Electric vehicle charging demand sensing method based on roadside camera

Abstract

The invention belongs to the field of electric vehicle charging, and relates to a roadside camera-based electric vehicle charging demand sensing method, which comprises the following steps of: calculating the traffic flow weight, the vehicle waiting time weight, the road width weight and the camera deployment cost of each intersection, and determining the intersection for deploying the camera in combination with a greedy algorithm and a genetic algorithm; according to the traffic flow weight, the vehicle waiting time weight and the road width weight, the definition and the number of the cameras at the intersection where the cameras are deployed are determined, the cameras are used for collecting video data, and the video data collected by the cameras are subjected to edge processing and then transmitted to the cloud; the cloud carries out comprehensive judgment and early warning according to the received data; the position of the road monitoring camera is determined according to the traffic flow weight, the vehicle waiting time weight, the road width weight and the camera deployment cost in combination with the improved greedy algorithm and the genetic algorithm, and the deployment cost is minimized while the camera meets the maximum road coverage rate.

Description

Technical Field

[0001] The present invention belongs to the field of electric vehicle charging and relates to a method for sensing electric vehicle charging demand based on a roadside camera. Background Art

[0002] The development of new energy vehicles is the only way for my country to move from an automobile power to an automobile powerhouse. It is also an important measure to cope with global climate change, promote green development and realize the dual-carbon strategy. At present, the number of new energy vehicles in my country is very large. Such a large-scale access to the power grid will have a great impact on the power grid load. Moreover, since the charging time of electric vehicles is longer than refueling, it may cause congestion at charging stations. In response to the above problems, road-side cameras are used to collect driver information and transmit it to traffic management departments, power companies and other departments. By analyzing the data, early warning and guidance are provided for areas where large-scale charging will occur, reducing the risk of power grid load being impacted in the area.

[0003] At present, roadside cameras are only deployed to deal with traffic accidents and traffic violations. Their coverage area is not wide enough, the deployment cost is not small enough, and the real-time and clarity of video transmission are low, which cannot meet the needs of early warning. Summary of the invention

[0004] In order to solve the above problems in the prior art, the present invention adopts a method for sensing charging demand of electric vehicles based on a roadside camera, comprising:

[0005] S1, calculate the traffic flow weight, vehicle waiting time weight, road width weight and camera deployment cost of each intersection, and determine the intersection where the camera is deployed according to the traffic flow weight, vehicle waiting time weight, road width weight and camera deployment cost of the intersection by combining the greedy algorithm and the genetic algorithm;

[0006] S2. Determine the clarity and number of cameras at the intersections where cameras are deployed according to the traffic flow weight, vehicle waiting time weight, and road width weight;

[0007] S3, using the deployed cameras to collect video data, and transmitting the video data collected by the cameras to the cloud after edge processing;

[0008] S4. The cloud makes comprehensive judgments and early warnings based on the received data.

[0009] The intersections where cameras are deployed include:

[0010] S11. Construct an undirected graph G of urban roads 0 =(P,S), construct the minimum vertex cover set; where P is the vertex set, the vertex is the intersection, S is the edge set, and the edge is the road connecting the intersection;

[0011] S12, calculate each vertex p according to the traffic flow weight, vehicle waiting time weight, road width weight and the cost of deploying cameras i The comprehensive weight of Where i is the index of the vertex;

[0012] S13. Calculate the undirected graph G of urban roads z-1 For each vertex p i The proximity The adjacent and comprehensive weight Combine to get the comprehensive proximity Where z is the number of iterations;

[0013] S14, if the comprehensive adjacency If there are multiple highest vertices, select the vertex with the highest comprehensive adjacency. The lowest vertex; otherwise, directly select the comprehensive adjacency the highest peak;

[0014] S15. Add the selected vertices to the minimum vertex cover set in the urban road undirected graph G. z-1 Delete the selected vertex and its associated edges to obtain the city road undirected graph G z ;

[0015] S16. Determine the undirected graph G of urban roads z Have all the edges in been deleted? If so, the minimum vertex cover set of the greedy algorithm is obtained, that is, the optimized camera deployment position; otherwise, return to step S13;

[0016] S17. The minimum vertex cover set of the greedy algorithm is used as an individual in the genetic algorithm. The genetic algorithm is used to process the individual to obtain the final minimum vertex cover set, that is, to determine the intersection where the camera is deployed.

[0017] Calculate intersection p i The traffic flow weights include: obtaining intersection p i According to the traffic flow, calculate each intersection p i Average traffic volume during peak hours and the average traffic volume per day Average traffic volume at all intersections Select the largest traffic flow as the maximum average traffic flow during the peak period Average traffic volume at all intersections Select the largest traffic flow as the maximum average traffic flow in a day Based on average traffic volume Get intersection p iThe traffic flow weight.

[0018] Calculate intersection p i The road width weights include: Get the intersection p i The width of the connected road, calculate the intersection p i The sum of the widths of all roads R i , select the largest road width r among all the roads connected by the intersection max , according to the sum of the width R i and the maximum road width r max Calculate intersection p i The road width weight.

[0019] Calculate intersection p i The vehicle waiting time includes: obtaining the intersection p i In each phase Green light time of traffic lights at and period C; the phase corresponds to a traffic flow direction at the intersection; obtain the intersection p i In each phase Saturation flow at Get intersection p i In each phase Traffic volume According to traffic flow Calculate intersection p i In each phase The weighted average traffic volume at Calculate weighted average traffic flow and saturation flow Ratio Based on the weighted average traffic volume time Period C and Ratio Calculate the vehicle passing through the intersection p i The total delay time t i , according to the total delay time t i Calculate intersection p i The vehicle waiting time weight.

[0020] Calculating the cost of deploying cameras includes: obtaining the road width at each intersection, obtaining the i Camera installation height h i And the horizontal viewing angle α of the camera h , vertical viewing angle α v and price, according to height h i , horizontal viewing angle α h and vertical viewing angle α v Calculate intersection p i The bottom width WB of the camera's monitoring rangei , according to the bottom width WB i and intersection p i The road width is calculated in p i The number of cameras installed at the intersection is calculated based on the number of cameras and the price. i The cost of deploying cameras.

[0021] Determining the clarity of the camera at the intersection where the camera is deployed includes: calculating the camera selection weight according to the traffic flow weight, vehicle waiting time weight and road width weight at the intersection where the camera is deployed, and determining the clarity of the camera at the intersection where the camera is deployed according to the camera selection weight.

[0022] The video data collected by the camera is edge processed and then transmitted to the cloud, including: building a system model, the system model includes: a cloud, a base station and its associated cameras; building a task offloading model and a computing resource allocation model according to the system model, building a video data task optimization model according to the task offloading model and the computing resource allocation model, solving the video data task optimization model to obtain the optimal task offloading decision and computing resource allocation decision, and transmitting the video data collected by the camera to the cloud after edge processing according to the optimal task offloading decision and computing resource allocation decision.

[0023] The video data task optimization model is:

[0024]

[0025]

[0026] Among them, C1 represents each camera u i Video can only be offloaded to one base station j Or cloud execution, C2 represents each base station e j is the camera u associated with it i Allocate a computing resource, and the computing resource allocated cannot exceed the base station e j The total resources of each base station e j Each subband k can be associated with at most one camera u. i For video transmission, C4 means that the previous layer of video data must be transmitted before the next layer of video data during video transmission; J(X, F) represents the target optimization function, For camera u i The computational offloading utility of Indicates the base station to the camera u i The preference degree, X and F represent the task offloading decision and computing resource allocation decision respectively, N(e j ) indicates the connection with base station e jThe associated camera set, E represents the base station set, Indicates the task offloading decision, indicating whether to offload from camera u i The mth layer of the video is offloaded to base station e via subband k j In the figure, M is the set of video levels, K is the set of subbands, Indicates base station e j Assigned to camera u i The amount of computing resources, f s is the available computing resources of the base station, Indicates camera u i Through subband k to base station e j The rate at which videos are uploaded to the m-layer.

[0027] Camera i The computational offloading utility is:

[0028]

[0029]

[0030] in, Respectively represent the preference for latency, energy consumption and cost, They represent the time characteristics, energy consumption characteristics, and cost characteristics of the video data task transmission completion, respectively. To reduce the latency of offloading video data tasks directly to the cloud, To set the camera u i The total delay of offloading the video data task to the base station, To offload the video data task directly to the cloud, To set the camera u i The total energy consumption of offloading the video data task to the base station, Indicates that the camera u i Video offload to base station j decision, To set the camera u i The transmission cost of offloading the video data task to the base station, Set the camera to i The transmission cost of transmitting video data directly to the cloud.

[0031] Beneficial effects:

[0032] 1. The present invention combines an improved greedy algorithm and a genetic algorithm to determine the position of a road monitoring camera according to the traffic flow weight, the vehicle waiting time weight, the road width weight and the cost of deploying the camera, so that the camera can maximize the road coverage while minimizing the deployment cost, so that the camera can obtain vehicle data more comprehensively, and can make more accurate and rapid comprehensive judgments and early warnings; 2. The present invention selects the clarity and number of cameras at the intersections where the cameras are deployed according to the traffic flow weight, the vehicle waiting time weight and the road width weight of the intersections where the cameras are deployed, thereby improving the clarity and effectiveness of the number of cameras at the intersections where the cameras are deployed; 3. The present invention transmits the video segmentation to the edge server for processing and then transmits it back to the cloud or the control center, thereby reducing the transmission delay of the video task, ensuring the real-time nature of the video transmission, and also ensuring the security of the driver's information. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A flow chart of a method for sensing charging demand of an electric vehicle based on a roadside camera provided in an embodiment of the present invention;

[0034] Figure 2 A flowchart combining an improved greedy algorithm and a genetic algorithm provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] like Figure 1 As shown, the present invention adopts a method for sensing charging demand of electric vehicles based on a roadside camera, comprising:

[0037] S1. Calculate the traffic flow weight, vehicle waiting time weight, road width weight and camera deployment cost of each intersection, and determine the camera deployment location based on the traffic flow weight, vehicle waiting time weight, road width weight and camera deployment cost of the intersection by combining the greedy algorithm and the genetic algorithm;

[0038] The urban road surveillance camera network is mapped into an undirected graph, so the problem of optimizing the deployment of the camera network can be transformed into the problem of finding the minimum vertex that can cover all the edges of the graph. The camera deployment problem is solved using an improved greedy algorithm combined with a genetic algorithm to obtain a deployment strategy for urban road side cameras. The purpose of the deployment strategy is to minimize the deployment cost while maximizing the coverage of all roads.

[0039] like Figure 2 As shown, the specific steps include:

[0040] S11. Initialization parameters: Constructing an undirected graph G of urban roads 0 =(P, S); where P = {p1, p2, ..., p n} is a vertex set, representing the intersection, S = {s1, s2, ..., s m} is an edge set, representing the roads connecting the intersections; the counting graph G 0 Initialize the minimum vertex cover set MPOS to be empty, and initialize the number of covered edges COS to 0;

[0041] S12, calculate each vertex p according to the traffic flow weight, vehicle waiting time weight, road width weight and the cost of deploying cameras i The comprehensive weight of Where i is the index of the vertex;

[0042] S13. Calculate the undirected graph G of urban roads obtained by the last iteration z-1 z-1 For each vertex p i The proximity The adjacent and comprehensive weight Combine to get the comprehensive proximity Where z is the number of iterations;

[0043] The specific steps include:

[0044] Calculate the adjacency of each vertex:

[0045] In the undirected graph G of urban roads z-1 Center and vertex p i The set of all adjacent points is N G (p i ), then vertex p i The proximity for:

[0046]

[0047] Among them, p j For the vertex p i Adjacent vertices, d i is the vertex p i The degree of a vertex is the number of edges connected to it.

[0048] Calculation of traffic flow weights at each intersection includes:

[0049] Traffic flow reflects the number of vehicles passing through a certain intersection or road section, which can be obtained through direct observation or calculation by traffic models. The higher the traffic flow, the more important it is to monitor. in It is through the intersection i of traffic volume, is the number of people passing through intersection p within the time period h i The traffic flow at intersection p is i The traffic flow weight W flow (p i )for:

[0050]

[0051] α1+β1=1

[0052] Among them, W flow (p i ) is the intersection p i The traffic flow weight, It is a junction i Average traffic volume during peak hours, It is a junction i Average traffic volume in a day; It represents the maximum average traffic flow at all intersections during peak hours. It represents the maximum average traffic flow of all intersections in a day; α1 and β1 are the weight coefficients of peak traffic flow and average traffic flow in a day, respectively.

[0053] Calculation of road width weights at each intersection includes:

[0054] Road width affects the coverage of surveillance cameras. The wider the road, the higher the resolution or the more cameras are needed to cover it. The road width weight is:

[0055]

[0056] Among them, W width (p i ) is the intersection p i The road width weight, R i For intersection p i The sum of the widths of the connected roads, r max The maximum road width among all roads connected by the intersection.

[0057] Calculation of waiting time weights for vehicles at each intersection includes:

[0058] The waiting time reflects the time a vehicle waits to pass through an intersection, which is usually related to the traffic lights and the degree of congestion at the intersection. The longer the waiting time, the more important the monitoring is. The set of vehicle travel time at each intersection is T = {t1, t2, ..., t n}, the set of traffic light time at each intersection is L = {l1, l2, ...l O}, then the vehicle passes through the intersection p i The total delay time t i for:

[0059]

[0060] Among them, O is the number of phases at the intersection. Each phase corresponds to a specific traffic flow direction (such as north-south straight, east-west straight, north-south left turn, east-west left turn, etc.). In one signal cycle, all phases get green light time in turn. For p i Phase at intersection The corresponding weighted average traffic flow, t i Indicates that the car is passing through the intersection p i The total delay time is , C is the cycle of the traffic light (seconds), For p i Phase at intersection The corresponding green light time (seconds), For p i Phase at intersection Traffic flow and saturation flow Q i The ratio of For intersection p i Peak hours in phase The average traffic volume at For intersection p i Phase within a day Average traffic volume at The calculation can be performed based on the above formula for calculating the average traffic flow during peak hours and the average traffic flow in a day in the traffic flow weights of each intersection, except that the intersection p i Traffic volume Replace with intersection p i In each phase Traffic volume Saturation flow It means that under ideal conditions, the intersection p i In phase The maximum number of vehicles that can pass in a unit time (usually one hour) is the vehicle waiting time weight:

[0061]

[0062] Among them, W wait (p i ) is the intersection p i The waiting time weight, t i Indicates that the vehicle is passing through the intersection p i The total delay time, t max Indicates the maximum total delay time among all intersections.

[0063] Calculate the intersection p i The costs of deploying cameras include:

[0064] The monitoring range is usually trapezoidal, with the bottom edge at the camera position and the top edge at the farthest point in the monitoring range.

[0065] Calculate the distance CD from the camera to the ground:

[0066]

[0067] Calculate the bottom width WB of the monitoring range:

[0068]

[0069] Among them, h represents the i The camera installation height is generally selected according to the actual location of the installation area. h Represents the horizontal viewing angle of the camera, α v Represents the vertical viewing angle of the camera, α h and α v Provided by the camera manufacturer.

[0070] Calculate the number of cameras:

[0071]

[0072] Calculate the cost of deploying cameras:

[0073] C deploy (p i )=C cam (p i )*CA i

[0074] Among them, CA i Indicates that in p i The number of cameras installed at r i，j Indicates intersection p j With intersection p i Width of connected roads, N G (p i ) indicates the intersection with p i A collection of connected intersections. represents rounding up to ensure that the number of cameras is sufficient to cover the entire intersection, C deploy (p i ) means at the intersection p i The cost of deploying cameras, C cam (p i ) means at the intersection p i Price to deploy a single camera.

[0075] The comprehensive weight calculation includes:

[0076]

[0077] α+β+γ=1

[0078] in, Represents vertex p i , α, β, γ, and λ are weight coefficients, reflecting the relative importance of traffic volume, waiting time, road width, and camera deployment cost in the comprehensive weight.

[0079] Calculate the comprehensive adjacency of each vertex Combining weight and proximity, we get:

[0080]

[0081] S14, if the comprehensive adjacency If there are multiple highest vertices, select the vertex with the highest comprehensive adjacency. The lowest vertex; otherwise, directly select the comprehensive adjacency the highest peak;

[0082] S15. Add the selected vertices to the minimum vertex cover set in the urban road undirected graph G. z-1 Delete the selected vertex and its associated edges, and update the degrees of the remaining vertices to obtain the city road undirected graph G z ;

[0083] S16. Determine the undirected graph G of urban roads z Have all the edges in been deleted? If so, the minimum vertex cover set of the greedy algorithm is obtained, that is, the optimized camera deployment position; otherwise, return to step S13.

[0084] Although the greedy algorithm can quickly generate an approximate solution, it may get stuck in a local optimum.

[0085] In one embodiment, the vertex cover set of the greedy algorithm is further optimized by combining the genetic algorithm to make it globally optimal; the specific steps include:

[0086] Step 1, individual coding: The minimum vertex cover set generated by the greedy algorithm is represented as a binary vector, with each vertex corresponding to one bit, 1 means that the vertex is in the cover set (i.e., the camera is deployed), and 0 means it is not.

[0087] Step 2, initial population: The minimum vertex cover set represented as a binary vector is used as the individuals in the population, and multiple individuals are randomly generated to increase the diversity of the population.

[0088] Step 3: Calculate the fitness function of each individual in the population: The fitness function evaluates the quality of each individual. The goal is to have the least number of vertices and cover all edges. The fitness function fitness(X) can be expressed as:

[0089]

[0090] Among them, μ is the penalty weight for controlling the number of vertices to prevent the individual from containing too many vertices, and X represents the current individual.

[0091] Step 4: Selection operation: According to the fitness value of the individual, use the roulette wheel selection or tournament selection method to select individuals with higher fitness values ​​from the population as the selected individual set.

[0092] Step 5: Crossover operation: Perform single-point or multi-point crossover on the individuals in the selected individual set to generate new individuals, and form all new individuals into a crossover individual set; the offspring after crossover will inherit some characteristics of the parent generation.

[0093] Step 6, mutation operation: Randomly select some genes of individuals in the crossover individual set for mutation to obtain a mutant individual set. The mutation operation can remove some vertices from the cover set (0 means no camera is deployed) or add vertices to the cover set (1 means camera is deployed) to increase population diversity and avoid local optimality.

[0094] Step 7, repair operation: After the mutation and crossover operations, individuals that do not meet the constraints may be generated (for example, a vertex cover set that does not cover all edges). At this time, it is necessary to perform a repair operation on the individuals in the mutated individual set and re-add the necessary vertices so that all edges in the urban road undirected graph are covered and a repaired individual set is obtained.

[0095] Step 8: Merge the selected individual set and the repaired individual set to obtain a new population.

[0096] Step 9, iteration and termination: Repeat steps 3 to 8 until the termination condition is met (such as reaching the maximum number of iterations or the fitness no longer increases significantly), and take the individual with the highest fitness in the current population as the final minimum vertex cover set.

[0097] S2. Determine the clarity and number of cameras at the intersections where cameras are deployed according to the traffic flow weight, vehicle waiting time weight, and road width weight;

[0098] Determining the clarity of the camera at the intersection where the camera is deployed includes:

[0099] Based on the minimum vertex cover set MPOS output above, the cameras at the deployment location are selected. The camera resolutions are divided into three types: ultra-high definition, high definition, and standard definition. Cameras with different resolutions are selected for intersections with different traffic volumes and vehicle waiting times. The larger the traffic volume and the longer the waiting time of vehicles, the higher the resolution of the camera deployed at the intersection, and vice versa. The camera selection weight W is camera for:

[0100] W camera (p i )=α2*W flow (p i )+β2*W wait (p i )+γ2*W width (p i )

[0101] α2+β2+γ2=1

[0102] Among them, W flow (p i ) means in p i The traffic flow weight at wait (p i ) means in p i The vehicle waiting time weight at width (p i ) means in p i The road width weight at , α2, γ2, and β2 represent the corresponding weight coefficients, reflecting the importance of traffic flow, road width, and vehicle waiting time for selecting camera clarity.

[0103] The camera resolution θ i for:

[0104]

[0105] Camera selection weight W camera The higher it is, the clearer the cameras deployed at the intersection will be.

[0106] The number of cameras at the intersection where cameras are deployed is determined by:

[0107] Determine the camera deployment method for the selected deployment location, including:

[0108] The deployment of cameras at intersections should cover oncoming vehicles as much as possible, and the camera shooting angle should be as close to the middle of the main road as possible. Mainly use the front angle and the top angle, avoid selecting the side angle, which will cause excessive distortion of the lane width and reduce the phenomenon of vehicles blocking each other. The camera is installed mainly to collect information about the driver, and also serves traffic violations and traffic safety incidents. This embodiment selects a directional camera, which has a fixed angle of view and its monitoring range is usually a trapezoidal or fan-shaped area.

[0109] Calculate the camera monitoring range, including:

[0110] The coverage area of ​​a camera is usually trapezoidal, with the bottom edge at the camera position and the top edge at the farthest point in the monitoring range.

[0111] Distance from camera to ground CD i :

[0112]

[0113] The bottom width WB of the monitoring range:

[0114]

[0115] Among them, h represents the i The camera installation height is generally selected according to the actual location of the installation area. h Represents the horizontal viewing angle of the camera, α v Represents the vertical viewing angle of the camera, α h and α v Provided by the camera manufacturer.

[0116] Calculate the number of cameras required CA i :

[0117]

[0118] Among them, CA i Indicates that in p i The sum of the number of cameras installed at i，j Represented by vertex p j With vertex p i The width of the connected road, N G (p i ) represents the vertex p i The set of connected vertices, It means rounding up to ensure that the number of cameras is sufficient to cover the entire intersection.

[0119] S3, using the deployed cameras to collect video data, and transmitting the video data collected by the cameras to the cloud after edge processing;

[0120] In order to ensure the validity and timeliness of video data, it is necessary to offload the video data collected by the camera to the surrounding edge servers for processing, using U = {u1, u2, u3...u U} represents the camera set deployed on the road side, E = {e1, e2, e3...e E} represents the edge server MEC set, M = {1, 2, 3...M} represents the set of video levels, and the clarity of videos at different levels is different.

[0121] Build a system model:

[0122] The system model includes: multiple cameras, multiple base stations BS and a cloud. j Associate multiple cameras, each camera corresponds to a base station; each camera u i There are video computing tasks Its performance depends on the task data size and computing resource requirements The server of each base station has certain computing resources (CPU cycles), and each video task can be offloaded to the MEC server of the associated base station for processing; orthogonal frequency division multiple access technology is applied to the uplink transmission system, and the transmission band B is divided into N equal sub-bands of size ω, that is, Each BS can receive tasks uploaded by at most N cameras simultaneously, and the set of subbands available to each BS is K = {1, 2, 3...N}.

[0123] Build a task offloading model:

[0124] Define the uninstall variable as Among them, u i ∈N(e j ), m∈M, e j ∈E,k∈K,N(e j ) indicates the connection with base station e j The set of associated cameras, E represents the base station set, and M represents the video level set; Indicates that from camera u i The mth layer of the video is offloaded to base station e via subband k j superior, The meaning is the opposite, that is:

[0125]

[0126] Then the task offloading strategy X is:

[0127]

[0128] Through feasibility analysis, we can get:

[0129]

[0130] Build a computing resource allocation model:

[0131] Define the computing resource allocation decision F as:

[0132]

[0133] Among them, f s For each base station’s available computing resources, Indicates base station e j Assigned to camera u i The amount of computing resources,

[0134] According to the task offloading model and computing resource allocation model, the total system delay and total energy consumption are calculated:

[0135] The following delays will occur during the offloading process: the delay of directly offloading the video task to the remote cloud; the delay of the camera offloading the video task to the MEC server in the base station through the uplink; the delay of the MEC server in the base station processing the upload task; and the time to transmit the processed data from the base station back to the cloud.

[0136] Delay in directly offloading video tasks to the remote cloud and energy consumption and transmission costs for:

[0137]

[0138]

[0139] Among them, R u Indicates camera u i The transmission rate of video tasks back to the cloud, Indicates camera u i The transmission power, Indicates camera u i The size of the camera data, θ i Indicates the clarity of the camera, F i Indicates the frame rate of the camera, T interval Indicates the transmission time interval of the camera (for example, every 1 second or every 5 seconds) to transmit the collected data to the cloud. η indicates the compression ratio, which is used to consider the amount of data after video compression (for example, H.264 encoding). This is usually provided by the camera or can be set manually according to the actual scene. From node u i The transmission cost of transferring data to the cloud, C cloud (ui ) is the transmission fee per bit transmitted to the cloud, which is related to the network type and transmission distance.

[0140] Offloading tasks to local edge servers MEC processing includes:

[0141] Calculate the latency of offloading the video task to the MEC server via the uplink:

[0142] Camera i and BS both have a single antenna for uplink transmission, then camera u i The signal-to-noise ratio of the video task layer m to subband k for:

[0143]

[0144] Among them, e l ∈E\{e j} represents base station e j All base stations except l ) indicates the connection with base station e l The set of associated cameras, u′ represents the base station e l The associated camera, σ 2 is the environmental noise variance; Indicates base station e j Associated with camera u i The channel gain coefficient transmitted on the kth subband, Indicates camera u i The power of transmitting tasks to the server, Indicates that camera u′ offloads the mth layer of the video to server e l The decision on p u′ represents the power of camera u′ to transmit tasks to the server; Indicates server e j Channel gain coefficient for transmission between camera u′.

[0145] The path loss model used in the present invention is: in, Indicates server e j With camera u i The distance between them, then the channel gain for:

[0146]

[0147] Among them, g is a parameter used to represent the combined influence of random effects such as small-scale fading and multipath gain of signals in wireless channels. In simulation, it is usually represented by a random variable generated by Rayleigh or Rice distribution.

[0148] Camera i To base station e j Upload video rate for:

[0149]

[0150] in, Indicates camera u i To base station e j Signal-to-noise ratio of uploaded video, camera u i Sending video data on uplink Transmission time for:

[0151]

[0152] in, Indicates camera u i Offload the mth layer of the video to server e via subband k j decision on.

[0153] Calculate the latency of the MEC server in the base station to process the upload task

[0154]

[0155] in, Indicates the computing resources required for video data; calculates the latency of transmitting processed data from the edge computing server back to the cloud:

[0156]

[0157] Transmission costs for:

[0158]

[0159] in, Indicates camera u i The size of the camera data, R mc is the transmission rate from the MEC server to the cloud, is the size of the task data after being processed by the edge server, σ represents the compression ratio of the video after being processed by the lightweight CNN processing algorithm deployed on the edge server, and C local (u i ) is the transmission cost per bit transmitted to the edge server, which is related to the network type and transmission distance.

[0160] On the edge server, the main task of the CNN processing algorithm is to analyze and extract video data to obtain key information, reduce the amount of data, and extract features related to vehicle charging requirements. The determination of the compression ratio σ mainly depends on factors such as the application goals of the system, the processing capabilities of the algorithm, network bandwidth, and computing resource limitations. Usually, the appropriate compression ratio is determined through experimental testing, dynamic adjustment, and optimization of the CNN model to achieve a balance between data volume and processing effect.

[0161] Camera i The total delay of offloading the task to the base station for processing is:

[0162]

[0163] Camera i The energy consumption of processing task uplink is:

[0164]

[0165] Base Station e j The energy consumption of transmitting processed data to the cloud is:

[0166]

[0167] Camera i The total energy consumption of offloading tasks to the base station for processing is:

[0168]

[0169] in, Camera u i The power of transmitting tasks to the server and the power of the server transmitting data to the cloud.

[0170] Then camera u i The optimization function is:

[0171]

[0172] in, Indicates camera u i The computational offloading utility of and represent the preference for latency and energy consumption respectively, The weight coefficient of the cost can be set through actual adjustment, and and They respectively represent the time characteristics, energy consumption characteristics, and cost characteristics of completing the video data task transmission.

[0173] The video data task optimization model is:

[0174]

[0175] Among them, C1 represents each camera u i Video can only be offloaded to one base station j Or cloud execution, C2 represents each base station e j The camera u associated with it i Allocate a meaningful computing resource, and the allocated computing resource cannot exceed the base station e j The total resources of each base station e j Each subband k can be associated with at most one camera u. i For video transmission, C4 means that the previous layer of video data must be transmitted before the next layer of video data during video transmission. The scalable video transmission feature is used to avoid the situation where only the upper layer data is received but not the lower layer data. J(X,F) is the target optimization function, which represents the weighted sum of system performance. For camera u i The computational offloading utility of Indicates the base station to the camera u i degree of preference.

[0176] For camera u i Through subband k to base station e j The rate of uploading videos at the m-layer:

[0177] The above optimization problem is solved by mixed integer linear programming optimization algorithm to obtain the unloading decision.

[0178] S4. The cloud makes comprehensive judgments and early warnings based on the received data.

[0179] The cloud can make a comprehensive judgment on the charging needs of electric vehicles based on the received data and send warning information. Specifically, the cloud can analyze the driver information collected by the camera and the data processed by the base station, and provide early warning and guidance for areas where large-scale charging will occur, reducing the risk of power grid load being impacted in the area.

[0180] The above embodiments further illustrate the purpose, technical solutions and advantages of the present invention in detail. It should be understood that the above embodiments are only preferred implementation modes of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made to the present invention within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for sensing charging demand of electric vehicles based on roadside cameras, characterized in that: include: S1, calculate the traffic flow weight, vehicle waiting time weight, road width weight and camera deployment cost of each intersection, and determine the intersection where the camera is deployed according to the traffic flow weight, vehicle waiting time weight, road width weight and camera deployment cost of the intersection by combining the greedy algorithm and the genetic algorithm; S2. Determine the clarity and number of cameras at the intersections where cameras are deployed according to the traffic flow weight, vehicle waiting time weight, and road width weight; S3, using the deployed cameras to collect video data, and transmitting the video data collected by the cameras to the cloud after edge processing; S4. The cloud makes comprehensive judgments and early warnings based on the received data.

2. The method for sensing charging demand of electric vehicles based on a roadside camera according to claim 1, characterized in that: The intersections where cameras are deployed include: S11. Construct an undirected graph G of urban roads 0 =(P, S), construct the minimum vertex cover set; where P is the vertex set, the vertex is the intersection, S is the edge set, and the edge is the road connecting the intersections; S12, calculate each vertex p according to the traffic flow weight, vehicle waiting time weight, road width weight and the cost of deploying cameras i The comprehensive weight of Where i is the index of the vertex; S13. Calculate the undirected graph G of urban roads z-1 For each vertex p i The proximity The adjacent and comprehensive weight Combine to get the comprehensive proximity Where z is the number of iterations; S14, if the comprehensive adjacency If there are multiple highest vertices, select the vertex with the highest comprehensive adjacency. The lowest vertex; otherwise, directly select the comprehensive adjacency the highest peak; S15. Add the selected vertices to the minimum vertex cover set in the urban road undirected graph G. z-1 Delete the selected vertex and its associated edges to obtain the city road undirected graph G z ; S16. Determine the undirected graph G of urban roads z Have all the edges in been deleted? If so, the minimum vertex cover set of the greedy algorithm is obtained, that is, the optimized camera deployment position; otherwise, return to step S13; S17. The minimum vertex cover set of the greedy algorithm is used as an individual in the genetic algorithm. The genetic algorithm is used to process the individual to obtain the final minimum vertex cover set, that is, to determine the intersection where the camera is deployed.

3. The method for sensing charging demand of electric vehicles based on a roadside camera according to claim 1, characterized in that: Calculate intersection p i The traffic flow weights include: obtaining intersection p i According to the traffic flow, calculate each intersection p i Average traffic volume during peak hours and the average traffic volume per day Average traffic volume at all intersections Select the largest traffic flow as the maximum average traffic flow during the peak period Average traffic volume at all intersections Select the largest traffic flow as the maximum average traffic flow in a day Based on average traffic volume Get intersection p i The traffic flow weight.

4. The method for sensing charging demand of electric vehicles based on a roadside camera according to claim 1, characterized in that: Calculate intersection p i The road width weights include: Get the intersection p i The width of the connected road, calculate the intersection p i The sum of the widths of all roads R i , select the largest road width r among all the roads connected by the intersection max , according to the sum of the width R i and the maximum road width r max Calculate intersection p i The road width weight.

5. The method for sensing charging demand of electric vehicles based on a roadside camera according to claim 1, characterized in that: Calculate intersection p i The vehicle waiting time includes: obtaining the intersection p i In each phase Green light time of traffic lights at and period C; the phase corresponds to a traffic flow direction at the intersection; obtain the intersection p i In each phase Saturation flow at Get intersection p i In each phase Traffic volume According to traffic flow Calculate intersection p i In each phase The weighted average traffic volume at Calculate weighted average traffic flow and saturation flow Ratio Based on the weighted average traffic volume time Period C and Ratio Calculate the vehicle passing through the intersection p i The total delay time t i , according to the total delay time t i Calculate intersection p i The vehicle waiting time weight.

6. The method for sensing charging demand of electric vehicles based on a roadside camera according to claim 1, characterized in that: Calculating the cost of deploying cameras includes: obtaining the road width at each intersection, obtaining the i Camera installation height h i And the horizontal viewing angle α of the camera h , vertical viewing angle α v and price, according to height h i , horizontal viewing angle α h and vertical viewing angle α v Calculate intersection p i The bottom width WB of the camera's monitoring range i , according to the bottom width WB i and intersection p i The road width is calculated in p i The number of cameras installed at the intersection is calculated based on the number of cameras and the price. i The cost of deploying cameras.

7. The method for sensing charging demand of electric vehicles based on a roadside camera according to claim 1, characterized in that: Determining the clarity of the camera at the intersection where the camera is deployed includes: calculating the camera selection weight according to the traffic flow weight, vehicle waiting time weight and road width weight at the intersection where the camera is deployed, and determining the clarity of the camera at the intersection where the camera is deployed according to the camera selection weight.

8. The method for sensing charging demand of electric vehicles based on a roadside camera according to claim 1, characterized in that: The video data collected by the camera is edge processed and then transmitted to the cloud, including: building a system model, the system model includes: a cloud, a base station and its associated cameras; building a task offloading model and a computing resource allocation model according to the system model, building a video data task optimization model according to the task offloading model and the computing resource allocation model, solving the video data task optimization model to obtain the optimal task offloading decision and computing resource allocation decision, and transmitting the video data collected by the camera to the cloud after edge processing according to the optimal task offloading decision and computing resource allocation decision.

9. The method for sensing charging demand of electric vehicles based on a roadside camera according to claim 8, characterized in that: The video data task optimization model is: Among them, C1 represents each camera u i Video can only be offloaded to one base station j Or cloud execution, C2 represents each base station e j is the camera u associated with it i Allocate a computing resource, and the computing resource allocated at the same time cannot exceed the base station e j The total resources of each base station e j Each subband k can be associated with at most one camera u. i For video transmission, C4 means that the previous layer of video data must be transmitted before the next layer of video data during video transmission; J(X, F) represents the target optimization function, For camera u i The computational offloading utility of Indicates the base station to the camera u i The preference degree of each task is represented by X and F, respectively, and N(e j ) indicates the connection with base station e j The associated camera set, E represents the base station set, Indicates the task offloading decision, indicating whether to offload from camera u i The mth layer of the video is offloaded to base station e via subband k j In the figure, M is the set of video levels, K is the set of subbands, Indicates base station e j Assigned to camera u i The amount of computing resources, f s is the available computing resources of the base station, Indicates camera u i Through subband k to base station e j The rate at which videos are uploaded to the m-layer.

10. The method for sensing charging demand of electric vehicles based on a roadside camera according to claim 9, characterized in that: Camera i The computational offloading utility is: in, Respectively represent the preference for latency, energy consumption and cost, They represent the time characteristics, energy consumption characteristics, and cost characteristics of the video data task transmission completion, respectively. To reduce the latency of offloading video data tasks directly to the cloud, To set the camera u i The total delay of offloading the video data task to the base station, To directly offload the video data task to the cloud, To set the camera u i The total energy consumption of offloading the video data task to the base station, Indicates that the camera u i Video offload to base station j decision, To set the camera u i The transmission cost of offloading the video data task to the base station, Set the camera to i The transmission cost of transmitting video data directly to the cloud.

Images