Expressway electric vehicle charging load prediction method and system

Abstract

The invention discloses a highway electric vehicle charging load prediction method and system. The method comprises the following steps: constructing a target highway entrance and exit vehicle OD matrix; preliminarily calculating the total number M of the electric vehicles running on the target highway; initializing initial parameters of all the electric vehicles; based on the initial parameters of the mth electric vehicle, generating operation parameters for the mth electric vehicle; mapping the operation parameters of the mth electric vehicle to the expressway road network model; the travel time, the state of charge and the remaining mileage of the mth electric vehicle in the expressway road network model are updated; estimating whether the mth electric vehicle in the expressway road network model needs to be charged in the service area, if so, calculating the charging duration of the mth electric vehicle, if not, judging whether the remaining mileage is zero, judging whether m is smaller than M or not, if not, calculating the total charging load of each expressway service area, and ending.

Description

Technical Field

[0001] The present invention relates to the technical field of charging load prediction, and in particular to a method and system for predicting charging load of electric vehicles on highways. Background Art

[0002] Driven by the global low-carbon transformation development goals, electric vehicles have led the direction of global automobile development due to their green and environmentally friendly characteristics. It can be foreseen that electric vehicles will gradually replace traditional fuel vehicles and become the mainstream means of transportation. At the same time, as an important supporting facility for electric vehicles, there is still a large gap between charging facilities and market demand. In addition, due to the lack of actual use data of electric vehicles as support and guidance, some charging facilities in the early stage have problems such as imperfect layout, unreasonable structure, and unbalanced services. These problems are particularly prominent on highways, resulting in the phenomenon of "one pile is hard to find". At the same time, due to the inherent deficiencies of electric vehicles in long-distance and high-speed driving, the resulting "mileage anxiety" problem has seriously hindered the popularization of electric vehicles on highways. Obviously, scientific and reasonable planning of charging facilities is the key to solving the above problems, and accurately grasping the laws of electric vehicle charging behavior on highways and the spatiotemporal distribution characteristics of charging demand is the top priority for scientific planning of highway charging facilities.

[0003] However, given the extremely high temporal and spatial freedom of electric vehicle travel, the strong heterogeneity of users' charging behavior and the complex influencing factors, the charging behavior of electric vehicles is highly random, and it is difficult to accurately predict the charging load and distribution of electric vehicles. Research on electric vehicle charging load prediction has achieved relatively fruitful results, but most existing research focuses on the problem of charging load prediction of electric vehicles in urban road networks, while research on highway charging load prediction is still relatively lacking. There are two main reasons for this situation. First, the travel behavior of electric vehicles on highways has greater uncertainty. In urban areas, a considerable proportion of electric vehicles have relatively regular travel and charging behaviors, while in highway scenarios, electric vehicle travel is usually sporadic and has strong randomness in time and space, making it more difficult to predict the charging load of electric vehicles on highways. Second, due to the limitations of power battery technology, electric vehicles have limited driving range and slow charging speed, resulting in a low proportion of electric vehicles on highways, and research on highway electric vehicle charging load prediction lacks sufficient basic data support. Summary of the invention

[0004] In order to solve the shortcomings of the existing technology, the present invention provides a method and system for predicting the charging load of electric vehicles on highways; a field survey of the charging behavior of electric vehicles on highways is conducted, and a charging behavior model of electric vehicles on highways is established based on the survey data. On this basis, a method for predicting the charging load of electric vehicles on highways based on Monte Carlo simulation is proposed. The research results of the present invention can provide a theoretical basis for the optimization of charging pile configuration in highway service areas, and have important theoretical significance and practical application value for promoting the development of electric vehicles.

[0005] On the one hand, a method for predicting charging load of electric vehicles on a highway is provided, comprising:

[0006] (1) Construct the OD matrix of the vehicles at the entrance and exit of the target highway; preliminarily calculate the total number M of electric vehicles on the target highway; calculate the hourly flow of each OD pair based on the OD matrix; and initialize the initial parameters of all electric vehicles;

[0007] (2) Based on the initial parameters of the mth electric vehicle, generate operating parameters for the mth electric vehicle; map the operating parameters of the mth electric vehicle to the highway network model; the initial value of m is 1;

[0008] (3) Update the travel time, state of charge, and remaining mileage of the mth electric vehicle in the highway network model;

[0009] (4) Estimate whether the mth electric vehicle in the highway network model needs to be charged in the service area. If it needs to be charged, calculate the charging time of the mth electric vehicle and return to (3); if it does not need to be charged, determine whether the remaining mileage is zero. If so, it means that the electric vehicle has arrived at the destination, record the end time and battery charge state, and enter (5); if the remaining mileage is greater than zero, return to (3);

[0010] (5) Determine whether m is less than M. If so, m=m+1 and return to (2). If not, calculate the total charging load of each highway service area and end.

[0011] On the other hand, a highway electric vehicle charging load prediction system is provided, comprising:

[0012] A construction module is configured to: construct an OD matrix of vehicles at the entrance and exit of the target highway; preliminarily calculate the total number M of electric vehicles traveling on the target highway; calculate the hourly flow of each OD pair based on the OD matrix; initialize the initial parameters of all electric vehicles;

[0013] A generating module is configured to: generate operating parameters for the mth electric vehicle based on the initial parameters of the mth electric vehicle; map the operating parameters of the mth electric vehicle to the highway network model; the initial value of m is 1;

[0014] An updating module is configured to: update the travel time, state of charge and remaining mileage of the mth electric vehicle in the highway network model;

[0015] The estimation module is configured to: estimate whether the mth electric vehicle in the highway network model needs to be charged in the service area; if charging is required, calculate the charging time of the mth electric vehicle and return to the update module; if charging is not required, determine whether the remaining mileage is zero; if so, it means that the electric vehicle has arrived at the destination, record the end time and battery charge state, and enter the judgment module; if the remaining mileage is greater than zero, return to the update module;

[0016] The judgment module is configured to: judge whether m is less than M, if so, m=m+1, and return to the generation module; if not, calculate the total charging load of each highway service area, and end.

[0017] On the other hand, there is also provided an electronic device, comprising:

[0018] a memory for non-transitory storage of computer-readable instructions; and

[0019] a processor for executing the computer readable instructions,

[0020] When the computer-readable instructions are executed by the processor, the method described in the first aspect is executed.

[0021] On the other hand, a storage medium is provided, which non-temporarily stores computer-readable instructions, wherein when the non-temporary computer-readable instructions are executed by a computer, the method described in the first aspect is executed.

[0022] On the other hand, a computer program product is provided, comprising a computer program, wherein the computer program is used to implement the method described in the first aspect when running on one or more processors.

[0023] The above technical solution has the following advantages or beneficial effects:

[0024] After obtaining the optimal fitting function through actual investigation, the optimal fitting function for vehicle range is Weibull fitting function; the optimal fitting function for the battery initial state of charge is beta fitting function; the optimal fitting function for the actual charging power is Weibull fitting function; the data obtained by sampling through the above fitting functions is more in line with travel behavior, and the prediction of highway electric vehicle charging load is more accurate, providing a theoretical basis for the optimization of charging pile configuration in highway service areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0026] Figure 1 is a flow chart of the method of embodiment 1;

[0027] Figure 2 It is the system structure diagram of ANFIS;

[0028] Figure 3(a) to Figure 3(f) It is the frequency distribution histogram of the survey data;

[0029] Figure 4(a) to Figure 4(b) is the traffic flow distribution map;

[0030] Figure 5(a) to Figure 5(c) It is the electric vehicle parameter fitting curve diagram;

[0031] Figure 6(a) to Figure 6(c) This is the distribution diagram of sampling results;

[0032] Figure 7 is the SOC membership function;

[0033] Figure 8 is the remaining mileage membership function;

[0034] Figure 9(a) to Figure 9(b) It is the error characteristic analysis diagram;

[0035] Fig.10 It is the fuzzy calculation result of charging probability;

[0036] Figure 11(a) to Figure 11(d) The spatiotemporal distribution of charging power for each service area;

[0037] Figure 12(a) to Figure 12(d) The spatiotemporal distribution map of charging demand in each service area. DETAILED DESCRIPTION

[0038] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0039] Embodiment 1

[0040] This embodiment provides a method for predicting charging load of electric vehicles on a highway;

[0041] like Figure 1 As shown, a method for predicting charging load of electric vehicles on a highway includes:

[0042] S101: construct the OD matrix of vehicles at the entrance and exit of the target highway; preliminarily calculate the total number M of electric vehicles traveling on the target highway; calculate the hourly flow of each OD pair based on the OD matrix; initialize the initial parameters of all electric vehicles;

[0043] S102: Based on the initial parameters of the mth electric vehicle, generate operating parameters for the mth electric vehicle; map the operating parameters of the mth electric vehicle to the highway network model; the initial value of m is 1;

[0044] S103: updating the travel time, state of charge, and remaining mileage of the mth electric vehicle in the highway network model;

[0045] S104: Estimate whether the mth electric vehicle in the highway network model needs to be charged in the service area. If it needs to be charged, calculate the charging time of the mth electric vehicle and return to S103; if it does not need to be charged, determine whether the remaining mileage is zero. If so, it means that the electric vehicle has arrived at the destination, record the end time and battery charge state, and enter S105; if the remaining mileage is greater than zero, return to S103;

[0046] S105: Determine whether m is less than M. If so, m=m+1 and return to S102; if not, calculate the total charging load of each highway service area and end.

[0047] Furthermore, the S101: constructing a target highway entrance and exit vehicle OD matrix, wherein the target highway entrance and exit vehicle OD matrix includes:

[0048] The rows of the OD matrix are the starting point numbers, and the columns of the OD matrix are the ending point numbers, where the elements of each cell of the OD matrix represent the traffic flow from the starting point to the ending point, OD (Origin-Destination).

[0049] Furthermore, the step S101: preliminarily calculating the total number M of electric vehicles traveling on the target highway includes:

[0050] Based on the penetration rate of electric vehicles on the target highway, the total number of electric vehicles M is calculated;

[0051] M = N × P;

[0052] Where: N is the total number of vehicles, P is the penetration rate of electric vehicles, and M is the total number of electric vehicles.

[0053] Furthermore, the S101: initializing the initial parameters of all electric vehicles includes:

[0054] Vehicle range, battery initial state of charge and actual charging power.

[0055] Exemplarily, the initial values ​​are initial_A=247.73 (vehicle range), initial_B=0.5 (battery initial charge), and initial_C=60 (charging power).

[0056] Further, the step S102: generating operating parameters for the mth electric vehicle based on the initial parameters of the mth electric vehicle, includes:

[0057] The Markov Monte Carlo algorithm is used to generate operating parameters for the mth electric vehicle based on the initial parameters of the mth electric vehicle.

[0058] Furthermore, the Markov Monte Carlo algorithm is used to generate operating parameters for the mth electric vehicle based on the initial parameters of the mth electric vehicle, including:

[0059] The optimal fitting function for vehicle range is the Weibull fitting function; the optimal fitting function for the battery initial state of charge is the beta fitting function; and the optimal fitting function for the actual charging power is the Weibull fitting function.

[0060] Furthermore, the operating parameters of the mth electric vehicle are mapped to a highway network model, wherein the highway network model is established according to the traffic planning software Transcad, and the highway network model data includes the highway section number, the highway section name, the number of lanes in the section, the section length, the lane capacity and the speed limit value.

[0061] Further, the S103: updating the travel time, state of charge and remaining mileage of the mth electric vehicle in the highway network model; wherein the state of charge is calculated by a formula, and the formula is as follows:

[0062]

[0063] Among them, SOC0 is the initial state of charge of the electric vehicle; d is the mileage of the electric vehicle; C is the maximum range of the electric vehicle, SOC r is the current state of charge of the electric vehicle; travel time and remaining mileage are real data in the OD matrix.

[0064] Furthermore, the step S104: estimating whether the m-th electric vehicle in the highway network model needs to be charged in the service area is based on an adaptive fuzzy neural network (ANFIS) to estimate whether the m-th electric vehicle in the highway network model needs to be charged in the service area, and specifically includes:

[0065] The battery SOC and remaining mileage of the m-th electric vehicle are input into the adaptive fuzzy neural network (ANFIS), and the adaptive fuzzy neural network outputs the charging probability of the m-th electric vehicle;

[0066] When the charging probability of the mth electric vehicle is greater than the set threshold, it means that the mth electric vehicle needs to be charged in the service area.

[0067] Further, in S104: if charging is required, the charging time of the mth electric vehicle is calculated, wherein the charging time t c :

[0068]

[0069] Where η is the charging efficiency; P n The charging power of the charging pile; E soc.full is the full-charge SOC value (the full-charge SOC threshold will be set to 0.9).

[0070] Furthermore, the calculation of the total charging load of each highway service area includes:

[0071]

[0072] P(t,s) represents the total charging load at time t and space point s; P i (t,s) represents the charging load of the i-th electric vehicle at time t and space point s; N is the total number of electric vehicles.

[0073] The present invention selects Highway A as the research object. According to the influencing factors of electric vehicle charging behavior, a questionnaire related to the influencing factors is formulated and distributed in the expressway service areas along Highway A, and then the collected data is analyzed to provide basic data for the later charging distribution prediction model.

[0074] The data used in this paper is based on the existing construction of Highway A and covers the basic data of Highway A. The data set contains multiple core variables, including lane capacity (Capacity), speed limit value (Fspeed), highway section number (ID), highway section name (Name), number of lanes in the section (Lanes), section length (Length, unit: kilometers) and travel time (Ftime, unit: minutes). These data provide necessary support for the research and lay a solid foundation for subsequent analysis.

[0075] The present invention adopts a link analysis method, selects highway A data, combines traffic distribution model and dynamic network model, deeply studies the traffic flow pattern in the network, analyzes the origin and destination (OD) distribution of vehicles on the road section, and provides scientific basis and data support for the accurate prediction of highway charging load.

[0076] The link analysis method is used to study the flow, speed, delay and other characteristics of each link (Link, i.e. road section) in the road network. It is often used in applications such as traffic status assessment, congestion analysis and optimization of signal control.

[0077] With the help of TransCAD software, the present invention constructs the highway network and toll station nodes of Highway A, and based on the OD matrix of Type I and Type II passenger cars of Highway A, adopts a balanced allocation method to realize the link analysis of key sections.

[0078] The balanced allocation method is a model for traffic flow allocation (Traffic Assignment), based on the principle of user equilibrium (UE) or system optimization (SO). Its core idea is:

[0079] User Equilibrium (UE): Each driver chooses the path with the shortest travel time, so that all used paths have the same travel cost (time) and no driver can unilaterally reduce travel time by changing the path.

[0080] System Optimization (SO): Traffic is distributed across the network to minimize total system travel time rather than individual shortest paths.

[0081] Among them, the user optimal model is:

[0082]

[0083] Where: A represents a set of directed arcs (road segments); r represents a starting node; s represents a terminal node; q rs represents the travel demand between the origin and destination points; ta represents the traffic impedance of road section a; K rs Represents the set of all paths between OD pairs; represents the traffic flow of path k in the OD pair; It represents the correlation between the segment and the path in the OD pair. If segment a is on path k, the value is 1.

[0084] System optimal model:

[0085]

[0086] In this process, traffic impedance, as an important indicator of traffic flow distribution, comprehensively considers factors such as travel time, distance, traffic signals, road conditions and traffic flow to reflect the "cost" of travelers in path selection, and conducts quantitative analysis through impedance function. Among them, the BPR function is used to study the relationship between traffic flow and speed, providing support for the analysis of highway networks.

[0087]

[0088] Where: t a represents the impedance on the road section a; t0 represents the free flow travel time; x a represents the traffic volume on road section a; c a It represents the number of vehicles that can actually pass through the road section per unit time, that is, the actual traffic capacity of the road section a; α and β represent the blocking coefficients, and the values ​​of α and β parameters are α=0.15 and β=4 respectively, which can also be obtained by regression analysis from actual data.

[0089] The present invention conducts statistical analysis on the vehicles entering and leaving each toll station, and further extracts the OD distribution matrix of Highway A, thereby revealing the traffic flow characteristics of the section. The statistical results show that the total number of vehicles passing through Highway A in a single day is 94,479. Among them, there are 219 toll stations with an entrance number of vehicles between 0 and 200, totaling 15,677 vehicles, accounting for 16%; there are 94 toll stations with an entrance number of more than 200 vehicles, totaling 78,802 vehicles, accounting for 84%. In terms of exit, there are 164 toll stations with an exit number of vehicles between 0 and 200, totaling 8,041 vehicles, accounting for 8.5%; there are 130 toll stations with an exit number of more than 200 vehicles, totaling 86,438 vehicles, accounting for 91.5%. The statistical results show that only about 20% of the toll stations carry about 80% of the traffic flow.

[0090] In order to deeply study the charging behavior and travel characteristics of electric vehicles in service areas, the present invention conducts a detailed investigation on the basic information of charging vehicles, focusing on collecting the electric vehicle model to determine the rated charging power and cruising range of the vehicle, the charging pile power (unit: kW), the power when entering the highway (i.e. the SOC value at the entrance of the highway toll station, unit: %), the power when charging (i.e. the remaining SOC value when entering the service area, unit: %), and the starting and ending points of the journey (OD information). At the same time, through a questionnaire survey in the charging station area of ​​the service area, the starting and ending point information of the driver and passengers is further collected to analyze the remaining mileage and driving characteristics of the electric vehicle. The comprehensive statistical and analytical results reveal the charging rules and travel characteristics of electric vehicles in highway service areas, providing a scientific basis for optimizing the layout of charging facilities and traffic planning.

[0091] Based on the survey data, the rated charging power and maximum range of electric vehicles are obtained by model. Figure 3(a) to Figure 3(f) The frequency distribution histogram of the rated charging power and driving range of electric vehicles in highway service areas was drawn. The analysis shows that the rated charging power of most electric vehicles is concentrated around 80kW, accounting for more than 2 / 3 of the total samples, and the rated power of some high-end models is as high as 300kW. The average rated charging power of vehicles in service areas is 102.7kW, which is significantly higher than the market average of 40kW, reflecting that most of the charging vehicles are mid-to-high-end models with higher charging power requirements.

[0092] The analysis of the cruising range shows that the cruising range of vehicles in the service area is mainly concentrated between 200-600km, and a few models exceed 800km, indicating that the service area vehicles are mainly used to meet the needs of long-distance travel, which is significantly different from short-distance vehicles in the city. From the distribution of SOC values, about 84% of the vehicles have an SOC value higher than 60% when they are on the highway, and the average starting SOC value is 83.29%, reflecting the characteristics of long-distance travel. 73% of the vehicles in the service area are charged when the SOC value is lower than 50%, and the average charging SOC value is 31.83%, indicating that the charging behavior is mostly based on the remaining power and travel requirements. The actual charging power is significantly related to the rated power of the vehicle. The average power of the charging pile is 64.62kW, accounting for 62% of the average rated power of the vehicle, but it is limited by the resource allocation state. For example, the power is significantly reduced when multiple vehicles share the charging pile. The remaining mileage is evenly distributed, with an average of 247.73km, which is mainly affected by travel behavior and individual charging strategies, and has a high randomness. The present invention provides an important reference for optimizing the layout of charging facilities in highway service areas and improving charging efficiency.

[0093] Since the battery SOC and remaining mileage affect the user's charging behavior, but these two factors cannot be calculated using a fixed mathematical expression, the present invention adopts ANFIS (Adaptive Neuro-Fuzzy Inference System) to fuzzify the data of the battery SOC and the remaining mileage as input parameters, adopts membership functions and expert experience fuzzy rules to train the fuzzy rules, normalizes the charging probability, and obtains the user's charging probability.

[0094] For a system with two inputs, battery SOCx1 and remaining mileage x2, and one output, charging probability y, the corresponding TS fuzzy neural network is first-order and has an if-then rule set that describes the characteristics of complex objects. Its structure is as follows: Figure 2 shown.

[0095] The fuzzy rules for the input variables in the figure are:

[0096] f x1 is A1 and x2 is B2, then y1=p1x1+q1x2+r1 (1)

[0097] if x1 is A2 and x2 is B2,theny2=p2x1+q2x2+r2 (2)

[0098] Among them, x1 is the battery charge state, x2 is the remaining mileage, and y i is the charging probability, A i , B i is the fuzzy set of input variables, p i ,q i 、r i is the conclusion parameter.

[0099] The nodes in each layer of the ANFIS structure represent the same series of functions, as described below:

[0100] The first layer is the fuzzification layer, which uses the membership function to fuzzify the model input. Determine the membership function u of x1 and x2 Ai 、u Bi , the first layer output is shown in formula (3) (4):

[0101]

[0102] Among them, x1 and x2 are the inputs of node i. i , O j is the value of the fuzzy set membership function. The model adopts the Gaussian membership function. The Gaussian function has good smoothness in the model, has no zero point and has clear physical meaning. The expression of the Gaussian membership function is shown in formula (5):

[0103]

[0104] Among them, a i , b i and c i is the Gaussian function membership parameter, b i The fixed value is 2, c i 、a i are the center value and width of the Gaussian function respectively.

[0105] The second layer is the incentive intensity layer. This layer calculates the incentive intensity of all fuzzy rules. The node calculates the membership function value of the input quantity corresponding to the rule and outputs the result by taking the form of small, algebraic product, bounded product, etc., as shown in formula (6):

[0106]

[0107] The third layer is the normalization layer. This layer calculates the percentage of each fuzzy rule's incentive strength to all rule incentive strengths. Its output represents the weight of the rule in all fuzzy rules, as shown in formula (7):

[0108]

[0109] The fourth layer is the function group layer, each node in this layer corresponds to a fuzzy rule f i , the function is the output of the fuzzy rule when the membership is determined. It is generally a linear function, and the number of function parameters depends on the number of input quantities. The output of the i-th node of this layer is shown in formula (8):

[0110]

[0111] in, represents the normalized excitation intensity of the third layer, p i ,q i 、r i is the conclusion parameter. This type of function is generally a linear function, and the number of conclusion parameters depends on the number of input quantities. In this example, there are only two input quantities x1, x2, so the parameters in the function group layer only contain p i ,q i and r i .

[0112] The fifth layer is the output layer, which is the result of each fuzzy rule f i The output sum of this layer is a fixed node, and the output is shown in formula (9):

[0113]

[0114] The present invention samples the parameters of electric vehicles through MCMC, and combines ANFIS to perform fuzzy modeling on battery SOC and remaining mileage, obtains the user charging probability through normalization, and finally obtains the overall charging load distribution of electric vehicles.

[0115] The classical Monte Carlo method is used to approximate the integral by obtaining a large number of samples from the original distribution according to the following formula.

[0116]

[0117] On this basis, the Markov Chain Monte Carlo method is a Monte Carlo method combined with the Markov chain, which can achieve the effect that the sampling distribution changes with the simulation process. It is especially suitable for scenarios with complex multidimensional distributions or likelihood functions. The so-called Markov chain is a sequence composed of different states of random variables.

[0118] The input values ​​of the Markov Chain Monte Carlo method are vehicle range, battery starting charge, and charging power;

[0119] initial_A = 247.73 (vehicle range), initial_B = 0.5 (battery initial charge), initial_C = 60 (charging power);

[0120] Distribution parameters: weibull_params_A = (2.768, 381.795) (Weibull distribution parameters for A);

[0121] beta_params_B = (1.2328, 0.5666) (Beta distribution parameters, for B);

[0122] weibull_params_C = (1.2546, 126.656) (Weibull distribution parameters for C);

[0123] Algorithm output value:

[0124] A_samples: vehicle range samples sampled from Weibull distribution;

[0125] B_samples: battery initial capacity samples sampled from Beta distribution;

[0126] C_samples: Charging power samples sampled from Weibull distribution;

[0127] For any time t, the conditional probability of the system state is:

[0128] p(X t+1 ∣X t,...,X1,X0)=p(X t+1 ∣X t )

[0129] Starting from the initial state X0, according to the one-step transition probability p(X t+1 ∣X t ) can continuously obtain new states, thus forming a Markov chain. If the transition probability is independent of time, the Markov chain is called time homogeneous. MCMC method usually uses time homogeneous Markov chain.

[0130] Markov chain has no memory, and the next state is only related to its previous state. Obviously, the state of the random variable at the beginning of sequence generation is closely related to the initial state, and when the number of transfers is large, the state of the random variable will be independent of the initial state. At this time, the Markov chain is said to have reached a stable state or converged. The core of the MCMC method is to construct a Markov chain whose stable distribution is the distribution to be sampled (called the target distribution). Assuming that M transfers have been made when the stable state is reached, and the total number of transfers is N (N>M and N is large), the expectation of h(x) can be estimated by the following formula:

[0131]

[0132] The MCMC model includes two sampling models, among which the Metropolis-Hastings algorithm formula and steps are as follows:

[0133] 1. Input any given Markov chain state transfer matrix Q, target stable distribution π(x), set the state transition number threshold n1, and the required number of samples n2;

[0134] 2. Get the initial state value x0 from any simple probability distribution;

[0135] 3.for t=0in n1+n2-1

[0136] (1) From the conditional probability distribution Q(x|x t ) to get the sample value x * ;

[0137] (2) Sample U~[0,1] from a uniform distribution;

[0138] (3) If Then accept x t →x * , that is, x t+1 =x * ;

[0139] (4) Otherwise, the transfer is not accepted, t = max{t-1,0}.

[0140] Often the Markov transition state matrix is ​​symmetric, so the acceptance rate can be written as:

[0141]

[0142] For the distribution p(x), we construct the transfer matrix P to satisfy the detailed stationary condition:

[0143] π(x)P(x→y)=π(y)P(y→x)

[0144] Here x is not required to be one-dimensional. For p(x) in high-dimensional space, if the detailed stability condition is satisfied

[0145] π(x)P(x→y)=π(y)P(y→x)

[0146] Then the above Metropolis-Hastings algorithm is equally effective.

[0147] Metropolis-Hastings algorithm, algorithm input values: initial_A = 247.73 (vehicle range), initial_B = 0.5 (battery initial charge), initial_C = 60 (charging power);

[0148] Metropolis-Hastings algorithm, distribution parameters: weibull_params_A = (2.768, 381.795) (Weibull distribution parameters, for A); beta_params_B = (1.2328, 0.5666) (Beta distribution parameters, for B); weibull_params_C = (1.2546, 126.656) (Weibull distribution parameters, for C);

[0149] Metropolis-Hastings algorithm, algorithm output values: A_samples: vehicle range samples sampled from Weibull distribution; B_samples: battery starting power samples sampled from Beta distribution; C_samples: charging power samples sampled from Weibull distribution.

[0150] The present invention preliminarily analyzes the ETC vehicle passing data of a certain high-speed gantry, and obtains that the penetration rate of electric vehicles in a certain year is 7%. It is further assumed that the penetration rate of electric vehicles in the target planning year is 20%. The OD daily traffic volume of the highway network toll station is obtained using the system optimal model. In order to infer the hourly traffic demand between each OD pair from the OD daily traffic volume, it is necessary to allocate it according to the typical daily traffic flow characteristics of the highway. Figure 4 (a) reflects the traffic flow distribution law in each period of the day; Figure 4 (b) is a diagram of the traffic flow of each service area of ​​Highway A over time, which shows that when the road is unobstructed, the traffic flow distribution of the highway presents a time delay, and the traffic flow reaches a peak at 12 noon, followed by a brief decline, which may be due to the fact that some vehicles stay in the service area or the driver rests. After a brief fluctuation, the traffic flow rebounded again and continued until 6 pm. The traffic flow fluctuation during this period reflects the typical peak characteristics.

[0151] In order to obtain the parameters of actual electric vehicles, the present invention obtains information such as the cruising range, power consumption when on the highway, and charging time of electric vehicles in the charging station area of ​​the service area through field surveys of highway service areas along Highway A. Taking electric vehicles as the research object, a highway service area electric vehicle model is established based on the field survey data. First, the present invention uses the Fitter library in Python to fit the probability distribution model of cruising range, initial state of charge, and actual charging power, and outputs the top five optimal distributions, and uses the evaluation indicators Sum of Squares due to Error (SSE), Akaike Information Criterion (AIC), KS statistic, and KS-P value to judge the probability fitting distribution curve, so as to select the optimal fitting function, and based on the optimal fitting function, determine the sampling model for the vehicle parameters in the later stage, and then sample the vehicle parameters according to the sampling model. The optimal fitting function is shown in the figure below. Figure 5(a) to Figure 5(c) shown.

[0152] Table 1 Fitting evaluation index table

[0153]

[0154] As shown in Table 1, the SSE of the five distributions of the cruising range probability is slightly different, and the AIC shows that the model complexity and goodness of fit are not much different; the KS test results show that only the KS-P value of the Weibull distribution is greater than 0.05, indicating that the data fits the distribution well. Therefore, it is assumed that the cruising range meets X1~Wb(2.768,381.795).

[0155] The SSE of the five distributions of the battery's initial SOC is slightly different. The AIC index shows that except for the genhyperbolic distribution, the other four distributions have a good balance between goodness of fit and complexity. The KS test results show that only the KS-P values ​​of the Beta, Pearson3 and Genpareto distributions are greater than 0.05, among which the Beta distribution has the smallest SSE. Therefore, assuming that the initial state of charge meets

[0156] The actual charging power of electric vehicles is different from the rated power. As shown in Table X, except for the Dgamma distribution, the KS-P values ​​of the other four distributions are all greater than 0.05, and the fit is good; among them, the SSE and AIC of the Weibull distribution are relatively small. Therefore, it is assumed that the actual charging power meets X3~Wb(1.254,126.656).

[0157] Furthermore, the optimal fitting function for the cruising range is the Weibull fitting function; the optimal fitting function for the initial state of charge is the beta fitting function; and the optimal fitting function for the actual charging power is the Weibull fitting function.

[0158] Furthermore, the charging time t c Indicates the duration of the charging state. In different charging demand scenarios, the charging duration can be expressed as:

[0159]

[0160] Where η is the charging efficiency; P n The charging power of the charging pile; E soc.full is the full-charge SOC value (the full-charge SOC threshold will be set to 0.9).

[0161] The vehicle's cruising range, battery starting SOC, and actual charging power need to be sampled and determined before the vehicle is driven. After obtaining the function through the above fitting probability function, the sampling results of the three parameters of the vehicle are as follows: Figure 6(a) to Figure 6(c) shown.

[0162] The dataset used in the present invention contains two features, where the target variable is the charging probability, and the input features include the battery SOC value and the remaining mileage. The dataset was preprocessed before modeling, and the specific operations included filling or deleting missing values, and all features were normalized to the interval range of [0,1].

[0163] In the fuzzy processing process, the fuzzy subsets of battery SOC include: SN (very low), N (low), MN (normal), F (high), FL (very high), these subsets represent different levels of battery power. Similarly, the remaining mileage is also represented by the fuzzy subsets SN, N, MN, F, FL, reflecting the degree of remaining mileage. For the membership functions of these fuzzy subsets, Gaussian membership function (GaussMF) and binary Gaussian membership function (Gauss2MF) are used to represent them, as shown in Figure 7 and Figure 8 As shown in the figure, SOC uses the Gaussian membership function, and the cruising range uses the binary Gaussian membership function. The fuzzy subsets include: SN (very low), N (low), MN (normal), F (high), and FL (very high), see formulas (1)-(5).

[0164] The data set is divided into 70% as a training set and 30% as a test set. The stopping error of the training process is set to 0.05 and the maximum number of iterations is 100. Figure 9(a) to Figure 9(b) The error local weighted regression diagram and error distribution diagram in the figure show the relationship between the model prediction error and the actual value and the probability distribution of the error, respectively. The local weighted regression line is mainly concentrated near the residual value of 0, indicating that there is no obvious systematic deviation between the residual and the predicted value, the model prediction does not show systematic overestimation or underestimation, and the residual distribution is uniform and has no significant trend. The kernel density estimation (KDE) curve in the error distribution diagram is close to the normal distribution, indicating that the prediction error is random and meets the normal distribution assumption, further verifying that the model error is disordered and unbiased.

[0165] The model performance was measured by the mean absolute percentage error (MAPE) and the coefficient of determination (R 2 ) was used for evaluation. The results showed that the MAPE of the model was 15.4%, indicating that the model prediction error was about 15.4% of the actual value, and the prediction effect was relatively ideal; R 2 The value is 0.702, which means that the model has a strong ability to explain the data and can explain about 70.2% of the data variation, indicating that it has a high degree of fit and can effectively capture the main trends and laws in the data.

[0166] The solution results of the charging probability model constructed based on the ANFIS method are as follows: Fig.10 As shown. The results show that the probability of charging gradually increases as the SOC value decreases, and when the SOC value drops to 0.4, the driver shows a clear demand for charging. This trend is consistent with expectations, indicating that when the battery power drops to a certain level, the driver is more inclined to charge. However, the results also show that a high remaining mileage does not significantly promote the probability of charging. This result is different from our initial hypothesis. The possible reason is that there are more service areas along the analyzed highway section, and drivers tend to choose to charge when the SOC value is low.

[0167] The charging demand data of different service areas in different time periods of the day are calculated based on the simulation, and finally the spatiotemporal distribution diagram of the charging demand and charging load of different service areas is obtained. Figure 11(a) to Figure 11(d) The spatiotemporal distribution of charging power for each service area; Figure 12(a) to Figure 12(d) The spatiotemporal distribution map of charging demand in each service area.

[0168] The charging demand of each service area varies in different time periods. Specifically, the charging demand peaks at 80 vehicles at 12:00 noon, 16:00 pm and 19:00 pm, respectively, with a total of 1,198 vehicles in the whole period; the charging demand of service area B reaches a peak of 70 vehicles at 16:00 pm, with a total of 934 vehicles in the whole period; the charging demand of service area C reaches a peak of 42 vehicles at 14:00 pm and 20:00 pm, respectively, with a total of 562 vehicles in the whole period; the charging demand peaks at 80 vehicles at 13:00 pm and 17:00 pm in service area D, similar to service area A, with a total of 1,068 vehicles in the whole period; since the distance between each service area is short, about 50 kilometers, most vehicles have completed charging when they arrive at service area A, so the charging demand of subsequent service areas B and C in the same time period is relatively small; and as the last service area, vehicles in service area D usually have low power when they arrive, especially vehicles with high mileage, which are more inclined to charge here. Further analysis of the total charging power of each service area in different time periods shows that the spatiotemporal trend demand is mainly concentrated from 10 am to 8 pm. The peak loads of the four service areas are 3000kW, 2500kW, 1700kW and 3000kW respectively.

[0169] Embodiment 2

[0170] This embodiment provides a highway electric vehicle charging load prediction system, including:

[0171] A construction module is configured to: construct an OD matrix of vehicles at the entrance and exit of the target highway; preliminarily calculate the total number M of electric vehicles traveling on the target highway; calculate the hourly flow of each OD pair based on the OD matrix; initialize the initial parameters of all electric vehicles;

[0172] A generating module is configured to: generate operating parameters for the mth electric vehicle based on the initial parameters of the mth electric vehicle; map the operating parameters of the mth electric vehicle to the highway network model; the initial value of m is 1;

[0173] An updating module is configured to: update the travel time, state of charge and remaining mileage of the mth electric vehicle in the highway network model;

[0174] The estimation module is configured to: estimate whether the mth electric vehicle in the highway network model needs to be charged in the service area; if charging is required, calculate the charging time of the mth electric vehicle and return to the update module; if charging is not required, determine whether the remaining mileage is zero; if so, it means that the electric vehicle has arrived at the destination, record the end time and battery charge state, and enter the judgment module; if the remaining mileage is greater than zero, return to the update module;

[0175] The judgment module is configured to: judge whether m is less than M, if so, m=m+1, and return to the generation module; if not, calculate the total charging load of each highway service area, and end.

[0176] It should be noted here that the above-mentioned construction module, generation module, update module, estimation module and judgment module correspond to steps S101 to S105 in the first embodiment, and the examples and application scenarios implemented by the above-mentioned modules and the corresponding steps are the same, but are not limited to the contents disclosed in the first embodiment. It should be noted that the above-mentioned modules as part of the system can be executed in a computer system such as a set of computer executable instructions. The description of each embodiment in the above-mentioned embodiments has different emphases. For the parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0177] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0178] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting charging load of electric vehicles on highways, characterized in that: include: (1) Construct the vehicle OD matrix at the entrance and exit of the target highway; preliminarily calculate the total number M of electric vehicles on the target highway; Calculate the hourly flow of each OD pair based on the OD matrix; initialize the initial parameters of all electric vehicles; (2) Based on the initial parameters of the mth electric vehicle, generate operating parameters for the mth electric vehicle; map the operating parameters of the mth electric vehicle to the highway network model; the initial value of m is 1; (3) Update the travel time, state of charge, and remaining mileage of the mth electric vehicle in the highway network model; (4) Estimate whether the mth electric vehicle in the highway network model needs to be charged in the service area. If it needs to be charged, calculate the charging time of the mth electric vehicle and return to (3); if it does not need to be charged, determine whether the remaining mileage is zero. If so, it means that the electric vehicle has arrived at the destination, record the end time and battery charge state, and enter (5); if the remaining mileage is greater than zero, return to (3); (5) Determine whether m is less than M. If so, m=m+1 and return to (2). If not, calculate the total charging load of each highway service area and end.

2. A method for predicting charging load of electric vehicles on highways as claimed in claim 1, characterized in that: Calculate the total number M of electric vehicles on the target highway; including: Based on the penetration rate of electric vehicles on the target highway, the total number of electric vehicles M is calculated; M = N × P; Where: N is the total number of vehicles, P is the penetration rate of electric vehicles, and M is the total number of electric vehicles.

3. A method for predicting charging load of electric vehicles on highways as claimed in claim 1, characterized in that: Based on the initial parameters of the mth electric vehicle, operating parameters are generated for the mth electric vehicle, including: The Markov Monte Carlo algorithm is used to generate operating parameters for the mth electric vehicle based on the initial parameters of the mth electric vehicle.

4. A method for predicting charging load of electric vehicles on a highway as claimed in claim 3, characterized in that: The Markov Monte Carlo algorithm is used to generate operating parameters for the mth electric vehicle based on the initial parameters of the mth electric vehicle, including: The optimal fitting function for vehicle range is the Weibull fitting function; the optimal fitting function for the battery initial state of charge is the beta fitting function; and the optimal fitting function for the actual charging power is the Weibull fitting function.

5. A method for predicting charging load of electric vehicles on a highway as claimed in claim 1, characterized in that: Update the travel time, state of charge, and remaining mileage of the mth electric vehicle in the highway network model; the state of charge is calculated by the formula, which is as follows: Among them, SOC0 is the initial state of charge of the electric vehicle; d is the mileage of the electric vehicle; C is the maximum range of the electric vehicle, SOC r is the current state of charge of the electric vehicle; travel time and remaining mileage are real data in the OD matrix.

6. A method for predicting charging load of electric vehicles on a highway as claimed in claim 1, characterized in that: The estimation of whether the mth electric vehicle in the highway network model needs to be charged in the service area is based on the adaptive fuzzy neural network ANFIS. Specifically, it includes: The battery SOC and remaining mileage of the m-th electric vehicle are input into the adaptive fuzzy neural network ANFIS, and the adaptive fuzzy neural network outputs the charging probability of the m-th electric vehicle; When the charging probability of the mth electric vehicle is greater than the set threshold, it means that the mth electric vehicle needs to be charged in the service area.

7. A method for predicting charging load of electric vehicles on a highway as claimed in claim 1, characterized in that: If charging is required, calculate the charging time of the mth electric vehicle, where the charging time t c : Where η is the charging efficiency; P n The charging power of the charging pile; E soc.full is the full-charge SOC value; The calculation of the total charging load of each highway service area includes: P(t,s) represents the total charging load at time t and space point s; P i (t,s) represents the charging load of the i-th electric vehicle at time t and space point s; N is the total number of electric vehicles.

8. A highway electric vehicle charging load prediction system, characterized in that it includes: A construction module is configured to: construct an OD matrix of vehicles entering and exiting a target expressway; preliminarily calculate the total number M of electric vehicles traveling on the target expressway; Calculate the hourly flow of each OD pair based on the OD matrix; initialize the initial parameters of all electric vehicles; A generating module is configured to: generate operating parameters for the mth electric vehicle based on the initial parameters of the mth electric vehicle; map the operating parameters of the mth electric vehicle to the highway network model; the initial value of m is 1; An updating module is configured to: update the travel time, state of charge and remaining mileage of the mth electric vehicle in the highway network model; An estimation module is configured to: estimate whether the mth electric vehicle in the highway network model needs to be charged in the service area, and if charging is required, calculate the charging time of the mth electric vehicle and return it to the update module; If charging is not required, it is determined whether the remaining mileage is zero. If so, it means that the electric vehicle has reached the destination, and the end time and battery charge state are recorded, and the judgment module is entered; if the remaining mileage is greater than zero, it returns to the update module; The judgment module is configured to: judge whether m is less than M, if so, m=m+1, and return to the generation module; if not, calculate the total charging load of each highway service area, and end.

9. An electronic device, comprising: a memory for non-transitory storage of computer readable instructions; as well as a processor for executing the computer readable instructions, When the computer-readable instructions are executed by the processor, the method described in any one of claims 1 to 7 is executed.

10. A storage medium, characterized in that: The computer-readable instructions are non-transitory stored, wherein when the non-transitory computer-readable instructions are executed by a computer, the method according to any one of claims 1 to 7 is performed.

Images