Electric vehicle load prediction method based on vehicle road station and trip chain fusion

Abstract

The invention discloses an electric vehicle load prediction method based on vehicle road station and trip chain fusion, belongs to the technical field of electric vehicle charging load prediction, and solves the problem of how to improve the accuracy of electric vehicle charging load prediction. Constructing an electric vehicle unit mileage energy consumption model; the road traffic information and the electric vehicle travel data are mined and analyzed by adopting a two-way long-short-term memory network, the road traffic flow is predicted, and the probability distribution of different types of travel chains of the electric vehicle is output; planning a driving path of the electric vehicle through an improved Floyd algorithm; analyzing the charging characteristics of the electric vehicles, and establishing an electric vehicle queuing-gravity model to calculate the total charging time of the electric vehicles; the method can achieve the prediction of the charging load of the electric vehicle in multiple types of functional areas, provides important information for the formulation of a power grid power scheduling strategy for each power distribution network node, and is of great significance for the load scheduling of a power system or the presetting of a charging station.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric vehicle charging load prediction, and relates to an electric vehicle load prediction method based on the integration of vehicle-road stations and travel chains. Background Art

[0002] Electric vehicles (EVs) are becoming a widespread part of people's lives as a clean, low-carbon mode of transportation. However, as the number of EVs increases, the impact of their charging load on power systems is becoming increasingly significant. The uncertainty and volatility of charging loads pose new challenges to power systems in terms of load balancing, energy allocation, and power supply stability. Therefore, accurately predicting EV charging loads has become a pressing need for intelligent energy management and intelligent transportation systems.

[0003] Current EV charging load forecasting lacks a detailed functional area classification model, making it difficult to accurately reflect the differences in EV load trends across different functional areas. Currently, EV travel chains are mostly described using statistical analysis methods, which struggle to capture the inherent characteristics of EV travel chains. The resulting travel chains are less accurate and cannot accurately determine the travel behavior of a specific type of travel chain, leading to significant errors in load forecasting. Furthermore, EV load forecasting currently lacks an EV charging station selection model that accurately reflects user psychology, resulting in large errors in the calculation of EV charging times. These shortcomings of existing forecasting methods result in low accuracy in EV charging load forecasting, significant deviations from reality, and their respective limitations. Summary of the Invention

[0004] The technical solution of the present invention is used to solve the problem of how to improve the accuracy of electric vehicle charging load prediction.

[0005] The present invention solves the above technical problems through the following technical solutions:

[0006] The electric vehicle load forecasting method based on the integration of vehicle-road-station and travel chain includes:

[0007] S1 uses graph theory to model the road network topology, establishes a traffic network model that takes into account dynamic traffic information and road travel time, uses Voronoi diagrams to define traffic node types and divide areas, and constructs an electric vehicle unit mileage energy consumption model that takes into account ambient temperature and real-time vehicle speed;

[0008] S2 Obtain the historical travel data of electric vehicles, extract the travel information and the characteristic data of historical road traffic flow from it, input the historical road traffic flow into a bidirectional long short-term memory network, output the predicted road traffic flow, input the travel information into a bidirectional long short-term memory network, output the predicted travel behavior of electric vehicles, extract the travel time, stay time, and arrival location travel information of electric vehicles, predict the relative rules of travel time and space of electric vehicles, and obtain the travel chain of electric vehicles;

[0009] S3 Use the improved Floyd algorithm to plan the shortest driving path of electric vehicles;

[0010] S4 Analyze the charging characteristics of different types of electric vehicles, and predict the charging load of electric vehicles through the queuing-gravitation model.

[0011] Furthermore, the road network topological structure is modeled using graph theory methods as follows:

[0012]

[0013] Among them, R N is the road network topology within the research area; N is the set of all road network traffic nodes within the road network scope except for the charging station nodes; C is the set of charging station nodes, indicating the locations where electric vehicles charge; S is the set of road segments within the road network; D is the set of distances of each road segment; W is the set of road impedances within the road network scope; n i is a road network node; c k is a charging station node; s l is a road segment; d l is the distance length of road segment l; w l is the road impedance of road segment l.

[0014] Furthermore, the area is divided by the Voronoi diagram as follows:

[0015] V(n i ) = {q|||q - n i || ≤ ||q - n k ||, i ≠ k, k ∈ {1, …, n}} (2)

[0016] Among them, V(n i ) is the Voronoi polygon formed with the road network node n i as the center point, q is any point within this Voronoi polygon, and ||q - n i || ≤ ||q - n k || means that the distance from any point in the Voronoi polygon V(n i ) to the road network node n i is the shortest.

[0017] Furthermore, construct a traffic road network model as follows:

[0018]

[0019] Among them, t l is the travel time of section l, i.e., the traffic impedance; c l is the initial designed traffic flow of section l during the traffic planning stage; q l is the traffic flow of section l; t0 is the travel time in the traffic relaxation state; α and β are fixed parameters.

[0020] Furthermore, the unit mileage energy consumption model of the electric vehicle considering the environmental temperature and the real-time vehicle speed is constructed as follows:

[0021]

[0022] Among them, Q T is the power consumption per unit mileage; K T is the air-conditioning power consumption per unit mileage; v l is the driving speed of the vehicle on section l; D is the driving distance; A L is the air-conditioning cooling power; A R is the air-conditioning heating power; T is the outdoor temperature; T max is the lower limit of the air-conditioning cooling temperature; T min is the upper limit of the air-conditioning heating temperature.

[0023] Furthermore, the method for planning the shortest driving path of the electric vehicle using the improved Floyd algorithm is as follows:

[0024] Obtain the distance between any two nodes according to the traffic road network topology model; correct the distance between two nodes using the road blockage degree between any two traffic nodes to generate the traffic road network adjacency matrix;

[0025] Generate and update the path distance storage matrix and the path node storage matrix with the same dimension as the adjacency matrix; traverse all traffic nodes to obtain the final shortest path storage matrix and the path node storage matrix;

[0026] According to the predicted electric vehicle travel chain, calculate the shortest path between each location in the travel chain through the improved Floyd algorithm, and the electric vehicle travels according to the shortest path to obtain the complete driving path of the electric vehicle traffic road network;

[0027] According to the predicted traffic flow of each road and the established travel time traffic impedance model, obtain the driving speed of the electric vehicle based on the traffic flow of the electric vehicle driving path, and obtain the power consumption per unit mileage of the electric vehicle during driving.

[0028] Furthermore, the update of the traffic road network adjacency matrix is as follows:

[0029] The traffic road network adjacency matrix is:

[0030]

[0031] The condition for iteratively updating the adjacency matrix Dist is as follows:

[0032]

[0033] Among them, Dist m (i, j) is the shortest distance between nodes i and j after m - times of iterative update; Dist m-1 (i, j) is the shortest distance between nodes i and j after m - 1 - times of iterative update; Dist m-1 (i, k)+Dist m-1 (k, j) is the distance from node i to node j passing through node k after m - 1 - times of iterative update; m is the number of updates.

[0034] Furthermore, the method for predicting the charging load of electric vehicles through the queuing - gravity model is as follows:

[0035] Construct an electric vehicle queuing - gravity model to determine the charging stations selected by electric vehicles:

[0036]

[0037] Among them, F ij is the attraction between charging station i and the j - th electric vehicle; T ij is the queuing time gravity coefficient for the j - th electric vehicle to reach charging station i; M i is the mass of the charging station "celestial body"; m j is the mass of the electric vehicle "planet"; r is the shortest distance between the electric vehicle and the charging station; among them, the queuing time coefficient T ij :

[0038]

[0039] Among them, is the average waiting time of electric vehicle j at the i - th charging station; E T , V T are the expectation and variance of the Poisson distribution with parameter λ that the electric vehicle process follows; C is the number of charging stations; ρ = λE T , λ is a parameter; the mass M i of the charging station "celestial body" is:

[0040] M i = τ i μ i (9)(6)

[0041] Among them, τ iis the number of charging piles in charging station i; μ i is the in-station charging efficiency; the mass m of the electric vehicle "Planet" ij is:

[0042]

[0043] where A ij is the shortest distance between charging station i and electric vehicle j; E j is the battery level of electric vehicle j; E th is the battery level threshold.

[0044] The present invention also provides an electronic device, including a memory and a processor. The memory is used to store a program that supports the processor to execute the above-mentioned electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain. The processor is configured to execute the program stored in the memory.

[0045] The present invention also provides a storage medium, on which a computer program is stored. When the computer program is run by a processor, it executes the steps of the above-mentioned electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain.

[0046] The beneficial effects of the present invention are as follows:

[0047] Aiming at the problem that the charging loads generated by electric vehicles in different types of functional areas are different, the present invention uses a Voronoi diagram combined with a traffic road network model to divide regions, defines traffic node types to refine the description of electric vehicle travel destinations, and constructs a travel time traffic impedance model and an electric vehicle energy consumption model considering temperature changes to characterize the driving speed and power consumption of electric vehicles.

[0048] Aiming at the problem that the travel information of electric vehicles is complex and difficult to accurately describe, the present invention uses a bidirectional long short-term memory neural network to mine and analyze electric vehicle travel data, mines and analyzes the relative laws of travel time and space of electric vehicles, predicts the probability distribution of various travel chains of electric vehicles, optimizes the problem that the existing electric vehicle travel chain model is too single, and improves the diversity and accuracy of electric vehicle travel chain prediction through the neural network.

[0049] Aiming at the problems of electric vehicle charging station selection and charging time calculation, the present invention uses a queuing-gravitational model to simulate the process of users selecting charging stations, and constructs a charging mode and behavior description model for different types of electric vehicles, fully considering the actual situation of electric vehicle charging, and effectively improving the accuracy of electric vehicle charging load prediction.

[0050] The present invention combines the actual data of electric vehicles and physical models, improves the accuracy of electric vehicle charging load prediction, provides a more accurate basis for power system dispatching, thereby optimizing the planning and operation of the distribution network, and ensuring the stability and security of the power grid. Description of the Drawings

[0051] Figure 1 It is a flowchart of the electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain in the first embodiment of the present invention;

[0052] Figure 2 It is a unit structure diagram of the bidirectional long short-term memory network (LSTM network) adopted by the electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain in the first embodiment of the present invention;

[0053] Figure 3 It is a load simulation verification diagram of each functional area of the electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain of the present invention. Detailed Embodiment

[0054] 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 in conjunction with the embodiments of the present invention. Apparently, 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.

[0055] The technical solutions of the present invention will be further described below in conjunction with the drawings of the specification and specific embodiments:

[0056] Embodiment 1

[0057] As Figure 1 shown, an electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain in the first embodiment of the present invention includes the following steps:

[0058] Step 1. Establish an electric vehicle traffic road network model and define traffic node types.

[0059] Step 1.1. Use graph theory method to model the road network topology structure to obtain the node and road connection graph of the traffic road network, which is convenient for subsequent electric vehicle path selection calculation:

[0060]

[0061] In formula (1): R N is the road network topology within the research area; N is the set of all road network traffic nodes except the charging station nodes within the road network range; C is the set of charging station nodes, indicating the location where electric vehicles charge; S is the set of road segments within the road network; D is the set of distances of each road segment; W is the set of road impedances within the road network range; n i is a road network node; c k is a charging station node; s lis a road segment; d l is the distance length of road segment l; w l is the road impedance of road segment l.

[0062] Step 1.2: Use the Voronoi diagram to divide the road network area with the set of road network traffic nodes as the center points. The Voronoi diagram is used to finely divide the area, which is convenient for combining with the subsequent travel chain to obtain more accurate travel behaviors of electric vehicles. At the same time, by dividing the area, the differences in loads under different types of functional areas are analyzed:

[0063] V(n i ) = {q|||q - n i || ≤ ||q - n k ||, i ≠ k, k ∈ {1, …, n}} (2)

[0064] In formula (2): V(n i ) is the Voronoi polygon formed with the road network node n i as the center point, q is any point in this Voronoi polygon, ||q - n i || ≤ ||q - n k || means that the distance from any point in the Voronoi polygon V(n i ) to the road network node n i is the shortest.

[0065] Step 1.3: Calculate the areas of various functional buildings in the Voronoi polygon area, and define the types of each traffic node according to the nature of the functional building with the largest area:

[0066]

[0067] In formula (3): represents the type of road network node n i , 1 is a commercial traffic point, 2 is a residential traffic point, and 3 is a work traffic point.

[0068] Step 1.4: Build a travel time traffic impedance model and calculate the EV driving speed:

[0069]

[0070] In formula (4): t l is the travel time of road segment l, that is, the traffic impedance; c l is the initial designed traffic flow of road segment l in the traffic planning stage; q l is the traffic volume of road segment l; t0 is the travel time in the traffic relaxation state; α and β are fixed parameters.

[0071] Step 1.5: Considering the environmental temperature change and driving speed, construct the power consumption model per unit mileage of an electric vehicle. The model takes into account the influence of the driving speed of the electric vehicle and the air conditioning equipment on the power consumption per unit mileage of the electric vehicle. When the temperature is higher than a certain value, the air conditioning equipment cools; when the temperature is lower than a certain value, the air conditioning equipment heats; in other cases, the air conditioning equipment is turned off:

[0072]

[0073] In Equation (5): Q T is the power consumption per unit mileage; K T is the power consumption of the air conditioner per unit mileage; v l is the driving speed of the vehicle on section l; D is the driving distance; A L is the cooling power of the air conditioner; A R is the heating power of the air conditioner; T is the outdoor temperature; T max is the lower limit of the air conditioner cooling temperature; T min is the upper limit of the air conditioner heating temperature.

[0074] Step Two: Prediction of the travel chain of electric vehicles based on the bidirectional long short-term memory network.

[0075] Step 2.1: Obtain the historical travel data of electric vehicles, process the missing values and outliers in the data, and complete data cleaning;

[0076] Step 2.2: Extract the characteristic data of the historical road traffic flow from the historical travel data;

[0077] Step 2.3: Input the historical road traffic flow into the bidirectional long short-term memory network and output the predicted road traffic flow. Specifically, as Figure 2 shown, the mathematical model of the LSTM network is shown in Equation (6):

[0078]

[0079] In Equation (6): x t is the input, f t is the forget gate value, g t is the input gate value, s t is the temporary state quantity, c t is the state quantity, o t is the intermediate output quantity, h t is the output gate value; W f , W g , W c , W o are the corresponding weight matrices respectively, b f , b g , b c , b oThey are the corresponding bias terms respectively. ⊙ represents array multiplication, that is, the elements of the vectors are multiplied. σ is the sigmoid activation function. The specific expressions of it and the tanh activation function are shown in Equation (7):

[0080]

[0081] The structure of the Bi-LSTM network consists of two LSTM networks, one in sequence and the other in reverse. One processes the input sequence forward in sequence, and the other processes it backward in reverse. The outputs of these two LSTM layers are concatenated at each time step and then passed to the subsequent layer as the output of the entire Bi-LSTM layer. This bidirectional processing method enables Bi-LSTM to utilize the past and future information in the electric vehicle travel sequence simultaneously, thereby capturing long-distance dependencies more effectively and enhancing the ability to understand and process data.

[0082] Step 2.4: According to the historical travel data of electric vehicles, the electric vehicle travel chains are pre-defined as n categories, and the electric vehicle travel chain is described as:

[0083]

[0084] In Equation (8): L i is the i-th category of electric vehicle travel chain, i = 1, 2, …, n; is the n-th stop position of the electric vehicle in the i-th travel chain; t in is the residence time of the electric vehicle at the n-th stop position in the i-th travel chain.

[0085] Step 2.5: Extract feature data such as the starting timestamp, start / end positions, residence duration, number of charging times, driving speed, SOC, etc. of electric vehicles from the historical travel data to form an electric vehicle historical travel chain sequence, and the travel chain sequence can be described as the transfer between different types of traffic nodes.

[0086] Step 2.6: Input the electric vehicle historical travel chain sequence into the bidirectional long short-term memory network, predict the probability distribution of the travel chain categories, and predict the probability distributions of the corresponding travel start time and residence time under different types of travel chains. After obtaining the probability distributions, use the Monte Carlo simulation method to extract the electric vehicle travel chain categories, residence time, and travel start time for subsequent charging load calculation.

[0087] Step Three: Use the improved Floyd algorithm to plan the shortest driving path of electric vehicles.

[0088] Step 3.1: Obtain the distance l′ between any two nodes according to the traffic road network topology model ij ;

[0089] Step 3.2: Define the road blockage degree ρ, and correct the distance between two nodes based on the road blockage degree. The corrected distance is related to the road travel time. The lower the road blockage degree, the shorter the corresponding road path compared to other paths.

[0090]

[0091] In Equation (9): k is the blockage coefficient; t l is the travel time of section l; t0 is the travel time in the traffic relaxation state; l ij is the corrected distance between two nodes; l′ ij is the initial distance between two nodes.

[0092] Step 3.3: Use the corrected distance l between two traffic nodes ij to generate the adjacency matrix Dist of the traffic road network. By continuously updating the initial adjacency matrix, the shortest distance between any two nodes in the road network can be obtained finally.

[0093]

[0094] In Equation (8): l ij is the corrected distance between two nodes.

[0095] Step 3.4: Iteratively update the adjacency matrix Dist:

[0096]

[0097] In Equation (9): Dist m (i,j) is the shortest distance between nodes i and j after m times of iterative update; Dist m-1 (i,j) is the shortest distance between nodes i and j after m - 1 times of iterative update; Dist m-1 (i,k)+Dist m-1 (k,j) is the distance from node i to node j passing through node k after m - 1 times of iterative update; m is the number of updates.

[0098] Step 3.5: According to the electric vehicle travel chain, determine the traffic nodes reached by the electric vehicle, look up the adjacency matrix of the traffic road network, and determine the shortest distance and specific path in each travel process of the electric vehicle.

[0099] [[ID=^48]]Step 3.6: According to the traffic impedance model of travel time, calculate the travel time of the electric vehicle, and calculate the power consumption during the driving process of the electric vehicle to obtain the state of charge SOC when reaching the destination.

[0100] Step 3.7: Judge whether charging is required according to the set charging SOC threshold;

[0101] Step 4. Establish a queuing-gravitational selection model for electric vehicle charging stations and calculate the charging time of electric vehicles:

[0102] Step 4.1. Construct a queuing-gravitational model for electric vehicles to determine the charging stations selected by electric vehicles. When users choose charging stations, they will consider influencing factors such as the queuing time for charging at the charging station, the scale efficiency of the charging station, and the EV battery level. Construct a queuing-gravitational model to describe these influencing factors and simulate the process of users' charging station selection;

[0103]

[0104] In Equation (12), F ij is the attraction between charging station i and the j-th electric vehicle; T ij is the queuing time gravitational coefficient for the j-th electric vehicle to reach charging station i; M i is the mass of the charging station "celestial body"; m j is the mass of the electric vehicle "planet"; r is the shortest distance between the electric vehicle and the charging station. Among them, the queuing time coefficient T ij is:

[0105]

[0106] In Equation (13), is the average waiting time of electric vehicle j at the i-th charging station; E T , V T are the expectation and variance of the Poisson distribution with parameter λ for the electric vehicle process; C is the number of charging stations; ρ = λE T , λ is a parameter; the mass M i of the charging station "celestial body" is:

[0107] M i = τ i μ i (14)

[0108] In Equation (14), τ i is the number of charging piles in charging station i; μ i is the charging efficiency inside the station; the mass m ij of the electric vehicle "planet" is:

[0109]

[0110] In Equation (15), A ij is the shortest distance between charging station i and electric vehicle j; E j is the battery level of electric vehicle j; E th is the battery level threshold.

[0111] Step 4.2: The electric vehicle selects the charging station with the greatest gravitational force for charging, and calculates the arrival time at the charging station according to the traffic impedance model;

[0112] Step 4.3: Assume that there are only private cars and taxis in the traffic road network. The initial state of charge (SOC) of the electric vehicle follows a normal distribution, and the battery capacities of different types of electric vehicles follow a gamma distribution. Combining the two can obtain the initial power of the electric vehicle.

[0113] Step 4.3.1: The charging method of private cars needs to be jointly determined according to the battery state of charge at the current traffic node, the residence time of users at the location, and the fast-slow power charging model;

[0114] Step 4.3.2: Taxis all adopt the "three-shift rotation" mode, that is, they all need to use time to exchange for benefits. At the same time, in order to pursue the shortest charging time as much as possible, it is assumed that taxis all adopt the fast charging method.

[0115] Step 4.4: Set two charging powers, fast charging and slow charging. Assume that the charging power is a constant power and there are only two charging modes:

[0116]

[0117] In formula (16): P f is the fast charging power, and P l is the slow charging power.

[0118] Step 4.5: Determine the charging switch decision according to the state of charge of the electric vehicle, and determine whether the electric vehicle needs to be charged after arriving at the destination. If the power is not enough to maintain the next trip, the electric vehicle needs to be charged:

[0119]

[0120] In formula (17): θ is the charging switch decision, 0 means not charging, and 1 means charging; E p is the current power of the current electric vehicle; E ij is the power required for the next trip behavior of the electric vehicle; E t is the power threshold.

[0121] Step 4.6: Select the charging mode through formula (18), and calculate the charging time through formula (19). The selection of the charging mode is jointly determined by the residence time of the electric vehicle and the current power. If the slow charging mode cannot meet the needs of the next trip within the residence time, then fast charging is selected. The charging time is the charging residence time or the full charge time:

[0122]

[0123] In formula (18): η is the charging efficiency; Ts is the rechargeable time; in Equation (19): E max is the maximum power of the electric vehicle.

[0124] Step 4.7, calculate the total charging time T of the electric vehicle all , if the electric vehicle does not need to be charged, extract the residence time at this location. If the electric vehicle needs to be charged, the residence time of the electric vehicle at this location is the total charging time:

[0125]

[0126] In Equation (18): T d is the time when the electric vehicle arrives at the charging station; is the charging queue time of the electric vehicle; T c is the charging time of the electric vehicle.

[0127] As Figure 3 shown, for the residential area load with a late peak, the charging load is relatively low during the day. For the work area load with a noon peak, and the commercial area load is mainly concentrated during its business operation period. By superimposing the loads of the three functional areas, the total regional load can be obtained. It can be seen from the figure that there are noon and late peaks in the charging load, and the load is at a low point in the early morning.

[0128] The present invention builds a traffic road network model considering dynamic traffic information and road travel time, uses a Voronoi diagram to define traffic node types and divide regions, and constructs an energy consumption model per unit mileage of an electric vehicle considering environmental temperature and vehicle real-time speed; uses a bidirectional long short-term memory neural network method to mine and analyze road traffic information and electric vehicle travel data, predict road traffic flow and output the probability distribution of different types of electric vehicle travel chains; plans the driving path of the electric vehicle by improving the Floyd algorithm to simulate the driving behavior of electric vehicle users; analyzes the charging characteristics of different types of electric vehicles, establishes an electric vehicle queuing-gravitational model to calculate the total charging time of the electric vehicle, and predicts the spatio-temporal distribution of the electric vehicle charging load through Monte Carlo simulation. The present invention can realize the prediction of the charging load of electric vehicles in multiple types of functional areas, provide important information for formulating grid power dispatching strategies for each distribution network node, and has important significance for research such as power system load dispatching or charging station presetting.

[0129] Embodiment 2

[0130] An electronic device, including a memory and a processor. The memory is used to store a program that supports the processor to execute the method for predicting the load of an electric vehicle based on the fusion of vehicle-road-station and travel chain in Embodiment 1, and the processor is configured to execute the program stored in the memory.

[0131] Embodiment 3

[0132] A storage medium stores a computer program, and when the computer program is run by a processor, it executes the steps of the electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain in the first embodiment.

[0133] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An electric vehicle load forecasting method based on the integration of vehicle-road-station and travel chain, characterized in that Including: S1: Use graph theory method to model the road network topology, establish a traffic road network model considering dynamic traffic information and road travel time, define traffic node types and divide regions using Voronoi diagram, and construct an energy consumption model per unit mileage of electric vehicles considering environmental temperature and vehicle real-time speed; S2: Obtain historical travel data of electric vehicles, extract the travel information and characteristic data of historical road traffic flow from it, input the historical road traffic flow into a bidirectional long short-term memory network to output the predicted road traffic flow, input the travel information into a bidirectional long short-term memory network to output the predicted travel behavior of electric vehicles, extract the travel information such as travel time, stay time, and arrival location of electric vehicles, predict the relative rules of travel time and space of electric vehicles, and obtain the travel chain of electric vehicles; S3: Use the improved Floyd algorithm to plan the shortest driving path of electric vehicles; S4: Analyze the charging characteristics of different types of electric vehicles and predict the charging load of electric vehicles through the queuing-gravitation model.

2. The electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain according to claim 1, wherein The method of using graph theory method to model the road network topology is as follows: Among them, R N is the road network topology within the research area; N is the set of all road network traffic nodes within the road network except for the charging station nodes; C is the set of charging station nodes, representing the locations where electric vehicles are charged; S is the set of road segments within the road network; D is the set of distances of each road segment; W is the set of road impedances within the road network; n i is a road network node; c k is a charging station node; s l is a road segment; d l is the distance length of road segment l; w l is the road impedance of road segment l.

3. The electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain according to claim 1, wherein The method of dividing regions by Voronoi diagram is as follows: V(n i ) = {q|||q - n i || ≤ ||q - n k ||, i ≠ k, k ∈ {1, …, n}} (2) Among them, V(n i ) is a Voronoi polygon formed with the road network node n i as the center point, q is an arbitrary point in the Voronoi polygon, ||q - n i || ≤ ||q - n k || means that the distance from any point in the Voronoi polygon V(n i ) to the road network node n i is the shortest.

4. The electric vehicle load forecasting method based on the integration of vehicle-road-station and travel chain according to claim 1, wherein The method of constructing a traffic road network model is as follows: Among them, t l is the travel time of section l, i.e., the traffic impedance; c l is the initial designed traffic flow of section l during the traffic planning stage; q l is the traffic flow of section l; t0 is the travel time in the traffic relaxation state; α and β are fixed parameters.

5. The electric vehicle load forecasting method based on the integration of vehicle-road-station and travel chain according to claim 1, wherein, The method of constructing an energy consumption model per unit mileage of electric vehicles considering environmental temperature and vehicle real-time speed is as follows: Among them, Q T is the power consumption per unit mileage; K T is the air-conditioning power consumption per unit mileage; v l is the driving speed of the vehicle on section l; D is the driving distance; A L is the air-conditioning cooling power; A R is the air-conditioning heating power; T is the outdoor temperature; T max is the lower limit of the air-conditioning cooling temperature; T min is the upper limit of the air-conditioning heating temperature.

6. The electric vehicle load forecasting method based on the integration of vehicle-road-station and travel chain according to claim 1, wherein The method of using the improved Floyd algorithm to plan the shortest driving path of electric vehicles is as follows: Obtain the distance between any two nodes according to the traffic road network topology model; correct the distance between two nodes using the road blockage degree between any two traffic nodes to generate the adjacency matrix of the traffic road network; Generate a path distance storage matrix and a path node storage matrix with the same dimension as the adjacency matrix and update them; traverse all traffic nodes to obtain the final shortest path storage matrix and path node storage matrix; According to the predicted travel chain of electric vehicles, calculate the shortest path between each location in the travel chain through the improved Floyd algorithm, and the electric vehicle travels according to the shortest path to obtain the complete driving path of the electric vehicle traffic road network; [[ID=(-13)]]According to the predicted traffic flow of each road and the established traffic impedance model of travel time, obtain the driving speed of the electric vehicle based on the traffic flow of the driving path of the electric vehicle, and obtain the power consumption per unit mileage of the electric vehicle during driving.

7. The electric vehicle load prediction method based on the integration of vehicle-road-station and travel chain according to claim 6, wherein The update of the adjacency matrix of the traffic road network is as follows: The adjacency matrix of the traffic road network is: The condition for iteratively updating the adjacency matrix Dist is: Among them, Dist m (i,j) is the shortest distance between nodes i and j after m iterations of update; Dist m-1 (i,j) is the shortest distance between nodes i and j after m - 1 iterations of update; Dist m-1 (i,k) + Dist m-1 (k,j) is the distance from node i to node j passing through node k after m - 1 iterations of update; m is the number of updates.

8. The electric vehicle load forecasting method based on the integration of vehicle-road-station and travel chain according to claim 1, wherein The method of predicting the charging load of electric vehicles through the queuing-gravitation model is as follows: Construct an electric vehicle queuing-gravitation model to determine the charging stations selected by electric vehicles for charging: Among them, F ij is the attraction between charging station i and the j-th electric vehicle; T ij is the queuing time gravitational coefficient for the j-th electric vehicle to reach charging station i; M i is the "celestial body" mass of the charging station; m j is the "planet" mass of the electric vehicle; r is the shortest distance between the electric vehicle and the charging station; among them, the queuing time coefficient T ij : Among them, is the average waiting time of electric vehicle j at the i-th charging station; E T , V T are the expectation and variance of the Poisson distribution with parameter λ that the electric vehicle process follows; C is the number of charging stations; ρ = λE T , λ is a parameter; the "celestial body" mass M of the charging station i is: M i = τ i μ i (9) Among them, τ i is the number of charging piles in charging station i; μ i is the in-station charging efficiency; the mass m of the electric vehicle "planet" ij is as follows: Among them, A ij is the shortest distance between charging station i and electric vehicle j; E j is the battery level of electric vehicle j; E th is the battery level threshold.

9. An electronic device, comprising a memory and a processor, characterized in that The memory is used to store the program for supporting the processor to execute the method for predicting the load of electric vehicles based on the integration of vehicle-road-station and travel chain according to any one of claims 1 to 8, and the processor is configured to execute the program stored in the memory.

10. A storage medium has a computer program stored thereon, characterized in that, When the computer program is run by the processor, it executes the steps of the method for predicting the load of electric vehicles based on the integration of vehicle-road-station and travel chain according to any one of claims 1 to 8. It should be noted that there seems to be an issue with the numbering in the original text where ID=(-13) is an incorrect numbering. This might be a typo in the original Chinese text. The translation is done as accurately as possible based on the provided content.

Images