Method for constructing electric vehicle user willingness model by fusing social tuning and sensing data
By integrating social survey and sensor data, a multi-agent model was constructed, which solved the problem of insufficient accuracy in modeling electric vehicle user intentions, achieved effective data integration and model optimization, and improved the ability to guide orderly charging.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING UNIV OF POSTS & TELECOMM
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-12
AI Technical Summary
Existing technologies struggle to effectively integrate social survey data with sensor data, resulting in insufficient accuracy and reliability in modeling electric vehicle user intentions, thus failing to meet the demand for orderly charging.
By integrating social survey data and sensor data, a multi-agent model is constructed. Monte Carlo random sampling and error analysis are used to achieve cross-validation and feature matching of the data, thereby optimizing the user intention model.
It improves the accuracy and feasibility of electric vehicle user intention modeling, solves the problem of data fusion modeling, and enhances the accuracy and reliability of the model.
Smart Images
Figure CN122199028A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of user behavior intention modeling technology, specifically involving a method for constructing an electric vehicle user intention model that integrates social surveys and sensor data. Background Technology
[0002] With the rapid growth of the electric vehicle industry, large-scale, unregulated charging of electric vehicles can easily lead to a sharp increase in regional power grid load and a widening of peak-valley differences, which in severe cases threaten the safe and stable operation of the power grid. Against this backdrop, guiding electric vehicles to participate in orderly charging has become an effective measure to ensure the safe and economical operation of the power grid. However, whether electric vehicles can effectively participate in orderly charging is related to the willingness of vehicle owners; therefore, constructing an accurate and reliable model of electric vehicle user behavior willingness has become an urgent problem to be solved.
[0003] Currently, existing research on user behavior intention modeling mainly falls into two categories: causal-driven and data-driven. However, both methods have significant limitations and cannot meet the actual needs of accurate modeling of electric vehicle user intentions. Causal-driven methods face the challenge of failing to analyze the factors influencing user decision-making intentions and their mechanisms of influence, thus failing to provide clear logical support for intention model construction. Data-driven methods, on the other hand, are highly dependent on data scale and quality; the limitations of the data directly affect the accuracy and reliability of the model. The data supporting the modeling of electric vehicle user behavior intentions mainly includes two categories: one is social survey data based on questionnaire design (referred to as social survey data). This type of data forms structured information through standardized questionnaire settings, which can be directly used to analyze the causal relationship between variables and provide a logical basis for model construction. However, due to the limitations of survey cost and scope, the sample size is limited and the data volume is insufficient, making it difficult to support the large-scale, high-precision intention modeling needs. The other category is sensor data collected by automated devices such as sensors. This type of data is mostly fused data, which can obtain objective physical quantities such as user vehicle behavior, vehicle operating status, and power system operating parameters. It has advantages such as strong data real-time performance and huge volume. However, the internal correlation of the data is complex, making it difficult to effectively quantify and analyze its relationship with user intentions, and it cannot be directly used to accurately build a model that reflects user intentions.
[0004] In summary, there is an urgent need for a technical solution that can integrate the advantages of both types of data and solve the challenges of data fusion modeling, so as to improve the feasibility and accuracy of user intention modeling and provide reliable support for orderly charging guidance. Summary of the Invention
[0005] To address the aforementioned technical problems, this invention provides a method for constructing an electric vehicle user intention model that integrates social survey data and sensor data. By integrating social survey data and sensor data, a multi-agent model of electric vehicle users' orderly charging intention is accurately constructed, effectively solving the problem of modeling with different types of data. Based on cross-validation of the two types of data, the multi-agent model is corrected and optimized, improving the feasibility and accuracy of behavioral intention modeling.
[0006] The method for constructing an electric vehicle user intention model that integrates social surveys and sensor data, as described in this invention, includes the following steps: Step 1: Social survey data collection: Design a standardized questionnaire to collect social survey data by considering the driving patterns, charging habits and willingness to charge in an orderly manner of electric vehicle users; Step 2: Extraction of deep correlation information from the questionnaire: Based on the social survey data, extract the deep correlation information of the social survey data in the form of multiple conditional probability distributions; Step 3: Multi-agent generation based on Monte Carlo random sampling: Using the extracted deep correlation information of the questionnaire as sampling data, a multi-agent model reflecting the orderly charging intention of the sampled population is constructed through the Monte Carlo stratified sampling method. Step 4: Validation of the surrogate model based on error analysis: Based on different numbers of surrogates, analyze the error sensitivity of the statistical distribution between the simulation results of the surrogate model and the questionnaire statistical results for each number of surrogates to verify the accuracy of the surrogate model; Step 5: Determining the required number of agents based on error analysis: Quantify the relationship between the number of agents and the model accuracy through error sensitivity analysis results, and determine the required number of agents; Step 6: Proxy attribute expansion based on consistency verification of social survey and sensor data: Extract the proxy feature attributes generated from the social survey data, match them with the sensor data collected by automated sensors to complete the consistency verification of the two types of data, and use the sensor data that has passed the verification to expand the proxy feature attributes to obtain the final electric vehicle user intention model.
[0007] Furthermore, in step 1, the survey of driving travel patterns specifically collects the following parameters: departure time to work t1, arrival time at work t2, departure time from get off work t3, arrival time home t4, and daily travel distance d. trip ; The survey on charging habits specifically collected data on users' charging methods. mode This includes charging at home slow charging stations, charging at company slow charging stations, charging at both home and company slow charging stations, and charging at public fast charging stations. The survey on willingness to charge in an orderly manner specifically collected data on users' acceptance of orderly charging. order Preferred subsidy methods modeAnd the expected subsidy discount / amount fee .
[0008] Furthermore, in step 2, the extraction of deep-seated correlation information from the questionnaire includes: Step 2-1: Based on the collected social survey data, statistically analyze and extract the probability distribution P1(c) of the charging methods of the sampled population. mode ); Step 2-2: Based on the obtained probability distribution of charging methods, extract the conditional probability distribution P2(t) of user travel time under slow charging method. trip |c mode ); Steps 2-3: Based on the conditional probability distribution of travel time, extract the conditional probability distribution P3(d) of user travel distance under different travel times. trip |t trip ), and the conditional probability distribution P4(c) of users accepting orderly charging under different travel times. order |t trip ); Steps 2-4: Under the condition that the user accepts orderly charging, extract the conditional probability distribution P5(s) of the user's choice of charging subsidy method. mode |c order ); Steps 2-5: Under the subsidy method selected by the user, extract the conditional probability distribution P6(s) of the user's expected subsidy discount / amount. fee |s mode ).
[0009] Further, in step 3, the construction of the multi-agent model based on Monte Carlo stratified sampling includes: Step 3-1: Number the agents using parameter i, i = 1, 2 ... N, where N is the total number of agents. Initially, i = 1. Step 3-2, Generate Given a uniformly distributed random number, determine the probability distribution interval of the charging method to which the random number falls, and use that charging method as the charging method c for agent i. i.mode ; Step 3-3, Generate Given a uniformly distributed random number, determine the probability distribution interval of the travel time (work hours and off-get off work hours) corresponding to that random number, and use that travel time as the travel time t for agent i. i.trip ; Steps 3-4: Generate A uniformly distributed random number is used to determine the probability distribution interval corresponding to the travel distance of the random number, and this travel distance is used as the proxy. Travel distance d i.trip ; Steps 3-5: Generate Given a uniformly distributed random number, determine the probability distribution interval within which the random number falls (corresponding to either accepting or not accepting ordered charging), and obtain the proxy. Does it accept the state of orderly charging? i.order ; Steps 3-6, if the agent The state is to receive orderly charging and generate Random numbers uniformly distributed between (c) order =1 indicates that it accepts ordered charging, c order =0 indicates that ordered charging is not accepted), determine the probability distribution interval corresponding to the subsidy method (choosing charging discount subsidy or cash subsidy) of the random number, and determine the agent. Desired orderly charging subsidy method i.mode If agent i's state is "do not accept orderly charging", then proceed to steps 3-8. Steps 3-7: Generate Given a uniformly distributed random number, determine the probability distribution interval of the random number corresponding to different subsidy discount / amount values, and obtain the agent. Expected orderly charging subsidy discount / amount i.fee ; Step 3-8: If i = i + 1, determine if i is greater than N. If not, jump to step 3-2 to continue generating the next agent; if yes, the agent model is completed and the sampling process ends.
[0010] Furthermore, in step 4, the relative mean and variance of the error between the simulation results of the agent model and the questionnaire statistics are analyzed when the number of agents is different. By analyzing the convergence of the error, the accuracy of the constructed agent model is verified.
[0011] Furthermore, in step 5, based on the error sensitivity data corresponding to different agent numbers, the relationship between agent number and model accuracy is quantified, and the agent number corresponding to the maximum acceptable error is taken as the minimum agent number requirement.
[0012] Furthermore, in step 6, the characteristic attributes of the agent generated based on the social survey data include: charging method, trip start time, trip end time, trip distance, willingness to accept orderly charging, subsidy method for accepting orderly charging, and expected subsidy discount / amount when accepting orderly charging; through feature matching, the social survey data and the sensor data collected by the automated equipment are checked for consistency, and after the check is completed, the sensor data is used to expand the characteristic attributes of the agent.
[0013] The beneficial effects of this invention are as follows: Traditional sensor data can only record user behavior data and it is difficult to obtain user intention data. This invention obtains structured data reflecting user behavior intentions through questionnaire design and surveys, and generates a multi-agent model containing multiple behavioral features (travel patterns, charging and discharging intentions, etc.) based on this structured data; it effectively solves the problem of missing and unobtainable user intention data in the traditional user behavior intention modeling process; it proposes a method for verifying the consistency between sensor data and social survey data through feature matching, and realizes the fusion analysis of multi-source data by utilizing the advantages of the two types of data. On the basis of effectively expanding the data sources, it further increases the feature attributes of the model built based on social survey data. Attached Figure Description
[0014] Figure 1 This is a flowchart of the method described in this invention; Figure 2 This is a flowchart of the process for extracting deep-seated correlation information from the questionnaire; Figure 3 This is a flowchart of the multi-agent model for generating electric vehicle user intentions; Figure 4 This is an error analysis chart of the charging method; Figure 5 This is a chart analyzing travel time errors when using home slow charging. Detailed Implementation
[0015] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.
[0016] like Figure 1 As shown, the method for constructing an electric vehicle user intention model that integrates social surveys and sensor data according to the present invention includes the following steps: Step 1: Social survey data collection: Design a standardized questionnaire to collect social survey data by considering the driving patterns, charging habits and willingness to charge in an orderly manner of electric vehicle users; Step 2: Extraction of deep correlation information from the questionnaire: Based on the social survey data, extract the deep correlation information of the social survey data in the form of multiple conditional probability distributions; Step 3: Multi-agent generation based on Monte Carlo random sampling: Using the extracted deep correlation information of the questionnaire as sampling data, a multi-agent model reflecting the orderly charging intention of the sampled population is constructed through the Monte Carlo stratified sampling method. Step 4: Validation of the surrogate model based on error analysis: Based on different numbers of surrogates, analyze the error sensitivity of the statistical distribution between the simulation results of the surrogate model and the questionnaire statistical results for each number of surrogates to verify the accuracy of the surrogate model; Step 5: Determining the required number of agents based on error analysis: Quantify the relationship between the number of agents and the model accuracy through error sensitivity analysis results, and determine the required number of agents; Step 6: Proxy attribute expansion based on consistency verification of social survey and sensor data: Extract the proxy feature attributes generated from the social survey data, match them with the sensor data collected by automated sensors to complete the consistency verification of the two types of data, and use the sensor data that has passed the verification to expand the proxy feature attributes to obtain the final electric vehicle user intention model.
[0017] In step 1, a questionnaire was designed considering driving patterns, charging habits, and willingness to charge in an orderly manner, as shown in Table 1.
[0018] Table 1 Questionnaire Questions
[0019] Based on the questionnaire questions, the information collected includes: (1-1) Daily commuting patterns of users: departure time to work t1, arrival time at work t2, departure time from get off work t3, arrival time home t4, and daily mileage d. trip ; (1-2) User's charging method c mode This includes: home slow charging stations, company slow charging stations, charging at both home and company slow charging stations, and public fast charging stations. (1-3) User willingness to participate in orderly charging: Whether users accept orderly charging order Subsidies tend to be offered after accepting orderly charging practices. mode Expected subsidy discount / amount fee .
[0020] like Figure 2 As shown, the deep correlation information extracted after the above questionnaire sampling includes: (2-1) Extract the probability distribution P1(c) of the charging methods of the respondents. mode ); (2-2) Extract the conditional probability distribution P2(t) of travel time under slow charging mode. trip |c mode ); (2-3) Extract the conditional probability distribution P3(d) of travel distance under different travel times. trip |t trip ); (2-4) Extract the conditional probability distribution P4(c) of whether users accept orderly charging under different travel times. order |t trip ); (2-5) For users who accept orderly charging, extract the conditional probability distribution P5(s) of the user's choice of charging subsidy method. mode |c order ); (2-6) Conditional probability distribution P6(s) of expected subsidy discount / amount extracted according to different subsidy methods fee |s mode ).
[0021] like Figure 3 As shown, after obtaining deep correlation information from the questionnaire, a proxy model reflecting user intentions is constructed using Monte Carlo stratified sampling, including: (3-1) Use parameter i (i=1, 2, ..., N) to number the agents, where N is the total number of agents, and the initial value of i is set to 1; (3-2) Generation Given a uniformly distributed random number, determine which charging method's probability distribution interval the random number falls into, and use that charging method as the charging method c for agent i. i.mode ; (3-3) Generation A uniformly distributed random number is generated. The probability distribution interval corresponding to which travel time (start of work and end of get off work) the random number falls is determined, and that travel time is used as the proxy. Travel time t i.trip ; (3-4) Generation Given a uniformly distributed random number, determine which probability distribution interval the random number falls into, and use this travel distance as the travel distance d of agent i. i.trip ; (3-5) Generation Given a uniformly distributed random number, determine whether the random number falls within the probability distribution interval corresponding to accepting or not accepting ordered charging, and obtain the proxy. Does it accept the state of orderly charging? i.order ; (3-6) such as agents Accept orderly charging, generate A uniformly distributed random number is used to determine whether the random number falls within the probability distribution interval corresponding to choosing between the charging discount subsidy and the cash subsidy. This interval is then used as a proxy. Desired orderly charging subsidy method i.mode If agent i does not accept orderly charging, jump to (3-8). (3-7) Generation A uniformly distributed random number is used to determine the probability distribution interval where the random number falls within different subsidy discount / amount values, and then the agent is obtained. Expected orderly charging subsidy discount / amounti.fee ; (3-8) i = i + 1, determine whether i is greater than the required number of electric vehicle user agents N. If not, jump to (3-2) and re-execute. If yes, end.
[0022] In step 4, different numbers of agent groups are generated. Based on the questionnaire statistics, the relative error and variance between the agent simulation results and the questionnaire statistics are compared to conduct error analysis.
[0023] The following analysis uses the relative error mean and variance of charging methods as an example.
[0024] (4-1) The probability distribution of charging methods in the questionnaire statistics is denoted as... , These represent four charging methods: home slow charging station charging, company slow charging station charging, using both home and company slow charging stations, and public fast charging station charging. (4-2) Calculate the probability distribution of the agent model with respect to the charging method when generating different numbers of agents. ; (4-3) Calculation and The percentage of each charging method is denoted as P1(c mode =A), P1(c mode =B), P1(c mode =C), P1(c mode =D) and , , , Calculate the relative error for each of the four charging methods (using P1(c) as the reference). mode (Based on the percentage of charging methods), the average and variance of the relative errors were statistically analyzed, and the results are as follows: Figure 4 As shown.
[0025] Similarly, the mean and variance of the relative error in the travel time distribution under different charging methods can be calculated. For example, the error results for travel time using home slow charging are as follows: Figure 5 As shown.
[0026] Depend on Figure 4 , Figure 5 As can be seen, the error gradually decreases as the number of agents increases. Based on the convergence characteristics of the error curve, a sufficient number of agents can be generated according to the actual accuracy requirements, that is, the minimum number of agents required is the number of agents corresponding to the maximum acceptable error.
[0027] The characteristic attributes of the agent generated based on social survey data include: charging method, trip start time, trip end time, trip distance, willingness to accept orderly charging, subsidy method for accepting orderly charging, and expected subsidy discount / amount when accepting orderly charging. The sensor data refers to data automatically recorded by electronic devices such as sensors, for example, charging history data recorded by charging piles. Consistency verification is performed between the social survey data and the sensor data through feature matching. For example, if the trip start time and trip end time of the agent model generated based on the social survey data are consistent with the charging end time and charging start time recorded by the sensor data, then the agent and the sensor data can be considered to belong to the same user. In this way, other characteristic data recorded by the sensor data can be used to expand the agent's characteristic attributes.
[0028] This invention constructs a proxy model that effectively represents the intentions of respondents by extracting deep information from social survey data. At the same time, it effectively expands the attributes of the proxy model by combining sensor data, thus solving the problem of difficulty in accurately modeling decision intentions caused by heterogeneity of user behavior and insufficient data.
[0029] The above description is merely a preferred embodiment of the present invention and is not intended to further limit the present invention. All equivalent changes made based on the description and drawings of the present invention are within the protection scope of the present invention.
Claims
1. A method for constructing an electric vehicle user intention model that integrates social surveys and sensor data, characterized in that, Includes the following steps: Step 1: Social survey data collection: Design a standardized questionnaire to collect social survey data by considering the driving patterns, charging habits and willingness to charge in an orderly manner of electric vehicle users; Step 2: Extraction of deep correlation information from the questionnaire: Based on the social survey data, extract the deep correlation information of the social survey data in the form of multiple conditional probability distributions; Step 3: Multi-agent generation based on Monte Carlo random sampling: Using the extracted deep correlation information of the questionnaire as sampling data, a multi-agent model reflecting the orderly charging intention of the sampled population is constructed through the Monte Carlo stratified sampling method. Step 4: Validation of the surrogate model based on error analysis: Based on different numbers of surrogates, analyze the error sensitivity of the statistical distribution between the simulation results of the surrogate model and the questionnaire statistical results for each number of surrogates to verify the accuracy of the surrogate model; Step 5: Determining the required number of agents based on error analysis: Quantify the relationship between the number of agents and the model accuracy through error sensitivity analysis results, and determine the required number of agents; Step 6: Proxy attribute expansion based on consistency verification of social survey and sensor data: Extract the proxy feature attributes generated from the social survey data, match them with the sensor data collected by automated sensors to complete the consistency verification of the two types of data, and use the sensor data that has passed the verification to expand the proxy feature attributes to obtain the final electric vehicle user intention model.
2. The method for constructing an electric vehicle user intention model that integrates social survey and sensor data according to claim 1, characterized in that, In step 1, the survey on driving travel patterns specifically collects the following parameters: departure time to work t1, arrival time at work t2, departure time from get off work t3, arrival time home t4, and daily travel distance d. trip ; The survey on charging habits specifically collected data on users' charging methods. mode This includes charging at home slow charging stations, charging at company slow charging stations, charging at both home and company slow charging stations, and charging at public fast charging stations. The survey on willingness to charge in an orderly manner specifically collected data on users' acceptance of orderly charging. order Preferred subsidy methods mode And the expected subsidy discount / amount fee .
3. The method for constructing an electric vehicle user intention model that integrates social survey and sensor data according to claim 1, characterized in that, Step 2, the extraction of deep-seated correlation information from the questionnaire, includes: Step 2-1: Based on the collected social survey data, statistically analyze and extract the probability distribution P1(c) of the charging methods of the sampled population. mode ); Step 2-2: Based on the obtained probability distribution of charging methods, extract the conditional probability distribution P2(t) of user travel time under slow charging method. trip |c mode ); Steps 2-3: Based on the conditional probability distribution of travel time, extract the conditional probability distribution P3(d) of user travel distance under different travel times. trip |t trip ), and the conditional probability distribution P4(c) of users accepting orderly charging under different travel times. order |t trip ); Steps 2-4: Under the condition that the user accepts orderly charging, extract the conditional probability distribution P5(s) of the user's choice of charging subsidy method. mode |c order ); Steps 2-5: Under the subsidy method selected by the user, extract the conditional probability distribution P6(s) of the user's expected subsidy discount / amount. fee |s mode ).
4. The method for constructing an electric vehicle user intention model that integrates social survey and sensor data according to claim 1, characterized in that, In step 3, a multi-agent model is constructed based on Monte Carlo stratified sampling, including: Step 3-1: Number the agents using parameter i, i = 1, 2 ... N, where N is the total number of agents. Initially, i = 1. Step 3-2, Generate Given a uniformly distributed random number, determine the probability distribution interval of the charging method to which the random number falls, and use that charging method as the charging method c for agent i. i.mode ; Step 3-3, Generate A random number is uniformly distributed between the given time intervals. The probability distribution interval corresponding to the travel time of this random number is determined, and this travel time is taken as the travel time t of agent i. i.trip ; Steps 3-4: Generate A uniformly distributed random number is used to determine the probability distribution interval corresponding to the travel distance of the random number, and this travel distance is used as the proxy. Travel distance d i.trip ; Steps 3-5: Generate Given a uniformly distributed random number, determine the probability distribution interval within which the random number falls (corresponding to either accepting or not accepting ordered charging), and obtain the proxy. Does it accept the state of orderly charging? i.order ; Steps 3-6, if the agent The state is to receive orderly charging and generate A random number, c, that is uniformly distributed between the two. order =1 indicates that it accepts ordered charging, c order =0 indicates that ordered charging is not accepted. Determine the probability distribution interval corresponding to the subsidy method of the random number and determine the agent. Desired orderly charging subsidy method i.mode If agent i's state is "do not accept orderly charging", then proceed to steps 3-8. Steps 3-7: Generate Given a uniformly distributed random number, determine the probability distribution interval of the random number corresponding to different subsidy discount / amount values, and obtain the agent. Expected orderly charging subsidy discount / amount i.fee ; Step 3-8: If i = i + 1, determine if i is greater than N. If not, jump to step 3-2 to continue generating the next agent; if yes, the agent model is completed and the sampling process ends.
5. The method for constructing an electric vehicle user intention model that integrates social survey and sensor data according to claim 1, characterized in that, In step 4, the mean and variance of the relative error between the simulation results of the agent model and the questionnaire statistics are analyzed when the number of agents is different. By analyzing the convergence of the error, the accuracy of the constructed agent model is verified.
6. The method for constructing an electric vehicle user intention model that integrates social survey and sensor data according to claim 1, characterized in that, In step 5, based on the error sensitivity data corresponding to different agent numbers, the relationship between agent number and model accuracy is quantified, and the agent number corresponding to the maximum acceptable error is taken as the minimum agent number requirement.
7. The method for constructing an electric vehicle user intention model that integrates social survey and sensor data according to claim 1, characterized in that, In step 6, the characteristic attributes of the agent generated based on the social survey data include: charging method, trip start time, trip end time, trip distance, willingness to accept orderly charging, subsidy method for accepting orderly charging, and expected subsidy discount / amount when accepting orderly charging. Through feature matching, the social survey data and the sensor data collected by the automated equipment are checked for consistency. After the check is completed, the sensor data is used to expand the characteristic attributes of the agent.