A method for electric vehicle charging path planning based on deep learning

Abstract

The present invention provides an electric vehicle charging path planning method based on deep learning, which belongs to the field of electric vehicle path planning. From the perspective of charging station operation, a deep belief network DBN model is used to predict the number of charging stations in a specific area, with the purpose of providing constraints for electric vehicle path planning; from the perspective of user interests, a long short-term memory neural network LSTM model is used to predict the average speed of electric vehicles on a specific road, and the prediction results of the LSTM model are used as the input of the electric vehicle driving energy consumption model, and then an electric vehicle comprehensive energy consumption model is established, and then a total travel time model and a user total cost model of the electric vehicle are established; an objective function with the shortest total travel time, the lowest total user cost, and the best overall total travel time and total user cost is designed, and finally an improved genetic algorithm is used to solve the path planning problem under different objectives.

Description

Technical Field

[0001] The present invention relates to the field of electric vehicle path planning, and in particular to an electric vehicle charging path planning method based on deep learning. Background Art

[0002] Many scholars in the field of electric vehicle path planning research have rarely combined deep learning prediction models to conduct systematic research from the perspectives of electric vehicles, charging stations, and users. In addition, for the problem of electric vehicle path planning, the Dijkstra algorithm is one of the most classic algorithms for solving path planning. This algorithm uses an extended method to traverse the shortest path from the starting point to each point and record it. The path obtained by finally reaching the end point is recorded as the shortest path. However, this traditional path planning algorithm and the single charging path planning goal can no longer meet the needs of actual life. Therefore, it is necessary to study the charging path planning of electric vehicles. Summary of the invention

[0003] In view of the above-mentioned defects, the present invention provides a method for planning charging paths for electric vehicles based on deep learning. The main content of the method is to establish a DBN model after analyzing the factors affecting the number of charging piles at the charging station, and predict the number of charging piles used at the charging station as one of the constraints for path planning. From the perspective of electric vehicles, the main content of the method is to establish a traffic network model, analyze the factors affecting the speed of electric vehicles, establish an LSTM model to predict the speed of electric vehicles, and then establish an electric vehicle driving energy consumption model. From the perspective of user interests, the path planning objective function is designed, and the path planning problem is solved by using an improved genetic algorithm.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A method for planning a charging path for an electric vehicle based on deep learning, comprising:

[0006] Step 1, obtaining the average speed of the electric vehicle, weather conditions, surrounding traffic congestion coefficient, remaining power, starting point, end point, charging station location and charging pile usage information;

[0007] Step 2: Based on the acquired information, the average speed of the electric vehicle at the next moment is predicted by the constructed long short-term memory (LSTM) model. The predicted average speed at the next moment is used as the input of the electric vehicle driving energy consumption model, and the driving energy consumption is calculated by the improved genetic algorithm. The number of charging piles used at the next moment is predicted by constructing a deep belief network (DBN) model, and the charging pile utilization rate is obtained, which provides constraints on the charging pile utilization rate in the improved genetic algorithm.

[0008] Step 3: Based on the results predicted by the LSTM model in step 2, a total travel time model of electric vehicles and a total cost model of electric vehicle users are constructed, and then a charging path planning objective function is constructed;

[0009] Step 4, when electric vehicle users have charging needs, the goals are to minimize the total travel time, minimize the total user cost, and optimize the total travel time and total user cost. Based on the predicted results in the deep belief network DBN model and the set constraints, the user goes from the starting point to the charging station for charging, and then reaches the destination from the charging station. The charging path planning is carried out in sequence through the improved genetic algorithm, and different optimal charging paths are obtained according to the objective function.

[0010] A further improvement of the technical solution of the present invention is that the average vehicle speed prediction method of the long short-term memory (LSTM) model aims to construct an electric vehicle driving energy consumption model, and the specific steps are as follows:

[0011] Step 11, for a certain area of ​​the city, establish a traffic network model to obtain the starting point and destination information of electric vehicles and the location information of charging stations;

[0012] Step 12, obtaining the electric vehicle speed data, weather temperature, traffic congestion coefficient, intersection density and traffic light density data, establishing the input and output of the long short-term memory (LSTM) model, and predicting the average speed of the electric vehicle at the next moment;

[0013] Step 13, based on the average vehicle speed predicted by the long short-term memory (LSTM) in step 12, the average vehicle speed is used as the input of the electric vehicle driving energy consumption model to construct the electric vehicle driving energy consumption model, and the output is the electric vehicle comprehensive energy consumption model; the electric vehicle comprehensive energy consumption model is the sum of the electric vehicle driving energy consumption model, the electric vehicle load equipment energy consumption model, and the electric vehicle battery degradation model; according to the electric vehicle comprehensive energy consumption model, the total travel time and total cost of the electric vehicle user in the objective function of the improved genetic algorithm are obtained.

[0014] A further improvement of the technical solution of the present invention is that the method for predicting the number of charging piles at a charging station based on the deep belief network DBN model aims to provide charging pile utilization rate constraints for path planning, that is, to improve the constraint conditions for path planning in the genetic algorithm. The prediction method steps are as follows:

[0015] Step 21, obtain the information of all charging stations in the traffic network area in step 1. For different charging stations, the charging station information includes the name of the charging station, the total number of charging piles, charging electricity fee, service fee, parking fee, and charging pile power; for a specific charging station, obtain the weather temperature, weather conditions, day type, traffic congestion coefficient every five minutes, and the number of charging piles used at the charging station every five minutes;

[0016] Step 22, using weather temperature, weather conditions, day type, traffic conditions every five minutes, and the number of charging stations used every five minutes as input information for the deep belief network DBN model;

[0017] Step 23, using the deep belief network DBN model to predict the number of charging piles at the charging station at the next moment of the current moment;

[0018] Step 24, calculate the usage rate of the charging pile and set the constraint conditions;

[0019] Step 25, determine whether all charging stations meet the path planning constraints. If so, execute step 4 to perform charging path planning according to different objective functions in the road network containing charging stations that meet the constraints, and obtain different optimal charging paths; if not, do not consider the charging station in the path planning.

[0020] A further improvement of the technical solution of the present invention is that the total travel time model of the electric vehicle is:

[0021] The total travel time model of electric vehicles includes the travel time t 1 , Charging waiting time t 2 , Charging time t 3 ; The distance between the starting point A and the end point B in the road network is called r i,j (n) , n represents different paths and two adjacent nodes in the road network are one path, m represents the number of paths; assuming that at time t, the electric vehicle user has a charging demand, and the driving time of the electric vehicle to reach the destination is t 1 , the travel time unit is h, t 1 The formula is as follows:

[0022]

[0023] In the formula, The average vehicle speed predicted by the LSTM model;

[0024] Charging waiting time t for electric vehicles 2 for:

[0025]

[0026] Where H is the number of charging stations in the road network area; X h is a decision variable. When charging at charging station h is chosen, X h =1, otherwise X h =0;

[0027] It is assumed that when the electric vehicle reaches the charging threshold, it can go to the destination; therefore, the charging time of the electric vehicle is t 3 for:

[0028]

[0029] In the formula, E t represents the threshold; E s Indicates the remaining battery power; P ch Indicates the charging power of the charging pile;

[0030] In summary, the total travel time T of an electric vehicle is:

[0031] T=t 1 +t 2 +t 3

[0032] A further improvement of the technical solution of the present invention is that the total cost model of the electric vehicle user is:

[0033] The total cost model of electric vehicle users includes the actual driving comprehensive energy consumption cost C 1 、Load equipment cost C 2 、Charging cost C 3 、Battery capacity decay cost after one charge C 4 ;

[0034]

[0035]

[0036] C 3 =P ch ·C s ·(t 2 +t 3 )

[0037]

[0038]

[0039] In the formula, C s The electricity price when the electric vehicle is connected to the grid for charging; when the air conditioner in the car is turned on, X 0 is 1, otherwise it is 0; t k Indicates the time of arrival at the charging station;

[0040] In summary, the total cost C of electric vehicle users is:

[0041] C=C 1 +C 2 +C 3 +C 4

[0042] A further improvement of the technical solution of the present invention is that the charging path objective function is:

[0043]

[0044] In the formula, T is the total travel time; C is the total cost of the user; T avg is the average total driving time of electric vehicles to different charging stations; C avg is the average total cost of electric vehicles going to different charging stations; 1 and λ 2 is the weight coefficient. When the optimization goal is to minimize the total travel time, λ 1 =1,λ 2 =0; when the optimization goal is to minimize the total cost of users, λ 1 =0,λ 2 =1; when the optimization goal is to optimize the total travel time and the total cost of the user, λ 1 =1,λ 2 =1.

[0045] Due to the adoption of the above scheme, the technical progress achieved by the present invention is: under the premise of considering the weather, traffic conditions, and the real-time operating status of the charging pile, the route planning goals are the shortest travel time and the best overall user cost. Compared with the route planning goal of taking the total travel time as the route planning goal, the total travel time is reduced; compared with the route planning goal of taking the lowest total user cost as the route planning goal, the total user cost is saved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 It is the LSTM model diagram;

[0048] Figure 2 Training structure for DBN model;

[0049] Figure 3 To improve the genetic algorithm flow chart. DETAILED DESCRIPTION

[0050] The present invention will be further described below:

[0051] The path method steps of the present invention are as follows:

[0052] Step 1, obtaining the average speed of the electric vehicle, weather conditions, surrounding traffic congestion coefficient, remaining power, starting point, end point, charging station location and charging pile usage information;

[0053] Step 2: Based on the information obtained, the long short-term memory (LSTM) model is constructed to predict the average speed of the electric vehicle at the next moment. The predicted average speed at the next moment is used as the input of the electric vehicle driving energy consumption model, and the electric vehicle driving energy consumption is calculated by the improved genetic algorithm. By constructing a deep belief network (DBN) model, the number of charging piles in use at the next moment is predicted, and then the charging pile utilization rate is obtained, which provides constraints on the charging pile utilization rate in the improved genetic algorithm.

[0054] The charging pile utilization rate constraints are as follows:

[0055]

[0056] Step 3: Based on the results predicted by the LSTM model in step 2, the total travel time model of electric vehicles and the total cost model of electric vehicle users are constructed, and then the charging path planning objective function is constructed. The objective function is as follows:

[0057]

[0058] In the formula, T is the total travel time; C is the total cost of the user; T avg is the average total driving time of electric vehicles to different charging stations; C avg is the average total cost of electric vehicles going to different charging stations; 1 and λ 2 is the weight coefficient. When the optimization goal is to minimize the total travel time, λ 1 =1,λ 2 =0; when the optimization goal is to minimize the total cost of users, λ 1 =0,λ 2 =1; when the optimization goal is to optimize the total travel time and the total cost of the user, λ 1 =1,λ 2 =1.

[0059] Step 4: When electric vehicle users have charging needs, the shortest total travel time, the lowest total user cost, and the best overall total travel time and total user cost are used as the objective functions. Based on the predicted results in the deep belief network DBN model, the constraints of the path planning are set. From the starting point to the charging station for charging, and then from the charging station to the destination, the charging path planning is carried out in sequence through the improved genetic algorithm, and different optimal charging paths are obtained according to the objective function.

[0060] The average vehicle speed prediction method of the long short-term memory (LSTM) model aims to build an electric vehicle driving energy consumption model. The specific steps are as follows:

[0061] Step 11, for a certain area of ​​the city, establish a traffic network model to obtain the starting point and destination information of electric vehicles and the location information of charging stations. Use graph theory to establish a directed traffic network model G = (F, R, H,). Among them, F = {1, 2, ..., f} is the set of road network nodes, and f is the total number of road network nodes selected in the area. H = {1, 2, ..., h} is the set of charging station nodes, and h is the total number of charging station nodes selected in the area. R is the set of road section distances between nodes, where the distance between road network nodes (i, j|i, j∈F) is represented by r i,j The distances between the initialization sections are shown below.

[0062]

[0063]

[0064] In the formula, inf means that the road segment formed by node i and node j is not directly connected; d ij is the travel distance between node i and node j.

[0065] Step 12, obtain the electric vehicle speed data, weather temperature, traffic congestion coefficient, intersection density and traffic light density data, establish the input and output of the long short-term memory (LSTM) model, and predict the average speed of the electric vehicle at the next moment.

[0066] The three inputs of the LSTM model to predict the average speed are the input value x of the network at the current moment t , the output value h of LSTM at the previous moment t-1 , the unit state C at the previous moment t-1 ; The two LSTM outputs are the LSTM output values ​​h at the current moment t and the current cell state C t . By the forget gate f t , input gate i t , output gate h t The input X(t) and output h of the LSTM model are 2 (t) are as follows:

[0067]

[0068] In the formula, is the average speed of the electric vehicle at the current moment, T is the weather temperature, B is the traffic congestion coefficient, ρ 1 is the intersection density, ρ 2 is the traffic light density, is the average vehicle speed at the next moment predicted by the LSTM model. 1 is the intersection density, ρ 2The calculation formula for traffic light density is as follows:

[0069]

[0070]

[0071] Where N 1 is the number of forks in the road, N 2 is the number of traffic lights, and L is the length of the road.

[0072] Step 13, based on the average vehicle speed predicted by the long short-term memory (LSTM) in step 12, the average vehicle speed is used as the input of the electric vehicle driving energy consumption model to construct the electric vehicle driving energy consumption model, and the output is the electric vehicle comprehensive energy consumption model; the electric vehicle comprehensive energy consumption model is the sum of the electric vehicle driving energy consumption model, the electric vehicle load equipment energy consumption model, and the electric vehicle battery degradation model; according to the electric vehicle comprehensive energy consumption model, the total travel time and total cost of the electric vehicle user in the objective function of the improved genetic algorithm are obtained.

[0073] The driving energy consumption model of electric vehicles is as follows:

[0074] Electric vehicle mechanical power P 1 for:

[0075]

[0076] In the formula, ρ air is the air density; m is the vehicle mass; ξ is the electric vehicle mass factor; a is the electric vehicle acceleration; g is the gravitational acceleration (take g = 9.8066m / s 2 );θ is the road slope value; c r is the rolling resistance parameter; is the average speed of the electric vehicle predicted by the LSTM model; A is the frontal windward area of ​​the electric vehicle; c d is the friction coefficient.

[0077] Assume that the road slope value of the electric vehicle is 0 during driving, that is, θ = 0°, and ignore the influence of air resistance and slope resistance. Considering the premise of regenerative braking energy recovery, the electric vehicle driving motion equation F 1 for:

[0078] F 1 =F f +F b (9)

[0079] F b =F 2 +F 3 (10)

[0080] F b=αF 2_max +βF 3_max (11)

[0081] Where α is the weighting coefficient of mechanical power; F 2 is the friction braking force; F 2_max is the maximum friction braking force; F 3 is the anti-drag braking force provided by the motor; F f is the rolling resistance; F b is the acceleration resistance; F 3_max is the maximum anti-drag braking force; β is the weighted coefficient of the motor braking force (where 0≤α≤1, 0≤β≤1).

[0082] The kinetic energy generated by the friction braking force of electric vehicles will be converted into heat energy, which will disappear in the air and is not conducive to recycling. 2_max = 0, the driving energy consumption of electric vehicles is derived from the braking equation of the anti-drag braking force as follows:

[0083] The instantaneous power P of an electric vehicle during driving 2 for:

[0084] P 2 (t) = F b ·v(t) (12)

[0085] If braking occurs during driving, the energy consumption caused by braking is:

[0086]

[0087] In the formula, v 0 and v 1 They represent the instantaneous speed at the beginning and end of braking respectively. According to the law of conservation of energy:

[0088] ΔE=∫F 1 ·vdt (14)

[0089] Let E b =∫F b ·vdt (15)

[0090] E f =∫F f vdt = F f ·l (16)

[0091] P 3 =Mω (17)

[0092] In the formula, E b Indicates the kinetic energy generated during braking; E fIt indicates that rolling resistance generates kinetic energy; l indicates braking distance; M indicates motor torque; ω indicates motor angular velocity.

[0093] Assume η 1 is the mechanical transmission efficiency, η 2 is the motor power generation efficiency, η 3 is the battery charging efficiency, then the energy recovered during braking of electric vehicles is E recov for:

[0094] E recov =η 1 η 2 η 3 (ΔE-F f l) (18)

[0095] The energy consumption of an electric car for d kilometers is E drive for:

[0096]

[0097] In summary, the actual electric energy consumed by electric vehicles during driving is E actual for:

[0098] E actual =E drive +E recov (20)

[0099] The energy consumption model of electric vehicle load equipment is as follows:

[0100] The power P required to maintain the current temperature inside the car air for:

[0101] Q 2 =Q a +Q d +Q f (twenty one)

[0102] Q a =δ·A·|T set -T srd | (22)

[0103]

[0104] In the formula, Q 2 is the heat load in the car; Q a is the thermal conduction load of the vehicle; Q d is the ventilation load; Q f To remove Q a and Q d Other air conditioning heat loads other than ; δ is the heat transfer coefficient of electric vehicles; A is the body area of ​​electric vehicles; T setThe temperature set for the air conditioner; T srd is the external ambient temperature of the car; σ is the air conditioning heat load transfer coefficient; τ is the mechanical transmission coefficient of the air conditioning compressor.

[0105] When the electric vehicle air conditioner is set to the temperature T set The amount of electricity consumed by the air conditioner during operation time t air for:

[0106] E air =P air ·t (24)

[0107] Without considering external load equipment, the sound equipment is on section d ij Power consumed for:

[0108]

[0109] The battery capacity degradation model is as follows:

[0110]

[0111] In the formula, SOC represents the battery state of charge; CRR represents the battery capacity retention rate; SOC mean represents the average value of SOC; ΔSOC represents the change of SOC. Then, the equivalent cycle number EFC can be derived as:

[0112]

[0113] Considering that retired batteries can be reused, the cost of each 1% loss of battery capacity is C 1％ for:

[0114]

[0115] In the formula, Q 1 Indicates the total capacity of the battery.

[0116] According to the retired battery price estimation model, the retired battery price p can be estimated u for:

[0117]

[0118] Where, T u Indicates the battery life in seconds; p n Represents the unit price of a new battery, set to 1.5 yuan / wh; r e represents the annual depreciation rate, which is set to 20%; DR represents the discount rate, which is set to 6%.

[0119] Therefore, the battery degradation cost C of an electric vehicle battery charge isbattery It can be calculated by formula (28):

[0120]

[0121] Where N ch is the total number of charge and discharge cycles of the battery; (100-CRR) is the decayed capacity of the battery.

[0122] The process of constructing the electric vehicle total travel time model and the electric vehicle user total cost model is as follows:

[0123] The total travel time model of electric vehicles includes the travel time t 1 , Charging waiting time t 2 , Charging time t 3 The distance between the starting point A and the end point B in the road network is called r i,j (n) , n represents different paths and two adjacent nodes in the road network are one path, and m represents the number of paths. Assuming that at time t, the electric vehicle user has a charging demand, then the driving time t of the electric vehicle to reach the destination 1 , the unit of travel time is h.

[0124]

[0125] In the formula, is the average vehicle speed predicted by the LSTM model.

[0126] Charging waiting time t for electric vehicles 2 for:

[0127]

[0128] Where H is the number of charging stations in the road network area; X h is a decision variable. When charging at charging station h is chosen, X h =1, otherwise X h =0.

[0129] It is assumed that when the electric vehicle reaches the charging threshold, it can go to the destination. Therefore, the charging time of the electric vehicle is t 3 for:

[0130]

[0131] In the formula, E t Indicates the threshold value; E s Indicates the remaining battery power; P ch Indicates the charging power of the charging pile.

[0132] In summary, the total travel time T of an electric vehicle is:

[0133] T=t 1 +t 2 +t 3 (34)

[0134] The total cost model of electric vehicle users includes the actual driving comprehensive energy consumption cost C 1 、Load equipment cost C 2 、Charging cost C 3 、Battery capacity decay cost after one charge C 4 .

[0135]

[0136]

[0137] C 3 =P ch ·C s ·(t 2 +t 3 ) (37)

[0138]

[0139]

[0140] In the formula, C s The electricity price when the electric vehicle is connected to the grid for charging; when the air conditioner in the car is turned on, X 0 is 1, otherwise it is 0; t k Indicates the time of arrival at the charging station.

[0141] In summary, the total cost C of electric vehicle users is:

[0142] C=C 1 +C 2 +C 3 +C 4 (40)

[0143] The deep belief network DBN model is used to predict the number of charging piles at a charging station. Its purpose is to provide charging pile utilization rate constraints for path planning, that is, to improve the constraints for path optimization in the genetic algorithm. The prediction method steps are as follows.

[0144] Step 21, obtain the information of all charging stations in the traffic network area of ​​step 1. For different charging stations, the charging station information includes the name of the charging station, the total number of charging piles, charging electricity charges, service fees, parking fees, and charging pile power; for a specific charging station, obtain the weather temperature, weather conditions, day type, traffic congestion coefficient every five minutes, and the number of charging piles used at the charging station every five minutes.

[0145] Step 22, weather temperature, weather conditions, day type, traffic conditions every five minutes, and the number of charging piles used at each charging station every five minutes are used as input information for the deep belief network DBN model; four characteristic data that affect the number of charging piles used are selected as data for the DBN model input layer for model training, as shown in the following table:

[0146]

[0147] Using the improved genetic algorithm for path planning, first input the EV's starting position and remaining power, determine whether there is a charging station within the remaining mileage of the EV, and if there is a charging station within the remaining mileage, determine whether the charging pile utilization rate is less than 90%. If it is less than 90%, initialize the algorithm parameters, encode the chromosomes of the traffic network nodes and EV vehicles, and plan the best path according to different objective functions. The specific process is as follows Figure 3 shown.

[0148] Step 23, using the deep belief network DBN model to predict the number of charging piles at the charging station at the next moment from the current moment;

[0149] Step 24, calculate the usage rate of the charging pile and set constraints;

[0150] Step 25, determine whether all charging stations meet the path planning constraints. If so, execute step 4 to obtain different optimal charging paths according to different objective functions in the road network containing charging stations that meet the constraints by improving the genetic algorithm; if not, do not consider the charging station in the path planning.

[0151] The embodiments described above are only descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the protection scope determined by the claims of the device of the present invention.

Claims

1. A method for electric vehicle charging path planning based on deep learning, It is characterized in that include: Step 1, obtaining the average speed of the electric vehicle, weather conditions, surrounding traffic congestion coefficient, remaining power, starting point, end point, charging station location and charging pile usage information; Step 2: Based on the acquired information, the average speed of the electric vehicle at the next moment is predicted by the constructed long short-term memory (LSTM) model. The predicted average speed at the next moment is used as the input of the electric vehicle driving energy consumption model, and the driving energy consumption of the electric vehicle is calculated by the improved genetic algorithm. The number of charging piles used at the next moment is predicted by constructing a deep belief network (DBN) model, and the charging pile utilization rate is obtained, which provides constraints on the charging pile utilization rate in the improved genetic algorithm. Step 3: Based on the results predicted by the LSTM model in step 2, a total travel time model of electric vehicles and a total cost model of electric vehicle users are constructed, and then a charging path planning objective function is constructed; Step 4: When electric vehicle users have charging needs, the shortest total travel time, the lowest total user cost, and the best overall total travel time and total user cost are used as objective functions, and the constraints of the path planning are set based on the predicted results of the deep belief network DBN model; From the starting point to the charging station for charging, and then from the charging station to the destination, the charging path planning is carried out in sequence through the improved genetic algorithm, and different optimal charging paths are obtained according to the objective function.

2. According to the deep learning-based electric vehicle charging path planning method of claim 1, It is characterized in that The purpose of the average vehicle speed prediction method of the long short-term memory (LSTM) model is to construct an electric vehicle driving energy consumption model. The specific steps are as follows: Step 11, for a certain area of ​​the city, establish a traffic network model to obtain the starting point and destination information of electric vehicles and the location information of charging stations; Step 12, obtaining the electric vehicle speed data, weather temperature, traffic congestion coefficient, intersection density and traffic light density data, establishing the input and output of the long short-term memory (LSTM) model, and predicting the average speed of the electric vehicle at the next moment; Step 13, based on the average vehicle speed predicted by the long short-term memory (LSTM) in step 12, the average vehicle speed is used as the input of the electric vehicle driving energy consumption model to construct the electric vehicle driving energy consumption model, and the output is the electric vehicle comprehensive energy consumption model; the electric vehicle comprehensive energy consumption model is the sum of the electric vehicle driving energy consumption model, the electric vehicle load equipment energy consumption model, and the electric vehicle battery degradation model; according to the electric vehicle comprehensive energy consumption model, the total travel time and total cost of the electric vehicle user in the objective function of the improved genetic algorithm are obtained.

3. According to the deep learning-based electric vehicle charging path planning method of claim 1, It is characterized in that The purpose of the method for predicting the number of charging piles in a charging station using the deep belief network DBN model is to provide charging pile utilization rate constraints for path planning, that is, to improve the constraint conditions for path optimization in the genetic algorithm. The steps of the prediction method are as follows: Step 21: Obtain all the charging station information in the traffic road network area of Step 1. For different charging stations, the charging station information includes the charging station name, total number of charging piles, charging electricity fee, service fee, parking fee, and charging pile power. For a specific charging station, obtain the weather temperature, weather condition, day type, traffic congestion coefficient at every five-minute moment, and the number of charging piles in use at the charging station at every five-minute moment. Step 22: Use the weather temperature, weather condition, day type, traffic condition at every five-minute moment, and the number of charging piles in use at the charging station at every five-minute moment as the input information of the Deep Belief Network (DBN) model. Step 23: Use the Deep Belief Network (DBN) model to predict the number of charging piles at the charging station at the next moment of the current moment. Step 24: Calculate the charging pile utilization rate and set the constraint conditions. Step 25: Determine whether all charging stations meet the path planning constraint conditions. If they meet, execute Step 4. Through the improved genetic algorithm, obtain different optimal charging paths according to different objective functions in the charging station road network that meets the constraint conditions. If not, do not consider this charging station in the path planning.

4. A method for planning an electric vehicle charging path based on deep learning according to claim 1, characterized in that, the total travel time model of the electric vehicle is: The total travel time model of electric vehicles includes the travel time t 1 , Charging waiting time t 2 , Charging time t 3 ; The distance between the starting point A and the end point B in the road network is called r i,j (n) , n represents different paths and two adjacent nodes in the road network are one path, m represents the number of paths; assuming that at time t, the electric vehicle user has a charging demand, and the driving time of the electric vehicle to reach the destination is t 1 , the travel time unit is h, t 1 The formula is as follows: In the formula, The average vehicle speed predicted by the LSTM model; Charging waiting time t for electric vehicles 2 for: Where H is the number of charging stations in the road network area; X h is a decision variable. When charging at charging station h is chosen, X h =1, otherwise X h =0; It is assumed that when the electric vehicle reaches the charging threshold, it can go to the destination; therefore, the charging time of the electric vehicle is t 3 for: In the formula, E t Indicates the threshold value; E s Indicates the remaining battery power; P ch Indicates the charging power of the charging pile; In summary, the total travel time T of the electric vehicle is: T=t 1 +t 2 +t 3 。 5. A method for planning an electric vehicle charging path based on deep learning according to claim 1, characterized in that, the total cost model for electric vehicle users is: The total cost model of electric vehicle users includes the actual driving comprehensive energy consumption cost C 1 、Load equipment cost C 2 、Charging cost C 3 、Battery capacity decay cost after one charge C 4 C 3 = P ch · C s · (t 2 + t 3 ) Where Cs is the electricity price when the electric vehicle is connected to the grid for charging; when the air conditioner in the car is turned on, X0 is 1, otherwise it is 0; t k Indicates the time of arrival at the charging station; In summary, the total cost C for electric vehicle users is: C=C 1 +C 2 +C 3 +C 4 。 6. A method for planning an electric vehicle charging path based on deep learning according to claim 1, characterized in that, the objective function for the charging path planning is: In the formula, T is the total travel time; C is the total cost of the user; T avg is the average total driving time of electric vehicles to different charging stations; C avg is the average total cost of electric vehicles going to different charging stations; 1 and λ 2 is the weight coefficient. When the optimization goal is to minimize the total travel time, λ 1 =1,λ 2 =0; when the optimization goal is to minimize the total cost of users, λ 1 =0,λ 2 =1; when the optimization goal is to optimize the total travel time and the total cost of the user, λ 1 =1,λ 2 =1.

Images