Data-model fusion driven electric vehicle travel chain construction method and system

Abstract

The invention provides a data-model fusion driven electric vehicle travel chain construction method and system, and the method comprises the steps: selecting a construction region, carrying out the modeling of a road topology into a directed weighted network based on a complex network theory, and obtaining a road topology network; the method comprises the following steps: embedding traffic infrastructures into a road topology network structure, constructing a road network model in a real traffic scene, and constructing vehicle state information and constraints as a model driving part; according to historical travel chain data of the selected construction area, travel chain travel data is generated by combining a Gaussian mixture model and the constructed road topology network and serves as a data driving part; the data driving part and the model driving part are fused, typical vehicle parameters and a vehicle energy consumption power model are combined, and complete stroke chain data are obtained through calculation. According to the method, the formation chain data with complete spatio-temporal dynamic characteristics can be generated, and the defects of EV simulation data, the economical efficiency problem of real data collection and the user privacy problem are relieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power systems, in particular to a data-model fusion driven electric vehicle trip chain construction method and system. BACKGROUND

[0002] Electric vehicles (EV) and their related infrastructure have been developing rapidly. According to the forecast data of the International Energy Agency, the electricity demand of electric vehicles will reach 12500 Gwh in 2035. The rapid growth of large-scale EVs brings significant challenges to the optimal operation of the power grid. However, due to privacy protection, variable scenarios, and other issues, it is challenging to monitor and obtain electric vehicle energy consumption information and charging information in reality.

[0003] Currently, academic research and project practice rely on high-quality electric vehicle trip, energy consumption, and charging data. Preprocessing of raw data is required, including outlier identification, missing value filling, and data standardization, to ensure the accuracy and effectiveness of the adjustable capacity. It is difficult to obtain comprehensive electric vehicle operation data covering all possible operating conditions in the real distribution network. The data used in existing EV research is mostly obtained from two sources. One is based on statistical sampling methods, and the other is real data. The data from these two sources has certain problems. The trip behavior and charging behavior of electric vehicles obtained by statistical sampling methods are usually used to verify the effectiveness of the method, but the data obtained in this way cannot guarantee the real-time characteristics, thus limiting the actual application. The method based on real data is subject to economic and privacy issues and cannot be applied on a large scale.

[0004] To address the lack of spatio-temporal characteristics of EV trip chain construction and the user privacy problem of real data, the present application proposes a data-model fusion driven electric vehicle trip chain construction method: a high-fidelity road network model under real traffic scenarios is constructed as the model-driven part; combined with the constructed road topology network, the starting and ending nodes (OD pairs), path length, and travel time of each trip chain are generated as the data-driven part, an electric vehicle trip chain energy model is constructed based on the data-model fusion mechanism, combined with typical vehicle parameters including battery capacity, unit energy consumption coefficient, and maximum charging power, an energy consumption power model of the vehicle is established, and the starting and ending state SOC values, required charging power, and time of each trip chain are calculated. SUMMARY

[0005] The present application aims to propose a data-model fusion driven electric vehicle trip chain construction method and system; generate trip chain data with complete spatio-temporal dynamic characteristics, and alleviate the defects of EV simulation data and the economic and user privacy problems of real data acquisition.

[0006] To achieve the above object, the technical scheme of the present application is as follows:

[0007] In the first aspect, the present application provides a data-model fusion driven electric vehicle trip chain construction method, which comprises:

[0008] Selecting a construction area, modeling the road topology into a directed weighted network based on complex network theory to obtain a road topology network; embedding the traffic infrastructure into the road topology network structure to construct a road network model in a real traffic scenario, and constructing vehicle state information and constraints as the model driven part;

[0009] According to the historical trip chain data of the selected construction area, generating trip chain travel data including the starting point, ending point, trip distance, starting time, trip time and parking time of each trip in the trip chain in combination with the Gaussian mixture model and the constructed road topology network as the data driven part;

[0010] Fusing the data driven part and the model driven part, and combining typical vehicle parameters and a vehicle energy consumption power model to calculate complete trip chain data including travel data, electric vehicle state data and power consumption data; the typical vehicle parameters include battery capacity, unit energy consumption coefficient and maximum charging power; the electric vehicle state data includes the speed, position and SOC value of the electric vehicle.

[0011] Preferably, the modeling of the road topology into a directed weighted network based on complex network theory is specifically:

[0012] For the selected construction area, the intersection is modeled as a node, the road is modeled as an edge, and the road passing direction and the number of lanes are modeled by the adjacency matrix of the directed weighted network to obtain a road matrix N r :

[0013] N r =[n ij ]

[0014] Wherein, n ij represents the number of lanes from node i to node j, n ij =0 represents that there is no lane from node i to node j, and n ii =2 represents that node i is a traffic island area.

[0015] Preferably, the traffic infrastructure includes charging stations and traffic lights, and the charging stations and traffic lights are embedded into the directed weighted network, specifically by embedding the charging stations into the edge attribute and embedding the traffic lights into the node attribute;

[0016] For the charging station, a vector s ch,m is used to record the position of the charging station:

[0017] s ch,m =[e m ,pos ch,m ]

[0018] where e m is the edge corresponding to the road where the mth charging station is located; pos ch,m is a 0-1 real number, which is used to represent the specific position of the mth charging station on the road;

[0019] For the traffic light, the traffic light position is set in the node, and an alternating red-green light signal TL = [tl ij ] is used to control all edges connected to the node; wherein tl ij is a 0-1 variable, which represents the traffic light from node i to node j, 0 means allowing passing, and 1 means prohibiting passing:

[0020]

[0021] In the formula, t 0,ij represents the time tl ij has passed in a cycle at 0 time, t tl,ij represents the total duration of tl ij a cycle, represents the floor function, and t ij,gr represents the duration of the green light in the cycle.

[0022] Preferably, the vehicle state information and constraints are constructed as follows:

[0023] The state matrix X k is used to represent the state of the electric vehicle:

[0024] X k = {x v,k ,x s,k ,x soc,k}

[0025] x s,k = [e k ,pos vh,k ]

[0026] where x v,k , x s,k and x soc,k represent the speed, position and SOC value of the electric vehicle k respectively; e k represents the edge corresponding to the road where the electric vehicle k is located; pos vh,k is a 0-1 variable, which represents the specific position of the electric vehicle k on the road;

[0027] The electric vehicle constraints include SOC constraints, vehicle distance constraints, vehicle speed constraints and traffic light constraints;

[0028] SOC constraint: SOC value of each electric vehicle at each time is greater than 0:

[0029] x soc,k > 0

[0030] Car distance constraint: car distance of each electric vehicle in the same lane is greater than safety distance d threshold :

[0031]

[0032] wherein, and respectively represent the car distance between electric vehicle k1 and electric vehicle k2;

[0033] Car speed constraint: speed of each electric vehicle must be less than the speed limit of the road:

[0034]

[0035] wherein, s is any road where electric vehicle k is located, x v,k,s is the speed of electric vehicle k on road s, and x max,s is the speed limit of road s;

[0036] Traffic light constraint: when the traffic light signal at the intersection prohibits traffic, the vehicle speed should be 0:

[0037]

[0038] tl ij represents the signal light from node i to node j.

[0039] Preferably, the generated trip chain travel data specifically comprises the following steps:

[0040] S1, randomize the trip chain first travel time t s , trip chain starting location l s and trip chain travel times N; and perform parameter initialization: let the first travel starting time t s1 = t s , the first travel starting location l s1 = l s , the remaining travel times n = N, and the i-th travel i = 1;

[0041] S2, according to the starting time t si , starting location l si and trip chain travel times N of the i-th travel, predict the termination location l ei and trip distance d i of the i-th travel, and calculate the travel time tdi and predict the parking time t pi ;

[0042] S3, update the remaining trip times: n = n-1; and output the starting location l si , the ending location l ei , the travel distance d i , the starting time t si , the travel time t di , and the parking time t pi ;

[0043] S4, determine whether the updated remaining trip times n is 0; if n≠0, calculate the starting time t s(i+1) and the starting location l s(i+1) of the next trip, then let i = i+1 and go to S2; otherwise, end the loop.

[0044] Preferably, the travel chain trip times N, the first trip time t s , the ending location l ei of the i-th trip, the travel distance d i , the travel time t di , and the parking time t pi are obtained as follows:

[0045] Travel chain trip times N:

[0046] The occurrence frequency of the trip times in the historical travel chain data is used to represent the occurrence probability of the trip times, and the travel chain trip times N are obtained by probability sampling;

[0047] The first trip time t s :

[0048] The first trip time t s is obtained by random sampling after fitting the first trip time in the historical travel chain data using a Gaussian mixture model GMM;

[0049] The first trip address:

[0050] The starting location coordinates are simulated using any random generation method;

[0051] The ending location l ei of the i-th trip:

[0052] The trip times, the starting location, and the starting time are used as initial variables, and the ending location l ei of the i-th trip is obtained based on the historical travel chain data using a Gaussian mixture model GMM after fitting the starting time t si and the starting location l si of the i-th trip.The number of trips in the trip chain N is used as the input variable of the Gaussian mixture model, and the posterior probability that the input variable of the i-th trip belongs to each terminal location is calculated. The terminal location l of the i-th trip is obtained by probability sampling. ei ;

[0053] The travel distance d of the i-th trip i :

[0054] Taking the ending location and the starting location as the initial variables, after fitting the historical itinerary chain data using the Gaussian mixture model GMM, the starting location l of the i-th trip is used as the si and end location ei As the input variable of the Gaussian mixture model, the posterior probability of the input variable belonging to each travel distance of the i-th trip is calculated, and the travel distance d of the i-th trip is obtained by probability sampling. i ;

[0055] The travel time t of the i-th trip di :

[0056] Travel time t di Use travel distance d i Divide by the average speed of the trip, the average speed of the trip is calculated based on the historical trip chain data;

[0057] The parking time t of the i-th trip pi :

[0058] Taking the end location and end time as the initial variables, after fitting the historical trip chain data using the Gaussian mixture model GMM, the end location l of the i-th trip is used as the ei and the termination time t ei As the input variable of the Gaussian mixture model, the posterior probability that the input variable of the i-th trip belongs to each parking time is calculated, and the parking time t of the i-th trip is obtained by probability sampling. pi , where the end time of the i-th trip is t ei According to the starting time t si and travel time t di Obtained by calculation.

[0059] Preferably, the data driven part and the model driven part are integrated, and typical vehicle parameters and vehicle energy consumption power model are combined to calculate the complete trip chain data as follows:

[0060] Based on the road network model and vehicle constraints constructed by the model-driven part, the initial SOC information of the electric vehicle is obtained, and the data of each trip in the trip chain is obtained in sequence according to the time series. Simulation calculations are performed based on typical vehicle parameters and the vehicle energy consumption power model to obtain the complete trip chain data, including electric vehicle status data, power consumption of each trip, and required charging power and time;

[0061] Wherein, the initial soc is obtained by random sampling, and the charging behavior is set as: when the SOC is lower than the first preset threshold, the electric vehicle goes to the nearest charging station for charging, and when the SOC reaches the second preset threshold, the charging is stopped and the electric vehicle goes to the destination.

[0062] Preferably, the construction of the vehicle energy consumption power model includes EV reference power consumption modeling and temperature-influenced power model correction:

[0063] The EV reference power consumption modeling is:

[0064] P1=(kv 3 +f rl mgv)K a K b

[0065] Wherein, P1 represents the EV reference power consumption, k represents the air resistance coefficient, v represents the speed, kv 3 represents the power loss generated by wind resistance; f rl represents the rolling resistance, m represents the vehicle mass, g represents the gravitational acceleration, f rl mgv represents the power loss generated by rolling resistance; K a represents the correction coefficient generated by ignoring the acceleration and deceleration process, K b represents the correction coefficient generated by motor loss;

[0066] The temperature-influenced correction power model is:

[0067]

[0068] Wherein, P2(T amb ) represents the corrected power consumption, T amb represents the temperature.

[0069] In the second aspect, the present application provides a data-model fusion driven electric vehicle trip chain construction system, comprising a processor, a memory and a computer program stored in the memory, and when the processor executes the computer program, the steps of the data-model fusion driven electric vehicle trip chain construction method described above are specifically executed.

[0070] In the third aspect, the present application provides a computer readable storage medium, which stores a computer program, and when the processor executes the computer program, the data-model fusion driven electric vehicle trip chain construction method described above is realized.

[0071] Compared with the prior art, the present application has the following beneficial effects:

[0072] 1. The authenticity and rationality of EV trip chain simulation are improved. The present application realizes high-fidelity restoration of electric vehicle travel behavior by fusing real data and model mechanism, overcomes the defects that traditional pure simulation-based methods are difficult to reflect the spatio-temporal dynamic characteristics of trip chain, and makes the generated trip chain more consistent with actual traffic rules and user behavior logic.

[0073] 2. Data acquisition problems and model generality are considered. Under the premise of lack of high-frequency real trip data, an EV trip feature generation method based on Gaussian mixture model is proposed, which effectively alleviates the privacy protection and economic problems faced by collecting high-quality EV trip data, and enhances the operability and universality of the method in practical application.

[0074] 3. Complex network modeling is innovatively introduced to enhance the topological consistency of trip chain structure. The present application models the urban traffic network as a directed and weighted graph using complex network theory, and embeds traffic infrastructure such as charging stations and traffic signals as the attributes of nodes and edges in the graph, realizing structured and expandable modeling of actual traffic scenarios, avoiding the simplification assumption of road topological structure in traditional models, and significantly improving the simulation accuracy.

[0075] 4. A dynamic chain modeling method for EV energy consumption evaluation is constructed. The present application proposes an energy consumption power model and an SOC updating mechanism based on the micro-operation characteristics of vehicles, combines environmental factors (such as temperature) for power correction, further introduces SOC threshold judgment and shortest path strategy to simulate charging behavior, and forms a closed-loop energy consumption chain generation framework, which can be widely used in electric vehicle flexibility evaluation, load modeling and dispatching simulation and other scenarios.

[0076] 5. It has good scalability and landing ability. The method of the present application can be applied to traffic networks in different regions and different types of electric vehicles, and can be adapted to various operating scenarios by adjusting model parameters, providing basic data support for city-level and regional-level electric vehicle flexible access and dispatching, and having good engineering promotion value. BRIEF DESCRIPTION OF DRAWINGS

[0077] Figure 1 is a data-model fusion driven electric vehicle trip chain flowchart;

[0078] Figure 2 is a directed and weighted network example of traffic network;

[0079] Figure 3 is a physical driving flowchart;

[0080] Figure 4 is a charging station distribution characteristic diagram;

[0081] Figure 5 is a vehicle trajectory heat map;

[0082] Figure 6 a vehicle average speed heat map;

[0083] Figure 7 a road passing vehicle number histogram;

[0084] Figure 8 a load summation graph of all charging stations;

[0085] Figure 9 a load of each charging station and its average value curve;

[0086] Figure 10 a generated energy consumption data graph. DETAILED DESCRIPTION

[0087] The technical solutions of the present application will be specifically described below in combination with the accompanying Figures 1-10 The technical solutions of the present application will be specifically described below in combination with the accompanying

[0088] The present application proposes a data-model fusion driven electric vehicle trip chain construction method, which comprises:

[0089] Select a construction area, model the road topology as a directed weighted network based on complex network theory, obtain the road topology network, embed the traffic infrastructure into the road topology network structure, construct the road network model under the real traffic scene, and construct the vehicle state information and constraints as the model driven part, and the physical driving flow chart reference Figure 3 ;

[0090] According to the historical trip chain data of the selected construction area, generate trip chain travel data including the starting point, ending point, trip distance, starting time, trip time and parking time of each trip in the trip chain in combination with the Gaussian mixture model and the constructed road topology network, as the data driven part;

[0091] Fuse the data driven part and the model driven part, and combine typical vehicle parameters and vehicle energy consumption power model to calculate complete trip chain data including travel data, electric vehicle state data and power energy consumption data; the typical vehicle parameters include battery capacity, unit energy consumption coefficient and maximum charging power; the electric vehicle state data includes the speed, position and SOC value of the electric vehicle.

[0092] The present application can generate high credible EV trip chain data with spatiotemporal consistency and reasonable energy logic under the condition of lacking high frequency original data and existing regional differences, and provide data support for electric vehicle adjustable capacity evaluation, load prediction and dispatching optimization and other applications.

[0093] In the present embodiment, the modeling of the road topology as a directed weighted network based on complex network theory is specifically:

[0094] For the selected construction area, the intersection is modeled as a node, the road is modeled as an edge, and the method of modeling the road direction and the number of lanes through the adjacency matrix of the directed weighted network obtains the road matrix N r :

[0095] N r =[n ij ]

[0096] Wherein, n ij represents the number of lanes from node i to node j, n ij =0 indicates that there is no lane from node i to node j, and n ii =2 indicates that node i is a traffic island area.

[0097] Referring to the example in Figure 2 , the value in the matrix represents the number of lanes connecting two intersections, n 12 =2 indicates that there are 2 connecting lanes from node 1 to node 2, and n 21 =0 indicates that there is no lane from node 2 to node 1, which is a one-way street in reality, and n 33 =2 indicates a traffic island area in reality.

[0098] Similarly, the directed weighted matrix can also be used to represent the length of each road, and the weight in the road matrix N r is replaced by the lane length, and the road length matrix N l is obtained.

[0099] In this embodiment, the traffic infrastructure includes charging stations and traffic lights, and the charging stations and traffic lights are embedded in the directed weighted network, specifically, the charging stations are embedded in the edge attribute, and the traffic lights are embedded in the node attribute.

[0100] For the charging station, the position of the charging station is recorded using the vector s ch,m :

[0101] s ch,m =[e m ,pos ch,m ]

[0102] Wherein, e m is the edge corresponding to the road where the mth charging station is located; pos ch,m is a 0-1 real number, which is used to represent the specific position of the mth charging station on the road.

[0103] For the signal light, the signal light position is set in the node, and the alternating red and green light signal TL=[tl ij ] is used to control all the edges connecting the nodes; wherein tl ijis a 0-1 variable, indicating the traffic light from node i to node j, 0 means allowing passing, 1 means prohibiting passing:

[0104]

[0105] In the formula, t 0,ij represents the traffic light tl ij at 0 time, t tl,ij represents the time that has passed in a cycle, t ij represents the total length of a cycle, represents the down rounding, t ij,gr represents the length of the green light in the cycle.

[0106] In this embodiment, the vehicle state information and constraints are constructed as follows:

[0107] The state matrix X k represents the state of the electric vehicle, the state matrix X k is composed of several finite time points arranged in time sequence, has a consistent measurement time interval, each row is each state variable at different time points, and the column represents the state vector:

[0108] X k ={x v,k ,x s,k ,x soc,k}

[0109] x s,k =[e k ,pos vh,k ]

[0110] Wherein, x v,k , x s,k and x soc,k represent the speed, position and SOC value of the electric vehicle k respectively; e k represents the edge corresponding to the road where the electric vehicle k is located; pos vh,k is a 0-1 variable, indicating the specific position of the electric vehicle k on the road;

[0111] Setting electric vehicle constraints includes SOC constraint, vehicle distance constraint, vehicle speed constraint and traffic light constraint;

[0112] SOC constraint: the SOC value of each electric vehicle at each time is greater than 0:

[0113] x soc,k > 0

[0114] Vehicle distance constraint: the vehicle distance of each electric vehicle in the same lane is greater than the safety distance d threshold :

[0115]

[0116] where, and denote the distance between electric vehicle k1 and electric vehicle k2 respectively;

[0117] Speed constraint: the speed of each electric vehicle must be less than the speed limit of the road:

[0118]

[0119] where, s is any road that electric vehicle k is on, x v,k,s is the speed of electric vehicle k on road s, x max,s is the speed limit of road s;

[0120] Traffic light constraint: when the traffic light at the intersection is red, the speed of the vehicle should be 0:

[0121]

[0122] tl ij denotes the traffic light from node i to node j.

[0123] The routing algorithm for each EV can be any algorithm, since the purpose of this patent is not to propose the optimal routing algorithm, therefore, Dijkstra algorithm is taken as an example. The above method can be applied to any region, where the modeling of the traffic network only needs to frame the geographical coordinates of the desired area vertex, based on the complex network theory to model the traffic roads and infrastructure within the region, for the modeling of the vehicle, the parameters of the vehicle can be modified according to the actual situation of the region, such as speed, acceleration, energy consumption, battery capacity, etc.

[0124] In this embodiment, referring to Figure 1 , the trip chain trip data is generated, specifically including the following steps:

[0125] S1, randomizing the trip chain first trip time t s , trip chain starting location l s and trip chain trip number N; and performing parameter initialization: letting the first trip starting time t s1 = t s , the first trip starting location l s1 = l s , the remaining trip number n = N, the i-th trip i = 1;

[0126] S2, according to the starting time t si , starting location l si and trip chain trip number N of the i-th trip, predicting the termination location l ei of the i-th trip.and the trip distance d i , the trip time t di of the i-th trip is calculated pi ;

[0127] S3, the remaining trip number n is updated: n=n-1; and the starting location l si , the ending location l ei , the trip distance d i , the starting time t si , the trip time t di and the parking time t pi of the i-th trip are outputted;

[0128] S4, it is judged whether the updated remaining trip number n is 0; if n≠0, the starting time t s(i+1) and the starting location l s(i+1) of the next trip are calculated, then i=i+1 and turn to S2; otherwise, the loop is ended.

[0129] In the embodiment, the trip chain trip number N, the first trip time t s , the ending location l ei of the i-th trip, the trip distance d i , the trip time t di and the parking time t pi are obtained as follows:

[0130] The trip chain trip number N:

[0131] The occurrence frequency of the trip number in the historical trip chain data is used to represent the occurrence probability of the trip number, and the trip chain trip number N is obtained by probability sampling; the occurrence probability P(N=n i ) of the trip number n i in the historical trip chain data is:

[0132]

[0133] In the formula, n(N=n i ) represents the number of trip chains with the trip number n i in the historical trip chain data, and N0 represents the total number of trip chains in the historical trip chain data.

[0134] The first trip time t s :

[0135] The first trip time in the historical trip chain data is fitted by using the Gaussian mixture model GMM and then is randomly sampled to obtain;

[0136] The first trip address:

[0137] Simulate the departure location coordinates using any random generation method;

[0138] the i-th trip termination location l ei :

[0139] Take the trip number, starting location, and starting time as initial variables, and use the Gaussian mixture model (GMM) to fit based on historical trip chain data. Then, take the starting time t si , starting location l si , and trip chain trip number N of the i-th trip as input variables of the Gaussian mixture model, calculate the posterior probability of each termination location to which the i-th trip input variable belongs, and obtain the termination location l ei of the i-th trip through probability sampling.

[0140] Take the starting location l s as home (H), the starting time t s as T s , and the trip chain trip number n' as N as input variables. For example, the probability P(l e = W | l s = H, t s = T s , n' = N) that the termination location is a work location (W) can be calculated.

[0141]

[0142] In the formula, p (H,N,W) (t s = T s | Θ) is the probability value of the GMM established using data in which the starting location is home, the termination location is a work location, and the trip number is N as data sources at the starting time T s , p (H,N) (t s = T s | Θ') is the probability value of the GMM established using data in which the starting location is home and the trip number is N as data sources at the starting time T s , p1 represents the probability of the starting location being home, the termination location being a work location, and the trip number being N in the historical trip chain data, and p2 represents the probability of the starting location being home and the trip number being N in the historical trip chain data. The expressions of p1 and p2 are as follows:

[0143] p1 = P(l s = H, n = N, l e = W)

[0144] p2 = P(l s = H, n = N)

[0145] The travel distance d of the i-th trip i :

[0146] The starting location l si and the ending location l ei of the i-th trip are taken as initial variables, and a Gaussian Mixture Model (GMM) is used to fit the historical trip chain data, and the travel distance d i of the i-th trip is obtained by probability sampling after the i-th trip input variable belongs to each travel distance posterior probability is calculated. di ;

[0147] The travel time t of the i-th trip di :

[0148] The travel time t is divided by the travel average speed to obtain i , and the travel average speed is calculated according to the historical trip chain data; the expression of the travel time is as follows:

[0149]

[0150] In the formula, t d represents the travel time, d represents the travel distance, and v represents the travel average speed.

[0151] The parking time t of the i-th trip pi :

[0152] The ending location l ei and the ending time t of the i-th trip are taken as initial variables, and a Gaussian Mixture Model (GMM) is used to fit the historical trip chain data, and the ending location l ei and the ending time t of the i-th trip are taken as input variables of the Gaussian Mixture Model, and the i-th trip input variable belongs to each parking time posterior probability is calculated, and the parking time t of the i-th trip is obtained by probability sampling. pi , wherein the i-th trip ending time t ei is calculated according to the starting time t si and the travel time t di .

[0153] Taking the ending location l e as the working location (W) and the travel ending time t e as T1, the probability p(t p = T | t p = T1, l e = W) that the parking time t e is T is calculated as an example of input variables:

[0154]

[0155] Among them, p(t p =T,t e =T1|l e =W) indicates the end point l e is the trip termination time t under the working location (W) condition e is T1, parking time t p is the probability of T, p(t e =T1|l e =W) indicates the end point l e is the trip termination time t under the working location (W) condition e is the probability of T1, Indicates the end time of the trip t e is the parking time t under T1 condition p is the probability of T; fit it using a two-dimensional Gaussian mixture model:

[0156]

[0157] Where, p(t p ,t e =T1|l e =W) indicates the end point l e Under the working location (W), the parking time t p and the trip end time t e The joint probability density of m represents the current Gaussian component number, M represents the number of Gaussian components (components) in the mixed Gaussian model, ω m is the weight of the mth Gaussian component, N m (t p ,t e ;μ m ,σ m ) represents a two-dimensional normal distribution, whose mean vector is μ m , the covariance matrix is ​​σ m , used to characterize the joint distribution characteristics between parking time and termination time, the mean vector μ m and the covariance matrix σ m They are:

[0158]

[0159] Where, They represent the mean of the parking time and the mean of the termination time of the mth component, Respectively represent the mth component t p The variance and t e The variance of denote the covariance of t p and t e in the mth component;

[0160] parking time probability is:

[0161]

[0162] where ω' m denotes the probability weight of generating t e under the condition that the termination time t e and the termination location l p are given, denotes the expected value of the mth Gaussian component generating the parking time t e under the condition that the termination time t p is known, denotes the variance of the mth Gaussian component generating the parking time t e under the condition that the termination time t p is known:

[0163]

[0164] In the formula, l denotes a summation index variable, which is used to enumerate all M Gaussian variables in turn in the conditional weight normalization calculation, ω l denotes the original mixing weight of the lth Gaussian component, denotes the one-dimensional normal probability density function value of the termination time t e under the lth Gaussian component, denotes the mean value of the termination time t e in the lth Gaussian component, denotes the variance of the termination time t e in the lth Gaussian component;

[0165] By substituting the mean value μ m , the covariance σ m and the t e value of the initial variable into the expressions of the weight ω' m , the expected value and the variance , the weight, the mean value and the variance of the corresponding mixed Gaussian model can be obtained, and a new one-dimensional GMM is established according to these parameters. Finally, the parking time t p of the current trip can be obtained through probability value sampling.

[0166] In this embodiment, the fusion data driving part and the model driving part are combined with typical vehicle parameters and vehicle energy consumption power models to calculate complete trip chain data, which is specifically:

[0167] Based on the road network model and vehicle constraints constructed by the model-driven part, the initial SOC information of the electric vehicle is obtained, and each trip data of the trip chain is obtained in time sequence, and the simulation operation is carried out combined with the typical vehicle parameters and the vehicle energy consumption power model, and the complete trip chain data is obtained, including the electric vehicle state data, the consumed electric quantity of each trip, the required charging power and time.

[0168] Wherein, the initial soc is obtained by random sampling, and the charging behavior is set as: when the SOC is lower than the first preset threshold, the electric vehicle goes to the nearest charging station for charging, and when the SOC reaches the second preset threshold, the charging is stopped and the destination is reached.

[0169] In the embodiment, the construction of the vehicle energy consumption power model includes EV reference power consumption modeling and temperature-influenced power model correction:

[0170] The EV reference power consumption modeling is:

[0171] P1=(kv 3 +f rl mgv)K a K b

[0172] Wherein, P1 represents the EV reference power consumption modeling, k represents the air resistance coefficient, v represents the speed, kv 3 represents the power loss generated by wind resistance; f rl represents the rolling resistance, m represents the vehicle mass, g represents the gravity acceleration, f rl mgv represents the power loss generated by rolling resistance; K a represents the correction coefficient generated by ignoring the acceleration and deceleration process, K b represents the correction coefficient generated by motor loss;

[0173] The temperature-influenced correction power model is:

[0174]

[0175] Wherein, P2(T amb ) represents the corrected power consumption, T amb represents the temperature.

[0176] The present application provides a data-model fusion driven electric vehicle trip chain construction system, comprising a processor, a memory and a computer program stored in the memory, when the processor executes the computer program, the steps of the above-mentioned data-model fusion driven electric vehicle trip chain construction method are specifically executed.

[0177] The present invention proposes a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for constructing an electric vehicle trip chain driven by data-model fusion is implemented.

[0178] A specific example is provided below;

[0179] Traffic network topology model construction and travel data generation:

[0180] Take an area in Tianjin city as the modeling object, such as Figure 4 As shown, the area is 6.1 km long and 2.5 km wide, covering approximately 15.25 square kilometers. The model includes 7,555 streets and 1,942 intersections, encompassing residential, commercial, and office areas. Charging station power levels were set at 8 kW, 18 kW, 30 kW, and 150 kW. 4,648 vehicle trajectories were generated, including 2,756 charging records. Vehicle parameters were based on those of the Xiaomi SU7 and BYD-qin, and the road speed limit was set to 60 km / h. The data timescale is 15 minutes, but a finer resolution is possible.

[0181] The generated vehicle trajectory heat map is as follows Figure 5 As shown. It can be observed that the vehicle density is higher on the main roads and less on the roads that are relatively away from the main roads. This is also consistent with the commercial, residential and office areas around the main roads in the region. Figure 6 The average vehicle speed in the red circle is lower than that in other streets, which is consistent with Figure 5 The vehicle density is high, and due to Figure 4 The area in the middle red circle is close to the river and has fewer connecting bridges, so the average vehicle speed is lower. Figure 7 As shown in , this shows that the generated vehicle trajectory conforms to the actual spatiotemporal distribution characteristics.

[0182] According to the charging behavior of the generated vehicles, the load conditions of each charging station are obtained as follows: Figure 8 、 Figure 9 As shown. Among them, Figure 8 This is the sum of the loads for all charging stations. It can be seen that the total power consumption of charging stations fluctuates over time, showing a certain trend. Power demand may be higher during the morning and evening peak hours, while lower at night. This fluctuation reflects the dynamic nature of charging station loads, which is related to factors such as user charging behavior and the frequency of charging device usage. Figure 9The load curves of each charging station and their average values are shown. It can be seen that the load curves of each charging station show fluctuation characteristics, and the load is high in some time periods, which is related to charging peak periods or specific demands. The average load curve is relatively flat, reflecting the balanced trend of the overall load. The highest load is close to 2500, and the lowest load is close to 0, indicating that there is a significant time difference in the use of charging stations, which helps to optimize the allocation and operation strategy of charging resources.

[0183] Energy consumption data generation:

[0184] Figure 10 Typical results of the electric vehicle energy consumption data generated by the present application are shown, including the average SOC (State of Charge) changes of electric vehicles at different time periods in multiple charging stations. Figure 10 Figure (a) shows the average SOC change trend of the vehicles connected to the five charging stations in the full time scale. It can be observed that the SOC data of each charging station shows obvious time-varying characteristics, reflecting the dynamic distribution of vehicle charging and discharging behavior within a day. The colors and line types in this figure represent the statistical results of different charging stations, respectively, indicating that the state of charge of each station shows differences on the same time axis, further verifying that the trip chain and energy chain model constructed by the present application has the ability to finely depict individual differences and station characteristics. Figure 10 Figure (b) is a local enlarged view of the black rectangular frame area in (a), to more clearly show the details of the SOC change process. It can be seen that the SOC of some vehicles shows a periodic downward and upward trend, reflecting the typical "travel-discharge-charge" behavior. Especially when the SOC value is low, some curves show a rapid rising charging process, indicating that the vehicle triggers the SOC threshold charging mechanism designed in the present application and automatically goes to the nearest charging station for energy replenishment until it reaches the expected SOC and leaves to continue the trip. This result fully illustrates that the energy chain modeling method proposed by the present application can simulate the energy consumption and replenishment process of electric vehicles, has high dynamic response ability and behavior consistency, and is suitable for various application scenarios such as flexible load modeling, dispatching strategy verification, and charging infrastructure layout optimization.

[0185] The above is the preferred embodiment of the present application. Any changes made in accordance with the technical solutions of the present application, as long as the resulting functional effects do not exceed the scope of the technical solutions of the present application, are within the scope of protection of the present application.

Claims

1. A data-model fusion-driven electric vehicle trip chain construction method, characterized in that: The method comprises: Select a construction area and model the road topology as a directed weighted network based on complex network theory to obtain a road topology network. Embed traffic infrastructure into the road topology network structure to construct a road network model under real traffic scenarios. Furthermore, vehicle status information and constraints are constructed as the model-driven component. Based on the historical trip chain data of the selected construction area, the trip chain travel data is generated by combining the Gaussian mixture model and the constructed road topology network, including the starting point, end point, trip distance, start time, travel time and parking time of each trip in the trip chain as the data-driven part; By integrating the data-driven part and the model-driven part, and combining typical vehicle parameters and the vehicle energy consumption and power model, the complete trip chain data is calculated, including travel data, electric vehicle status data and power consumption data; the typical vehicle parameters include battery capacity, unit energy consumption coefficient and maximum charging power; the electric vehicle status data includes the speed, position and SOC value of the electric vehicle.

2. The method for constructing an electric vehicle trip chain driven by data-model fusion according to claim 1, characterized in that: The road topology is modeled as a directed weighted network based on complex network theory as follows: For the selected construction area, the intersection is modeled as a node, the road is modeled as an edge, and the road traffic direction and the number of lanes are modeled by the adjacency matrix of the directed weighted network to obtain the road matrix N r : N r =[n ij ] Among them, n ij Indicates the number of lanes from node i to node j, n ij =0 means there is no lane from node i to node j, n ii =2 means that node i is a traffic circle area.

3. The electric vehicle trip chain construction method driven by data-model fusion according to claim 2 is characterized in that: The transportation infrastructure includes charging stations and traffic lights, which are embedded in a directed weighted network. Specifically, the charging stations are embedded in edge attributes, and the traffic lights are embedded in node attributes. For charging stations, use the vector s ch,m Record the location of charging stations: s ch,m =[and m ,pos ch,m ] Among them, e m is the edge of the road where the mth charging station is located; pos ch,m is a real number between 0 and 1, used to indicate the specific location of the mth charging station on the road; For traffic lights, set the traffic light position in the node and use alternating traffic light signals TL=[tl ij ] controls the edges of all connected nodes; where tl ij Is a 0-1 variable, representing the traffic light from node i to node j, 0 means allowing passage, 1 means prohibiting passage: Where, t 0,ij Indicates that the signal light tl at time 0 ij The time elapsed in a cycle, t tl,ij Indicates tl ij The total duration of a cycle, Indicates rounding down, t ij,gr Indicates the duration of the green light in the cycle.

4. The method for constructing an electric vehicle trip chain driven by data-model fusion according to claim 1, characterized in that: The construction of vehicle status information and constraints is as follows: Using the state matrix X k Indicates the status of the electric vehicle: X k ={x v,k ,x s,k ,x soc,k } x s,k =[and k ,pos vh,k ] Among them, x v,k 、x s,k and x soc,k They represent the speed, position and SOC value of electric vehicle k respectively; e k represents the edge of the road where electric car k is located; pos vh,k is a 0-1 variable that represents the specific location of electric vehicle k on the road; Setting electric vehicle constraints includes SOC constraints, vehicle distance constraints, vehicle speed constraints and traffic light constraints; SOC constraint: The SOC value of each electric vehicle at each moment is greater than 0: x soc,k ＞0 Distance constraint: The distance between each electric vehicle in the same lane must be greater than the safety distance d threshold : in, and They represent the distance between electric vehicle k1 and electric vehicle k2 respectively; Speed ​​constraint: The speed of each electric vehicle must be less than the speed limit of the road: Where s is any road where electric vehicle k is located, x v,k,s is the speed of electric car k on road s, x max,s is the speed limit of road s; Traffic light constraint: When the traffic light at the intersection prohibits passing, the vehicle speed should be 0: tl ij A semaphore that represents the link from node i to node j.

5. The method for constructing an electric vehicle trip chain driven by data-model fusion according to claim 1, characterized in that: Generating the trip chain travel data specifically includes the following steps: S1, the first travel time t of randomized trip chain s 、Starting point of the trip chain l s and the number of trips in the itinerary chain N; and initialize the parameters: let the starting time of the first trip t s1 =t s 、Starting point of first trip s1 =l s , the remaining number of trips n = N, the i-th trip i = 1; S2, based on the starting time t of the i-th trip si 、Starting point si and the number of trips in the itinerary chain N, predict the end location l of the i-th trip ei and travel distance d i , calculate the travel time t of the i-th trip di And predict the parking time t of the i-th trip pi ; S3. Update the remaining number of trips: n = n-1; and output the starting point l of the i-th trip si 、Ending location ei , travel distance d i , start time t si , travel time t di and parking time t pi ; S4. Determine whether the updated remaining number of trips n is 0; If n≠0, calculate the starting time t of the next trip s(i+1) and starting point s(i+1) Then, set i=i+1 and go to S2; otherwise, end the loop.

6. The method for constructing an electric vehicle trip chain driven by data-model fusion according to claim 1, characterized in that: Number of trips in the itinerary chain N, first trip time t s , the ending location l of the i-th trip ei , travel distance d i , travel time t di and parking time t pi The details of obtaining are as follows: Number of trips in the trip chain N: The frequency of trips in the historical itinerary chain data is used to represent the probability of occurrence of trips, and the number of trips in the itinerary chain N is obtained through probability sampling; First trip time t s : The Gaussian mixture model (GMM) is used to fit the first travel time in the historical itinerary chain data and then randomly sampled; First trip address: Use any random generation method to simulate the coordinates of the departure location; The ending location l of the i-th trip ei : Taking the number of trips, starting location and starting time as the initial variables, the Gaussian mixture model GMM is used to fit the historical itinerary chain data, and the starting time t of the i-th trip is used as the starting time. si 、Starting point si The number of trips in the trip chain N is used as the input variable of the Gaussian mixture model, and the posterior probability that the input variable of the i-th trip belongs to each terminal location is calculated. The terminal location l of the i-th trip is obtained by probability sampling. ei ; The travel distance d of the i-th trip i : Taking the ending location and the starting location as the initial variables, after fitting the historical itinerary chain data using the Gaussian mixture model GMM, the starting location l of the i-th trip is used as the si and end location ei As the input variable of the Gaussian mixture model, the posterior probability of the input variable belonging to each travel distance of the i-th trip is calculated, and the travel distance d of the i-th trip is obtained by probability sampling. i ; The travel time t of the i-th trip di : Travel time t di Use travel distance d i Divide by the average speed of the trip, the average speed of the trip is calculated based on the historical trip chain data; The parking time t of the i-th trip pi : Taking the end location and end time as the initial variables, after fitting the historical trip chain data using the Gaussian mixture model GMM, the end location l of the i-th trip is used as the ei and the termination time t ei As the input variable of the Gaussian mixture model, the posterior probability that the input variable of the i-th trip belongs to each parking time is calculated, and the parking time t of the i-th trip is obtained by probability sampling. pi , where the end time of the i-th trip is t ei According to the starting time t si and travel time t di Obtained by calculation.

7. The method for constructing an electric vehicle trip chain driven by data-model fusion according to claim 1, characterized in that: The fusion of the data-driven part and the model-driven part, combined with typical vehicle parameters and vehicle energy consumption power model, calculates the complete trip chain data as follows: Based on the road network model and vehicle constraints constructed by the model-driven part, the initial SOC information of the electric vehicle is obtained, and the data of each trip in the trip chain is obtained in sequence according to the time series. Simulation calculations are performed based on typical vehicle parameters and the vehicle energy consumption power model to obtain the complete trip chain data, including electric vehicle status data, power consumption of each trip, and required charging power and time; Among them, the initial SOC is obtained by random sampling, and the charging behavior is set as follows: when the SOC is lower than the first preset threshold, the electric vehicle goes to the nearest charging station for charging, and when the SOC reaches the second preset threshold, it stops charging and goes to the destination.

8. The method for constructing an electric vehicle trip chain driven by data-model fusion according to claim 1, characterized in that: The construction of the vehicle energy consumption power model includes EV baseline power consumption modeling and power model correction considering temperature effects: The EV baseline power consumption is modeled as: P1=(kv 3 +f rl mgv)K a K b Among them, P1 represents the EV benchmark power consumption, k represents the air resistance coefficient, v represents the speed, kv 3 Indicates the power loss caused by wind resistance; f rl represents rolling resistance, m represents vehicle mass, g represents acceleration due to gravity, and f rl mgv represents the power loss caused by rolling resistance; K a Indicates the correction factor caused by ignoring the acceleration and deceleration process, K b Indicates the correction factor caused by motor loss; The modified power model considering the temperature effect is: Among them, P2(T amb ) represents the corrected power consumption, T amb Indicates temperature.

9. The electric vehicle trip chain construction system driven by data-model fusion is characterized by: The method comprises a processor, a memory and a computer program stored in the memory. When the processor executes the computer program, the method specifically performs the steps of the electric vehicle trip chain construction method driven by data-model fusion as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the data-model fusion-driven electric vehicle trip chain construction method as described in any one of claims 1 to 8 is implemented.

Images