-domain collaborative modeling framework is constructed in combination with user historical behavior data and real-time environment data, and evolution trend of user charging and battery swapping behavior in time and space is simulated.

[0012] The multi-dimensional characteristic space-time quantification method for user charging and battery swapping behavior considering cross-domain factors in the embodiment of the application can also have the following additional technical features:

[0013] In an embodiment of the application, the S1 comprises:

[0014] S11, vehicle state parameters including battery capacity, current remaining power, acceptable minimum charge threshold and whether supporting battery swapping are defined, so as to quantitatively represent urgency of energy supplement of user in different power states;

[0015] S12, charging facility service level parameters including rated charging power of charging pile, queue length and queue time are constructed, so as to reflect sensitivity of user to charging efficiency and service accessibility when selecting charging facility.

[0016] In an embodiment of the application, the S2 comprises:

[0017] S21, a time stamp alignment algorithm is used to match user travel trajectory data and power grid time-of-use electricity price data in time dimension, so as to ensure synchronization of different time domain data in behavior modeling;

[0018] S22, a spatial interpolation method is used to perform standardized processing of traffic congestion index and meteorological observation data in spatial dimension, so as to unify expression forms of data in different geographical positions.

[0019] In an embodiment of the application, the S3 comprises:

[0020] S31, the regression coefficients and intercept terms in the multinomial logistic regression model are fitted by using a nonlinear least squares method to improve the accuracy of the model in describing the user behavior response characteristics;

[0021] S32, based on the user historical charging and battery swapping behavior data, real-time electricity price, traffic congestion, and queuing time, the model parameters are dynamically updated to adapt to the evolution trend of user behavior.

[0022] In an embodiment of the present application, the S4 comprises:

[0023] S41, a travel chain path accessibility analysis module is introduced to evaluate the path feasibility and traffic efficiency of the user from the current location to each candidate charging station;

[0024] S42, a charging station selection model based on a utility function is constructed, wherein the utility function includes distance, electricity price, queuing time, traffic congestion index, and weighted coefficients of each influencing factor.

[0025] In an embodiment of the present application, it further comprises:

[0026] S5, based on the cross-domain collaborative modeling framework, a spatio-temporal heat map of user charging and battery swapping behavior is generated to predict the charging and battery swapping demand density and power grid load change trend in each region in a specific time period.

[0027] To achieve the above purpose, another aspect of the present application provides a user charging and battery swapping behavior multi-dimensional feature spatio-temporal quantification device considering cross-domain factors, comprising:

[0028] A multi-dimensional feature construction module is used to construct a multi-dimensional behavior feature index system including vehicle state, charging facility service level, traffic operating condition, power grid electricity price level, and meteorological factor to quantify the influence of each dimension on user charging and battery swapping behavior.

[0029] A data fusion and spatio-temporal alignment module is used to fuse user travel trajectory, charging pile operation data, traffic congestion index, time-of-use electricity price information, and meteorological observation data based on the multi-dimensional behavior feature index system to perform standardized processing and spatio-temporal alignment of multi-source heterogeneous data.

[0030] A behavior probability modeling module is used to use a multinomial logistic regression model and a nonlinear least squares method to calibrate and model the probability of user selecting charging and battery swapping in different time periods and the probability of charging station selection to realize dynamic prediction of user behavior.

[0031] A cross-domain collaborative modeling module is used to combine user historical behavior data and real-time environmental data to construct a cross-domain collaborative modeling framework to simulate the evolution trend of user charging and battery swapping behavior in time and space.

[0032] The user charging and battery swapping behavior multi-dimensional feature space-time quantification method and device considering cross-domain factors can systematically fuse multi-source heterogeneous data of'vehicle-pile-road-network-weather', accurately quantify and dynamically model user charging and battery swapping behavior in the space-time dimension, and improve the behavior prediction accuracy and the collaborative optimization capability of the transportation and energy system.

[0033] Additional aspects and advantages of the present application will be set forth in part in the description that follows, and in part will become apparent to those having ordinary skill in the art upon examination of the following or can be learned from practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0034] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description, taken in conjunction with the accompanying drawings, in which:

[0035] Figure 1 is a flowchart of the user charging and battery swapping behavior multi-dimensional feature space-time quantification method according to an embodiment of the present application;

[0036] Figure 2 is another flowchart of the user charging and battery swapping behavior multi-dimensional feature space-time quantification method according to an embodiment of the present application;

[0037] Figure 3 is a structural diagram of the user charging and battery swapping behavior multi-dimensional feature space-time quantification device according to an embodiment of the present application. DETAILED DESCRIPTION

[0038] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0039] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0040] The user charging and battery swapping behavior multi-dimensional feature space-time quantification method and device considering cross-domain factors proposed according to the embodiments of the present application will be described below with reference to the accompanying drawings.

[0041] Embodiment 1

[0042] Figure 1 is a flowchart of the user charging and battery swapping behavior multi-dimensional feature space-time quantification method according to an embodiment of the present application, as Figure 1as shown, including:

[0043] S1, a multi-dimensional behavior characteristic index system is constructed, including vehicle state, charging facility service level, traffic operating condition, power grid price level and meteorological factor, which quantitatively represents the influence of each dimension on user charging and battery swapping behavior.

[0044] Specifically, this step aims to construct a multi-dimensional behavior characteristic index system covering vehicle state, charging facility service level, traffic operating condition, power grid price level and meteorological factor, to quantify the influence of each dimension on user charging and battery swapping behavior. By systematically integrating the five key elements of "vehicle-pile-road-grid-weather", this system realizes multi-dimensional and spatio-temporal dynamic modeling of user behavior, thereby improving the accuracy of behavior prediction and decision support capability.

[0045] This step first defines and extracts key indicators of each dimension. For example, the vehicle state dimension includes battery capacity (Ah), current state of charge (SOC, percentage), acceptable minimum state of charge threshold (SOC_min, usually set to 15%-20%), and whether battery swapping is supported (binary variable, 1 for not supported, 2 for supported). The charging facility service level is quantified by charging pile rated power (kW), queue length (vehicles) and queue time (minutes), etc. The queue time can be calculated based on queuing theory models such as the M / M / n model, which derives the system idle probability, average queue length and average queue time based on arrival rate (λ, vehicles / hour) and service rate (μ, vehicles / hour).

[0046] In terms of traffic operating condition, traffic congestion coefficient (TTI) is used as the core indicator, which is defined as the ratio of peak period travel time to off-peak period travel time, to represent the accessibility and travel cost of users going to charging stations. The power grid price level is modeled by the time-of-use (TOU) mechanism, introducing the price sensitivity index to calculate the change in user charging behavior at different price periods, to reveal their economic driving characteristics. In terms of meteorological factors, an extreme weather occurrence probability model is constructed, using a logistic regression method with input variables including temperature (℃), humidity (%RH), etc., and output as the occurrence probability (0-1 interval) of weather types such as rainstorm and typhoon.

[0047] This step provides a structured and standardized data foundation for subsequent behavior modeling and strategy optimization by constructing a multi-dimensional index system, realizing collaborative analysis of cross-domain factors, significantly improving the quantification accuracy and model generalization ability of user charging and battery swapping behavior, and has important engineering application value and theoretical research significance.

[0048] Further, S1 includes:

[0049] S11 defines vehicle status parameters including battery capacity, current remaining charge, minimum acceptable charge threshold, and whether battery swapping is supported, in order to quantify the urgency of recharging for users in different charge states.

[0050] Specifically, this step aims to quantitatively assess the urgency of recharging for users under different battery charge states by defining vehicle state parameters (including battery capacity, current remaining charge, acceptable minimum state of charge threshold SOC_min, and whether battery swapping is supported). This step is a key input module in the entire multidimensional feature spatiotemporal quantification method, providing fundamental data support for subsequent charging and swapping behavior modeling and decision support.

[0051] The collection and modeling of vehicle status parameters requires combining onboard OBD data, user historical charging records, and a vehicle type database. Battery capacity is typically measured in ampere-hours (Ah) or kilowatt-hours (kWh), reflecting the vehicle's maximum energy storage capacity. State of Charge (SOC) is expressed as a percentage, such as 80% SOC indicating 80% battery capacity remaining. The acceptable minimum charge threshold (SOC_min) is the lowest tolerable charge level for the user, typically set between 15% and 20% SOC to avoid operational interruptions due to insufficient charge during driving. Whether battery swapping is supported is encoded using a binary variable, such as "1" indicating no swapping and "2" indicating support. This parameter directly affects the user's choice of charging methods and response speed.

[0052] The SOC_min setting needs to be dynamically adjusted based on the vehicle's range and the user's travel habits. For example, for a user with an average daily mileage of 50km and a vehicle range of 300km, SOC_min can be set to 15%; while for users who travel frequently and over long distances, SOC_min may need to be increased to 25% to ensure travel continuity. Furthermore, the difference between SOC and SOC_min (ΔSOC) can serve as a key indicator of the urgency of refueling; the smaller the ΔSOC, the stronger the user's need for refueling.

[0053] This step is widely applicable to scenarios such as electric vehicle user behavior prediction, charging load scheduling, and battery swapping station site optimization. For example, in urban traffic management systems, by monitoring the difference between a user's State of Charge (SOC) and SOC_min in real time, their charging demand at different road sections and time periods can be dynamically predicted, thereby optimizing charging resource allocation and traffic guidance strategies.

[0054] S12, constructing a charging facility service level parameter including a rated charging power of a charging pile, a queue length and a queue time to reflect the sensitivity of a user to charging efficiency and service accessibility when selecting a charging facility.

[0055] Specifically, this step aims to construct a charging facility service level parameter, specifically including a rated charging power of a charging pile, a queue length and a queue time to quantify the sensitivity of a user to charging efficiency and service accessibility when selecting a charging facility. In some implementations, this step is based on queuing theory and service system modeling principles, combined with charging station operation data and user behavior characteristics, to construct a multi-dimensional evaluation index system, thereby realizing dynamic evaluation of the service performance of a charging facility.

[0056] The rated charging power (unit: kW) of a charging pile is a core parameter for measuring its service capability, which is usually set according to the type (such as fast charging, slow charging) and equipment specifications of the charging pile. For example, the power range of a fast charging pile is 60-360 kW, and that of a slow charging pile is 7-21 kW. The queue length (unit: vehicles) represents the number of vehicles currently waiting to charge in the charging station, which can be collected by a real-time monitoring system or predicted based on historical data. The queue time (unit: minutes) reflects the average waiting time of a user at a charging station, which depends on the arrival rate (λ, unit: vehicles / hour) and service rate (μ, unit: vehicles / hour), and is analyzed for system stability in combination with the number of service counters (n).

[0057] This step can be widely applied in urban charging facility planning, charging station operation optimization and user behavior prediction systems. For example, during peak traffic periods or holidays, the queue length and time of a charging station significantly increase, and the sensitivity of users to charging efficiency also rises, at which time the system can dynamically adjust charging strategies or guide users to alternative sites to improve overall service experience.

[0058] By introducing quantitative indicators, the service capability of a charging facility and user perception can be accurately reflected, providing key support for subsequent user behavior modeling and cross-system collaborative optimization, and improving the intelligent level and operation efficiency of a charging service system.

[0059] The present application first systematically introduces the five key elements of "vehicle-pile-road-network-weather", integrates user travel trajectory, charging pile status, electricity price curve, traffic flow and weather factors, and realizes unified modeling of multi-source heterogeneous data. It breaks through the limitations of traditional single-dimensional modeling, realizes comprehensive quantification, dynamic analysis and spatio-temporal prediction of user charging / swapping behavior, and improves modeling accuracy.

[0060] S2, based on the multi-dimensional behavior characteristic index system, integrating user travel trajectory, charging pile operation data, traffic congestion index, time-of-use electricity price information and weather observation data, performing standardized processing and spatio-temporal alignment of multi-source heterogeneous data.

[0061] The traffic congestion index generally adopts the real-time traffic index provided by Baidu or Gaode Map, with a value range of 0-10, and the higher the value, the more congested the road. The time-of-use electricity price information is based on the time-of-use electricity price standard published by the State Grid or local power companies (such as the peak-valley electricity price ratio of 1.6:1:0.4), and is mapped into time series data. Meteorological data includes temperature, rainfall, wind speed, etc., usually sourced from the China Meteorological Administration or OpenWeatherMap API, and needs to be timestamped and spatially interpolated to match the spatio-temporal resolution of user behavior data.

[0062] Further, this step accurately associates user travel trajectories with charging behavior events in the time and space dimensions through a spatio-temporal alignment algorithm (such as a time window-based sliding matching or a GIS-based spatial buffer matching). For example, set the time window to 15 minutes and the spatial buffer to 500 meters to identify the user's charging intention at a specific time and place. Through the above processing, a unified spatio-temporal feature vector can be constructed, providing a high-quality, structured data foundation for subsequent user behavior modeling and decision analysis, significantly improving the generalization ability and prediction accuracy of the model.

[0063] Further, S2 includes:

[0064] S21, using a timestamp alignment algorithm to match user travel trajectory data with power grid time-of-use electricity price data in the time dimension, ensuring the synchronization of different time domain data in behavior modeling.

[0065] Specifically, this step uses a timestamp alignment algorithm to accurately match user travel trajectory data with power grid time-of-use electricity price data in the time dimension, to ensure the synchronization and consistency of traffic behavior and energy consumption behavior in the time series during behavior modeling. This step is a key link in the realization of "car-pole-road-grid-weather" multi-dimensional feature fusion modeling, and its technical implementation is based on interpolation, resampling, and time window matching strategies for time series data.

[0066] In some implementations, the user travel trajectory data is usually in the form of a sequence of spatio-temporal points recorded by GPS or OBU (on-board unit), with millisecond-level timestamp accuracy and a sampling frequency of 1 second to 10 seconds. The time-of-use electricity price data is released at fixed time intervals (such as 15 minutes, 30 minutes, or 1 hour), with a relatively low time resolution. To align the two in the time dimension, the present application uses an interpolation alignment algorithm based on a sliding time window to map each time point in the user trajectory to the nearest electricity price period, and performs linear interpolation or nearest neighbor interpolation according to the time difference to obtain the electricity value at that time. For example, if the user charges at 14:07, and the electricity price data is divided into 15-minute units, the time point can be mapped to the 14:00-14:15 period, and interpolation processing is performed according to the time offset (7 minutes).

[0067] Further, to improve alignment accuracy, a timestamp calibration mechanism can be introduced to remove or correct abnormal data through timestamp consistency checks (such as timestamp continuity and sampling interval stability). At the same time, a time tolerance threshold (such as ± 30 seconds) is set to deal with data collection errors or system delay problems.

[0068] This step is widely applicable in practical applications such as user behavior-based charging and battery swapping prediction, load scheduling optimization, and traffic-energy collaborative control scenarios. Through timestamp alignment, the user's charging behavior response under different electricity price periods can be effectively identified, providing high-quality time synchronization data support for subsequent construction of user behavior models based on electricity price sensitivity, thereby improving the prediction accuracy and decision support capability of the model.

[0069] S22, the traffic congestion index and meteorological observation data are standardized in the spatial dimension by a spatial interpolation method to unify the expression forms of data at different geographical locations.

[0070] Specifically, this step standardizes the traffic congestion index and meteorological observation data in the spatial dimension by a spatial interpolation method, aiming to unify the expression forms of data at different geographical locations, and provides a consistent spatial data foundation for subsequent multi-dimensional feature modeling of user charging and battery swapping behavior. In some implementations, this step uses spatial statistical methods such as Kriging interpolation or inverse distance weighting interpolation (IDW, Inverse Distance Weighting) to convert discrete traffic congestion index and meteorological observation point data (such as temperature, rainfall, wind speed, etc.) into continuous spatial distribution fields, thereby eliminating the data expression differences caused by uneven geographical distribution.

[0071] This step is widely used in urban traffic and energy system optimization, especially in electric vehicle user behavior prediction, charging station site selection evaluation, emergency dispatch under extreme weather, etc. By spatial standardization of traffic and meteorological data, the influence factors of user charging and changing behavior in different regions and different time periods can be uniformly modeled, and the spatial adaptability and generalization ability of the prediction model can be improved.

[0072] This step effectively solves the inconsistency problem of multi-source heterogeneous data in the spatial dimension, enhances the sensitivity of the model to geographical space characteristics, provides key support for building a "vehicle-pile-road-network-weather" collaborative modeling framework, and improves the accuracy and reliability of the spatio-temporal quantitative analysis of user charging and changing behavior.

[0073] The embodiment of the application constructs multi-dimensional user behavior characteristic indicators including vehicle state, charging facility service level, price sensitivity, traffic operation state, meteorological influence factors, etc. Quantitative characterization of the behavior patterns of different user groups in different situations helps to classify charging behavior portraits and develop individualized scheduling strategies.

[0074] S3, using a multinomial logistic regression model and a nonlinear least squares method, the probability of user selection of charging and changing in different time periods and the probability of charging station selection are parameterized and modeled, and the dynamic prediction of user behavior is realized.

[0075] Specifically, the core of this step is to use a multinomial logistic regression model and a nonlinear least squares method to parameterize and model the probability of user selection of charging and changing in different time periods and the probability of charging station selection, thereby realizing the dynamic prediction of user behavior. In some implementations, the modeling process first constructs a quantitative index system of behavior influence factors based on user vehicle state (such as SOC, battery capacity, whether supporting changing), charging facility service level (such as charging pile power, queuing time, number of service counters), traffic operation state (such as congestion index), power grid price level (such as time-of-use electricity price, user sensitivity), and meteorological factors (such as temperature, rainfall, extreme weather probability), etc.

[0076] Further, a multinomial logistic regression model (Multinomial Logistic Regression, MNL) is used to model the selection behavior of users in multiple charging and changing time periods. The model input is the above-mentioned multi-dimensional characteristic variables, and the output is the probability of user selection of charging or changing in time period T.

[0077] To improve the fitting accuracy and generalization ability of the model, the invention further uses a nonlinear least squares method (Nonlinear Least Squares, NLS) to calibrate the model parameters. This method minimizes the sum of squares of errors between the predicted probability and the actual user behavior data to realize the optimal estimation of the regression coefficients.

[0078] In practical applications, this step can be deployed in a city-level charging and battery swapping scheduling system, combined with real-time traffic flow, power grid load, weather forecast and other data, to dynamically predict user charging and battery swapping behavior at different times and places, providing decision support for charging station load management, traffic resource optimization scheduling and power grid demand response. This modeling method significantly improves the spatio-temporal resolution and accuracy of user behavior prediction, providing key technical support for the realization of "vehicle-pile-road-grid-weather" multi-system collaborative optimization.

[0079] Further, S3 comprises:

[0080] S31, fitting the regression coefficients and intercept terms in the multinomial logistic regression model using a nonlinear least squares method to improve the accuracy of the model in describing the user behavior response characteristics.

[0081] Specifically, in the present application, the nonlinear least squares method is used to fit the regression coefficients and intercept terms in the multinomial logistic regression model, which is one of the key steps to improve the accuracy of multi-dimensional feature description of user charging and battery swapping behavior. Based on the nonlinear relationship between user historical charging and battery swapping behavior data and multi-source heterogeneous influencing factors (such as vehicle state, charging facility service level, power grid price, traffic congestion index and weather conditions), this step optimizes the model parameters to improve the goodness of fit and prediction ability of the behavior response model.

[0082] In terms of technical implementation, this step first constructs a multinomial logistic regression model, with the dependent variable being the probability of user charging and battery swapping at different time periods T, and the independent variables including key features such as battery remaining SOC, charging price, queuing time, traffic congestion index, weather coefficient, etc.

[0083] This step is widely used in scenarios such as city-level charging and battery swapping facility site selection optimization, user behavior profiling and power grid load forecasting. The regression coefficients obtained by fitting can reflect the sensitivity of users to variables such as price, queuing time, traffic congestion, etc., providing quantitative basis for formulating differentiated service strategies and dynamic pricing mechanisms. This step significantly improves the accuracy of the model in describing user behavior response characteristics. Compared with traditional linear regression or simple logistic regression methods, the nonlinear least squares method can more accurately capture the nonlinear interaction between variables, improve the goodness of fit Coefficient of Determination (R 2 ) and prediction accuracy Area Under the ROC Curve (AUC), thereby enhancing the system's prediction ability and regulatory response level for user behavior.

[0084] S32, based on the user historical charging and battery swapping behavior data, real-time electricity price, traffic congestion, queuing time and other data, dynamically update the model parameters to adapt to the evolution trend of user behavior.

[0085] Specifically, the core of this step is to dynamically update the model parameters based on the user historical charging and battery swapping behavior data and real-time electricity price, traffic congestion, queuing time and other multi-source dynamic data to adapt to the evolution trend of user behavior. In some implementations, this process uses an online learning mechanism and a multi-source data fusion strategy to construct a dynamic feature vector input into the behavior prediction model by collecting and processing data from the vehicle OBD system, charging station operation platform, power grid dispatching system, traffic flow monitoring system and weather database in real time. Specifically, the user historical data includes but is not limited to charging time, location, SOC change curve, vehicle type, battery swapping support status, etc., while the real-time data covers the current period electricity price (such as time-of-use electricity price, peak electricity price), traffic congestion index (such as link travel time ratio based on probe vehicle data), charging station queue length and waiting time (calculated by queuing theory model, such as average queue time and system idle probability in M / M / n model), etc.

[0086] The present application adopts a nonlinear least squares method or an online gradient descent algorithm to dynamically fit the regression coefficients (such as charging price sensitivity coefficient, traffic congestion influence coefficient, queuing time weight, etc.) in the model. The model input features include: whether the remaining SOC is below a set threshold (0 or 1), the current electricity price (yuan / kWh), the traffic congestion index (such as 0.8 for slight congestion and 1.5 for severe congestion), the queuing time (minutes), etc., and the output is the probability of user choosing charging or battery swapping in a specific period. By continuously updating the model parameters, the system can capture the dynamic changes of user behavior over time, space and external environment, and improve the prediction accuracy and response ability.

[0087] This step can be deployed in a city-level charging and battery swapping scheduling platform or a smart grid management system in practical applications, combined with real-time data stream processing technology (such as Apache Kafka, Flink) and distributed computing framework (such as Spark) to realize minute-level or even second-level model parameter updating. Its technical effect lies in significantly improving the adaptability and robustness of the user behavior prediction model, providing data support for charging facility dynamic scheduling, load prediction and user response mechanism optimization, thereby enhancing the collaborative control and regulation capability of the traffic-energy system.

[0088] The embodiments of the present application establish a unified collaborative modeling framework to collaboratively couple the modeling modules of the traffic, energy, power and weather subsystems. From the perspective of collaborative control and linkage scheduling in a complex urban system, the user charging / battery swapping behavior is quantified to support charging infrastructure site selection and user-level response optimization.

[0089] S4, combine user historical behavior data and real-time environment data, build a cross-domain collaborative modeling framework, simulate the evolution trend of user charging and battery swapping behavior in time and space.

[0090] Specifically, the core of this step is to combine user historical behavior data and real-time environment data, build a cross-domain collaborative modeling framework, and simulate the evolution trend of user charging and battery swapping behavior in time and space. This step realizes dynamic prediction and spatio-temporal quantification of user behavior through fusion modeling of multi-source heterogeneous data, which is a key link for the invention to realize multi-dimensional feature-driven behavior modeling.

[0091] In terms of technical implementation, this step first extracts key features from user historical behavior data, including vehicle battery capacity, current remaining power (SOC), acceptable minimum charge threshold (SOC_min), and whether to support battery swapping (Binary Flag). At the same time, integrate real-time environment data such as traffic congestion index (TTI), charging pile service rate (λ), queuing time (Wq), time-of-use electricity price (TOU), meteorological parameters (temperature T, humidity H, extreme weather probability P_weather), etc. Through the construction of a five-dimensional influence factor quantification system of "vehicle-pile-road-network-weather", the above data is mapped to a unified feature vector, which is input into the cross-domain collaborative modeling framework.

[0092] In terms of parameter index, the SOC threshold is usually set to 20%~30% to reflect the user's sensitivity to range anxiety; the electricity price sensitivity ε is fitted through historical data, with a value range of 0.5~1.2; the traffic congestion index TTI uses real-time data provided by Gaode Map or Baidu Map API, with a value range of 0~10, the higher the value, the more serious the congestion; the extreme weather probability P_weather is based on historical data from the Meteorological Bureau and predicted by the LSTM model, with an accuracy of over 90%.

[0093] This step can be widely applied to urban-level charging and battery swapping facility layout optimization, power grid load forecasting, traffic flow regulation, etc. For example, during peak hours or extreme weather conditions, the system can dynamically adjust the charging pile scheduling strategy based on the evolution trend of user behavior, improving overall service efficiency and power grid stability.

[0094] Further, S4 includes:

[0095] S41, introduce a trip chain path accessibility analysis module to evaluate the path feasibility and traffic efficiency of users from the current location to each candidate charging station.

[0096] Specifically, the step of "introducing a trip chain path accessibility analysis module to evaluate the path feasibility and travel efficiency of the user from the current location to each candidate charging station" is one of the key steps in the implementation of the spatiotemporal quantitative modeling of user charging and battery swapping behavior in the present application. The technical implementation principle is based on traffic network topology analysis and path planning algorithms, combined with real-time traffic flow, road restriction rules, geographic information system (GIS) data, etc., to dynamically evaluate the accessibility between the user's current location and the candidate charging stations, thereby providing basic input for subsequent charging station selection probability modeling.

[0097] In some implementations, the module first obtains the real-time location coordinates of the user through GPS or a vehicle positioning system, and combines the geographic coordinate information in the charging station database to construct the spatial topological relationship between the user and each candidate charging station. Subsequently, Dijkstra algorithm or A* algorithm is used for path planning to calculate the shortest path or optimal path from the user's current location to each charging station. During path planning, factors such as road type (e.g., expressway, trunk road, branch road), restricted period, traffic congestion index (e.g., real-time travel time based on floating car data or traffic camera collection), and possible weather influence (e.g., rain causing some roads to be closed) need to be considered.

[0098] Further, path accessibility evaluation includes two core indicators: path feasibility and travel efficiency. Path feasibility is represented by a Boolean variable, if there is an impassable section in the path (such as construction, restriction, closure, etc.), the path is not accessible, recorded as 0; otherwise, 1. Travel efficiency is quantified by the ratio of path travel time (unit: minutes) to historical average travel time, i.e., traffic congestion coefficient, which takes value in the range [1, ∞), the larger the value, the lower the travel efficiency.

[0099] Optionally, the module can also introduce time window constraints, i.e., whether the user can arrive at the charging station on time within a certain time period. For example, if the user plans to complete the trip within 30 minutes, only paths with travel time less than 30 minutes are retained as feasible paths. In addition, energy accessibility can also be judged in combination with vehicle range and state of charge (SOC) to ensure that the user can complete the arrival path without charging.

[0100] This step can be deployed in an intelligent charging and battery swapping scheduling platform or a vehicle networking system in practical applications, supporting users to obtain real-time optimal charging station recommendations during movement. Through the introduction of this module, the spatiotemporal accuracy of user behavior modeling can be effectively improved, and the prediction ability of the system for user actual travel and charging behavior can be enhanced, providing reliable data support for subsequent multi-dimensional feature fusion and collaborative optimization.

[0101] S42, a charging station selection model based on the utility function is constructed, wherein the utility function includes distance, electricity price, queuing time, traffic congestion index and weighted coefficients of each influencing factor.

[0102] Specifically, this step constructs a charging station selection model based on the utility function, the core of which is to quantify the multi-dimensional influencing factors considered by users when selecting charging stations, including distance, electricity price, queuing time, traffic congestion index, etc., and introduce the weighted coefficients of each factor to reflect their relative importance. This model uses the Multinomial Logit Model (MNL) in the discrete choice theory for modeling, and probabilistically models the selection behavior of users among multiple candidate charging stations through the utility function, thereby achieving accurate prediction and optimized guidance of user behavior.

[0103] This model can be embedded in a smart charging scheduling system, a user behavior prediction platform or a city traffic-energy collaborative management system, and used for real-time recommendation of optimal charging stations, prediction of charging demand distribution, optimization of charging pile scheduling strategy, etc. Especially in peak hours or extreme weather conditions, the model can dynamically adjust the weights to improve the adaptability and response efficiency of the recommendation system. By introducing multi-dimensional influencing factors and weighting mechanism, the modeling accuracy and explanatory power of user charging station selection behavior are significantly improved, providing key support for subsequent charging load prediction, facility site optimization and traffic-energy collaborative regulation.

[0104] The embodiment of the application introduces a spatiotemporal trajectory analysis technology, combines user historical trajectory and real-time location, dynamically identifies potential charging demand points and accessibility. Enhance the time-varying sensitivity and geographical distribution adaptability of the prediction model to user behavior, realize more accurate charging / swap behavior simulation.

[0105] Also includes:

[0106] S5, based on the cross-domain collaborative modeling framework, a spatiotemporal heat map of user charging and swapping behavior is generated, which is used to predict the charging and swapping demand density and power grid load change trend of each region in a specific time period in the future.

[0107] Specifically, this step generates a spatiotemporal heat map of user charging and swapping behavior based on a cross-domain collaborative modeling framework, which is used to predict the charging and swapping demand density and power grid load change trend of each region in a specific time period in the future. Its technical implementation principle integrates multi-source heterogeneous data such as traffic trajectory, power price, charging and swapping facility status, traffic congestion index and meteorological environment, and realizes dynamic characterization and load prediction of user behavior through spatiotemporal modeling and behavior prediction algorithm.

[0108] In some implementations, this step first constructs a spatio-temporal behavior feature vector based on user historical travel and charging data, including user location coordinates, timestamps, SOC state, charging power, queuing time, traffic congestion index, etc. Subsequently, a spatial clustering method based on grid division (such as the H3 hexagonal grid system) is used to divide the city into several spatial units, aggregate user behavior features in each unit, and form a regional-level charging behavior heat map. In the time dimension, a sliding time window (such as 15 minutes or 1 hour) is used to dynamically update user behavior, combined with external factors such as time-of-use electricity prices and weather changes, to construct a time series prediction model (such as LSTM, ARIMA, or Prophet).

[0109] At the parameter index level, the generation of the heat map relies on key indicators such as the SOC threshold (usually set to 20%), charging power (generally 7kW to 120kW), queuing time (in minutes), traffic congestion index (0-10 interval, 10 for severe congestion), and electricity price fluctuation coefficient (such as a peak-valley electricity price ratio of 1.6:1:0.4). In addition, the probability of extreme weather events (such as heavy rain, typhoon) is predicted by a meteorological model (such as random forest or XGBoost) and used as an input variable for the behavior response model.

[0110] In practical application scenarios, this step can be deployed in a city-level charging station scheduling platform or a power grid load forecasting system, combined with real-time traffic flow, power grid operation data, and weather forecasts, to dynamically predict the charging demand of each region the next day or in the next few hours. In terms of technical effects, this step significantly improves the spatio-temporal resolution and accuracy of user behavior prediction, providing data support for power grid load scheduling, charging station site optimization, and traffic resource coordination, enhancing the collaborative control capability of urban energy and transportation systems.

[0111] The embodiment of the present application designs a preprocessing and fusion strategy that adapts to different data structures and granularities of "vehicle-pole-road-grid-weather". It realizes the comparability and linkage analysis capability of data in different fields, and provides a unified data basis for subsequent user charging behavior derivative model training and evaluation.

[0112] The embodiment of the application proposes a user charging and discharging behavior multi-dimensional feature space-time quantification method which can comprehensively consider cross-domain factors such as travel behavior, power grid price level, charging and discharging infrastructure service level, road traffic network condition and meteorological environment. By fusing multi-source heterogeneous data, accurate description and dynamic modeling of user charging and discharging behavior in time and space dimensions are realized, and the limitations of existing methods in dependence on a single factor and insufficient description of coupling relationship are broken through. Through the application, data support and decision basis can be provided for scientific layout of charging and discharging facilities, charging station load management, traffic and energy system collaborative optimization and vehicle-grid interaction mechanism design, the collaborative control ability of vehicle-pile-road-network-weather multi-system is improved, and the deep integration development of intelligent traffic and smart grid is promoted.

[0113] Embodiment 2

[0114] The application proposes a user charging and discharging behavior multi-dimensional feature space-time quantification method considering cross-domain factors, and the calculation process is as shown in Figure 2 The method first constructs a multi-dimensional behavior feature index system, systematically integrates multi-source heterogeneous data such as user vehicle state, charging facility service level, power grid price level, traffic running condition and meteorological elements, comprehensively quantifies the feature performance of user charging and discharging behavior in time and space dimensions, and extracts key indicators such as battery capacity, queuing waiting time and traffic congestion coefficient. On this basis, a cross-domain collaborative modeling strategy is introduced, a charging selection behavior model is constructed combined with user travel state and environmental constraint conditions, and dynamic simulation and fine description of the space-time evolution trend of charging and discharging behavior are realized. Through the above process, a systematic and expandable user behavior quantification evaluation method is formed, which provides high-quality data support and decision basis for charging and discharging facility site optimization, charging station load management, traffic resource scheduling and vehicle-grid collaborative control, and promotes the deep integration and intelligent collaborative development of traffic and energy.

[0115] Preferably, the user charging and discharging behavior multi-dimensional feature space-time quantification method considering cross-domain factors should first establish a quantitative characterization system of “vehicle-pile-road-network-weather” multi-dimensional influencing factors, and systematically evaluate the key factors affecting user charging and discharging behavior. The evaluation system mainly covers the following dimensions: user vehicle state, charging facility service level, traffic running condition, power grid price level and meteorological elements, to comprehensively describe the external driving mechanism of user decision behavior.

[0116] Preferably, in terms of user vehicle state, the influence on its charging and discharging selection behavior mainly reflects in the battery remaining state and its related characteristics. For this purpose, a battery capacity (E ba ), current remaining power (SOC ini ), acceptable minimum charge threshold (SOC minand whether to support battery replacement (indicated by "1" for not supporting and "2" for supporting) and other parameters. The above parameters can not only reflect the urgency of the user to supplement energy under different remaining power, but also provide basic data support for subsequent charging behavior modeling.

[0117] Preferably, in terms of charging facility service level, in order to accurately evaluate its influence on the user's charging and battery replacement behavior decision, it is necessary to quantitatively model from two dimensions of facility operation efficiency and service experience. Specifically, the rated charging power P char of the charging pile (reflecting the charging efficiency per unit time), the queue length (indicating the number of vehicles currently waiting to charge), and the queue time (measuring the average waiting cost of the user at the charging site) and other key parameters can be selected as core evaluation indicators. The above indicators can comprehensively reflect the user's sensitivity to charging speed and service accessibility when selecting a charging facility, and provide effective support for subsequent behavior modeling and system optimization.

[0118] In the system performance analysis based on queuing theory, the system idle probability P0 represents the probability that the system is in an idle state (i.e., no task in the system); the average queue length L q represents the average number of tasks waiting for service in the system; and the average queue time W q represents the average time a task waits in the queue. The calculation formulas of the above parameters are as follows:

[0119] The system idle probability P0 satisfies the following relationship.

[0120]

[0121] In the formula, λ: arrival rate, number of vehicles arriving at the charging station per unit time (vehicles / hour), subject to Poisson distribution, need to multiply by a coefficient k = 2-3 for holiday expressway service area, and 1 for ordinary time.

[0122] μ: service rate, number of vehicles served per unit time by a single charging pile (vehicles / hour), service time subject to exponential distribution.

[0123] c: number of service counters, here representing the total number of charging piles.

[0124] ρ: service counter utilization rate, needs to satisfy ρ < 1 to ensure system stability

[0125] P0: system idle probability, representing the initial state probability that all charging piles are not occupied.

[0126] The average queue length L q satisfies the following relationship.

[0127]

[0128] L q The average number of vehicles in the waiting queue in the charging station is strongly related to the arrival rate and the number of service counters.

[0129] The average queuing time W q The following relationship is satisfied.

[0130]

[0131] W q The average waiting time in the waiting queue in the charging station is related to the arrival efficiency and the number of service counters.

[0132] Preferably, in terms of traffic conditions, to comprehensively evaluate the influence between road traffic conditions and user charging and battery swapping behaviors, a key parameter reflecting road traffic efficiency should be introduced. The traffic congestion coefficient as an important indicator to measure the road operation state can effectively represent the saturation degree of road traffic in different time periods and regions. Through dynamic monitoring and analysis of the traffic congestion coefficient, the traffic obstacles that users may face during travel can be identified, so that their selection behaviors in charging and battery swapping decisions can be more accurately described.

[0133]

[0134] t h : peak time period, charging station adjacent road traffic time;

[0135] t p : flat peak time period, charging station adjacent road traffic time;

[0136] Preferably, in terms of the level of electricity price, to quantify the influence of the electricity price mechanism on user charging and battery swapping behaviors, a key parameter reflecting the change of charging behavior and charging cost should be introduced. As an important means to reflect the relationship between power supply and demand and dispatching strategy, the price fluctuation of different time periods directly affects the charging cost of users and their behavior preferences. By quantitatively analyzing the sensitivity of time-of-use electricity price to user charging and battery swapping behaviors, the economic driving factors of users choosing charging time in different time periods can be revealed.

[0137] The user sensitivity index satisfies the following relationship.

[0138]

[0139] Δp: represents the percentage change in electricity price, Δp = p1 x (p2-p1), p1 represents the charging price in the first time period, and p2 represents the charging price in the second time period.

[0140] ΔE: represents the percentage change of the charging amount, ΔE=E1*(E2-E1), E1 represents the charging price of the first time period, and E2 represents the charging price of the second time period.

[0141] Preferably, in terms of meteorological elements, in order to comprehensively evaluate the influence of environmental conditions on the charging and battery swapping behavior of users, key indicators such as temperature, rainfall and the probability of occurrence of extreme weather events should be introduced. A quantitative model of the probability of occurrence of extreme weather types is constructed through indicators such as temperature and humidity, and the probability of occurrence of extreme weather events (such as typhoon, high temperature, heavy rain) determines the avoidance characteristics of user travel behavior and the urgency of charging and battery swapping behavior to a certain extent. The introduction of the above meteorological parameters helps to construct a more realistic and adaptive behavior response model, and improves the accuracy and robustness of user behavior modeling, so the weather occurrence probability is quantified.

[0142]

[0143] In the formula: y represents the type of extreme weather, 0=normal weather, 1=heavy rain, and 2=typhoon.

[0144] K: the number of days of weather observation data samples;

[0145] X=[X1,X2,…,X p ] T is the independent variable (feature vector), including temperature, humidity, etc.

[0146] β i =[β i0 ,β i1 ,…,β ip ] T is the coefficient vector of the i-th type, which represents the influence strength of the variable X i in the i-th type of weather.

[0147] Preferably, in order to realize accurate modeling and dynamic evaluation of user charging and battery swapping behavior, the present application constructs a multi-dimensional feature-driven user behavior representation method from the five key dimensions of "vehicle-pile-road-network-weather". This method fully considers factors such as electric vehicle technical parameters, charging infrastructure service level, road traffic operating state, power grid price level and external meteorological environment, as shown in Table 1:

[0148] Table 1: Factors affecting electric vehicle charging

[0149]

[0150]

[0151] By systematic index design, the user behavior influencing mechanism is fully described. Based on vehicle state, charging service level, power grid price level and weather conditions, a multiple logistic regression model is used to quantify the user charging time T selection probability as shown in the following formula.

[0152]

[0153] In the formula: k soc represents whether the remaining SOC is lower than the SOC threshold SOC min or it is difficult to meet the driving demand, if it is met, it is 1, otherwise it is 0; P t represents the charging price of period t, including electricity fee and service fee; K w represents the weather coefficient, the higher the probability of extreme weather, the smaller the value, K w = 1-p(y = w ext ); β kj represents the regression coefficient, which represents the influence degree of the independent variable on β kj , β k0 represents the intercept term.

[0154] Desired charging duration:

[0155]

[0156] In the formula: SOC end represents the desired SOC after charging; SOC end represents the initial SOC of charging; E ba represents the electric capacity; P char represents the charging power.

[0157] Desired battery replacement duration:

[0158] T swap = t char

[0159] In the formula: T swap represents the actual time needed for battery replacement.

[0160] Preferably, the charging station selection probability needs to be combined with the user trip chain, traffic congestion and charging and battery replacement facility queuing conditions to build a user charging location selection probability model.

[0161] Probability of user selecting charging station i:

[0162]

[0163] In the formula: U i represents the utility function of charging station i Attractive model of charging station i to users: A i = k1 / di +k2 / p t +k3 / t q +k4 / TTI;

[0164] wherein: d i : distance from user to charging station; p t : charging price of time period t; t q : represents the length of queuing time; TTI: traffic congestion index; k i : influence coefficient.

[0165] Preferably, by comprehensively considering the user's instant charging demand and the selection preference of each candidate charging station, the probability of the user selecting the i-th charging station for charging can be further determined. This probability, as one of the core parameters for measuring user charging behavior, provides a basis for subsequent charging behavior modeling and station power load management. The calculation method is shown below.

[0166] P = P(T = k) x P(A = i)

[0167] wherein: P(T = k): probability of the user selecting charging and battery swapping within time period T; P(A = i): charging probability of the user at the i-th charging station.

[0168] Preferably, to more accurately quantify the key factors affecting user charging and battery swapping behavior, based on the probability model of the user selecting charging and battery swapping at different time periods T and the user's historical charging behavior data, a nonlinear least squares method is used to fit the model parameters. By calibrating the model parameters, the user charging characteristics can be quantified by parameter values, which not only describes the user's preference in the time dimension, but also reflects the response characteristics of the user to factors such as electricity price and convenience, thereby providing quantitative basis for user behavior modeling and strategy optimization.

[0169]

[0170] wherein: x i : represents the independent variable in the model, x i = [x1, x2…x n ]; y i : represents the dependent variable in the model; b: represents the parameters to be calibrated in the model, β i = [β k0 , β k1 … β k4 ]; k i = [k1, k2…k4].

[0171] In summary, the application first systematically introduces the "vehicle-pile-road-network-weather" multi-source core elements, fuses multi-source heterogeneous data such as user travel trajectory, charging infrastructure, power grid price level, traffic congestion and meteorological environment, realizes the spatio-temporal dynamic simulation and quantitative analysis of user charging and changing behavior by constructing a multi-dimensional behavior characteristic index system and a cross-domain collaborative modeling framework. Compared with the existing research, this method not only improves the quantitative precision of user charging behavior characteristics, but also provides a unified modeling basis and decision support for charging facility layout optimization, charging station load regulation, traffic resource coordination and extreme weather response, etc. Practical applications, significantly promoting the evolution of intelligent sensing and collaborative control technology under the background of deep integration of user charging behavior and power grid, traffic and other multi-systems.

[0172] Embodiment 3

[0173] In order to realize the above-mentioned embodiments, as shown in the embodiment, a multi-dimensional feature spatio-temporal quantization device 10 considering cross-domain factors is also provided, comprising: Figure 3

[0174] The multi-dimensional feature construction module 100 is configured to construct a multi-dimensional behavior characteristic index system comprising vehicle state, charging facility service level, traffic operating condition, power grid price level and meteorological factor, and to quantify the influence of each dimension on user charging and changing behavior.

[0175] The data fusion and spatio-temporal alignment module 200 is configured to fuse user travel trajectory, charging pile operation data, traffic congestion index, time-of-use price information and meteorological observation data based on the multi-dimensional behavior characteristic index system, and to perform standardization processing and spatio-temporal alignment of multi-source heterogeneous data.

[0176] The behavior probability modeling module 300 is configured to use a multinomial logistic regression model and a nonlinear least squares method to calibrate and model the probability of user selecting charging and changing in different time periods and the probability of selecting a charging station, and to realize dynamic prediction of user behavior.

[0177] The cross-domain collaborative modeling module 400 is configured to combine user historical behavior data and real-time environmental data to construct a cross-domain collaborative modeling framework, and to simulate the evolution trend of user charging and changing behavior in time and space.

[0178] Further, the multi-dimensional feature construction module is further configured to:

[0179] Define vehicle state parameters including battery capacity, current remaining power, acceptable minimum charge threshold and whether to support changing, to quantify the energy replenishment urgency of users in different power states.

[0180] ​The charging facility service level parameters include the rated charging power of the charging pile, the queue length and the queue time, so as to reflect the sensitivity of the user to the charging efficiency and service accessibility when selecting the charging facility.

[0181] Further, the data fusion and space-time alignment module is also used for:

[0182] The timestamp alignment algorithm is adopted to match the user travel trajectory data and the power grid time-of-use price data in the time dimension, so as to ensure the synchronism of different time domain data in behavior modeling.

[0183] The spatial interpolation method is adopted to perform the standardization processing on the traffic congestion index and the meteorological observation data in the space dimension, so as to unify the expression forms of the data in different geographical positions.

[0184] Further, the behavior probability modeling module is also used for:

[0185] The nonlinear least square method is adopted to fit the regression coefficients and the intercept term in the multinomial logistic regression model, so as to improve the description accuracy of the user behavior response characteristics by the model.

[0186] Based on the user historical charging and battery swapping behavior data and the real-time price, the traffic congestion and the queue time, the model parameters are dynamically updated, so as to adapt to the evolution trend of the user behavior.

[0187] The user charging and battery swapping behavior multidimensional feature space-time quantization device considering cross-domain factors can systemically fuse the multi-source heterogeneous data of'vehicle-pile-road-grid-weather', and realize the accurate quantization and dynamic modeling of the user charging and battery swapping behavior in the space-time dimension, and improve the behavior prediction accuracy and the multi-system collaborative optimization capability.

[0188] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in combination with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples without contradiction.

[0189] Furthermore, the terms "first", "second", "third", "fourth", "fifth" and "sixth" are used herein for descriptive purposes only and are not to be construed as indicating or implying relative importance or a significant nature of so described technical features. It is to be understood that a technical feature described with the "first", "second", "third", "fourth", "fifth" or "sixth" can implicitly or explicitly include at least one of the technical features described with the "first", "second", "third", "fourth", "fifth" or "sixth". In the description of the present application, the meaning of "a plurality" is at least two, for example, two, three, etc., unless otherwise specifically defined.

Claims

1. A multi-dimensional feature space-time quantization method for user charging and swapping behavior considering cross-domain factors, characterized in that, Comprise: S1, construct a multi-dimensional behavior characteristic index system including vehicle state, charging facility service level, traffic operating condition, power grid price level and meteorological factors, and quantitatively represent the influence of each dimension on user charging and changing behavior; S2, based on the multi-dimensional behavior characteristic index system, fuse user travel trajectory, charging pile operation data, traffic congestion index, time-of-use electricity price information and meteorological observation data, and perform standardization processing and space-time alignment of multi-source heterogeneous data; S3, use a multiple logistic regression model and a nonlinear least squares method to calibrate parameters and model the probability of user selecting charging and changing in different time periods and the probability of selecting a charging station, and realize dynamic prediction of user behavior; S4, combine user historical behavior data and real-time environmental data to construct a cross-domain collaborative modeling framework, and simulate the evolution trend of user charging and changing behavior in time and space.

2. The method of claim 1, wherein, The S1 comprises: S11, define vehicle state parameters including battery capacity, current remaining power, acceptable minimum charge threshold and whether to support changing, to quantitatively represent the energy supplement urgency of users in different power states; S12, construct charging facility service level parameters including rated charging power of charging piles, queue length and queue time, to reflect the sensitivity of users to charging efficiency and service accessibility when selecting charging facilities.

3. The method of claim 1, wherein, The S2 comprises: S21, use a timestamp alignment algorithm to match user travel trajectory data and power grid time-of-use electricity price data in the time dimension, to ensure the synchronicity of different time domain data in behavior modeling; S22, use a spatial interpolation method to perform standardization processing on traffic congestion index and meteorological observation data in the spatial dimension, to unify the expression form of data at different geographical locations.

4. The method of claim 1, wherein, The S3 comprises: S31, use a nonlinear least squares method to fit the regression coefficients and intercept terms in the multiple logistic regression model, to improve the description accuracy of the model to the response characteristics of user behavior; S32, based on user historical charging and changing behavior data and real-time electricity price, traffic congestion and queue time, dynamically update model parameters to adapt to the evolution trend of user behavior.

5. The method of claim 1, wherein, The S4 comprises: S41, introduce a travel chain path accessibility analysis module to evaluate the path feasibility and passing efficiency of users from the current location to each candidate charging station; S42, construct a charging station selection model based on a utility function, wherein the utility function includes distance, electricity price, queue time, traffic congestion index and weighted coefficients of each influencing factor.

6. The method of claim 1, wherein, Further comprise: S5, based on the cross-domain collaborative modeling framework, generate a space-time heat map of user charging and changing behavior, to predict the charging and changing demand density and power grid load change trend of each region in a specific future time period.

7. A device for spatiotemporal quantification of multi-dimensional features of user charging and swapping behavior considering cross-domain factors, characterized in that, Comprise: A multi-dimensional feature construction module is configured to construct a multi-dimensional behavior characteristic index system including vehicle state, charging facility service level, traffic operating condition, power grid price level and meteorological factors, and quantitatively represent the influence of each dimension on user charging and changing behavior; The data fusion and space-time alignment module is configured to fuse user travel trajectory, charging pile operation data, traffic congestion index, time-of-use electricity price information and meteorological observation data based on the multi-dimensional behavior characteristic index system, and perform standardization processing and space-time alignment of multi-source heterogeneous data. The behavior probability modeling module is configured to calibrate and model the probability of user selection of charging and battery swapping in different time periods and the probability of charging station selection by using a multinomial logistic regression model and a nonlinear least squares method, and to realize dynamic prediction of user behavior. The cross-domain collaborative modeling module is configured to combine user historical behavior data and real-time environmental data to construct a cross-domain collaborative modeling framework, and to simulate the evolution trend of user charging and battery swapping behavior in time and space.

8. The apparatus of claim 7, wherein, The multi-dimensional feature construction module is further configured to: define vehicle state parameters including battery capacity, current remaining power, acceptable minimum charge threshold and whether to support battery swapping, to quantify the urgency of user power supplement in different power states; construct charging facility service level parameters including rated charging power of charging piles, queue length and queue time, to reflect the sensitivity of users to charging efficiency and service accessibility when selecting charging facilities.

9. The apparatus of claim 7, wherein, The data fusion and space-time alignment module is further configured to: use a timestamp alignment algorithm to match user travel trajectory data and power grid time-of-use electricity price data in the time dimension, to ensure the synchronicity of different time domain data in behavior modeling; perform standardization processing of traffic congestion index and meteorological observation data in the space dimension by a spatial interpolation method, to unify the expression form of data in different geographical locations.

10. The apparatus of claim 7, wherein, The behavior probability modeling module is further configured to: use a nonlinear least squares method to fit the regression coefficients and intercept terms in the multinomial logistic regression model, to improve the accuracy of the model in describing the response characteristics of user behavior; update the model parameters based on user historical charging and battery swapping behavior data and real-time electricity price, traffic congestion and queue time, to adapt to the evolution trend of user behavior.

Images