Electric vehicle charging demand prediction method based on Markov dynamic path decision

Abstract

The invention discloses an electric vehicle charging demand prediction method based on Markov dynamic path decision, and the method comprises the steps: S1, collecting a map of a target region, a vehicle travel origin-destination matrix, daily temperature, a traffic jam index, and electric vehicle parameters; s2, analyzing the parameters of the electric vehicles, dividing the electric vehicles into household electric vehicles and rented electric vehicles, and constructing a trip chain model for the household electric vehicles to obtain trip data of the rented electric vehicles; s3, constructing a Markov dynamic path optimization model to obtain a path of a time sequence and an energy consumption change data set; s4, constructing a charging demand judgment model to obtain the charging demand of the electric vehicle; and S5, charging load superposition and node load calculation: superposing the charging demands of the household electric vehicles and the taxis to corresponding road network nodes, and generating a final space-time distribution diagram of the charging demands of the electric vehicles. According to the design, the flexibility of path planning is improved, and the prediction accuracy of the traffic jam influence is effectively improved.

Description

Technical Field

[0001] The present invention relates to an electric vehicle charging demand prediction method based on Markov dynamic path decision-making, and is specifically applicable to improving the accuracy of charging demand prediction, so as to reasonably plan charging stations. Background Art

[0002] With the improvement of environmental protection and energy conservation awareness, electric vehicles (EVs) have gradually become the mainstream direction of future transportation. Compared with traditional fuel vehicles, electric vehicles have the advantages of zero emissions, low noise, and high energy efficiency. Many countries and regions promote the development of the electric vehicle industry through policy support and subsidies. However, the popularization of electric vehicles has also brought new problems, one of which is the management of charging demand, especially how to efficiently and conveniently find charging stations and reasonably plan charging routes during driving. Solving this problem is crucial for improving the user experience of electric vehicle users and promoting the further development of electric vehicles. Existing electric vehicle charging demand prediction methods usually assume fixed paths and road network conditions, and it is difficult to adapt to the dynamic path selection requirements caused by traffic changes during actual driving. Traditional models often do not consider real-time traffic congestion when planning charging demand, resulting in inflexible route selection, which may increase driving time and power consumption. Many methods fail to consider the psychological factors of drivers when choosing paths, ignoring the behavior that drivers may choose sub-optimal paths in congested situations. Therefore, by introducing a traffic congestion index and a path penalty coefficient, the tolerance of drivers to congested sections and the psychological factors in path selection are incorporated into the model, improving the problem of ignoring the impact of traffic congestion in traditional methods. Using the Markov decision process and combining real-time traffic information, the path selection can be dynamically adjusted according to traffic changes, thereby improving the flexibility of path planning, better adapting to real-time changing traffic conditions, and reasonably planning charging paths. Summary of the Invention

[0003] The purpose of the present invention is to overcome the problem of unreasonable planning of electric vehicle charging stations in the prior art, and provide an electric vehicle charging demand prediction method based on Markov dynamic path decision-making that can be reasonably planned.

[0004] To achieve the above purpose, the technical solution of the present invention is as follows:

[0005] In a first aspect, the present invention provides an electric vehicle charging demand prediction method based on Markov dynamic path decision-making, including the following steps:

[0006] S1. Collect the map of the target area, the vehicle travel origin-destination matrix, the daily temperature T, the traffic congestion index and electric vehicle parameters;

[0007] S2. Construct a road-traffic node map of the target area based on the target area map; analyze the parameters of electric vehicles, classify electric vehicles into household electric vehicles and rental electric vehicles, construct a travel chain model for household electric vehicles, calculate travel time parameters, and obtain household electric vehicle travel data; for rental electric vehicles, sample the departure location, destination, start time, and parking duration from the vehicle travel origin-destination matrix to obtain rental electric vehicle travel data.

[0008] S3. Construct a Markov dynamic path optimization model, input the household electric vehicle travel data and rental electric vehicle travel data into the Markov dynamic path optimization model, select the optimal path for each electric vehicle, calculate the energy consumption change during the journey, and obtain a time-series path and energy consumption change data set.

[0009] S4. Construct a charging demand determination model. For users of household electric vehicles, divide the users into demand types to obtain the charging demand of household electric vehicles; for users of rental electric vehicles, construct a charging demand selection model to obtain the charging demand of rental electric vehicles.

[0010] S5. Charge load superposition and node load calculation. Superpose the charging demands of household electric vehicles and taxis on the corresponding road network nodes to calculate the total node charging load, output the spatio-temporal distribution of the charging demand, and generate the final spatio-temporal distribution map of the electric vehicle charging demand based on the charging load data of all nodes.

[0011] Construct a road-traffic node map of the target area based on the target area map. Extract the road map of the target area from the target area map, determine multiple traffic nodes according to the distribution of the road map, and obtain the road-traffic node map of the target area.

[0012] Construct a travel chain model for household electric vehicles, calculate travel time parameters, and obtain household electric vehicle travel data.

[0013] Describe the travel chain containing various user information with spatial feature quantities and time feature quantities; spatial feature quantities include vehicle trip origin and destination, intermediate stop locations, and driving path information; time feature quantities include vehicle start departure time, destination stay time, driving time, and trip end time information; divide the city into four regions: residential area, working area, entertainment area, and other areas, and use this to represent the coordinate partition of the spatial feature quantities.

[0014] Calculation method of time feature quantities:

[0015] Vehicle start departure time: The initial time when the vehicle leaves home within a day follows a normal distribution, and the probability density function:

[0016]

[0017] Among them, μ T is the mean value of the first time leaving home; σ T is the variance of the first time leaving home, and x is the first time leaving home;

[0018] The residence time at different destination places: The residence time of the vehicle at different destination places varies, so it is represented by multiple probability density functions;

[0019] The probability density function of the vehicle residence time in the residential area satisfies the Weibull distribution:

[0020]

[0021] Among them, k is the shape parameter of the Weibull distribution; λ is the scale parameter of the Weibull distribution;

[0022] The probability density function of the vehicle residence time in the places within the work area satisfies the type III extreme value distribution:

[0023]

[0024] The probability density function of the vehicle residence time in the places within the entertainment area and other areas satisfies the type II extreme value distribution:

[0025]

[0026] Among them, μ is the location parameter of the generalized extreme value distribution; δ is the scale parameter of the generalized extreme value distribution, δ > 0; ε is the shape parameter of the generalized extreme value distribution, when ε > 0, a = 1 / ε, when ε < 0, a = -1 / ε;

[0027] The start time of the next trip: After the electric vehicle arrives at the destination place and ends its stay, the start time of the next trip:

[0028]

[0029] Among them, J is the total number of trips in the travel chain; t j+1 is the time when the vehicle starts the (j + 1)-th trip; t 0 is the first time the vehicle leaves home; is the total time consumed for the vehicle to complete the previous j trips; t stay,j is the residence time of the vehicle at the destination place of the j-th trip; The travel data of household electric vehicles is calculated using the travel chain model.

[0030] In the above S3, a Markov dynamic path optimization model is constructed. The travel data of household electric vehicles and the travel data of rental electric vehicles are input into the Markov dynamic path optimization model to select the optimal path for each electric vehicle.

[0031] Introduce the traffic congestion index I tp , which quantitatively describes the operating state of road traffic. The value range of the road traffic congestion index is [0, 100], and the larger the value, the more congested the road is;

[0032]

[0033] Among them, V tp is the traffic flow of road p at time t; C tp is the maximum traffic capacity of road p at time t; A penalty coefficient γ is introduced to make the user's decision not limited to the shortest path:

[0034]

[0035] Among them, γ k,t represents the penalty coefficient of road k at decision-making time t; is the traffic congestion index of road k at decision-making time t; I tp,max is the historical maximum traffic congestion index of the reference section; The steps for path decision for each car are as follows:

[0036] Pre-departure planning: Before departure, the user selects the travel destination and plans a shortest path; The pre-planned shortest path is denoted as π d , and each section of this path is called a reference section L d , and the path sets of the unselected road network nodes are selected as the alternative path set, denoted as A k ;

[0037] Arrival at the decision node: When the user arrives at road node i, enter the decision-making state, select the reference section L d , the traffic congestion index I tp,d and the penalty coefficient γ as references, and for each section k in the alternative path set A k , judge whether the following conditions are met: If then regard this section as a highly congested section and remove it from the alternative path set;

[0038] Calculate the state transition probability: When calculating the state transition probability for each section k in the remaining alternative paths, comprehensively consider the length L k and the penalty coefficient γ k,t 's influence:

[0039]

[0040] Among them, P(k|i) represents the probability of transferring from the current node i to the alternative path k; L k is the length of the alternative path k; γ k,tThe penalty coefficient of the alternative path k at the decision-making moment t; α and β are the weight factors of the path length and congestion penalty, which adjust the relative influence of the two on the decision-making.

[0041] Select the next section of the path: Select the next section of the road k with the largest state transition probability among the alternative paths according to the state transition probability P(k|i). * , and add it to the final path π. * ; Update the current node i to k. * , and repeat the above steps until the target node is reached.

[0042] Calculate the energy consumption change during the journey in S3.

[0043] Establish an energy consumption model for electric vehicles:

[0044]

[0045] In the formula, ΔE is the total power consumption during a single trip of the electric vehicle; P m is the power consumption per km of the electric vehicle under free traffic in the traffic network; L p is the length of the selected path p; P e is the additional power consumption caused by the start and stop of the electric vehicle due to traffic congestion per unit time; t e is the additional time caused by traffic congestion; C EV is the battery capacity of the electric vehicle; Considering that the environmental temperature will affect the opening and closing of the electric vehicle air conditioner, the energy consumption model is corrected:

[0046]

[0047] In the formula, E is the total power consumption during a single trip of the electric vehicle; is the temperature energy consumption at temperature T and road grade v k , and use the energy consumption model of the electric vehicle to calculate the energy consumption change of the electric vehicle during the journey.

[0048] In S3, according to the air conditioner usage curve, perform a normal distribution fitting on the start probability of the air conditioner to obtain the corresponding normal distribution function of the air conditioner start as follows:

[0049]

[0050] Among them, K open (T) represents the probability of the air conditioner turning on at temperature T; μ c and δ c are the mean and variance of the heating start parameters respectively; μ h and δ h are the mean and variance of the cooling start parameters; When the temperature T is between 5 and 21 °C, it is expressed as T c, denoted as T at 24 to 35 °C h ;

[0051] Analyze the air - conditioning energy consumption of the vehicle at different temperatures T and its proportion in the driving process, and the relationship between temperature and the proportion of air - conditioning energy consumption is obtained as follows:

[0052] E ac (T)=4.09×10 -6 T 3 +7.28×10 -5 T 2 -6.58×10 -3 T + 0.0873

[0053] Determine the temperature - related energy consumption at different temperatures T and road grades v k under the condition

[0054]

[0055] wherein, represents the energy - consumption factor of different road grades at time t, corresponding to the speeds of different road grades: v 1 , v 2 , v 3 , representing the driving speeds of first - class, second - class and third - class roads respectively; E p represents the initial energy consumption per unit mileage of the electric vehicle; K oc represents the start - stop state of the air conditioner, K oc =1 indicates that the air conditioner is on, K oc =0 indicates that the air conditioner is off: where the value of K oc is as follows:

[0056]

[0057] wherein, r is a uniformly - distributed random number in the interval [0,1], and its purpose is to compare with the probability K open (T) of the air conditioner being on, and determine whether the air conditioner is on through this comparison.

[0058] In step S4, for the users of home electric vehicles, classify the user demand types, and obtain the charging - mode demand results according to the classification types:

[0059] By collecting and analyzing the user's parking time, SOC, and electricity - price expectation information, classify the users into ordinary rigid users, emergency rigid users, high - anchor elastic users or low - anchor elastic users;

[0060] If the SOC of the electric vehicle does not meet the demand for the next trip and the parking time is greater than or equal to 6 hours, it is determined as an ordinary rigid user, the charging mode is slow charging, and the charging time is immediate;

[0061] If the SOC of an electric vehicle does not meet the demand for the next trip and the parking time is less than 6 hours, it is determined as an emergency rigid user, the charging mode is fast charging, and the charging time is immediate;

[0062] If the SOC of an electric vehicle meets the demand for the next trip and the user's expected value of the electricity price is higher than the actual price, it is a high-anchor elastic user:

[0063] If the parking time is greater than or equal to 6 hours, the charging mode is slow charging, and the charging time is immediate;

[0064] If the parking time is less than 6 hours, the charging mode is fast charging, and the charging time is immediate;

[0065] If the SOC of an electric vehicle meets the demand for the next trip and the user expects the electricity price to be lower than the actual price, it is determined as a low-anchor elastic user: the charging mode is slow charging, and the charging time is when the electricity price is low;

[0066] The charging demand results of household electric vehicle users are obtained according to the classification type.

[0067] For the charging demand of taxis: According to the current time t and the actual electricity price P of the charging station c c Calculate the charging cost and evaluate it in combination with the driver's revenue demand:

[0068]

[0069] R c is the revenue trade-off index; F t The expected total revenue after running from a full charge; E t The energy increment demand during charging, that is, the full charge amount minus the remaining charge amount; P c is the electricity price of the charging station;

[0070] If the current SOC is sufficient to support the next trip, the driver will further consider the revenue: When R c >R 阈值 the driver will choose the charging station with the corresponding electricity price P c to charge, otherwise, the charging will be delayed or other stations will be selected;

[0071] If the current SOC is not enough to support the next trip, it is judged whether the driver's parking time allows slow charging: If the parking time is greater than or equal to 6 hours, the driver will give priority to slow charging in combination with the electricity price and the charging cost, and the charging time is immediate; if the parking time is less than 6 hours, the driver will choose fast charging;

[0072] The charging demand results of rental electric vehicle users are obtained according to the above determination method.

[0073] Second aspect, the present invention provides an electric vehicle charging demand prediction system based on Markov dynamic path decision-making, specifically including: a data collection module, a travel data construction module, a path optimization module, a charging demand determination module, and a spatio-temporal distribution map generation module;

[0074] The data collection module: is used to collect the map of the target area, the vehicle travel origin-destination matrix, the daily temperature T, the traffic congestion index and electric vehicle parameters;

[0075] The travel data construction module: is used to construct a road-traffic node map of the target area based on the target area map; analyze the electric vehicle parameters, classify the electric vehicles into household electric vehicles and rental electric vehicles, construct a travel chain model for the household electric vehicles, and calculate the travel time parameters to obtain the travel data of the household electric vehicles; for the rental electric vehicles, sample the departure place, destination, start time, and parking duration from the vehicle travel origin-destination matrix to obtain the travel data of the rental electric vehicles;

[0076] The path optimization module: is used to construct a Markov dynamic path optimization model, input the travel data of the household electric vehicles and the travel data of the rental electric vehicles into the Markov dynamic path optimization model, select the optimal path for each electric vehicle, and calculate the energy consumption change during the journey to obtain a time-series path and energy consumption change data set;

[0077] The charging demand determination module: is used to construct a charging demand determination model. For the users of household electric vehicles, divide the users into demand types to obtain the charging demand of the household electric vehicles; for the users of rental electric vehicles, construct a charging demand selection model to obtain the charging demand of the rental electric vehicles;

[0078] The spatio-temporal distribution map generation module: is used for charging load superposition and node load calculation, superimpose the charging demands of the household electric vehicles and taxis on the corresponding road network nodes to calculate the total node charging load, output the spatio-temporal distribution of the charging demand, and generate the final spatio-temporal distribution map of the electric vehicle charging demand according to the charging load data of all nodes.

[0079] In the travel data construction module, a road-traffic node map of the target area is constructed based on the target area map. The road map of the target area is extracted from the target area map, and multiple traffic nodes are determined according to the distribution of the road map to obtain the road-traffic node map of the target area;

[0080] Construct a travel chain model for the household electric vehicles and calculate the travel time parameters to obtain the travel data of the household electric vehicles;

[0081] Describe the trip chain containing various user information using spatial feature quantities and temporal feature quantities; the spatial feature quantities include the starting and ending points of the vehicle trip, the stopping points along the way, and the driving path information; the temporal feature quantities include the starting departure time of the vehicle, the staying time at the destination, the driving time, and the trip end time information; divide the city into four regions: residential area, working area, entertainment area, and other areas, and use this to represent the coordinate partition of the spatial feature quantities;

[0082] Calculation method of temporal feature quantities:

[0083] Starting departure time of the vehicle: The initial time of the vehicle leaving home within a day follows a normal distribution, and the probability density function:

[0084]

[0085] where, μ T is the mean value of the initial time of leaving home; σ T is the variance of the initial time of leaving home, and x is the initial time of leaving home;

[0086] Staying time at different destination places: The staying time of the vehicle at different destination places varies, so it is represented by multiple probability density functions;

[0087] The probability of the staying duration of the vehicle in the residential area satisfies the Weibull distribution, and the probability density function:

[0088]

[0089] where, k is the shape parameter of the Weibull distribution; λ is the scale parameter of the Weibull distribution;

[0090] The probability of the staying duration of the vehicle at the places in the working area satisfies the type III extreme value distribution, and the probability density function:

[0091]

[0092] The probability of the staying duration of the vehicle at the places in the entertainment area and other areas satisfies the type II extreme value distribution, and the probability density function:

[0093]

[0094] where, μ is the location parameter of the generalized extreme value distribution; δ is the scale parameter of the generalized extreme value distribution, δ > 0; ε is the shape parameter of the generalized extreme value distribution, a = 1 / ε when ε > 0, and a = -1 / ε when ε < 0;

[0095] Starting time of the next trip: After the electric vehicle arrives at the destination place and ends the stay, it completes the current trip, and the starting time of the next trip:

[0096]

[0097] Among them, J is the total number of trips in the trip chain; t j+1 is the time when the vehicle starts the (j + 1)-th trip; t 0 is the time when the vehicle first leaves home; is the total time consumed for the vehicle to complete the first j trips; t stay,j is the residence duration of the vehicle at the destination of the j-th trip; The travel chain model is used to calculate the household electric vehicle travel data.

[0098] The path optimization module constructs a Markov dynamic path optimization model, inputs the household electric vehicle travel data and the rental electric vehicle travel data into the Markov dynamic path optimization model, and selects the optimal path for each electric vehicle.

[0099] Introduce the traffic congestion index I tp , which quantitatively describes the operation state of the road traffic. The value range of the road traffic congestion index is [0, 100], and the larger the value, the more congested the road is;

[0100]

[0101] Among them, V tp is the traffic flow of road p at time t; C tp is the maximum traffic capacity of road p at time t; The penalty coefficient γ is introduced to make the user's decision not limited to the shortest path:

[0102]

[0103] Among them, γ k,t represents the penalty coefficient of road k at decision time t; is the traffic congestion index of road k at decision time t; I tp,max is the historical maximum traffic congestion index of the reference section; The steps for path decision for each vehicle are as follows:

[0104] Pre-departure planning: Before departure, the user selects the travel destination and plans a shortest path; The pre-planned shortest path is denoted as π d , and each section of this path is called a reference section L d , and the path set of the road network nodes not selected is selected as the alternative path set, denoted as A k ;

[0105] Arrival at the decision node: When the user arrives at road node i, enter the decision state, select the reference section L d , the traffic congestion index I tp,d and the penalty coefficient γ as references, and for each section k in the alternative path set A k , judge whether the following conditions are satisfied: If Then this section of the road is regarded as a highly congested section and removed from the set of alternative paths;

[0106] Calculate the state transition probability: When calculating the state transition probability for each section k in the remaining alternative paths, comprehensively consider the length L k and the penalty coefficient γ k,t for its influence:

[0107]

[0108] Among them, P(k|i) represents the probability of transferring from the current node i to the alternative path k; L k is the length of the alternative path k; γ k,t is the penalty coefficient of the alternative path k at the decision-making moment t; α and β are the weight factors of the path length and congestion penalty, adjusting the relative influence of the two on the decision-making;

[0109] Select the next section of the path: Select the next section k with the largest state transition probability in the alternative paths according to the state transition probability P(k|i) * , and add it to the final path π * ; Update the current node i to k * , and repeat the above steps until the target node is reached.

[0110] Calculate the energy consumption change during the journey;

[0111] Establish an energy consumption model for electric vehicles:

[0112]

[0113] In the formula, ΔE is the total power consumption during a single trip of the electric vehicle; P m is the power consumption per km of the electric vehicle under free traffic in the transportation network; L p is the length of the selected path p; P e is the additional power consumption caused by the start and stop of the electric vehicle due to traffic congestion per unit time; t e is the additional time caused by traffic congestion; C EV is the battery capacity of the electric vehicle; Considering that the ambient temperature affects the opening and closing of the air conditioner of the electric vehicle, correct the energy consumption model:

[0114]

[0115] In the formula, E is the total power consumption during a single trip of the electric vehicle; is the temperature energy consumption at temperature T and road grade v k , and use the energy consumption model of the electric vehicle to calculate the energy consumption change of the electric vehicle during the journey.

[0116] According to the air conditioner usage curve, perform a normal distribution fitting on the startup probability of the air conditioner to obtain the corresponding normal distribution function for the air conditioner startup as follows:

[0117]

[0118] Among them, K open (T) represents the probability of the air conditioner being turned on at temperature T; μ c and δ c are respectively the mean and variance of the heating startup parameters; μ h and δ h are the mean and variance of the cooling startup parameters; the temperature T is denoted as T c when it is between 5 and 21 °C, and is denoted as T h when it is between 24 and 35 °C;

[0119] Analyze the air conditioner energy consumption at different temperatures T and its proportion during the journey to obtain the relationship between temperature and the proportion of air conditioner energy consumption as follows:

[0120] E ac (T) = 4.09×10 -6 T 3 + 7.28×10 -5 T 2 - 6.58×10 -3 T + 0.0873

[0121] Determine the temperature energy consumption at different temperatures T and road grades v k as follows

[0122]

[0123] Among them, represents the energy consumption factor of different road grades at time t, corresponding to the speeds of different road grades: v 1 , v 2 , v 3 , respectively representing the driving speeds of first-class, second-class, and third-class roads; E p represents the initial energy consumption per unit mileage of the electric vehicle; K oc represents the start-stop state of the air conditioner, K oc = 1 indicates that the air conditioner is turned on, K oc = 0 indicates that the air conditioner is turned off: among them, the value of K oc is as follows:

[0124]

[0125] Among them, r is a uniformly distributed random number in the interval [0, 1], and its purpose is to be used to compare with the probability K open(T), and determine whether to turn on the air conditioner based on this comparison.

[0126] In the charging demand determination module, for users of household electric vehicles, the user demand types are classified, and the charging mode demand results of users are obtained according to the classification types:

[0127] By collecting and analyzing the user's parking time, SOC, and expected electricity price information, the user is classified as an ordinary rigid user, an emergency rigid user, a high-anchor elastic user, or a low-anchor elastic user;

[0128] If the SOC of the electric vehicle does not meet the demand for the next trip and the parking time is greater than or equal to 6 hours, it is determined as an ordinary rigid user, the charging mode is slow charging, and the charging time is immediate;

[0129] If the SOC of the electric vehicle does not meet the demand for the next trip and the parking time is less than 6 hours, it is determined as an emergency rigid user, the charging mode is fast charging, and the charging time is immediate;

[0130] If the SOC of the electric vehicle meets the demand for the next trip and the user's expected electricity price is higher than the actual price, it is a high-anchor elastic user:

[0131] If the parking time is greater than or equal to 6 hours, the charging mode is slow charging, and the charging time is immediate;

[0132] If the parking time is less than 6 hours, the charging mode is fast charging, and the charging time is immediate;

[0133] If the SOC of the electric vehicle meets the demand for the next trip and the user's expected electricity price is lower than the actual price, it is determined as a low-anchor elastic user: the charging mode is slow charging, and the charging time is when the electricity price is low;

[0134] Obtain the charging demand results of household electric vehicle users according to the classification types.

[0135] For taxi charging demand: According to the current time t and the actual electricity price P of the charging station c c Calculate the charging cost and evaluate it in combination with the driver's revenue demand:

[0136]

[0137] R c is the revenue trade-off index; F t is the expected total revenue from running after being fully charged; E t is the energy increment demand during charging, that is, the full charge capacity minus the remaining capacity; P c is the electricity price of the charging station;

[0138] If the current SOC is sufficient to support the next trip, the driver will further consider the revenue: When R c >R 阈值Only when the electricity price is reached will the driver choose the corresponding electricity price P c to charge at the charging station, otherwise charging will be delayed or another station will be selected;

[0139] If the current SOC is not sufficient to support the next trip, it is judged whether the driver's parking time allows slow charging: if the parking time is greater than or equal to 6 hours, the driver will give priority to slow charging considering the electricity price and charging cost, and the charging time is immediate; if the parking time is less than 6 hours, the driver will choose fast charging;

[0140] The charging demand results of rental electric vehicle users are obtained according to the above determination method.

[0141] In a third aspect, the present invention provides an electric vehicle charging demand prediction device based on Markov dynamic path decision-making, including a memory and a processor. The memory is used to store computer program codes and transmit the computer program codes to the processor;

[0142] The processor is used to execute the aforementioned electric vehicle charging demand prediction method based on Markov dynamic path decision-making according to the instructions in the computer program codes.

[0143] In a fourth aspect, the present invention provides a computer program product, including a computer program, and the computer program is executed by the processor to perform the aforementioned electric vehicle charging demand prediction method based on Markov dynamic path decision-making.

[0144] Compared with the prior art, the beneficial effects of the present invention are:

[0145] 1. In the electric vehicle charging demand prediction based on Markov dynamic path decision-making of the present invention, for the travel of electric vehicles, it is usually assumed that the path and road network conditions are fixed, which is difficult to adapt to the dynamic path selection requirements caused by traffic changes during actual driving. Traditional models often do not consider real-time traffic congestion when planning charging demands, resulting in inflexible route selection and possible problems such as increased driving time and power consumption. By introducing dynamic path selection, using the Markov decision process, and combining real-time traffic information, the path selection can be dynamically adjusted according to traffic changes, thereby improving the flexibility of path planning. By introducing the traffic congestion index and path penalty coefficient, the driver's tolerance for congested sections and psychological factors in path selection are incorporated into the model, improving the problem of ignoring the impact of traffic congestion in traditional methods.

[0146] 2. In the electric vehicle charging demand prediction based on Markov dynamic path decision of the present invention, it performs excellently in considering path dynamics and driver decision-making psychology, can better adapt to real-time changing traffic conditions, and reasonably plan the charging path. This improvement enhances the accuracy and adaptability of electric vehicle charging demand prediction, helps relieve urban traffic pressure, and promotes the more efficient popularization and use of electric vehicles.

[0147] 3. An electric vehicle charging demand prediction system based on Markov dynamic path decision of the present invention includes: a data acquisition module, a travel data construction module, a path optimization module, a charging demand determination module, and a spatio-temporal distribution map generation module. This system is used to implement the steps of the electric vehicle charging demand prediction method based on Markov dynamic path decision provided in any of the above technical solutions. Therefore, this system simultaneously includes all the beneficial effects of the electric vehicle charging demand prediction method based on Markov dynamic path decision provided in any of the above technical solutions, which will not be elaborated here.

[0148] 4. An electric vehicle charging demand prediction device based on Markov dynamic path decision of the present invention includes a processor and a memory. The memory is used to store computer program code and transmit the computer program code to the processor. The processor is used to execute the electric vehicle charging demand prediction method based on Markov dynamic path decision provided in any of the above technical solutions according to the instructions in the computer program code. Therefore, this device simultaneously includes all the beneficial effects of the electric vehicle charging demand prediction method based on Markov dynamic path decision provided in any of the above technical solutions, which will not be elaborated here.

[0149] 5. A computer program product of the present invention, when the computer program is executed by a processor, it implements the steps of the electric vehicle charging demand prediction method based on Markov dynamic path decision provided in any of the above technical solutions. Therefore, this computer program product simultaneously includes all the beneficial effects of the electric vehicle charging demand prediction method based on Markov dynamic path decision provided in any of the above technical solutions, which will not be elaborated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0150] Figure 1 is the flowchart of the method of the present invention.

[0151] Figure 2 is the system diagram of the present invention.

[0152] Figure 3 is the device diagram of the present invention.

[0153] Figure 4 is the flowchart of the steps of Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0154] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0155] Embodiment 1:

[0156] Refer to Figure 1 、 Figure 4 , a method for predicting the charging demand of electric vehicles based on Markov dynamic path decision-making, comprising the following steps:

[0157] S1. Collect the map of the target area, the origin-destination matrix of vehicle trips, the daily temperature T, the traffic congestion index and the parameters of electric vehicles;

[0158] S2. Based on the map of the target area, construct a road-traffic node map of the target area; analyze the parameters of electric vehicles, classify electric vehicles into household electric vehicles and rental electric vehicles, construct a travel chain model for household electric vehicles, and calculate the travel time parameters to obtain the travel data of household electric vehicles; for rental electric vehicles, sample the departure place, destination, start time and parking duration from the origin-destination matrix of vehicle trips to obtain the travel data of rental electric vehicles;

[0159] In S2, based on the map of the target area, construct a road-traffic node map of the target area. Extract the road map of the target area from the map of the target area, determine multiple traffic nodes according to the distribution of the road map, and obtain the road-traffic node map of the target area;

[0160] Construct a travel chain model for household electric vehicles, and calculate the travel time parameters to obtain the travel data of household electric vehicles;

[0161] Describe the travel chain containing various user information with spatial feature quantities and time feature quantities; the spatial feature quantities include the start and end points of the vehicle journey, the stopover locations along the way and the driving path information; the time feature quantities include the vehicle start time, the destination stay time, the driving time and the journey end time information; divide the city into four regions: residential area, work area, entertainment area and other areas, and use this to represent the coordinate partition of the spatial feature quantities;

[0162] Calculation method of time feature quantity:

[0163] Vehicle start time: The initial departure time of the vehicle within a day follows a normal distribution, and the probability density function:

[0164]

[0165] where μ T is the mean value of the initial departure time, taking 7.32; σ T is the variance of the initial departure time, taking 1.34, and x is the initial departure time;

[0166] Residence time at different destination locations: The residence time of the vehicle at different destination locations varies, so it is represented by multiple probability density functions;

[0167] The probability of the residence time of vehicles in residential areas satisfies the Weibull distribution, and the probability density function:

[0168]

[0169] where k is the shape parameter of the Weibull distribution, taking 14.05; λ is the scale parameter of the Weibull distribution, taking 7.16;

[0170] The probability of the residence time of vehicles at locations within the work area satisfies the type III extreme value distribution, and the probability density function:

[0171]

[0172] The probability of the residence time of vehicles at locations within the entertainment area and other areas satisfies the type II extreme value distribution, and the probability density function:

[0173]

[0174] where μ is the location parameter of the generalized extreme value distribution; δ is the scale parameter of the generalized extreme value distribution, δ > 0; ε is the shape parameter of the generalized extreme value distribution, a = 1 / ε when ε > 0, and a = -1 / ε when ε < 0;

[0175] Start time of the next trip: After the electric vehicle arrives at the destination location, the completion of the stay means the end of this trip, and the start time of the next trip:

[0176]

[0177] where J is the total number of trips in the travel chain; t j+1 is the time when the vehicle starts the (j + 1)-th trip; t 0 is the time when the vehicle leaves home for the first time; is the total time taken for the vehicle to complete the first j trips; t stay,j is the residence time of the vehicle at the destination location of the j-th trip; The travel data of household electric vehicles is calculated using the travel chain model.

[0178] S3. Construct a Markov dynamic path optimization model, input the travel data of household electric vehicles and rental electric vehicles into the Markov dynamic path optimization model, select the optimal path for each electric vehicle, and calculate the energy consumption change during the trip to obtain a time-series dataset of paths and energy consumption changes;

[0179] In S3, a Markov dynamic path optimization model is constructed. The household electric vehicle travel data and rental electric vehicle travel data are input into the Markov dynamic path optimization model to select the optimal path for each electric vehicle.

[0180] A detailed description of travel choices is as follows: During the travel process, when choosing a path, the driver mainly considers the length of the driving distance and the degree of traffic congestion. Therefore, the Dijkstra algorithm is usually used to select the path with the shortest driving distance. For the traffic situation, under the same congestion state, due to the complexity of the urban road network, people may have different feelings under different road network conditions, and the traffic congestion index I is introduced. tp , which quantitatively describes the operating state of the road traffic. The value range of the road traffic congestion index is [0, 100], and the larger the value, the more congested the road is.

[0181]

[0182] Among them, V tp is the traffic flow of road p at time t (number of vehicles per hour); C tp is the maximum traffic capacity of road p at time t (number of vehicles per hour);

[0183] The penalty coefficient γ is introduced to prevent the user's decision from being limited to the shortest path:

[0184]

[0185] Among them, γ k,t represents the penalty coefficient of road k at decision time t; is the traffic congestion index of road k at decision time t; I tp,max is the historical maximum traffic congestion index of the reference section. The steps for path decision for each vehicle are as follows:

[0186] Pre-trip planning: Before departure, the user selects the travel destination and plans a shortest path; the pre-planned shortest path is denoted as π d , and each section of this path is called a reference section L d , and the path set of the unselected road network nodes is selected as the alternative path set, denoted as A k ;

[0187] Arrival at the decision node: When the user arrives at road node i, enter the decision state, select the reference section L d , the traffic congestion index I tp,d and the penalty coefficient γ as references, and for each section k in the alternative path set A k , judge whether the following conditions are met: If Then this section of the road is regarded as a highly congested section and removed from the set of alternative paths;

[0188] Calculate the state transition probability: For each section k in the remaining alternative paths, when calculating its state transition probability, comprehensively consider the length L k and the penalty coefficient γ k,t 's influence:

[0189]

[0190] Among them, P(k|i) represents the probability of transferring from the current node i to the alternative path k; L k The length of the alternative path k; γ k,t The penalty coefficient of the alternative path k at the decision-making moment t; α, β are the weight factors of the path length and congestion penalty, adjusting their relative influence on the decision-making;

[0191] Select the next section of the path: Select the next section k with the largest state transition probability in the alternative paths according to the state transition probability P(k|i) * , and add it to the final path π * ; Update the current node i to k * , and repeat the above steps until the target node is reached.

[0192] Calculate the energy consumption change during the journey in S3;

[0193] Establish an energy consumption model for electric vehicles:

[0194]

[0195] In the formula, ΔE is the total power consumption during a single trip of the electric vehicle; P m is the power consumption per km of the electric vehicle under free traffic in the traffic network, taking 0.2 kWh / km; L p is the length of the selected path p, in km; P e is the additional power consumption caused by the start and stop of the electric vehicle due to traffic congestion per unit time; t e is the additional time caused by traffic congestion; C EV is the battery capacity of the electric vehicle; Considering that the ambient temperature will affect the opening and closing of the air conditioner of the electric vehicle, the energy consumption model is corrected:

[0196]

[0197] In the formula, E is the total power consumption during a single trip of the electric vehicle; is the temperature energy consumption at temperature T and road grade v k Under the condition, use the energy consumption model of the electric vehicle to calculate the energy consumption change of the electric vehicle during the journey.

[0198] In S3,

[0199] Since the air conditioner is one of the main energy-consuming devices of an electric vehicle, the start-stop state and temperature change of the air conditioner directly affect the charging demand of the electric vehicle. The energy consumption of the on-vehicle air conditioner is directly extracted from the battery, and its energy consumption changes significantly with temperature and start-stop state. Under different temperature conditions, the desired temperature value required inside the vehicle will also be different. According to the air conditioner usage curve, a normal distribution fitting is performed on the start probability of the air conditioner to obtain the corresponding normal distribution function of the air conditioner start as follows:

[0200]

[0201] where K open (T) represents the probability of the air conditioner turning on at temperature T; μ c and δ c are the mean and variance of the heating start parameters respectively, μ c takes the value of 35.15, and δ c takes the value of 6.13; μ h and δ h are the mean and variance of the cooling start parameters respectively, μ h takes the value of 4.56, and δ h takes the value of 7.54. When the temperature T is between 5 and 21 °C, it is denoted as T c , and when it is between 24 and 35 °C, it is denoted as T h ; By analyzing the air conditioner energy consumption at different temperatures T and its proportion in the journey, the relationship between temperature and the proportion of air conditioner energy consumption is obtained as follows:

[0202] E ac (T) = 4.09×10 -6 T 3 + 7.28×10 -5 T 2 - 6.58×10 -3 T + 0.0873

[0203] Determine the temperature energy consumption at different temperatures T and road grades v k as follows

[0204]

[0205] where, represents the energy consumption factor at time t for different road grades, corresponding to the speeds of different road grades, which are v 1 , v 2 , v 3 (unit: km / h), representing first-class, second-class, and third-class roads respectively; E p represents the initial energy consumption per unit mileage of the electric vehicle; K oc represents the start-stop state of the air conditioner, Koc = 1 indicates that the air conditioner is turned on, K oc = 0 indicates that the air conditioner is turned off: where K oc takes the following values:

[0206]

[0207] where r is a uniformly distributed random number in the interval [0, 1], and its purpose is to be used to compare the probability K open (T) of the air conditioner being turned on, and to determine whether the air conditioner is turned on based on this comparison.

[0208] S4. Build a charging demand determination model. For users of household electric vehicles, classify the users' demand types to obtain the charging demands of household electric vehicles; for users of rental electric vehicles, build a charging demand selection model to obtain the charging demands of rental electric vehicles;

[0209] In S4, for users of household electric vehicles, classify the users' demand types, and obtain the charging mode demand results of the users according to the classified types:

[0210] By collecting and analyzing the users' parking time, SOC, and electricity price expectation information (obtained from electric vehicle parameters), classify the users into ordinary rigid users, emergency rigid users, high-anchor elastic users, or low-anchor elastic users; according to the user type and parameters such as parking time, the system matches an appropriate charging mode (fast charging or slow charging) for them to maximize the charging efficiency and meet the travel needs. By default, at the same time period, the fast charging price is higher than the slow charging price;

[0211] Introduce the high-anchor and low-anchor concepts in the psychological concept of the anchoring effect to judge the users' acceptance of electricity prices; high-anchor means that the users' expected value of the electricity price is higher than the actual price, and at this time the users' charging willingness is higher; low-anchor means that the users expect the electricity price to be lower than the actual price;

[0212] If the SOC of the electric vehicle does not meet the next travel demand and the parking time is greater than or equal to 6 hours, it is determined as an ordinary rigid user, the charging mode is slow charging, and the charging time is immediate;

[0213] If the SOC of the electric vehicle does not meet the next travel demand and the parking time is less than 6 hours, it is determined as an emergency rigid user, the charging mode is fast charging, and the charging time is immediate;

[0214] If the SOC of the electric vehicle meets the next travel demand and the users' expected value of the electricity price is higher than the actual price, it is a high-anchor elastic user:

[0215] If the parking time is greater than or equal to 6 hours, the charging mode is slow charging, and the charging time is immediate;

[0216] The parking time is less than 6 hours, the charging mode is fast charging, and the charging time is instant;

[0217] The SOC of the electric vehicle meets the demand for the next trip, and the user's expected electricity price is lower than the actual price. It is determined as a low-anchor elasticity user: the charging mode is slow charging, and the charging time is when the electricity price is low;

[0218] According to the classification type, the charging demand results of household electric vehicle users are obtained.

[0219] Taxi charging demand: According to the actual electricity price P of the current time t and the charging station c c Calculate the charging cost and evaluate it in combination with the driver's revenue requirements:

[0220]

[0221] R c is the revenue trade-off index; F t The total expected revenue after running from a full charge (such as the revenue from the order starting point to the end point); E t The energy increment demand during charging, that is, the full charge amount minus the remaining charge amount; P c is the electricity price of the charging station;

[0222] If the current SOC is sufficient to support the next trip, the driver will further consider the revenue: when R c >R 阈值 the driver will choose the charging station with the corresponding electricity price P c to charge, otherwise, the charging will be delayed or other stations will be selected;

[0223] If the current SOC is not enough to support the next trip, it is judged whether the driver's parking time allows slow charging: if the parking time is greater than or equal to 6 hours, the driver will give priority to slow charging in combination with the electricity price and the charging cost, and the charging time is instant; if the parking time is less than 6 hours, the driver chooses fast charging;

[0224] During the low electricity price period (such as at night), even if the SOC is sufficient, the driver may actively perform slow charging to reduce the charging cost during the future peak period. During the peak period or high electricity price period, the driver may delay charging and give priority to completing the operation task;

[0225] According to the above determination method, the charging demand results of rental electric vehicle users are obtained.

[0226] S5. Charging load superposition and node load calculation. Superimpose the charging demands of household electric vehicles and taxis on the corresponding road network nodes to calculate the total node charging load, output the spatio-temporal distribution of the charging demand, and generate the final spatio-temporal distribution map of the electric vehicle charging demand based on the charging load data of all nodes.

[0227] The present invention realizes dynamic route selection through real-time traffic data and route penalty coefficients. First, the method adjusts the route in real time during driving to avoid congested sections, thereby reducing power consumption during the journey. At the same time, through the Markov decision process, the current traffic conditions, the state of charge (SOC) of the vehicle, and the actual decision-making behavior of the driver are comprehensively evaluated, simulating the "adverse selection" tendency of users in a complex traffic environment, making the route decision more in line with reality. In addition, the method intelligently recommends charging modes according to the parking time, choosing the slow charging mode to extend the battery life when there is sufficient time, and giving priority to fast charging in case of emergency to meet the needs. This method not only improves the flexibility of route planning and the accuracy of prediction, but also optimizes the user charging experience, can save charging costs, extend the battery life, and improve the adaptability of electric vehicles in complex urban traffic environments

[0228] Embodiment 2:

[0229] An electric vehicle charging demand prediction system based on Markov dynamic route decision-making, which is used to execute the aforementioned electric vehicle charging demand prediction method based on Markov dynamic route decision-making, specifically including: a data acquisition module, a travel data construction module, a route optimization module, a charging demand determination module, and a spatio-temporal distribution map generation module;

[0230] Data acquisition module: used to collect maps of the target area, vehicle travel origin-destination matrices, daily temperature T, traffic congestion indices and electric vehicle parameters;

[0231] Travel data construction module: used to construct a road-traffic node map of the target area based on the target area map; analyze electric vehicle parameters, classify electric vehicles into household electric vehicles and rental electric vehicles, construct a travel chain model for household electric vehicles, and calculate travel time parameters to obtain household electric vehicle travel data; for rental electric vehicles, sample the departure place, destination, start time, and parking duration from the vehicle travel origin-destination matrix to obtain rental electric vehicle travel data;

[0232] Route optimization module: used to construct a Markov dynamic route optimization model, input the household electric vehicle travel data and rental electric vehicle travel data into the Markov dynamic route optimization model, select the optimal route for each electric vehicle, and calculate the energy consumption change during the journey to obtain a time-series dataset of routes and energy consumption changes;

[0233] Charging demand determination module: used to construct a charging demand determination model, for users of household electric vehicles, divide the user demand types to obtain the charging demand of household electric vehicles; for rental electric vehicle users, construct a charging demand selection model to obtain the charging demand of rental electric vehicles;

[0234] Space-time distribution map generation module: used for charging load superposition and node load calculation, superposing the charging demands of household electric vehicles and taxis onto the corresponding road network nodes to calculate the total node charging load, outputting the space-time distribution of the charging demand, and generating the final space-time distribution map of the electric vehicle charging demand based on the charging load data of all nodes.

[0235] Example 3:

[0236] An electric vehicle charging demand prediction device based on Markov dynamic path decision-making, comprising a memory and a processor. The memory is used for storing computer program code and transmitting the computer program code to the processor; the processor is used for executing the aforesaid electric vehicle charging demand prediction method based on Markov dynamic path decision-making according to the instructions in the computer program code.

[0237] Example 4:

[0238] A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the aforesaid electric vehicle charging demand prediction method based on Markov dynamic path decision-making.

[0239] A computer program product, comprising a computer program, and the computer program is used to execute the aforesaid electric vehicle charging demand prediction method based on Markov dynamic path decision-making.

Claims

1. A method for predicting electric vehicle charging demand based on Markov dynamic path decision, characterized by: The steps include: S1. Collect maps of the target area, vehicle travel origin and destination matrix, daily temperature T, and traffic congestion index I tpk,t and electric vehicle parameters; S2. Construct a road-traffic node diagram of the target area based on the target area map; analyze electric vehicle parameters, divide electric vehicles into household electric vehicles and rental electric vehicles, construct a travel chain model for household electric vehicles, and calculate travel time parameters to obtain household electric vehicle travel data; for rental electric vehicles, sample the departure place, destination, start time, and parking duration from the vehicle travel start and end point matrix to obtain rental electric vehicle travel data; S3, construct a Markov dynamic path optimization model, input the travel data of family electric vehicles and rental electric vehicles into the Markov dynamic path optimization model, select the optimal path for each electric vehicle, and calculate the energy consumption change during the trip to obtain a time series path and energy consumption change data set; S4. Construct a charging demand determination model. For users of home electric vehicles, divide the users into demand types to obtain the charging demand of home electric vehicles; for users of rental electric vehicles, construct a charging demand selection model to obtain the charging demand of rental electric vehicles; S5. Charging load superposition and node load calculation: superimpose the charging demands of household electric vehicles and taxis on the corresponding road network nodes to calculate the total node charging load, output the spatiotemporal distribution of charging demands, and generate the final spatiotemporal distribution diagram of electric vehicle charging demands based on the charging load data of all nodes.

2. The method for predicting electric vehicle charging demand based on Markov dynamic path decision according to claim 1 is characterized in that: In S2 A road-traffic node graph of the target area is constructed based on the target area map, a road graph of the target area is extracted from the target area map, a plurality of traffic nodes are determined according to the distribution of the road graph, and a road-traffic node graph of the target area is obtained; Construct a travel chain model for family electric vehicles, calculate travel time parameters, and obtain family electric vehicle travel data; Use spatial and temporal features to describe the travel chain containing multiple user information; spatial features include the starting and ending points of the vehicle trip, the places where the vehicle stops along the way, and the driving path information; temporal features include the vehicle's starting time, the destination stay time, the driving time, and the end time of the trip; divide the city into four areas: residential area, work area, entertainment area, and other areas, and use this to represent the coordinate partition of the spatial features; Calculation method of time characteristic quantity: Vehicle departure time: The first departure time of a vehicle in a day satisfies the normal distribution, and the probability density function is: Among them, μ T is the mean time of leaving home for the first time; σ T is the variance of time of first leaving home, x is the time of first leaving home; Dwell time at different destinations: The time a vehicle spends at different destinations varies, so it is represented by a variety of probability density functions; The probability of vehicle stay time in residential areas satisfies the Weibull distribution, and the probability density function is: Where k is the shape parameter of the Weibull distribution; λ is the scale parameter of the Weibull distribution; The probability of the length of time a vehicle stays in a work area satisfies the type III extreme value distribution, and the probability density function is: The probability of the length of stay of vehicles in entertainment areas and other places in the area meets the type II extreme value distribution, and the probability density function is: Wherein, μ is the location parameter of the generalized extreme value distribution; δ is the scale parameter of the generalized extreme value distribution, δ>0; ε is the shape parameter of the generalized extreme value distribution, when ε>0, a=1 / ε, when ε<0, a=-1 / ε; The start time of the next trip: When the electric vehicle arrives at the destination, the trip is completed and the next trip starts at: Where J is the total number of trips in the travel chain; t j+1 is the time when the vehicle starts the j+1th trip; t0 is the time when the vehicle leaves home for the first time; is the total time taken by the vehicle to complete the first j segments of the journey; t stay,j is the length of time the vehicle stays at the destination of the jth segment of the journey; the travel chain model is used to calculate the travel data of family electric vehicles.

3. The method for predicting electric vehicle charging demand based on Markov dynamic path decision according to claim 1 is characterized in that: In S3, Construct a Markov dynamic path optimization model, input the travel data of household electric vehicles and rental electric vehicles into the Markov dynamic path optimization model, and select the optimal path for each electric vehicle. Traffic congestion index I tp , quantitatively describes the operating status of road traffic. The value range of the road traffic congestion index is [0,100]. The larger the value, the more congested the road. The penalty coefficient γ is introduced so that the user's decision is not limited to the shortest path: Among them, γ k,t represents the penalty coefficient of road k at decision time t; is the traffic congestion index of road k at decision time t; I tp,max is the historical maximum traffic congestion index of the reference road section; the steps for each car path decision are as follows: Pre-departure planning: Before departure, the user selects the travel destination and plans a shortest path; the pre-planned shortest path is denoted as π d , each segment of the path is called a reference segment L d , the path set of the unselected road network nodes is selected as the candidate path set, denoted as A k ; Arrival at a decision node: When the user reaches road node i, he enters the decision state and selects reference road segment L d , Traffic congestion index I tp,d and penalty coefficient γ as a reference, for the alternative path set A k For each road section k in , determine whether the following conditions are met: like The road section is regarded as a highly congested road section and removed from the set of alternative routes; Calculate the state transition probability: For each section k in the remaining candidate paths, when calculating its state transition probability, take the length L into consideration k and the penalty coefficient γ k,t Impact: Among them, P(k|i) represents the probability of transferring from the current node i to the alternative path k; L k The length of the alternative path k; γ k,t The penalty coefficient of alternative path k at decision time t; α, β weight factors of path length and congestion penalty, regulating the relative influence of the two on the decision; Select the next path: According to the state transition probability P(k|i), select the next path k with the highest state transition probability among the alternative paths. * , add the final path π * ; Update the current node i to k * , repeat the above steps until reaching the target node.

4. The method for predicting electric vehicle charging demand based on Markov dynamic path decision according to claim 3 is characterized by: In S3, the energy consumption change during the trip is calculated; Build an electric vehicle energy consumption model: Where ΔE is the total power consumption of an electric vehicle during a trip; P m L is the power consumption per km when electric vehicles travel freely in the traffic network; p is the length of the selected path p; P e is the additional power consumption caused by the start and stop of electric vehicles due to traffic congestion per unit time; t e The extra time consumed due to traffic congestion; C EV is the battery capacity of the electric vehicle; considering that the ambient temperature will affect the opening and closing of the electric vehicle air conditioner, the energy consumption model is modified: In the formula, E is the total power consumption of an electric vehicle during a trip; For the temperature T and road grade v k The energy consumption under different temperature conditions is calculated, and the energy consumption change of the electric vehicle during the journey is calculated using the electric vehicle energy consumption model.

5. The method for predicting electric vehicle charging demand based on Markov dynamic path decision according to claim 4 is characterized in that: In S3, According to the air conditioner usage curve, the startup probability of the air conditioner is fitted with a normal distribution to obtain the corresponding normal distribution function of the air conditioner startup as follows: Among them, K open (T) represents the probability of the air conditioner being turned on at temperature T; μ c and δ c are the mean and variance of the heating start parameters respectively; μ h and δ h is the mean and variance of the cooling startup parameters; when the temperature T is between 5 and 21 °C, it is expressed as T c , at 24 to 35°C, it is expressed as T h ; By analyzing the air conditioning energy consumption at different temperatures T and its proportion in the journey, the relationship between temperature and air conditioning energy consumption is as follows: E ac (T)=4.09×10 -6 T 3 +7.28×10 -5 T 2 -6.58×10 -3 T+0.0873 Determine the different temperatures T and road grades v k Energy consumption under temperature in, It represents the energy consumption factor of different road grades at time t, corresponding to the speed of different road grades: v1, v2, v3, representing the driving speed of primary, secondary and tertiary roads respectively; E p Indicates the initial energy consumption per unit mileage of electric vehicles; K oc Indicates the start and stop status of the air conditioner, K oc =1 means the air conditioner is on, K oc =0 means the air conditioner is off: K oc The values ​​of are as follows: Among them, r is a uniformly distributed random number in the interval [0,1], and its purpose is to compare the probability K of the air conditioner being turned on. open (T), and decide whether to turn on the air conditioner through this comparison.

6. The method for predicting electric vehicle charging demand based on Markov dynamic path decision according to claim 1 is characterized in that: In S4, for users of home electric vehicles, the users are classified into demand types, and the charging mode demand results of the users are obtained according to the classified types: By collecting and analyzing users' parking time, SOC, and expected electricity price information, users are classified into ordinary rigid users, emergency rigid users, high anchor elastic users, or low anchor elastic users; If the SOC of the electric vehicle does not meet the next travel demand and the parking time is greater than or equal to 6 hours, it is judged as a normal rigid user, the charging mode is slow charging, and the charging time is instant; If the SOC of the electric vehicle does not meet the next travel demand and the parking time is less than 6 hours, it is judged as an emergency rigid user, the charging mode is fast charging, and the charging time is instant; The SOC of electric vehicles meets the next travel demand, and the user's expected electricity price is higher than the actual price, which is a high anchor elasticity user: If the parking time is greater than or equal to 6 hours, the charging mode is slow charging and the charging time is instant; If the parking time is less than 6 hours, the charging mode is fast charging and the charging time is instant; The SOC of the electric vehicle meets the next travel demand, and the user's expected electricity price is lower than the actual price, which is determined as a low anchor elastic user: the charging mode is slow charging, and the charging time is when the electricity price is low; The charging demand results of household electric vehicle users are obtained based on the classification type.

7. The method for predicting electric vehicle charging demand based on Markov dynamic path decision according to claim 1 is characterized in that: Taxi charging demand: based on the current time t and the actual electricity price P of the charging station c c Calculate the charging cost and evaluate it in combination with the driver's revenue needs: R c is the benefit trade-off index; F t Total revenue expected from running on a full charge; E t The energy increment requirement during charging is the full charge minus the remaining charge; P c is the electricity price of the charging station; The current SOC is sufficient to support the next trip, and the driver will further consider the benefits: when R c >R 阈值 When the driver chooses the corresponding electricity price P c Otherwise, charging will be delayed or other stations will be selected; If the current SOC is insufficient to support the next trip, determine whether the driver's parking time allows slow charging: If the parking time is greater than or equal to 6 hours, the driver will give priority to slow charging based on the electricity price and charging cost, and the charging time is instant; If the parking time is less than 6 hours, the driver chooses fast charging; The charging demand results of rental electric vehicle users are obtained according to the above determination method.

8. An electric vehicle charging demand prediction system based on Markov dynamic path decision, characterized in that: The system is used to execute the electric vehicle charging demand prediction method based on Markov dynamic path decision as described in any one of claims 1 to 7, specifically comprising: a data acquisition module, a travel data construction module, a path optimization module, a charging demand determination module, and a spatiotemporal distribution map generation module; Data collection module: used to collect maps of the target area, vehicle travel origin and destination matrix, daily temperature T, traffic congestion index I tpk,t and electric vehicle parameters; Travel data construction module: used to construct the road-traffic node map of the target area based on the target area map; analyze electric vehicle parameters, divide electric vehicles into household electric vehicles and rental electric vehicles, build a travel chain model for household electric vehicles, and calculate travel time parameters to obtain household electric vehicle travel data; for rental electric vehicles, sample the departure point, destination, start time, and parking time from the vehicle travel start and end point matrix to obtain rental electric vehicle travel data; Path optimization module: used to build a Markov dynamic path optimization model, input the travel data of family electric vehicles and rental electric vehicles into the Markov dynamic path optimization model, select the optimal path for each electric vehicle, and calculate the energy consumption change during the trip to obtain the time series path and energy consumption change data set; Charging demand determination module: used to build a charging demand determination model. For users of home electric vehicles, users are divided into demand types to obtain the charging demand of home electric vehicles; for users of rental electric vehicles, a charging demand selection model is built to obtain the charging demand of rental electric vehicles; Spatiotemporal distribution diagram generation module: used for charging load superposition and node load calculation, superimposing the charging demand of household electric vehicles and taxis on the corresponding road network nodes to calculate the total node charging load, output the spatiotemporal distribution of charging demand, and generate the final spatiotemporal distribution diagram of electric vehicle charging demand based on the charging load data of all nodes.

9. An electric vehicle charging demand prediction device based on Markov dynamic path decision, characterized in that: The method comprises a memory and a processor, wherein the memory is used to store computer program code and transmit the computer program code to the processor; The processor is used to execute the electric vehicle charging demand prediction method based on Markov dynamic path decision as described in any one of claims 1 to 7 according to the instructions in the computer program code.

10. A computer program product, comprising a computer program, characterized in that The computer program is executed by a processor as the electric vehicle charging demand prediction method based on Markov dynamic path decision as described in any one of claims 1 to 7.

Images