A Method for Locating Electric Vehicle Charging Stations Based on User Behavior

Abstract

The present invention provides a method for locating electric vehicle charging stations based on user behavior, including: Step 1, mining based on vehicle data, time data, location data, trajectory data, POI data, and SOC data to obtain the probability distribution of feature quantities, and using the Monte Carlo simulation method to predict the charging demand of electric vehicles; Step 2, based on the charging demand of electric vehicles predicted in Step 1, constructing an electric vehicle charging station location model using a multi-objective decision-making method; Step 3, based on the constructed electric vehicle charging station location model, using the simulation method and the improved moth-flame algorithm to solve the model to obtain specific location decisions. The method for locating electric vehicle charging stations proposed by the present invention makes full use of the behavior data of electric vehicle users to mine specific user needs, can provide a basis for the planning of electric vehicle charging stations in actual scenarios, and provide more flexible location decisions for charging stations based on specific needs.

Description

Technical Field

[0001] The present invention relates to the technical field of big data, and specifically to a method for locating electric vehicle charging stations based on user behavior. Background Art

[0002] According to the statistics of the International Energy Agency, the energy consumption and related carbon emissions of the transportation system account for about 29% and 23% of the world respectively, which are one of the important roots of the current energy crisis and environmental pollution problems. To achieve energy conservation and emission reduction in the transportation system, China has invested a large amount of funds to vigorously promote the development of the electric vehicle industry. Driven by the national dual-carbon strategy and the "new infrastructure" policy, the scale of electric vehicles will expand rapidly. In this context, it is necessary to study data-driven electric vehicle charging station location technology to guide relevant electric vehicle services and provide theoretical and technical support for improving the user experience of electric vehicles.

[0003] Current related research mostly starts from the perspectives of the power grid and operators, mainly considering the distribution network or cost. For example, the application (patent) number: 202011051740.X, title: A method for modeling the location of electric vehicle charging stations based on an improved whale algorithm. This patent adds the influence of service capacity factors to the traditional research on charging station location, constructs a location model for electric vehicle charging stations that minimizes cost and restricts service capacity, but does not optimize the location from the perspective of electric vehicle users. The application (patent) number: 202110291281.0, title: A method and system for optimizing the location of electric vehicle charging stations. This patent mainly considers factors such as population distribution, regional function, and geographical location, constructs a charging station location model considering construction cost and distance condition constraints for location selection, and it cannot better conduct location selection based on the specific needs of users.

[0004] Some research starts from the user perspective, mainly predicting user needs based on the usage data of charging stations, and then further considering the charging station location problem. For example, the application (patent) number: 201910167083.6, title: A method and system for locating electric vehicle charging piles. The data mainly based on this patent includes the map of the target area, population density distribution data, information point data, and multiple potential charging event data, and then conducts location selection based on achieving the optimal satisfaction rate of random charging events and the utilization rate of charging piles, lacking consideration based on user usage behavior and user convenience and charging station cost.

[0005] In addition, some studies have begun to introduce the travel data of electric vehicles to calculate the charging load, so as to guide the site selection. However, there is still room for improvement in some aspects such as the sufficiency of feature dimensions. For example, the application (patent) number: 201910820177.9, the name: Charging station planning method considering user charging experience and distribution network operation risk. The feature dimensions considered in this patent mainly include: the charging station address and the electric vehicle capacity in each area, charging process parameters, distribution network, etc. Based on this, the charging load is calculated, and the charging station planning is carried out by integrating the needs of user charging experience, grid operation conditions and charging station construction costs. This patent introduces data in dimensions such as the historical travel data and state of charge of electric vehicles, and can consider from the user's perspective to a certain extent, but lacks the consideration of features such as spatio-temporal flow characteristics, geographical factors (such as POIs, road segments, etc.), and speed, and the comprehensiveness is insufficient. The application (patent) number: 201910534374.4, the name: A charging station site selection and capacity determination strategy considering the transfer of user charging needs. This patent mainly aims at the randomness of user charging behavior and the impact of charging behavior on the power quality of the power grid. Based on the historical driving trajectory data of electric vehicles and the service profit model of the construction, operation and maintenance costs of charging piles, according to the mileage of electric vehicles and objective geographical environment factors, the site selection and capacity determination of electric vehicle charging stations are carried out to maximize the profit of charging service providers and reduce the impact on the power quality of the power grid. However, this patent does not predict the charging behavior of individual electric vehicles, but uses the historical driving trajectory of vehicles to formulate the optimal site selection and capacity determination plan, and the static data will affect the results to a certain extent.

[0006] In summary, the existing research rarely makes full use of the behavior data of electric vehicle users to mine specific user needs for charging station site selection. Summary of the Invention

[0007] In view of the above deficiencies in the existing technology, the present invention provides an electric vehicle charging station site selection method based on user behavior, which can provide a more flexible charging station site selection decision based on specific needs.

[0008] The technical solution adopted by the present invention is as follows:

[0009] Step 1: Mine based on vehicle data, time data, location data, trajectory data, POI data, and SOC data to obtain the probability distribution of feature quantities, and use the Monte Carlo simulation method to predict the charging demand of electric vehicles;

[0010] Step 2: Based on the charging demand of electric vehicles predicted in Step 1, use the multi-objective decision-making method to construct an electric vehicle charging station site selection model;

[0011] Step 3: Based on the constructed site selection model for electric vehicle charging stations, use the method of simulation and the improved moth - flame algorithm to solve the model and obtain specific site selection decisions. Further, the specific content of Step 1 includes:

[0012] A. Obtain data, including vehicle data, time data, location data, trajectory data, POI data, and SOC data of electric vehicle users;

[0013] B. Pre - process the data obtained in Step A;

[0014] C. Mine the data pre - processed in Step B, including map matching, generating regenerated data, spatial modeling through grid division, POI identification, and mining of spatio - temporal flow characteristics of functional areas;

[0015] D. Based on the data mining results in Step C, conduct simulation of the travel process and charging decision of electric vehicle users, calculate the charging demand of users based on the travel process simulation and charging decision simulation, and obtain the theoretical charging demand of each region.

[0016] Further, in Step 2, based on the predicted charging demand of electric vehicles, use the multi - objective decision - making method to construct an electric vehicle charging station site selection model, specifically:

[0017] (1) Take meeting the charging demand of electric vehicle users with as high a service level as possible as the goal, and set constraints for meeting the charging demand of electric vehicles. The constraints for meeting the charging demand of electric vehicles include the maximum charging distance constraint and the charging demand satisfaction rate constraint:

[0018] Suppose there is a set of demand points M in the research area, the number of demand points is m, the maximum number of charging stations is n, the set of charging stations is N, and the maximum charging distance acceptable to users is D. Then the average charging distance of charging stations can be expressed by the following formula:

[0019]

[0020] where, d i is the average charging distance of the i - th charging station, (x i , y i ) are the coordinates of the i - th charging station, (x j , y j ) are the coordinates of the demand point, M i represents the set of demand points served by the i - th charging station, and m i is the number of demand points served by the i - th charging station;

[0021] Among them, the maximum charging distance constraint is:

[0022] The demand satisfaction rate of the charging station includes the satisfaction rates of fast charging and slow charging, which is determined by the demand satisfaction rate during the time period with the highest charging demand in the area. The charging demand is the number of electric vehicles that need to be charged at the same moment. The constraints for the charging station demand satisfaction rate are as follows:

[0023]

[0024]

[0025] Among them, SC i is the slow charging satisfaction rate of the i-th charging station; FC i is the fast charging satisfaction rate of the i-th charging station; m i1 is the number of electric vehicles with slow charging demand; m i2 is the number of electric vehicles with fast charging demand; n i1 is the number of n i1 slow charging piles at the i-th charging station; n i2 is the number of n i2 fast charging piles at the i-th charging station; CL is the minimum level of charging demand satisfaction;

[0026] (2) With the goal of minimizing the total construction cost of the charging station, set up the objective function: mincost 总 = cost 固定 + cost 可变 ;

[0027] (3) Combining step (1) and step (2), the charging station location model is obtained as follows: Objective function:

[0028] Minimize the average charging distance:

[0029]

[0030] Minimize the charging station cost:

[0031] mincost 总 = cost 固定 + cost 可变

[0032] Constraints:

[0033]

[0034]

[0035]

[0036] Further, in step three, based on the established siting model of electric vehicle charging stations, the simulation method and the improved moth - flame algorithm are used to solve the model, and specific siting decisions are obtained, that is, the number of charging stations and the siting locations, specifically as follows:

[0037] According to the charging demand, the improved moth - flame algorithm is used to simulate the siting and capacity determination, find the dominant solution set, and obtain the final siting decision. The specific steps are as follows:

[0038] ① Input the charging demand data;

[0039] ② Initialize the parameters of the maximum number of stations N, population size S, maximum number of iterations K, maximum charging distance, and maximum service capacity of the charging station;

[0040] ③ Randomly generate the initial generation of moth populations among the existing demand points;

[0041] ④ Start the iteration, and calculate the number of flames f in this iteration. If it is the first iteration, calculate the fitness of the first - generation moth population, sort the population based on this, and use the sorted population as the flame matrix, then jump to step ⑦. If it is not the first iteration, go to step ⑤;

[0042] ⑤ Combine the new population and the unsorted population in the previous iteration, calculate the population fitness on this basis, and sort the new population based on the individual fitness;

[0043] ⑥ For the sorted population, take the first - half of the individuals as the optimal population in this iteration, and take the first f individuals in this part as the flame matrix;

[0044] ⑦ Let the moths gradually approach the flames, and update the moth population based on this;

[0045] ⑧ Judge whether to end the iteration according to the termination condition. If the termination condition is met, terminate the iteration and output the data; otherwise, go back to step ④ and continue the iteration;

[0046] As the number of iterations increases, the values of the two objective functions: the average charging distance and the construction cost of the charging station gradually converge, and all Pareto - dominant solutions are obtained. The surface composed of the dominant solutions is the Pareto - front surface. When the breadth and uniformity of the Pareto - front surface distribution are good, the method is effective. There is no distinction between good and bad among the dominant solutions. When comparing two Pareto - dominant solutions, if one solution is better than the other in a certain objective, its value in the other objective must be worse than the other solution. Select the final siting plan according to the specific situation;

[0047] The calculation process of population fitness is as follows: first, input the moth population data and calculate the objective function values ​​of each individual; then, perform non-dominated stratification according to the individual objective function values; then, calculate the crowding degree of individuals at each level. The specific formula is as follows:

[0048]

[0049] In the formula, i+1 and i-1 are two individuals adjacent to the i-th individual; d j i is the crowding distance of the jth objective function of the ith individual; f j i+1 is the value of the jth objective function of the i+1th individual; f j max and f j min are the maximum and minimum values ​​of the jth objective function respectively; furthermore, individuals are sorted based on the levels after non-dominated stratification and the crowding degree of each level. If the levels are different, the smaller the level, the better the individual; if the levels are the same, the greater the crowding degree, the better the individual. Finally, the fitness of each individual is calculated, the best is 1, the worst is 0, and the values ​​in between are equally spaced between 0 and 1.

[0050] Furthermore, the vehicle data described in step A includes vehicle ID; the time data includes the start time of the journey, the arrival / departure time of each destination along the way, and the stay time; the location data includes the starting / end location and the stop location; the track data includes track GPS latitude, track GPS longitude, speed data, and direction data.

[0051] Furthermore, in step B, the data obtained in step A is preprocessed, specifically:

[0052] First, the data of a single day is divided into several sets;

[0053] Then, bad data are deleted, including: data that is not within the study area; multiple repeated data within a short distance of the same vehicle in the same time period; data with abnormal speed; data with abnormal deviation;

[0054] Finally, based on specific needs, the POI data is classified into: residential areas, commercial areas, industrial areas, and public service areas.

[0055] Furthermore, in step C, the data preprocessed in step B is mined, specifically:

[0056] First, map matching is performed, that is, matching the GPS trajectory with the road network model;

[0057] Secondly, generate the regenerative feature data: OD travel set, road network travel speed set, actual driving path set;

[0058] Then, conduct spatial modeling, select an appropriate scale to divide the research area into grids. The division method is as follows: ① Divide the spatial range of the research at different scales and number the grids: A u (j) = {A u (1), A u (2),... A u (m)}, where u is the division scale, u ∈ [50, 500], with an interval of 50 m; A u (j) is the number of the jth grid at scale u; m is the total number of grids divided; ② Calculate the amount of valid data within the grids at different spatial scales: where E u (i) is the number of valid OD points in the ith grid. Through the above two steps, obtain the optimal u value and conduct grid division based on this;

[0059] Furthermore, conduct POI recognition: Combine the actual situation to classify POIs. On this basis, identify the functional areas based on the POI data, that is, calculate the proportion of the density of various POIs within the grids to determine the functional nature of the grids and the functional types of the grids: ① Calculate the number of various POIs in the OD point set of each grid: where N i k is the number of k-type POIs in the ith grid; n o (i, m, k) is the starting point of the mth k-type in the ith grid; n d (i, n, k) is the ending point of the nth k-type in the ith grid; ② Calculate the POI type situation in each grid: N i = ∑n(i, k), where N i is the total number of POIs in the ith grid, and n(i, k) is the number of k-type POIs in the ith grid; ③ Aggregate the grids as needed;

[0060] Finally, mine the spatio-temporal flow characteristics of the functional areas, and combine the OD set and POI recognition to mine the spatio-temporal flow characteristics of the functional areas.

[0061] Furthermore, step D specifically includes:

[0062] First, conduct the simulation of the electric vehicle travel process: Based on the data mining results in step C, determine the probability distribution of each travel characteristic quantity of the electric vehicle. For each characteristic quantity, use the Monte Carlo simulation method to extract random numbers that conform to its probability distribution to realize the simulation of the travel process of each electric vehicle;

[0063] Then, conduct a simulation of the charging decision of electric vehicle users: Based on three factors, namely, the existing remaining battery power of the electric vehicle, the distance of the next trip, and the staying duration at the current destination, judge whether the electric vehicle will charge and the charging duration after arriving at a certain location. Use the Monte Carlo simulation method to extract random numbers that follow the probability distributions of these three characteristic quantities and conduct simulations on a computer to obtain the charging data of each electric vehicle each time. Based on this, make a decision judgment;

[0064] Furthermore, integrate the simulation of the electric vehicle travel process and the simulation of the charging decision process to obtain the charging demand data of the electric vehicle in a day, including the location where the charging demand occurs, the time when it occurs, and the charging duration;

[0065] Finally, conduct the calculation of the charging demand: Divide the research area into grids; the demands in the same small grid area are regarded as occurring at the same location; call the center of each small grid area the theoretical charging demand occurrence point of this area; divide the time of a day into 24 time periods per hour, and count the fast charging demand FQ need-t and the slow charging demand SQ need-t in each small area for each time period; select the demand in the time period with the largest demand as the theoretical charging demand Q need of this area, then the following formula can be obtained:

[0066] Q need =max{FQ need-t +SQ need-t}(t=1,2,...24).

[0067] Further, the decision judgment specifically includes:

[0068] Fourth, judge whether to charge: When SOC<SOC min +Q next , then choose to charge; when SOC≥SOC min +Q next , then choose not to charge, where SOC is the current state of charge of the electric vehicle, SOC min is the lowest state of charge acceptable to the user, and Q next is the power required for the next trip;

[0069] Fifth, judge which charging mode to choose: When SOC+Q 慢 ≥SOC min +Q next , choose slow charging; when SOC+Q 慢 <SOC min +Q next , choose fast charging, where Q 慢 is the slow charging amount during the staying time;

[0070] Sixth, determine the charging amount and charging duration: If it can be fully charged within the stay time, then Q 充 = SOC max - SOC, If it cannot be fully charged within the stay time, then t = t 停留 , Q 充 = t * P, where Q 充 is the charging amount, SOC max is the power of the fully charged state, t is the charging duration, P is the charging power, and t 停留 is the stay duration.

[0071] Furthermore, the travel characteristic quantities include travel location, travel time, travel destination, arrival time at the destination, travel distance, and stay duration at the destination.

[0072] The present invention provides a method for locating an electric vehicle charging station based on user behavior. From the user's perspective, using data from multiple dimensions such as the trajectory data, geographical data, and battery data of electric vehicle users, the real needs of users are fully explored. Based on this, the location of the electric vehicle charging station is selected. In the specific location selection, optimization is carried out simultaneously from the user's perspective and the charging station's perspective, using a multi-objective decision-making method: constructing an objective function based on minimizing the average charging distance and minimizing the charging station cost, constructing constraint conditions based on constraints such as charging distance and demand satisfaction rate, using the Monte Carlo method to simulate the location selection, and finally using an improved moth-flame algorithm to solve, which can provide a more flexible charging station location decision based on specific needs. The present invention can provide a basis for the planning of electric vehicle charging stations in actual scenarios. Description of the Drawings

[0073] Figure 1 is a flowchart of an embodiment of a method for locating an electric vehicle charging station based on user behavior according to the present invention;

[0074] Figure 2 is the relationship between the real charging demand and the distance from the planned charging station. Detailed Embodiments

[0075] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0076] Please refer to Figure 1, an embodiment of the present invention provides a method for locating an electric vehicle charging station based on user behavior. First, predict the charging demand based on electric vehicle travel data; secondly, construct a location model for the electric vehicle charging station using a multi-objective decision-making method; finally, solve the model using a simulation method and an improved moth-flame algorithm to obtain a specific location decision. The specific steps are as follows:

[0077] A. Obtain the following data from the data source: vehicle data (vehicle ID); time data (start time of the trip, end time of the trip); location data (starting location of the trip, ending location of the trip); trajectory data (latitude of the trajectory GPS, longitude of the trajectory GPS, speed data); POI data; SOC data.

[0078] B. Preprocess the obtained original data, specifically:

[0079] First, divide the data by single-day data into several sets.

[0080] Then, delete the bad point data, where the bad point data includes: data outside the research area; multiple duplicate data within a short distance for the same vehicle within the same time period; trajectory data with abnormal speeds such as instantaneous speeds exceeding 120 km / h; abnormally offset data.

[0081] Finally, based on specific requirements, determine the classification of the POI data, specifically classified as: residential area, industrial area, commercial area, public service area.

[0082] C. Mine the data preprocessed in step B, specifically:

[0083] First, perform map matching, that is, match the GPS trajectory with the road network model.

[0084] Secondly, generate regenerated feature data: OD trip set, road network travel speed set, actual driving path set.

[0085] Then, perform spatial modeling. Select an appropriate scale to divide the research area into grids. The specific division method is as follows: ① Divide the research spatial range at different scales and number the grids: L u (i) = {L u (1), L u (2),... L u (n)}, where u is the division scale, the range of u is 50 - 500 m, and values are taken at intervals of 50 m; L u (i) is the number of the jth grid at scale u; n is the total number of grids divided. ② Calculate the effective data volume within the grid at different spatial scales: Among them, V u(i) is the number of valid OD points in the i-th grid. Through the following two steps, the optimal u value is obtained, and grid division is carried out based on this.

[0086] Furthermore, POI recognition is carried out. Considering the actual situation, POI classification is performed. Based on this, the identification of functional areas is carried out based on POI data, that is, the functional nature of the grid is determined by calculating the proportion of the density of various types of POIs in the grid, and the functional type of the grid is determined: ① Calculate the number of various types of POIs in the OD point set in each grid: Among them, N i k is the number of POIs of type k in the i-th grid; n o n(i,k) is the number of starting points of type k in the i-th grid; n d n(i,k) is the number of ending points of type k in the i-th grid. ② Calculate the POI type situation in each grid: Among them, N i is the total number of POIs in the i-th grid, N i k is the number of POIs of type k in the i-th grid. ③ Aggregate the grids as needed.

[0087] Finally, the spatio-temporal flow characteristics of the functional areas are mined. Combining OD set POI recognition, the spatio-temporal flow characteristics of the functional areas are mined.

[0088] D. Based on the data mining results in step C, the construction of an electric vehicle charging demand model is carried out, specifically as follows:

[0089] First, the travel process of electric vehicles is simulated. Based on the travel data mined in step C, the probability distributions of various travel characteristic quantities (travel location, travel time, travel destination, arrival time at the destination, travel distance, destination stay duration) of electric vehicles are determined. For each characteristic quantity, the Monte Carlo simulation method is used to extract random numbers that conform to its probability distribution, and the travel process of each electric vehicle can be simulated. Among them, the determination of the mathematical expectation of the random variable q in the Monte Carlo simulation is obtained by performing N repeated samplings on the random variable ε to obtain n observed values, and then further calculated according to the following formula:

[0090]

[0091] Then, the charging decision simulation of electric vehicle users is carried out. The judgment on whether to charge and the charging duration after an electric vehicle arrives at a certain location is mainly based on three factors: the existing remaining power of the electric vehicle battery, the distance of the next trip, and the staying duration at the current destination. The Monte Carlo simulation method is used to extract random numbers that obey the probability distributions of these three characteristic quantities and perform simulations on a computer to obtain the charging data of each electric vehicle each time. On this basis, three main decision-making judgments are carried out:

[0092] First, judge whether to charge. When SOC p <SOC min +N next , then choose to charge; when SOC p ≥SOC min +N next , then choose not to charge. Among them, SOC p is the current state of charge of the electric vehicle, SOC min is the minimum state of charge acceptable to the user, and N next is the power required for the next trip.

[0093] Second, judge which charging mode to choose. When SOC p +N 慢 ≥SOC min +N next , choose slow charging; when SOC p +N 慢 <SOC min +N next , choose fast charging. Among them, N 慢 is the slow charging amount during the staying time.

[0094] Third, judge the charging amount and charging duration. If it can be fully charged during the staying time, then N 充 =SOC max -SOC p , If it cannot be fully charged during the staying time, then t=t 停留 , N 充 =t*P. Among them, N 充 is the charging amount, SOC max is the power in the fully charged state, t is the charging duration, P is the charging power, and t 停留 is the staying duration.

[0095] Furthermore, the simulation of the electric vehicle travel process and the simulation of the charging decision process are integrated to obtain the charging demand data of the electric vehicle for one day, including the location where the charging demand occurs, the time when it occurs, and the charging duration.

[0096] Finally, the charging demand calculation is carried out. The research area is divided into grids; the demands in the same small grid area are regarded as occurring at the same location; the center of each small grid area is called the theoretical charging demand occurrence point of this area; the time of a day is divided into 24 time periods per hour, and the fast charging demand FN t and the slow charging demand SN t in each small area for each time period are counted; the demand in the time period with the largest demand is selected as the theoretical charging demand T need of this area, and the following formula can be obtained:

[0097] T need =max{FN t +SN t}(t=1,2,...24)

[0098] E. Based on the electric vehicle charging volume demand predicted in steps A - D, a charging station location model is constructed: with the goal of minimizing the average charging distance and the total construction cost, setting the maximum charging distance constraint of electric vehicles and the service capacity constraint of charging stations, and finally using the multi - objective optimization method to solve the Pareto - optimal solution of the charging station location and capacity determination, specifically:

[0099] (1) With the goal of meeting the charging needs of electric vehicle users with as high a service level as possible, set the charging demand satisfaction constraint. The charging demand satisfaction constraint mainly includes two points: the maximum charging distance constraint and the charging demand satisfaction rate constraint:

[0100] First, the maximum charging distance constraint:

[0101] Let the average charging distance be: where s avg is the average charging distance of a certain charging station, s n is the distance from a certain demand point served by this charging station to the charging station, n is the total demand, and it is necessary to first determine which demand points are served by the charging station: First, from the perspective of demand points, according to the charging distance constraint, find the nearest charging station for each demand point; then, from the perspective of charging stations, divide the demand points within the service range of each charging station and calculate the average service distance of the charging station. The specific steps are as follows:

[0102] ① When setting the maximum value of the number of charging stations, in the first iteration, randomly select the same number of demand points as the number of stations from the demand points as the initial positions of the charging stations;

[0103] ② Determine which charging stations exist within the maximum charging distance range of the demand points according to the maximum charging distance constraint. If there are no charging stations within the maximum charging distance range of the demand points, add a penalty value to the corresponding objective function, and the farther the distance, the greater the penalty value;

[0104] ③ Select the charging station closest to the demand point from the set of charging stations obtained in the first step as the station for it to receive charging services according to the minimum charging distance constraint;

[0105] ④ After each demand point finds the station for it to receive charging services. Using reverse thinking, from the perspective of the charging stations, determine the set of demand points served by each charging station;

[0106] ⑤ Calculate the average charging distance of each charging station and the average charging distance of the entire area to be laid out;

[0107] Suppose there is a set of demand points M in the research area, the number of demand points is m, the maximum number of charging stations is n, the set of charging stations is N, and the maximum charging distance acceptable to users is D. Then the average charging distance of the charging stations can be expressed by the following formula:

[0108]

[0109] Among them, d i is the average charging distance of the i-th charging station, (x i , y i ) are the coordinates of the i-th charging station, (x j , y j ) are the coordinates of the demand points, M i represents the set of demand points served by the i-th charging station, m i is the number of demand points served by the i-th charging station;

[0110] The average charging distance of the research area is the average value of the average charging distances of all charging stations in this area. Then the formula for minimizing the average charging distance of the research area is as follows:

[0111]

[0112] The maximum charging distance constraint is:

[0113]

[0114] Second, the charging demand satisfaction rate constraint:

[0115] The demand satisfaction rate of the charging stations includes the satisfaction rates of fast charging and slow charging, which is determined by the demand satisfaction rate during the time period with the largest charging demand in this area. The charging demand is the number of electric vehicles that need to be charged at the same moment. The specific charging station demand satisfaction rate constraint is as follows:

[0116]

[0117]

[0118] Among them, SC i is the slow charging satisfaction rate of the i-th charging station; FC i is the fast charging satisfaction rate of the i-th charging station; m i1 is the number of electric vehicles with slow charging demand; m i2 is the number of electric vehicles with fast charging demand; n i1 is the number of slow charging piles in the i-th charging station with n i1 ; n i2 is the number of fast charging piles in the i-th charging station with n i2 ; CL is the minimum level of charging demand satisfaction;

[0119] (2) With the goal of minimizing the total construction cost of the charging station, set up the objective function: mincost 总 = cost 固定 + cost 可变 ;

[0120] (3) Combining step (1) and step (2), the charging station site selection model is obtained as follows: Objective function:

[0121] Minimize the average charging distance:

[0122]

[0123] Minimize the charging station cost:

[0124] mincost 总 = cost 固定 + cost 可变 .

[0125] Constraints:

[0126]

[0127]

[0128]

[0129] F. Based on the electric vehicle charging station site selection model constructed in step E, use simulation and the improved moth - flame algorithm to solve the model, and obtain the final site selection decision, specifically:

[0130] First, use the Monte Carlo method to simulate the charging demand prediction process and obtain the charging demand data;

[0131] Then, according to the charging demand, the improved moth-flame algorithm is used to simulate the site selection and capacity determination, find the Pareto optimal solution, and obtain the final site selection decision. The specific steps are as follows:

[0132] ① Input the charging demand data;

[0133] ② Initialize fixed parameters such as the maximum number of stations N, population size S, maximum number of iterations K, maximum charging distance, and maximum service capacity of the charging station;

[0134] ③ Randomly generate a primary generation of moth populations among the existing demand points;

[0135] ④ Start the iteration and calculate the number of flames f in this iteration. If it is the first iteration, calculate the fitness of the first generation of moth populations, sort the populations based on this, use the sorted populations as the flame matrix, and then jump to step ⑦. If it is not the first iteration, go to step ⑤;

[0136] ⑤ Combine the new population and the unsorted population in the previous iteration, calculate the population fitness based on this, and sort the new population based on the individual fitness;

[0137] ⑥ For the sorted population, take the first half of the individuals as the optimal population in this iteration, and take the first f of these individuals as the flame matrix;

[0138] ⑦ Let the moths gradually approach the flames and update the moth populations based on this;

[0139] ⑧ Judge whether to end the iteration according to the termination condition. If the termination condition is met, terminate the iteration and output the data; otherwise, go to step ④ and continue the iteration.

[0140] Among them, the calculation process of the population fitness is as follows: First, input the moth population data and calculate the objective function values of each individual; then, perform non-dominated sorting according to the individual objective function values. Next, calculate the crowding degree of individuals in each layer. The specific formula is as follows:

[0141]

[0142] In the formula, i + 1 and i - 1 are two individuals adjacent to the i-th individual; d j i , is the crowding distance of the i-th individual on the j-th objective function; f j i+1 is the value of the j-th objective function of the (i + 1)-th individual; f j max and f j minThey are respectively the maximum and minimum values of the j-th objective function. Furthermore, individuals are sorted based on the levels after non-dominated sorting and the crowding degree of each level. For different levels, the smaller the level, the better the individual; for the same level, the larger the crowding degree, the better the individual. Finally, the fitness of each individual is calculated, with the best being 1, the worst being 0, and the intermediate values being evenly spaced between 0 and 1.

[0143] Finally, the selected location, the number of charging facilities, and the cost in the research area that meet the electric vehicle charging station location model of the present invention are shown in Table 1.

[0144] Table 1

[0145]

[0146] In the verification of the planning results, the effectiveness is verified by comparing the number of charging demands during the peak period of the charging station with the shortest distance between the planned charging stations. The comparison results are shown in Table 2.

[0147] Table 2

[0148]

[0149] On this basis, the relationship between the actual charging demand and the distance of the planned charging stations is more intuitively reflected through a line chart, as Figure 2 shown. The results show that the actual charging demand of charging stations farther away from the planned charging stations is less; conversely, the actual charging demand of charging stations closer to the planned charging stations is more, which can to a certain extent verify the rationality of the planning model and results.

[0150] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Anyone can still modify or equivalently replace the specific implementation manners of the present invention, and any modification or equivalent replacement without departing from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A method for locating electric vehicle charging stations based on user behavior, characterized in that it includes the following steps: Step 1: Mine based on vehicle data, time data, location data, trajectory data, POI data, and SOC data to obtain the probability distribution of feature quantities, and use the Monte Carlo simulation method to predict the charging demand of electric vehicles; Step 2: Based on the predicted charging demand of electric vehicles in Step 1, use the multi-objective decision-making method to construct an electric vehicle charging station location model; Step 3: Based on the constructed electric vehicle charging station location model, use the simulation method and the improved moth-flame algorithm to solve the model to obtain specific location decisions; In Step 2, based on the predicted charging demand of electric vehicles, use the multi-objective decision-making method to construct an electric vehicle charging station location model, specifically: (1) Taking the goal of meeting the charging needs of electric vehicle users, set the constraints for meeting the charging needs of electric vehicles. The constraints for meeting the charging needs of electric vehicles include the maximum charging distance constraint and the charging demand satisfaction rate constraint: Suppose there is a set of demand points M in the research area, the number of demand points is m, the maximum number of charging stations is n, the set of charging stations is N, and the maximum charging distance acceptable to users is D. Then the average charging distance of the charging stations can be expressed by the following formula: ； Among them, d i is the average charging distance of the i-th charging station, (x i , y i ) are the coordinates of the i-th charging station, (x j , y j ) are the coordinates of the demand point, M i represents the set of demand points served by the i-th charging station, and m i is the number of demand points served by the i-th charging station; Among them, the maximum charging distance constraint is: ； The demand satisfaction rate of the charging station includes the satisfaction rates of fast charging and slow charging, which is determined by the demand satisfaction rate in the time period with the largest charging demand in this area. The charging demand is the number of electric vehicles that need to be charged at the same moment. The charging station demand satisfaction rate constraint is as follows: ； ； Among them, SC i is the slow charging satisfaction rate of the i-th charging station; FC i is the fast charging satisfaction rate of the i-th charging station; m i1 is the number of electric vehicles with slow charging demand; m i2 is the number of electric vehicles with fast charging demand; n i1 is the number of slow charging piles at the i-th charging station with n i1 ; n i2 is the number of fast charging piles at the i-th charging station with n i2 ; CL is the lowest level of charging demand satisfaction rate; (2) Taking the goal of minimizing the total construction cost of the charging station, set the objective function: ； (3) Combining Step (1) and Step (2), the following charging station location model is obtained: Objective function: Minimize the average charging distance: minimize ; Minimize the charging station cost: ； Constraint conditions: ； ； 。 2. The method for locating electric vehicle charging stations based on user behavior according to claim 1, characterized in that: The specific content of Step 1 includes: A. Obtain data, including vehicle data, time data, location data, trajectory data, POI data, and SOC data of electric vehicle users; B. Preprocess the data obtained in Step A; C. Mine the data preprocessed in Step B, including map matching, generating regenerated data, performing spatial modeling through grid division, POI recognition, and mining the spatio-temporal flow characteristics of functional areas; D. Based on the data mining results in Step C, perform the simulation of the travel process of electric vehicle users and the simulation of charging decisions, calculate the charging demand of users based on the travel process simulation and charging decision simulation, and obtain the theoretical charging demand in each area.

3. The method for locating electric vehicle charging stations based on user behavior according to claim 1, characterized in that: Step 3: Based on the constructed electric vehicle charging station location model, use the simulation method and the improved moth-flame algorithm to solve the model to obtain specific location decisions, that is, the number of charging stations and the location of the sites, specifically: According to the charging demand, the improved moth flame algorithm is used to simulate the site selection and capacity setting, find the dominant solution set, and obtain the final site selection decision. The specific steps are as follows: ① Input charging demand data; ② Initialize the parameters of the maximum number of stations N, population size S, maximum number of iterations K, maximum charging distance, and maximum service capacity of charging stations; ③ Randomly generate the primary generation of moth populations in the existing demand points; ④ Start iteration and calculate the number of flames f in this iteration. If it is the first iteration, calculate the fitness of the first generation of moth population, sort the population based on it, use the sorted population as the flame matrix, and then jump to step ⑦. If it is not the first iteration, go to step ⑤. ⑤Merge the new population with the unsorted population in the previous iteration, calculate the population fitness on this basis, and sort the new population based on individual fitness; ⑥ For the sorted population, take the first half of the individuals as the optimal population in this iteration, and take the first f of these individuals as the flame matrix; ⑦ Let the moths gradually approach the flame, and update the moth population based on this; ⑧Determine whether to end the iteration according to the termination condition. If the termination condition is met, terminate the iteration and output the data; otherwise, go to step ④ and continue the iteration; As the number of iterations increases, the values ​​of the two objective functions: average charging distance and charging station construction cost gradually converge, and all Pareto dominant solutions are obtained. The surface composed of dominant solutions is the Pareto frontier surface. When the breadth and uniformity of the Pareto frontier surface distribution are good, the method is effective. There is no difference between superior and inferior dominant solutions. When two Pareto dominant solutions are compared, if the value of one solution on a certain objective is better than that of another solution, its value on another objective must be worse than that of another solution. The final site selection plan is selected according to the specific situation. The calculation process of population fitness is as follows: first, input the moth population data and calculate the objective function values ​​of each individual; then, perform non-dominated stratification according to the individual objective function values; then, calculate the crowding degree of individuals at each level. The specific formula is as follows: ； where \(i + 1\) and \(i - 1\) are two individuals adjacent to the \(i\)-th individual; \(d\) j i is the crowding distance on the \(j\)-th objective function of the \(i\)-th individual; \(f\) j i+1 is the value of the \(j\)-th objective function of the \((i + 1)\)-th individual; \(f\) j max and \(f\) j min are the maximum and minimum values of the \(j\)-th objective function respectively; furthermore, the individuals are sorted based on the levels after non-dominated ranking and the crowding degree of each level. If the levels are different, the smaller the level, the better the individual; if the levels are the same, the larger the crowding degree, the better the individual. Finally, the fitness of each individual is calculated, with the best being 1, the worst being 0, and the intermediate values being evenly spaced between 0 and 1.

4. The method for selecting a site for an electric vehicle charging station based on user behavior as claimed in claim 2, Features: The vehicle data in step A includes the vehicle ID; the time data includes the start time of the journey, the arrival / departure time of each destination along the way, and the stay time; the location data includes the starting / end location and the stop location; the track data includes the track GPS latitude, track GPS longitude, speed data, and direction data.

5. The method for selecting a site for an electric vehicle charging station based on user behavior as claimed in claim 4, Features: In step B, the data obtained in step A is preprocessed, specifically: First, the data of a single day is divided into several sets; Then, bad data are deleted, including: data that is not within the study area; multiple repeated data within a short distance of the same vehicle in the same time period; data with abnormal speed; data with abnormal deviation; Finally, based on specific needs, the POI data is classified into: residential areas, commercial areas, industrial areas, and public service areas.

6. The method for locating an electric vehicle charging station based on user behavior according to claim 5, characterized in that: in step C, the data preprocessed in step B is mined, specifically: First, map matching is performed, that is, the GPS trajectory is matched with the road network model; Secondly, regenerative feature data is generated: OD trip set, road network co-travel speed set, actual driving path set; Then, perform spatial modeling, select an appropriate scale to divide the research area into grids. The division method is as follows: ① Divide the spatial range of the research at different scales and number the grids: , where u is the division scale, u ∈ [50, 500], with an interval of 50 m; A u (j) is the number of the j-th grid at scale u; m is the total number of grids divided; ② Calculate the amount of valid data within the grids at different spatial scales: , where E u (i) is the number of valid OD points in the i-th grid. Through the above two steps ① and ②, the optimal value of u is obtained, and grid division is carried out based on this; Furthermore, perform POI recognition: Combine the actual situation to classify POIs. On this basis, identify the functional areas based on the POI data, that is, calculate the proportion of the density of various POIs in the grid to determine the functional nature of the grid and determine the functional type of the grid: ① Calculate the number of various POIs in the OD point set in each grid: , where N i k is the number of POIs of type k in the i-th grid; n o (i, m, k) is the starting point of the m-th type k in the i-th grid; n d (i, n, k) is the ending point of the n-th type k in the i-th grid; ② Calculate the POI type situation in each grid: , where N i is the total number of POIs in the i-th grid, and n(i, k) is the number of POIs of type k in the i-th grid; ③ Aggregate the grids as needed; Finally, the spatio-temporal flow characteristics of the functional area are mined, and combined with the OD set POI recognition, the spatio-temporal flow characteristics of the functional area are mined.

7. The method for locating an electric vehicle charging station based on user behavior according to claim 6, characterized in that: step D specifically includes: First, simulate the electric vehicle travel process: based on the data mining results in step C, determine the probability distribution of each travel characteristic quantity of the electric vehicle. For each characteristic quantity, use the Monte Carlo simulation method to extract random numbers that conform to its probability distribution to realize the simulation of the travel process of each electric vehicle; Then, simulate the charging decision of the electric vehicle user: based on three factors, namely, the existing remaining battery power of the electric vehicle, the distance of the next trip, and the stay duration at the current destination, judge whether the electric vehicle will charge and the charging duration after arriving at a certain location. Use the Monte Carlo simulation method to extract random numbers that conform to the probability distribution of these three characteristic quantities and perform simulations on the computer to obtain the charging data of each electric vehicle each time. On this basis, make a decision judgment; Furthermore, integrate the simulation of the electric vehicle travel process and the simulation of the charging decision process to obtain the charging demand data of the electric vehicle in one day, including the location where the charging demand occurs, the time when it occurs, and the charging duration; Finally, the charging demand calculation is carried out: the research area is divided into grids; the demands in the same small grid area are regarded as occurring at the same location; the center of each small grid area is called the theoretical charging demand occurrence point of this area; the time of a day is divided into 24 time periods per hour, and the fast charging demand FQ need-t and slow charging demand SQ need-t in each small area for each time period are counted; the demand in the time period with the largest demand is selected as the theoretical charging demand Q need of this area, and the following formula can be obtained: 。 8. The method for locating an electric vehicle charging station based on user behavior according to claim 7, characterized in that: the decision judgment specifically includes: First, determine whether to charge: When occurs, then select charging; when occurs, then select not to charge, where SOC is the current state of charge of the electric vehicle, SOC min is the lowest state of charge acceptable to the user, and Q next is the amount of electricity required for the next trip. Second, determine which charging mode to select: When occurs, select slow charging; when occurs, select fast charging, where Q 慢 is the slow charging amount during the residence time; Third, determine the charging amount and charging duration: If it can be fully charged within the stay time, then , ; If it cannot be fully charged within the stay time, then , , where Q 充 is the charging amount, SOC max is the amount of electricity in the fully charged state, t is the charging duration, P is the charging power, and t 停留 is the stay duration.

9. The method for locating an electric vehicle charging station based on user behavior according to claim 7, characterized in that: the travel characteristic quantities include travel location, travel time, travel destination, arrival time at the destination, travel distance, and stay duration at the destination.

Images