Electric vehicle charging station load forecasting method considering multi-source information and decision correction

Abstract

The application provides a method for electric vehicle charging station load prediction considering multi-source information and decision correction, comprising the following steps: step one, simulating the multi-source information that can be collected by users in the region; step two, when the user encounters low power and makes a decision to go to the charging station, simulating through the prospect theory; step three, during the charging process of the user, the decision is corrected through the prospect theory according to the real-time information change, so that the load distribution of the charging station in the region is more uniform and conforms to the prediction of the actual demand of the user for the vehicle; the application proposes a plurality of traffic characteristic values influencing the decision of the electric vehicle owner according to the multi-source information collected in the road network, constructs a decision model of the prospect theory to simulate the decision of the vehicle owner, and corrects the decision according to the real-time information change at special nodes, so that the application scene of the electric vehicle load prediction is more fitted to the actual vehicle use of the vehicle owner.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric vehicle charging station load prediction, and particularly to an electric vehicle charging station load prediction method considering multi-source information and decision correction. BACKGROUND

[0002] With the gradual cultivation and perfection of new energy automobile supporting industry chain, electric vehicles are increasingly recognized in the fields of private cars, urban transportation, logistics, etc. The improvement of the comprehensive endurance mileage of electric vehicles improves the practicability of high-power direct current charging piles, and high-power charging piles of 60kW, 120kW and even 480kW are gradually put into use. Short charging time and small power loss make direct current charging piles the mainstream in the current expansion and layout of charging piles.

[0003] However, the charging load of electric vehicles has certain volatility and randomness. When the power grid line accesses the electric vehicle charging station, the high-power charging load will increase the line load rate and load loss rate, accompanied by a decrease in no-load loss rate. At the same time, when the penetration rate of electric vehicles in the same area is too large, or when too many vehicles are charging at the same time, the voltage at the end node will drop sharply, and in severe cases, even the lower limit will occur. Therefore, how to predict the electric vehicle load to set measures to deal with it is an important research direction.

[0004] At present, the scene of electric vehicle load prediction can be divided into two categories according to whether there is historical data: one category is the prediction under the condition that the charging station has historical data, which is commonly used in the scene of charging station built for future weather, day type, etc. Based on historical data, future data is predicted; the other category is when there is no historical data for verification and model training, through analyzing user characteristics, electric vehicle penetration rate, etc. Factors are used for charging station location selection, charging pile scale planning, etc. in the region.

[0005] At the same time, with the development of Internet of Vehicles technology, electric vehicle owners can more intuitively obtain various information before and during driving through mobile phone app and other ways. These information will have a great influence on the specific decision of electric vehicle owners to choose charging sites, thereby affecting the distribution of charging load to some extent.

[0006] Based on the above background, the present application proposes an electric vehicle charging station load prediction method considering multi-source information and decision correction. SUMMARY

[0007] The application provides a method for predicting load of an electric vehicle charging station, which takes into account multi-source information and decision correction.

[0008] The application adopts the following technical solutions.

[0009] The method for predicting load of the electric vehicle charging station taking into account multi-source information and decision correction comprises the following steps.

[0010] Step one, first simulate the multi-source information that can be collected by users in the region.

[0011] Step two, when the user encounters low power and makes a decision to go to the charging station, simulate it through the prospect theory.

[0012] Step three, during the charging process of the user, the decision is corrected through the prospect theory according to the real-time information change, so that the load distribution of the charging station in the region is more uniform and meets the actual demand of the user.

[0013] Step one is specifically divided into the following steps.

[0014] Step S1: divide the user's travel model into private cars and taxis, and generate a travel chain through an OD matrix and a travel chain model.

[0015] Step S2: collect the multi-source information available to users in the road network, such as road conditions, weather, and charging station queuing conditions.

[0016] Step S3: count the traffic volume of each road in the road network, generate the speed of the road network according to the Greenshield traffic flow fitting model, so as to obtain the driving speed of each road vehicle, and further obtain the driving distance and energy consumption.

[0017] Step S4: determine the opening state of the vehicle air conditioner according to the predicted air temperature, determine whether it is cold wind, warm wind or off, so as to calculate the energy consumption of the vehicle air conditioner; deduct the power consumption at each moment from the vehicle power; steps two and three are specifically divided into the following steps.

[0018] Step S5: for a taxi, when the vehicle power is lower than the expected power of the owner, evaluate the charging of each charging station in the road network that can be accessed under the current scenario using the prospect theory, obtain the optimal charging station, and make a decision to go to the charging station for charging.

[0019] Step S6: Whenever the low battery vehicle reaches a node or intersection in the road network on the way to the charging station, jump to step S5 for reevaluation of the charging station selection; if the optimal charging station has changed, modify the charging station destination in the charging decision; if the result remains unchanged, continue to the charging station;

[0020] Step S7: After the vehicle arrives at the charging station, it enters the queuing state. After queuing and entering the charging state, the vehicle's charging amount is estimated using different estimated charging powers according to the range of the vehicle's remaining power.

[0021] Step S8: The charging amount of each vehicle at each charging station at each time is accumulated throughout the day to obtain the charging load of each charging station in a single day.

[0022] The step S5 specifically includes the following steps:

[0023] Step S51: Considering the factors of electricity price, time required to go to the charging station, and charging station queuing situation, three corresponding traffic characteristic values, namely, payment cost, time cost, and expected income, are calculated.

[0024] Step S52: The prospect value corresponding to each charging station is calculated.

[0025] The step S51 specifically includes the following steps:

[0026] Step S511: Calculate the payment cost, which is the electricity fee required for charging after the estimated queuing time. The calculation method is: Where W charging (t) is the charging amount of the vehicle at time t; E price (t) is the time-of-use electricity price of the charging station at time t; t EVCS is the time when the vehicle starts charging after the estimated queuing time; t charging is the estimated charging duration.

[0027] Step S512: Calculate the time cost, which is the income loss converted after considering the waiting charging duration calculated by the vehicle owner according to the SOC state corresponding to each charging gun and the charging station queuing situation. The calculation method is: x2=(t queue +t charging )·v ave ·εEquation Two,

[0028] Where t queue is the queuing time; v ave is the average speed of all roads in the road network fitted by step S3; and ε is the estimated profit per kilometer.

[0029] Step S513: Expected income, taking into account the difficulty of subsequent passenger pickup and order acceptance by the operating electric vehicle owner, and the estimated order destination. The calculation method is: Where num dest The number of areas covered by the charging station; OD dest t represents the probability of traveling to the corresponding destination in the OD matrix; wait This represents the time required to stay in the current area.

[0030] Step S52 specifically includes the following steps:

[0031] Step S521: Define the original decision matrix, the formula is as follows

[0032]

[0033] Where, x t,ij This refers to the j-th traffic characteristic value of the vehicle traveling to the i-th charging station at time t. This represents the various factors that the vehicle owner needs to consider and converts them into profit and loss values, which has been calculated in step S51.

[0034] Step S522: Iterate through all the foreground values ​​obtained from all charging stations, arrange them from beneficial to detrimental, and name them x. c1 ,x c2 ,…,x c,TCV_Valid The standard value of traffic characteristics acceptable to car owners is calculated using the following formula:

[0035]

[0036] Some charging stations may be inaccessible due to insufficient range of electric vehicles, TCV Valid s1 represents the number of accessible charging stations; s2 and s1 represent the weighted intervals, calculated using the formula s1 = floor(TCV). Valid / 3)-1、

[0037] s2=TCV Valid -[floor(TCV Valid Formula 6;

[0038] floor is rounded down to the nearest integer.

[0039] Step S523: Calculate the prospective value function of vehicle k at time t for selecting charging station i based on the j-th traffic characteristic value. The formula is as follows:

[0040]

[0041] Where α and β are risk sensitivity coefficients, respectively, λ is the loss aversion coefficient, and d kij The difference between the traffic characteristic value j for vehicle k choosing charging station i for charging and the vehicle owner's internal evaluation standard, which is the S calculated in step S522;

[0042] Step S524: determining the proportion of each traffic characteristic value affecting the decision of the vehicle owner by analytic hierarchy process;

[0043] Step S525: calculating the weight function of the vehicle owner facing the benefits and losses, the formula being

[0044]

[0045] wherein γ is the risk benefit attitude coefficient, δ is the risk loss attitude coefficient, ω tj is the weight of the jth traffic characteristic value;

[0046] Step S526: calculating the prospect value of the kth vehicle selecting the ith charging station at the tth time, the formula being:

[0047]

[0048] The step S524 specifically comprises the following steps:

[0049] Step S5241: listing the importance correlation degree of each two traffic characteristic values in step S51, listing the matrix A, wherein 1 represents that two traffic characteristic values are equally important, 2 represents that one traffic characteristic value is obviously more important than the other, and 3 represents that one traffic characteristic value is much more important than the other; the reciprocal of each number is the opposite meaning;

[0050] Step S5242: setting j as the number of traffic characteristic values, calculating the value W of the product of each row element in the matrix A multiplied by j;

[0051] Step S5243: normalizing W to obtain W * ;

[0052] Step S5244: solving the maximum eigenvalue λmax, wherein W * i is the ith row of the matrix W * ;

[0053] Step S5245: performing consistency check, calculating Particularly, when j=3, RI is taken as 0.58 here; if CR>1, it does not meet the requirements, indicating that the importance correlation degree needs to be reconsidered.

[0054] The present application relates to a kind of electric vehicle charging station load forecasting method considering multi-source information and decision correction, since when in electric vehicle load forecasting scene, any change of reality will influence the decision of user when simulating user decision, therefore the present application is directed to the problems existing in traditional electric vehicle charging station load forecasting, the improvement of user decision simulation process is carried out.Firstly, the multi-source information that user can collect in region is simulated, then when user encounters low power and makes the decision to go to charging station, it is simulated by prospect theory, and simultaneously, decision correction is carried out by prospect theory according to real-time information change in the way to charging station.In the case of decision correction, it can achieve more uniform load distribution of charging station in region, and more in line with the prediction of actual vehicle demand of user.

[0055] Compared with prior art, the present application has the following two outstanding advantages.

[0056] 1, the present application can carry out load prediction to electric vehicle charging station by simulating electric vehicle user behavior, and the best charging pile / car ratio can be estimated.

[0057] 2, the present application can make more accurate decision and decision correction by simulating electric vehicle user to obtain real-time information and make comprehensive consideration, and more accurate electric vehicle charging station load forecasting effect is obtained. DETAILED DESCRIPTION

[0058] The present application is further described in detail below in combination with the drawings and specific embodiments:

[0059] Figure 1 is a flow chart of private car behavior; Figure 1 Figure 2 is a flow chart of taxi behavior and electric vehicle charging station load forecasting.

[0060] Figure 1 is a flow chart of private car behavior; Figure 2 Figure 2 is a flow chart of taxi behavior and electric vehicle charging station load forecasting. DETAILED DESCRIPTION

[0061] As shown in the figure, the electric vehicle charging station load forecasting method considering multi-source information and decision correction includes the following steps:

[0062] Step one, firstly, the multi-source information that user can collect in region is simulated;

[0063] Step two, when user encounters low power and makes the decision to go to charging station, it is simulated by prospect theory;

[0064] Step three, in the process of user charging, decision correction is carried out by prospect theory according to real-time information change, through the correction of decision, the load distribution of charging station in region is more uniform, and the prediction of actual vehicle demand of user is met.

[0065] Step one is specifically divided into the following steps:

[0066] Step S1: divide the user's travel model into private cars and taxis, generate a travel chain by a travel chain model and an OD matrix;

[0067] Step S2: collect multi-source information available to users in the road network, such as road conditions, weather, and charging station queue conditions;

[0068] Step S3: count the traffic volume of each road in the road network, generate the speed of the road network according to the Greenshield traffic flow fitting model, thereby obtaining the driving speed of each road vehicle, and further obtaining the driving distance and energy consumption;

[0069] Step S4: determine the opening state of the vehicle air conditioner according to the predicted air temperature, determine whether it is cold wind, warm wind or closed, thereby calculating the energy consumption of the vehicle air conditioner; deduct the power consumption at each time from the vehicle power; Step two and step three are specifically divided into the following steps;

[0070] Step S5: for a taxi, when the vehicle power is lower than the expected power of the owner, use the prospect theory to evaluate the charging of each charging station in the road network that can be reached under the current scenario, obtain the optimal charging station, and make a decision to go to the charging station for charging;

[0071] Step S6: whenever a vehicle with low power arrives at a node or intersection in the road network during the journey to the charging station, jump to step S5 for reevaluation of charging station selection; if the optimal charging station changes, modify the charging station destination to be reached in the charging decision; if the result does not change, continue to go to the charging station;

[0072] Step S7: after the vehicle arrives at the charging station, it enters the queuing state, and after queuing and entering the charging state, the charging amount of the vehicle is estimated using different estimated charging powers according to the range of the power stored in the vehicle;

[0073] Step S8: accumulate the charging amount of each vehicle at each charging station at each time to obtain the charging load of each charging station in a single day.

[0074] The step S5 specifically includes the following steps:

[0075] Step S51: consider the three factors of electricity price, time required to go to the charging station, and charging station queue condition, and calculate three corresponding traffic characteristic values, namely, payment cost, time cost, and expected income;

[0076] Step S52: calculate the prospect value corresponding to each charging station.

[0077] The step S51 specifically includes the following steps:

[0078] Step S511: Calculate the cost of payment, i.e. the electricity cost required after the estimated queuing time; the calculation method is: Wherein W charging (t) is the amount of electricity charged by the vehicle at time t; E price (t) is the time-of-use electricity price at time t of the charging station; t EVCS is the time when the vehicle starts charging after the estimated queuing time; t charging is the estimated charging duration;

[0079] Step S512: Calculate the time cost, i.e. the income loss converted after considering the waiting time for charging calculated by the vehicle owner according to the SOC state corresponding to each charging gun and the queuing situation of the charging station; the calculation method is: x2=(t queue +t charging )·v ave ·εEquation Two,

[0080] Wherein t queue is the time required for queuing; v ave is the average speed of all roads in the road network fitted by step S3; and ε is the estimated profit per kilometer;

[0081] Step S513: Expected income, taking into account the difficulty of subsequent pick-up and order for the operator of the electric vehicle, and the estimated destination of the order. The calculation method is: Wherein num dest is the number of regions covered by the charging station; OD dest is the OD matrix probability to the corresponding destination; t wait is the time required to stay in the current region.

[0082] The step S52 specifically includes the following steps:

[0083] Step S521: Define the original decision matrix, the formula is

[0084]

[0085] Wherein, x t,ij is the jth traffic characteristic value of the vehicle to the ith charging station at time t, i.e. the vehicle needs to consider various factors and convert them into profit and loss values. The amount has been calculated in step S51;

[0086] Step S522: Traverse all the prospect values obtained by the charging station, arrange them from beneficial to non-beneficial, and name them as x c1 ,x c2 ,…,x c,TCV_Valid , calculate the standard value of the traffic characteristic value acceptable to the vehicle owner, the formula is

[0087]

[0088] wherein part of the charging stations can not be reached due to the insufficient endurance of the electric vehicle, TCV Valid is the number of reachable charging stations; s1 and s2 are weight interval, the calculation formula is s1 = floor(TCV Valid / 3)-1、

[0089] s2 = TCV Valid -[floor(TCV Valid / 3)+1] Formula six;

[0090] floor is the floor function.

[0091] Step S523: calculate the prospect value function of the charging station i selected by the kth vehicle at the tth moment for the jth traffic characteristic value, the formula is

[0092]

[0093] wherein a and b are risk sensitivity coefficients, and d kij is the difference between the jth traffic characteristic value selected by the kth vehicle at the tth moment and the inner measurement standard of the vehicle owner, and the inner measurement standard of the vehicle owner is S calculated through step S522;

[0094] Step S524: determine the proportion of each traffic characteristic value affecting the decision of the vehicle owner through the analytic hierarchy process;

[0095] Step S525: calculate the weight function of the vehicle owner facing the benefits and losses, the formula is

[0096]

[0097] wherein g is the risk benefit attitude coefficient, and d tj is the weight of the jth traffic characteristic value;

[0098] Step S526: calculate the prospect value of the kth vehicle selecting the ith charging station at the tth moment, the formula is:

[0099]

[0100] The step S524 specifically includes the following steps:

[0101] Step S5241: list the importance correlation degree of each two traffic characteristic values in step S51, list the matrix A, wherein 1 represents that two traffic characteristic values are equally important, 2 represents that one traffic characteristic value is obviously more important than the other, and 3 represents that one traffic characteristic value is much more important than the other; the reciprocal of each number is the opposite meaning.

[0102] Step S5242: Set j as the number of traffic characteristic values, calculate the value W of the product of the elements of each row in matrix A multiplied by j times;

[0103] Step S5243: Normalize W to obtain W * ;

[0104] Step S5244: Solve the maximum eigenvalue, wherein W * i is the i-th row of matrix W * ;

[0105] Step S5245: Perform consistency check, calculate In particular, when j=3, RI is taken as 0.58 here; if CR>1, it does not meet the requirements, indicating that the importance correlation degree needs to be reconsidered.

[0106] The above only describes the preferred embodiments of the present application, and any equivalent changes and modifications made within the scope of the patent application of the present application shall be included in the scope of the present application.

Claims

1. A load forecasting method for electric vehicle charging stations that takes into account multi-source information and decision correction, characterized in that: Includes the following steps; Step 1: First, simulate the multi-source information that users can collect within the region; Step 2: Simulate the user's situation using prospect theory when they encounter low battery and decide to go to a charging station; Step 3: During the user's charging process, based on real-time information changes, decision-making is corrected using prospect theory. By correcting the decisions, the load distribution of charging stations in the region is made more even, and the load distribution is consistent with the predicted actual demand of users. For taxis, when the vehicle's battery level is lower than the owner's expected level, prospect theory is used to evaluate the charging stations within the road network that are accessible in the current scenario, to obtain the optimal charging station, and to make a decision to go to that charging station for charging. Specifically, the following steps are included: Step S51: Considering the three factors of electricity price, time required to reach the charging station, and queuing situation at the charging station, summarize and calculate three corresponding traffic characteristic values, namely, payment cost, time cost, and expected benefit. Step S52: Calculate the foreground value corresponding to each charging station; Step S51 specifically includes the following steps: Step S511: Calculate the cost of charging, i.e., the electricity cost required after the estimated queuing time; the calculation method is as follows: Among them W charging (t) represents the amount of electricity charged by the vehicle at time t; E price (t) represents the time-of-use electricity price at the charging station at time t; t EVCS The time after the estimated queuing time for the vehicle to begin charging; t charging To estimate charging time; Step S512: Calculate time cost, which means considering the revenue loss after the car owner calculates the waiting time for charging based on the SOC status of each charging gun and the queuing situation at the charging station; the calculation method is as follows: x2=(t queue +t charging )·v ave Formula 2 for ε Where t queue The time required to queue; v ave ε is the average speed of all roads within the road network; ε is the estimated profit per kilometer. Step S513: Expected revenue, taking into account the ease with which electric vehicle owners can wait for and accept subsequent orders, and the expected destination of the orders; the calculation method is as follows: Where num dest The number of areas covered by the charging station; OD dest t represents the probability of traveling to the corresponding destination in the OD matrix; wait This refers to the time required to stay in the current area. Step S52 specifically includes the following steps: Step S521: Define the original decision matrix, the formula is as follows Where, x t,ij This refers to the j-th traffic characteristic value of the vehicle traveling to the i-th charging station at time t. This represents the various factors that the vehicle owner needs to consider and converts them into profit and loss values, which has been calculated in step S51. Step S522: Iterate through all the foreground values ​​obtained from all charging stations, arrange them from beneficial to detrimental, and name them x. c1 ,x c2 ,…,x c,TCV_Valid The standard value of traffic characteristics acceptable to car owners is calculated using the following formula: Some charging stations may be inaccessible due to insufficient range of electric vehicles, TCV Valid s1 represents the number of accessible charging stations; s2 and s1 represent the weighted intervals, calculated using the formula s1 = floor(TCV). Valid / 3)-1、 s2=TCV Valid -[floor(TCV Valid Formula 6; floor is rounded down to the nearest integer. Step S523: Calculate the prospective value function of vehicle k at time t for selecting charging station i based on the j-th traffic characteristic value. The formula is as follows: Where α and β are risk sensitivity coefficients, respectively, λ is the loss aversion coefficient, and d kij The difference between the traffic characteristic value j for vehicle k choosing charging station i for charging and the vehicle owner's internal evaluation standard, which is the S calculated in step S522; Step S524: Determine the proportion of each traffic characteristic value that influences the car owner's decision-making using the analytic hierarchy process; Step S525: Calculate the weighting function for the benefits and losses faced by the car owner, the formula is as follows: Where γ is the risk-return attitude coefficient, δ is the risk-loss attitude coefficient, and ω tj The weight of the j-th traffic characteristic value; Step S526: Calculate the foreground value of vehicle k selecting the i-th charging station at time t, using the formula:

2. The electric vehicle charging station load forecasting method considering multi-source information and decision correction according to claim 1, characterized in that: Step one is specifically divided into the following steps; Step S1: Divide the user's travel model into private cars and taxis, and generate travel chains through the travel chain model and OD matrix; Step S2: Collect multi-source information available to users within the road network, including road conditions, weather, and charging station queuing status; Step S3: Calculate the traffic flow of each road in the road network, generate the road network speed based on the Greenshield traffic flow fitting model, and thus obtain the driving speed of vehicles on each road, and then obtain the driving distance and energy consumption. Step S4: Determine the on / off status of the vehicle's air conditioning based on the predicted temperature, and determine whether it is operating in cold air, warm air, or off mode, thereby calculating the energy consumption of the vehicle's air conditioning; deduct the energy consumption at each moment from the vehicle's battery charge. Steps two and three are specifically divided into the following steps; Step S5: For taxis, when the vehicle's battery level is lower than the owner's expected battery level, prospect theory is used to evaluate the charging stations within the road network that are accessible in the current scenario, to obtain the optimal charging station, and to make a decision to go to that charging station for charging. Step S6: Whenever a vehicle with low battery arrives at a node or intersection in the road network on its way to a charging station, it jumps to step S5 to re-evaluate the charging station selection; if the optimal charging station changes, the destination of the charging station to be traveled to in the charging decision is modified; if the result does not change, it continues to travel to the charging station. Step S7: After the vehicle arrives at the charging station, it enters the queuing state. After queuing and entering the charging state, the amount of electricity to be charged for the vehicle is estimated using different estimated charging powers based on the range of the vehicle's remaining power. Step S8: Accumulate the charging amount of each vehicle at each charging station at each time point throughout the day to obtain the charging load of each charging station on a single day.

3. The electric vehicle charging station load forecasting method considering multi-source information and decision correction according to claim 1, characterized in that: Step S524 specifically includes the following steps: Step S5241: List the degree of importance correlation between each pair of traffic characteristic values ​​in step S51, and list matrix A, where 1 indicates that the two traffic characteristic values ​​are equally important, 2 indicates that one traffic characteristic value is significantly more important than the other, and 3 indicates that one traffic characteristic value is very important than the other; the reciprocal of each number has the opposite meaning. Step S5242: Let j be the number of traffic characteristic values, and calculate W, the j-th root of the product of the elements in each row of matrix A; Step S5243: Normalize W to obtain W * ; Step S5244: Solve for the largest eigenvalue. Among them, W * i For matrix W * The i-th row; Step S5245: Perform a consistency check and calculate... Specifically, when j=3, RI is set to 0.58 here; if CR>1, it does not meet the requirements, indicating that the importance correlation needs to be reconsidered.

Images