Electric vehicle charging decision-making method considering individual demands of users

Abstract

The invention discloses an electric vehicle charging decision-making method considering user personalized demands. The method comprises the following steps: processing traffic network data and charging station data by using grid sampling and a cellular automaton model, constructing the cellular automaton model, and setting functional area labels; the method comprises the following steps: constructing a driving speed model which is constructed by considering the travel jerky degree of a user and the speed limit of a functional area where the user is located, a charging cost model which is constructed by considering the time cost and the expense cost, and a charging budget model which is constructed by considering the travel purposes of different users; a personalized demand model of the electric vehicle, which is a charging satisfaction model considering a marginal benefit decline principle; building a CA-LSTM integrated model to predict the traffic flow on the road; and acquiring an optimal travel path by using an improved A * algorithm, simulating the optimal path for multiple times in combination with CA-LSTM to acquire a travel risk, converting the travel risk into a risk of the power supply scheme, constructing an improved foreground theory decision model to analyze all the schemes, and finally obtaining an optimal power supply scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle charging and swapping decision-making, and particularly relates to an electric vehicle charging and swapping decision-making method considering user personalized needs. Background Art

[0002] In recent years, due to their great advantages in energy conservation and emission reduction, electric vehicles (EVs) have been widely used. The global sales volume of electric vehicles shows an irresistible trend, and the publicly available charging points have also increased rapidly. At present, the energy replenishment of electric vehicles in urban areas mainly adopts two methods, namely, the fast charging method of going to charging stations (FCST) and the battery swapping method of going to battery swapping stations (BSST).

[0003] However, due to the random travel behaviors of more and more users and different power supply methods, the charging load has the problems of uncertainty and peak values in a short period of time. At the same time, when users search for charging stations, problems such as high travel costs, low charging efficiency, long waiting times, and low satisfaction have attracted the close attention of users, especially during the charging peak period. The existing charging demand prediction methods considering travel behaviors focus on the influence of heterogeneity among vehicles, stations, and road networks, but ignore the influence of user personalized needs on travel and the differences in traffic conditions in each functional area, and do not quantify user personalized needs, resulting in great difficulty in charging demand. The existing decision-making methods build a rational assumption path planning model of the Expected Utility Theory (EUT), ignoring their irrational decision-making behaviors and risk attitudes, and at the same time do not analyze the different impacts of charging and swapping modes on decision-making. Therefore, it is crucial to propose an effective charging and swapping decision-making scheme while taking into account factors such as traffic jams, charging costs, and personalized needs that affect decision-making on the premise of ensuring that users complete their charging demands. Summary of the Invention

[0004] 1. Technical problems to be solved:

[0005] In view of the above technical problems, the present invention provides an electric vehicle charging and swapping decision-making method considering user personalized needs, which realizes the model coupling of traffic road networks, charging stations, and electric vehicles through cellular automata, takes into account the influence of urban functional zoning and user travel personalized needs on charging demand prediction, and takes into account the bounded rational decision-making behaviors of users and their risk attitudes towards decision-making schemes, so as to maximize the satisfaction of electric vehicles and help electric vehicle users select an energy acquisition mode for energy replenishment and a driving route.

[0006] 2. Technical solutions:

[0007] An electric vehicle charging and swapping decision-making method considering users' personalized needs, characterized in that it includes:

[0008] Step 1: Use grid sampling and cellular automata model to process traffic road network data and charging station data to construct a cellular automata model; combine the POI data of the traffic network to set functional area labels for the cells in the model;

[0009] Step 2: Construct a personalized demand model for electric vehicles based on the constructed cellular automata model; the personalized demand model includes a driving speed model built considering the user's travel anxiety level and the speed limit in the functional area, a charging and swapping cost model built considering time cost and cost, a charging and swapping budget model built considering different users' travel purposes, and a charging and swapping satisfaction model considering the principle of diminishing marginal benefit;

[0010] Step 3: Based on the constructed cellular automata model, build a CA-LSTM integrated model to predict the future traffic flow on the road; in the CA-LSTM integrated model, the cellular automata CA is used as the upper framework to express the interaction between cells on the spatial road network, and LSTM is used as the bottom layer to process the change of vehicle data in cells at different times, and traffic flow prediction is realized through the simulation of the research area;

[0011] Step 4: Based on the constructed personalized demand model, use the improved A* algorithm to obtain the optimal travel path with the lowest travel cost, combine the CA-LSTM integrated model to simulate the optimal path multiple times to obtain its corresponding travel risk and convert it into the risk of the corresponding power replenishment plan; use the improved prospect theory decision-making model to analyze all charging and swapping plans, and finally obtain the optimal charging and swapping plan.

[0012] Further, step 1 specifically includes:

[0013] S11: Construct a cellular automata model; divide the traffic network data of the area to be decided by grids; where the traffic network of the area to be decided is the cellular space; each grid is used as a cell of the cellular space; define the state of each cell according to whether there is an electric vehicle in each cell; mark the cell containing the charging station in the cellular space as a special cell; use the following formula to describe the road driving rule of the cell with speed:

[0014]

[0015] In the above formula, v n (t) represents the speed of the nth vehicle at time t; v max represents the maximum speed of the electric vehicle; gap n (t) represents the distance between the nth vehicle and the vehicle immediately in front at time t; △t represents the unit time;

[0016] S12: Conduct a functional area analysis on the cells; obtain the POI data on each cell based on the POI data in the electronic map. After preprocessing through longitude and latitude import and projection, the following formula is used to calculate according to the frequencies of different types of POI data appearing in the corresponding cells, and then obtain the weights of different types of POI data in the corresponding cells:

[0017]

[0018] In the above formula, n i,j represents the number of times the i-th type of POI data appears in cell j; represents the total number of POI data appearing in cell j; |E| represents the number of cells in the cell space; |{j∈E}| represents the number of cells containing the i-th type of POI data; tfidf i,j represents the weight of the i-th type of POI data in cell j;

[0019] S13: Determine the functional area label of each cell; according to the weight and frequency of each POI data in the cell where it is located, determine the functional area label corresponding to each cell, as shown in formula (3):

[0020]

[0021] In the above formula, D i,j is the weight of the i-th type of POI data in cell j; as shown in the above formula, the POI category with the largest weight in cell j is taken as the functional area type D j .

[0022] Furthermore, the driving speed model established in step two considering the user's travel anxiety level and the speed limit in the functional area is to model the electric vehicle in combination with its travel anxiety level, speed preference, and the speed limit in the functional area. The specific construction process includes:

[0023] S21: Define the minimum speed limit in each functional area as v min and the maximum speed limit as v max ; take the anxiety level as the personalized feature of the electric vehicle; since users with different anxiety levels have different preferences for driving speed, the maximum speed limit for users with a lower anxiety level is defined as v' max , where v' max < v max ; and most users prefer higher speeds during actual driving. Assume that the actual maximum speed limit v' max in each functional area conforms to a skewed distribution, and the probability density function of this skewed distribution is shown in the following formula:

[0024]

[0025] In the above formula: σ - is the lower standard deviation of the functional area corresponding to the cell where the vehicle is preset, and is negatively correlated with the anxiety level; σ + is the upper standard deviation of the functional area corresponding to the cell where the vehicle is preset, and is positively correlated with the anxiety level; the variable σ + and σ - are judged according to the historical driving statistical records of the vehicle;

[0026] S22: Divide the vehicles driving in the functional area into two categories. One category is for users with a lower anxiety level and a speed range of [v min , v' max . For this type of user, the actual maximum speed v' max is less than the maximum speed limit v max of this functional area; the other category is for users with a stronger anxiety level and a speed range of [v min , v max . For this type of user, the maximum speed is the maximum speed limit v max of this functional area; users with different anxiety levels have different speeds in the same functional area, and the preferred driving speeds of electric vehicles are different.

[0027] Furthermore, the charging and swapping cost model constructed by considering the time cost and the cost cost in step two is to consider the time cost and the cost cost in the charging and swapping process, judge the time benefit according to the value of the electric vehicle, and construct the charging and swapping cost model in the same dimension for the time cost and the cost cost, specifically including:

[0028] S23: Model the charging cost and the swapping cost separately;

[0029] Among them, the charging cost of the electric vehicle choosing to charge is as follows:

[0030]

[0031] In the above formula: C c,exp represents the cost of the i-th electric vehicle choosing the charging mode; c cha (t) is the real-time electricity price at different times; SOC i,j is the battery power of the i-th electric vehicle after driving to the j-th cell;

[0032] The swapping cost of the electric vehicle choosing to swap is as follows:

[0033]

[0034] In the above formula: C s,exp represents the swapping cost; c bat is the cost required to purchase a fully charged battery; c serService fees required for installing the battery; c sub Price subsidy for battery swapping;

[0035] S24: Model the charging time and battery swapping time separately;

[0036] The queuing rule for charging or battery swapping follows the first-come, first-served rule; and the number of batteries in the battery swapping station meets the user's needs; assume there are N k charging piles or battery swapping bins in the fast charging station or battery swapping station; the k-th fast charging station or battery swapping station has electric household vehicles; if the electric vehicle chooses the charging method, the charging time of the i-th electric vehicle in the k-th charging station is the sum of the service charging duration and the waiting duration ;

[0037]

[0038] Among them, the service charging duration is shown as follows:

[0039]

[0040] In the above formula: C is the rated battery capacity of the electric vehicle; P c is the charging power during charging; η is the charging efficiency of charging;

[0041] The waiting duration is as follows:

[0042]

[0043] In the above formula: represents rounding down; represents the total number of vehicles queuing for charging at the k-th station, S represents the SOC set of the vehicles at the station arranged in descending order of SOC; the queuing time of the i-th electric vehicle waiting for charging in the k-th charging station depends on the number of electric vehicles waiting in line. If indicates that there are idle charging piles in the fast charging station, then the electric vehicle does not need to wait;

[0044] If the electric vehicle chooses the battery swapping method, the time t ser,s for the i-th electric vehicle to replace the battery is fixed. The battery swapping time of the i-th electric vehicle in the k-th battery swapping station in the battery swapping mode

[0045]

[0046] S25: Construct a time value model for the vehicle based on the market price of the electric vehicle, as follows:

[0047] λ i = M·δ i (11)

[0048] In the above formula: λ i is the time value of the i-th electric vehicle user, which is preset by the owner of the electric vehicle according to his specific situation in advance; M is the market price of the electric vehicle, and the unit is in millions of yuan; δ i is the anxiety level of the electric vehicle, which is preset by the owner of the electric vehicle according to his specific situation in advance.

[0049] Furthermore, the charging and swapping budget model established by considering the travel purposes of different electric vehicles in step two is as follows:

[0050]

[0051] In the above formula: C i,s is the travel cost of the i-th electric vehicle; t i is the time spent during the driving process of the i-th electric vehicle; M i is the vehicle model price of the i-th electric vehicle, and the unit is: millions of yuan; δ i is the anxiety level of the i-th electric vehicle, η i is a binary quantity, η i = 1 indicates that the i-th electric vehicle selects the swapping mode, η i = 0 indicates that the i-th electric vehicle selects the charging mode.

[0052] Furthermore, the charging and swapping satisfaction model considering the principle of diminishing marginal benefit in step two is to construct the satisfaction of charging and swapping decision-making by using the satisfaction Richards function model, as follows:

[0053]

[0054] In the above formula, r represents the satisfaction rate coefficient of the user; C i represents the expected spending price of the user for charging and swapping; C r represents the maximum intolerable cost of the user.

[0055] Further, the CA-LSTM integrated model built in step three combines the long short-term memory network LSTM with the cellular automaton CA; the nth cell where the electric vehicle travels is used as the exploration cell, and the time-series data of the traffic flow on the exploration cell and its neighbor cells before the moment t is extracted. The extracted time-series data is divided into multiple time-series data according to the time series order, and the time-series data with a preset time step is used as an input sample; the input sample is input into the input layer of the LSTM, and the memory unit in the hidden layer of the LSTM captures the characteristics of the traffic flow data on the exploration cell; at each time step, the hidden layer calculates based on the current input sample and the hidden state of the previous moment, and controls the flow and update of information through the input gate, forget gate, and output gate. The output layer predicts the vehicle change data on the cell at the next time step t+1; finally, the probability k that there is traffic flow on the exploration cell at the moment t+1 is calculated.

[0056] Further, step four specifically includes:

[0057] S41: With the goal of minimizing the travel cost, use the improved A* algorithm to obtain the optimal travel path; in the A* algorithm, each searched cell is first evaluated to obtain the best cell, and then search from this cell until the goal; when exploring the nth cell to the target, record the cost required to travel from the starting point to the nth cell and the cost required for the nth cell to travel to the end point, and judge whether there is a vehicle on the nth cell. If there is, add the congestion waiting cost generated by the vehicle; the dynamic measurement heuristic of the specific improved A* search algorithm is as follows

[0058]

[0059] In the above formula: g represents the distance of the traffic network contained in the corresponding exploration cell; v n represents the speed at the nth cell, following the road driving rules in formula (1); h n represents the Euclidean distance between the nth cell and the target cell; represents the preferred speed of different users in different functional areas; w is the evaluation weight coefficient, the closer to the target, the larger the unit space, and vice versa; k represents the possibility that there is a vehicle at the nth exploration cell when traveling to it; σ represents the time lost by one vehicle; f(n) represents the value of node n;

[0060] S42: Construct an improved prospect theory decision-making model; in this model, a value function is constructed using the charging and swapping budget of the electric vehicle and the travel cost under decision simulation. The value function of this decision-making model is shown as follows:

[0061]

[0062] In the above formula, α and β represent the sensitivity of electric vehicles' attention to the risks of charging and swapping, and their values are both preset; λ represents the loss aversion coefficient, and this value is also preset; C s is the travel cost under decision simulation, which is the actual simulated travel cost from the starting point to the end point; C i is the travel budget cost of the electric vehicle, which is the travel cost obtained from the straight-line distance from the starting point to the end point;

[0063] S43: Due to the uncertainty of the travel road conditions of each electric vehicle, there is a random slowdown rule. Therefore, for an electric vehicle making a decision, the specific time it takes to drive to the destination is uncertain. The CA-LSTM integrated model is used to perform multiple travel simulations on the optimal travel plan obtained by the decision-making electric vehicle to obtain the travel time and its probability, that is, the risk probability of this plan; assume that the electric vehicle has m + n + 1 possible arrival times, C s(-m) <···<C i <···<C s(n) , and their occurrence probabilities are p -m ···p n , which are represented by C s =(C s(-m) ···C s(n) ) and P=(p -m ···p n ); then the cumulative decision weight function is shown as follows:

[0064]

[0065] In the above formula: and are the cumulative decision weight values when facing gains and losses respectively, and their calculation methods are represented by the functions ω + (·) and ω - (·) respectively, as shown in the following formula:

[0066]

[0067] In the above formula, γ and δ are the probability weight coefficients when the vehicle owner faces gains or losses respectively, and their preset values are: γ﹦0.61, δ﹦0.69;

[0068] Furthermore, calculate the cumulative prospect value of different electric vehicles for charging decision-making, as shown in the following formula:

[0069]

[0070] In the above formula, CPV j + represents the positive cumulative prospect value under plan j; CPV j－ Denote the negative cumulative prospect value under scenario j; CPV j Denote the actual comprehensive cumulative prospect value under scenario j, where j = 1, 2, 3...;

[0071] When making the decision of charging or swapping batteries, the electric vehicle selects the charging and swapping battery scenario with the largest comprehensive cumulative prospect value as the optimal charging and swapping battery choice, as shown in the following formula:

[0072] CPV i = MAX(CPV i,1 , CPV i,2 ,..., CPV i,j ), j ∈ {1, 2, 3,...} (20)

[0073] In the above formula, CPV denotes the cumulative prospect value.

[0074] 3. Beneficial effects:

[0075] (1) A method for making charging and swapping battery decisions for electric vehicles considering user personalized needs provided by the present invention uses a cellular automaton model to process the traffic network and combines POI data to achieve refined functional zoning of urban areas. It takes into account the traffic differences between functional areas and the distribution of user origin and destination points, thereby making the travel behavior of users more realistic and reducing the error in predicting the power replenishment demand of electric vehicle users.

[0076] (2) A method for making charging and swapping battery decisions for electric vehicles considering user personalized needs provided by the present invention fully considers the impact of user personalized travel needs on charging demand prediction, and combines the driving speed and power replenishment cost of electric vehicles to model, establishing the calculation of different user travel differences and different dimension costs. And the power replenishment budget and satisfaction of users are quantitatively modeled to reflect the different personalized needs among users, which can improve the pertinence of the power replenishment decision of electric vehicles.

[0077] (3) A method for making charging and swapping battery decisions for electric vehicles considering user personalized needs provided by the present invention takes into account the bounded rational decision-making behavior of users and their risk attitudes towards decision-making scenarios, and proposes a unified decision-making method for selecting paths and power replenishment methods, which can improve the power replenishment satisfaction of electric vehicles. In order to verify the practicability of this method, based on the traffic network of a certain city, the data of the comparison results of path and power replenishment mode selection under different scenarios show that the simulation results verify the effectiveness of the proposed decision-making method. Description of the Drawings

[0078] Figure 1 It is a comparison schematic diagram of the cost difference between the charging and swapping battery modes involved in this method;

[0079] Figure 2Schematic diagram of building a CA-LSTM integrated model in this method;

[0080] Figure 3 Distribution map of functional areas of urban traffic road network in specific embodiments;

[0081] Figure 4 Real-time charging price curve graph of charging and swapping stations in specific embodiments;

[0082] Figure 5 Route selection map of electric vehicles in different scenarios in specific embodiments;

[0083] Figure 6 Overall flowchart of this method. Specific implementation manners

[0084] The present invention will be specifically described below with reference to the accompanying drawings.

[0085] As shown in the Figure 6 accompanying drawings, an electric vehicle charging and swapping decision-making method considering user personalized needs is characterized by comprising:

[0086] Step 1: Process traffic road network data and charging station data using grid sampling and cellular automata model to construct a cellular automata model; set functional area labels for cells in the model in combination with POI data of the traffic network;

[0087] Step 2: Construct a personalized demand model for electric vehicles based on the constructed cellular automata model; the personalized demand model includes a driving speed model built considering the user's travel anxiety level and speed limits in the functional area, a charging and swapping cost model built considering time cost and cost cost, a charging and swapping budget model built considering different users' travel purposes, and a charging and swapping satisfaction model considering the principle of diminishing marginal benefit;

[0088] Step 3: Based on the constructed cellular automata model, build a CA-LSTM integrated model to predict future traffic flow on the road; in the CA-LSTM integrated model, the cellular automata CA is used as the upper framework to express the interaction between cells on the spatial road network, and LSTM is used as the bottom layer to process the change of vehicle data in cells at different times, and traffic flow prediction is realized through simulation of the research area;

[0089] Step 4: Based on the constructed personalized demand model, use the improved A* algorithm to obtain the optimal travel path with the lowest travel cost, combine the CA-LSTM integrated model to simulate the optimal path multiple times to obtain its corresponding travel risk and convert it into the risk of the corresponding power replenishment plan; use the improved prospect theory decision-making model to analyze all charging and swapping plans, and finally obtain the optimal charging and swapping plan.

[0090] Further, step one specifically includes:

[0091] S11: Construct a cellular automaton model; divide the traffic network data of the area to be decided using grids; where the traffic network of the area to be decided is the cellular space; each grid serves as a cell of the cellular space; define the state of each cell according to whether there is an electric vehicle in each cell; mark the cells containing charging stations in the cellular space as special cells; use the following formula to describe the road driving rules of the cells with speed:

[0092]

[0093] In the above formula, v n (t) represents the speed of the nth vehicle at time t; v max represents the maximum speed of the electric vehicle; gap n (t) represents the distance between the nth vehicle and the vehicle immediately in front at time t; △t represents the unit time.

[0094] S12: Conduct a functional area analysis on the cells; obtain the POI data on each cell based on the POI data in the electronic map. After import and projection preprocessing of the longitude and latitude, use the following formula to calculate according to the frequency of different types of POI data appearing in the corresponding cells, and then obtain the weights of different types of POI data in the corresponding cells:

[0095]

[0096] In the above formula, n i,j represents the number of times the ith type of POI data appears in cell j; represents the total number of times POI data appears in cell j; |E| represents the number of cells in the cellular space; |{j∈E}| represents the number of cells containing the ith type of POI data; tfidf i,j represents the weight of the ith type of POI data in cell j.

[0097] S13: Determine the functional area label of each cell; determine the functional area label corresponding to each cell according to the weight and frequency of each POI data in the cell where it is located, as shown in formula (3):

[0098]

[0099] In the above formula, D i,j is the weight of the ith type of POI data in cell j; as shown in the above formula, take the POI category with the largest weight in cell j as the functional area type D j .

[0100] Further, the driving speed model established in step two by considering the user's travel anxiety level and the speed limit in the functional area is to model the electric vehicle by combining its travel anxiety level, speed preference, and the speed limit in the functional area. The specific construction process includes:

[0101] S21: Define the minimum speed limit for each functional area as v min and the maximum speed limit as v max ; Use the anxiety level as the personalized feature of the electric vehicle. Since users with different anxiety levels have different preferences for driving speed, the maximum speed limit for users with a lower anxiety level is defined as v' max , where v' max < v max ; And most users prefer higher speeds during actual driving. Assume that the actual maximum speed limit v' max in each functional area conforms to a skewed distribution. Then the probability density function of this skewed distribution is shown as follows:

[0102]

[0103] In the above formula: σ - is the lower standard deviation of the functional area corresponding to the cell where the vehicle is located, which is negatively correlated with the anxiety level; σ + is the upper standard deviation of the functional area corresponding to the cell where the vehicle is located, which is positively correlated with the anxiety level; The values of variables σ + and σ - are judged according to the historical driving statistical records of this vehicle;

[0104] S22: Divide the vehicles driving in the functional area into two categories. One category is users with a lower anxiety level, and their speed range is [v min , v' max . The actual maximum speed limit v' max of this type of user is less than the maximum speed limit v max of this functional area; The other category is users with a stronger anxiety level, and their speed range is [v min , v max . The maximum speed limit of this type of user is the maximum speed limit v max of this functional area; Users with different anxiety levels have different speed limits in the same functional area, and the driving speeds preferred by electric vehicles are different.

[0105] Further, the charging and swapping cost model established in step two by considering the time cost and cost is to consider the time cost and cost in the charging and swapping process, judge the time benefit according to the value of the electric vehicle, and construct the time cost and cost into a charging and swapping cost model in the same dimension, specifically including:

[0106] S23: Model the charging cost and the swapping cost separately;

[0107] Among them, the charging cost of an electric vehicle choosing charging is as follows:

[0108]

[0109] In the above formula: C c,exp represents the cost of the i-th electric vehicle choosing the charging mode; c cha (t) is the real-time electricity price at different times; SOC i,j is the battery power of the i-th electric vehicle after driving to the j-th cell;

[0110] The cost of battery swapping for an electric vehicle choosing battery swapping is as follows:

[0111]

[0112] In the above formula: C s,exp represents the cost of battery swapping; c bat is the cost required to purchase a fully charged battery; c ser is the service cost required to install the battery; c sub is the electricity price subsidy for battery swapping;

[0113] S24: Model the charging time and battery swapping time respectively;

[0114] The queuing rule for charging or battery swapping follows the first-come, first-served rule; and the number of batteries in the battery swapping station meets the user's needs; assume that there are N k charging piles or battery swapping bins in the fast charging station or battery swapping station; the k-th fast charging station or battery swapping station has electric household vehicles; if the electric vehicle chooses the charging method, the charging time of the i-th electric vehicle in the k-th charging station is the sum of the service charging duration and the waiting duration ;

[0115]

[0116] Among them, the service charging duration is as follows:

[0117]

[0118] In the above formula: C is the rated battery capacity of the electric vehicle; P c is the charging power during charging; η is the charging efficiency of charging;

[0119] The waiting duration is as follows:

[0120]

[0121] In the above formula: represents rounding down; represents the total number of vehicles queuing for charging at the k-th charging station, S represents the set of SOCs of the vehicles at the station arranged in descending order of SOC; the queuing time of the i-th electric vehicle waiting for charging at the k-th charging station depends on the number of electric vehicles queuing. If indicates that there are idle charging piles at the fast charging station, then this electric vehicle does not need to wait;

[0122] If the electric vehicle chooses the battery swapping method, the time t for the i-th electric vehicle to swap the battery ser,s is fixed. The battery swapping time of the i-th electric vehicle in the battery swapping mode at the k-th battery swapping station is as shown in the following formula:

[0123]

[0124] S25: Construct a time value model of the vehicle according to the market price of the electric vehicle, as shown in the following formula:

[0125] λ i = M·δ i (11)

[0126] In the above formula: λ i is the time value of the i-th electric vehicle user, which is preset by the owner of the electric vehicle according to his specific situation in advance; M is the market price of this electric vehicle, and the unit is in millions of yuan; δ i is the anxiety level of this electric vehicle, which is preset by the owner of the electric vehicle according to his specific situation in advance.

[0127] Furthermore, the charging and battery swapping budget model established by considering the travel purposes of different electric vehicles in step two is as follows:

[0128]

[0129] In the above formula: C i,s is the travel cost of the i-th electric vehicle; t i is the time spent by the i-th electric vehicle during the driving process; M i is the model price of the i-th electric vehicle, and the unit is: millions of yuan; δ i is the anxiety level of the i-th electric vehicle, η i is a binary quantity, η i = 1 indicates that the i-th electric vehicle chooses the battery swapping mode, η i = 0 indicates that the i-th electric vehicle chooses the charging mode.

[0130] As attached Figure 1Schematic diagram for comparing the cost differences between the charging and battery swapping modes involved in this method.

[0131] Furthermore, the charging and swapping satisfaction model considering the principle of diminishing marginal benefits in step two is to construct the satisfaction of charging and swapping decisions using the satisfaction Richards function model, as follows:

[0132]

[0133] In the above formula, r represents the satisfaction rate coefficient of this user; C i represents the expected spending price of this user for charging and swapping; C r represents the maximum intolerable cost of this user.

[0134] Furthermore, the CA-LSTM integrated model built in step three combines the long short-term memory network LSTM with the cellular automaton CA; the nth cell where the electric vehicle travels is used as the exploration cell, and the time series data of the traffic flow on this exploration cell and its neighbor cells before the moment t is extracted. The extracted time series data is divided into multiple time series data in the order of time series, and the time series data with a preset time step is used as an input sample; the input sample is input into the input layer of the LSTM, and the memory unit in the hidden layer of the LSTM captures the characteristics of the traffic flow data on the exploration cell; at each time step, the hidden layer calculates based on the current input sample and the hidden state of the previous moment, and controls the flow and update of information through the input gate, forget gate, and output gate. The output layer predicts the vehicle change data on this cell at the next time step t + 1; finally, the probability k of the existence of traffic flow on the exploration cell at the moment t + 1 is calculated. Figure 2 That is, the schematic diagram of building the CA-LSTM integrated model in this method.

[0135] Furthermore, step four specifically includes:

[0136] S41: With the goal of minimizing the travel cost, use the improved A* algorithm to obtain the optimal travel path; in the A* algorithm, first evaluate each searched cell to get the best cell, and then search from this cell until the goal; when exploring the goal to the nth cell, record the cost required to travel from the starting point to the nth cell and the cost required for the nth cell to travel to the end point, and judge whether there is a vehicle on the nth cell. If there is, add the congestion waiting cost generated by the vehicle; the dynamic measurement heuristic of the specific improved A* search algorithm is as follows

[0137]

[0138] In the above formula: g represents the distance of the traffic network contained in the corresponding exploration cell; v nrepresents the speed of the nth cell, following the driving rules on the road in formula (1); h n represents the Euclidean distance between the nth cell and the target cell; represents the preferred speeds of different users in different functional areas; w is the evaluation weight coefficient, the closer to the target, the larger the unit space, and vice versa; k represents the possibility that there is a vehicle in the nth exploration cell when driving to it; σ represents the time lost by having a vehicle; f(n) represents the value of node n;

[0139] S42: Construct an improved prospect theory decision model; in this model, a value function is constructed using the charging and swapping budget of electric vehicles and the travel cost under decision simulation. The value function of this decision model is shown as follows:

[0140]

[0141] In the above formula, α and β represent the sensitivity of electric vehicles to the risk of charging and swapping, and their values are both preset; λ represents the loss aversion coefficient, and this value is also preset; C s is the travel cost under decision simulation, which is the actual simulated travel cost from the starting point to the ending point; C i is the travel budget cost of the electric vehicle, which is the travel cost obtained from the straight-line distance from the starting point to the ending point;

[0142] S43: Due to the uncertainty of the travel road conditions of each electric vehicle, there is a random slowdown rule. Therefore, for an electric vehicle making a decision, the specific time it takes to drive to the destination is uncertain. Use the CA-LSTM integrated model to conduct multiple travel simulations on the optimal travel plan obtained by the decision-making electric vehicle to obtain the travel time and its probability, that is, the risk probability of this plan; assume that the electric vehicle has m + n + 1 possible arrival times, C s(-m) <···<C i <···<C s(n) , and their occurrence probabilities are p -m ···p n , which are represented by C s =(C s(-m) ···C s(n) ) and P=(p -m ···p n ); then the cumulative decision weight function is shown as follows:

[0143]

[0144] In the above formula: and are the cumulative decision weight values when facing gains and losses respectively, and their calculation methods are determined by the function ω +(·) and ω - (·) represents, as shown in the following formula:

[0145]

[0146] In the above formula, γ and δ are the probability weight coefficients when the vehicle owner faces gains or losses, and their preset values are: γ﹦0.61, δ﹦0.69;

[0147] Furthermore, calculate the cumulative prospect value of different electric vehicles for charging decision-making, as shown in the following formula:

[0148]

[0149] In the above formula, CPV j + represents the positive cumulative prospect value under scenario j; CPV j － represents the negative cumulative prospect value under scenario j; CPV j represents the actual comprehensive cumulative prospect value under scenario j, j = 1, 2, 3…;

[0150] When making a decision on charging or battery swapping, the electric vehicle selects the charging and swapping plan with the largest comprehensive cumulative prospect value as the optimal charging and swapping choice, as shown in the following formula:

[0151] CPV i = MAX(CPV i,1 , CPV i,2 ,..., CPV i,j ), j ∈ {1, 2, 3,...} (20)

[0152] In the above formula, CPV represents the cumulative prospect value. Specific embodiment:

[0154] To verify the effectiveness of this method, this embodiment takes a city traffic network with a total of 55 nodes and 90 roads as an example. The city has a traffic road functional area based on the POI data type, see Figure 4 (the lower right corner of the POI map).

[0155] The number of charging piles, rated fast charging power, and charging efficiency of the charging and battery swapping stations in this embodiment are shown in Table 1; the battery capacity, maximum driving distance, and power consumption per kilometer of the electric vehicle are shown in Table 2; the relevant parameters of different functional areas are shown in Table 3. The real-time charging price curve is as Figure 4 shown. The cost of battery swapping includes the energy purchase cost and the service cost, which are 40 yuan and 10 yuan respectively.

[0156] Table 1

[0157]

[0158]

[0159]

[0160] Table 2

[0161]

[0162] Table 3

[0163] Parameter Name Residential House Large Shopping Mall Medical and Healthcare Company Science, Education and Culture Scenic Spot Government Agency Others Maximum Speed (km / h) 40 40 50 50 45 60 30 60 Random Slowing Probability 0.1 0.1 0.07 0.07 0.06 0.04 0.12 0.04 Power Consumption (km / h) 0.15 0.15 0.1 0.1 0.12 0.09 0.17 0.09

[0164] To illustrate the superiority of the proposed method, the following three cases are established in this embodiment:

[0165] Case 1: The electric vehicle completes the charging behavior through the shortest driving path.

[0166] Case 2: According to the expected utility theory EUT of the existing technology, the electric vehicle is helped to select the shortest driving path to complete the charging behavior.

[0167] Case 3: According to the electric vehicle charging and swapping decision-making method considering user personalized needs of this method, the private car is helped to select the optimal path to complete the charging behavior.

[0168] Set the same variables, select 8:00 as the decision-making moment, select the longitude and latitude (15, 65) as the decision-making initial position, and select 0.4 as the decision-making initial SOC. According to the charging and swapping budget model in this method, the travel budget of this private car is 78.55 yuan. In Case 1, the user selects the charging mode at the charging station in (28, 44). In Cases 2 and 3, the user selects the swapping mode at the battery swapping station in (48, 32). The three energy replenishment methods are compared, and the results are shown in Figure 5 and Table 4. Figure 5 In, the depth of color represents the degree of congestion. The driving distance in Case 1 is the shortest, but due to the lack of consideration of the decision-making behavior of bounded rationality and risk attitude, the user satisfaction is low. Although Case 2 rationally analyzes the user's energy replenishment decision by comparing the shortest paths of each site, it ignores the user's risk attitude towards traffic congestion. In Case 3, the best driving path obtained according to the improved A* algorithm and the optimal path of the improved prospect theory decision model avoid the possibility of traffic jams, the actual travel cost is close to the budget, and the user satisfaction is high.

[0169] Table 4

[0170] Case Satisfaction Degree Arrival Time at Charging Station Driving Distance (km) Actual Cost Charging Mode 3 100％ 8:52 17.75 79 Battery Swap 2 18.9960％ 9:02 16.50 99 Battery Swap 1 <1％ 8:40 7.25 114.2426 Charging

[0171] Although the present invention has been disclosed above in preferred embodiments, they are not intended to limit the present invention. Any person skilled in this art can make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be defined by the protection scope of the claims of this application.

Claims

1. A decision-making method for charging and replacing electric vehicles considering personalized needs of users, characterized by: include: Step 1: Use grid sampling and cellular automaton model to process traffic network data and charging station data, and build a cellular automaton model; set functional area labels for cells in the model in combination with POI data of the traffic network; Step 2: construct a personalized demand model for electric vehicles based on the constructed cellular automaton model; the personalized demand model includes a driving speed model that takes into account the user's travel anxiety and the speed limit of the functional area where the user is located, a charging and swapping cost model that takes into account time cost and expense cost, a charging and swapping budget model that takes into account the travel purposes of different users, and a charging and swapping satisfaction model that takes into account the principle of diminishing marginal benefits; Step 3: Based on the constructed cellular automaton model, build a CA-LSTM integrated model to predict future traffic flow on the road; In the CA-LSTM integrated model, the cellular automaton CA is used as the upper framework to express the interaction between cells on the spatial road network, and the LSTM is used as the bottom layer to process the changes in vehicle data in cells at different times, so as to realize traffic flow prediction by simulating the study area. Step 4: Based on the constructed personalized demand model, the improved A* algorithm is used to obtain the optimal travel path with the lowest travel cost. The CA-LSTM integrated model is used to simulate the optimal path multiple times to obtain the corresponding travel risk and convert it into the risk of the corresponding power replenishment plan; the improved prospect theory decision model is used to analyze all charging and swapping plans, and finally the optimal charging and swapping plan is obtained.

2. According to claim 1, a decision-making method for charging and replacing electric vehicles considering the personalized needs of users is characterized by: Step 1 specifically includes: S11: Construct a cellular automation model; divide the traffic network data of the area to be decided into grids; the traffic network of the area to be decided is the cellular space; each grid is a cell in the cellular space; define the state of each cell according to whether there is an electric car in each cell; mark the cells in the cellular space that contain charging stations as special cells; use the speed to describe the road driving rules of the cell as follows: In the above formula, v n (t) represents the speed of the nth vehicle at time t; v max Indicates the maximum speed of an electric vehicle; gap n (t) represents the distance between the nth vehicle and the vehicle immediately in front at time t; △t represents the unit time; S12: Perform functional area analysis on the cells; obtain POI data on each cell based on the POI data in the electronic map, and after longitude and latitude import and projection preprocessing, calculate the weights of different types of POI data in the corresponding cells according to the following formula based on the frequency of occurrence of different types of POI data in each corresponding cell: In the above formula, n i,j Represents the number of times the i-th type of POI data appears in cell j; represents the total number of POI data appearing in cell j; |E| represents the number of cells in the cellular space; |{j∈E}| represents the number of cells containing the i-th type of POI data; tfidf i,j represents the weight of the i-th type of POI data in cell j; S13: Determine the functional area label of each cell; determine the functional area label corresponding to each cell according to the weight and frequency of each POI data in the cell, as shown in formula (3): In the above formula, D i,j is the weight of the i-th type of POI data in cell j; as shown in the above formula, the POI category with the largest weight in cell j is the functional area type D of the cell. j .

3. The electric vehicle charging and replacement decision-making method considering the personalized needs of users according to claim 1 is characterized by: The driving speed model constructed in step 2 considering the user's travel anxiety and the speed limit of the functional area is to model the electric vehicle in combination with its travel anxiety, speed preference and the speed limit of the functional area. The specific construction process includes: S21: Define the minimum pace of each functional area as v min , the maximum speed is v max ; Anxiety level is used as the personalized feature of electric vehicles; Since users with different anxiety levels have different preferences for driving speed, the maximum speed of users with lower anxiety levels is defined as v' max , where v' max <v max ; and most users prefer higher speeds in actual driving, assuming that the actual maximum speed v' in each functional area max If it conforms to the skewed distribution, the probability density function of the skewed distribution is as follows: In the above formula: - is the lower standard deviation of the functional area corresponding to the preset cell where the car is located, which is negatively correlated with the degree of anxiety; + is the upper standard deviation of the functional area corresponding to the preset cell where the car is located, which is positively correlated with the degree of anxiety; variable σ + With σ - The value of is determined based on the historical driving statistics of the vehicle; S22: The cars driving in the functional area are divided into two categories: one is the user speed range with low anxiety level [v min ,v' max ], the actual maximum pace v' of this type of user max Less than the maximum speed limit v of the functional area max ; Another type of user with a strong sense of anxiety speed range [v min ,v max ], the maximum speed of this type of user is the maximum speed limit of this functional area v max ; Users with different anxiety levels have different paces in the same functional area, and electric vehicles prefer different driving speeds.

4. The electric vehicle charging and replacement decision-making method considering the personalized needs of users according to claim 3 is characterized by: The charging and swapping cost model constructed by considering time cost and expense cost in step 2 is to consider the time cost and expense cost of the charging and swapping process. According to the time benefit of the value judgment of electric vehicles, the time cost and expense cost are constructed into a charging and swapping cost model of the same dimension, which specifically includes: S23: Model charging cost and battery replacement cost separately; Among them, the charging cost of electric vehicles is shown in the following formula: In the above formula: C c,exp represents the cost of the i-th electric vehicle choosing the charging mode; c cha (t) is the real-time electricity price at different times; SOC i,j is the power of the i-th electric car after it reaches the j-th cell; The cost of battery replacement for electric vehicles is as follows: In the above formula: C s,exp represents the cost of battery replacement; c bat The cost of purchasing a fully charged battery; ser Service charges for battery installation; sub Subsidy for electricity price for battery swapping; S24: Model charging time and battery replacement time separately; The queuing rules for charging or battery swapping follow the first-come-first-served rule; and the number of batteries in the battery swapping station meets user needs; there are N fast charging stations or battery swapping stations k charging piles or battery swap stations; the kth fast charging station or battery swap station has Electric family car; If the electric vehicle chooses a charging method, the charging time of the i-th electric vehicle at the k-th charging station Charging time for service And waiting time sum; Among them, service charging time As shown below: In the above formula: C is the rated battery capacity of the electric vehicle; P c is the charging power during charging; η is the charging efficiency during charging; Waiting time As follows: In the above formula: Indicates rounding down; represents the total number of vehicles waiting to charge at the kth station, S represents the SOC set of vehicles in the station arranged from large to small. The waiting time for the i-th electric vehicle to charge in the k-th charging station depends on the number of electric vehicles waiting in line. If This means that there are idle charging piles at the fast charging station, so the electric car does not need to wait; If the electric vehicle chooses to replace the battery, the time t for the i-th electric vehicle to replace the battery ser , s is fixed, the battery swap time of the i-th electric vehicle at the k-th battery swap station in the battery swap mode As shown below: S25: Construct a time value model of electric vehicles based on the market price of electric vehicles, as follows: In the above formula: i is the time value of the i-th electric car user, which is pre-set by the owner of the electric car according to his specific situation; M is the market price of the electric car, in units of one million yuan; δ i The anxiety level of the electric vehicle is pre-set by the owner of the electric vehicle according to his specific situation.

5. The electric vehicle charging and replacement decision-making method considering the personalized needs of users according to claim 4 is characterized by: In step 2, the charging and swapping budget model considering the travel purposes of different electric vehicles is as follows: In the above formula: C i,s is the travel cost of the i-th electric car; t i is the time spent by the i-th electric car in the driving process; M i is the model price of the i-th electric car, in million yuan; δ i is the anxiety level of the i-th electric car, η i is a binary quantity, η i =1 means that the i-th electric vehicle selects the battery replacement mode, η i =0 indicates that the i-th electric vehicle selects the charging mode.

6. The electric vehicle charging and replacement decision-making method considering the personalized needs of users according to claim 5 is characterized by: The charging and swapping satisfaction model considering the principle of diminishing marginal benefits in step 2 is to construct the satisfaction of charging and swapping decisions using the satisfaction Richards function model, as shown in the following formula: In the above formula, r represents the user's satisfaction rate coefficient; C i represents the expected cost of charging and swapping for the user; C r Indicates the maximum intolerable cost for the user.

7. The electric vehicle charging and replacement decision-making method considering the personalized needs of users according to claim 6 is characterized by: The CA-LSTM integrated model built in step 3 is a combination of the long short-term memory network LSTM and the cellular automaton CA; the nth cell that the electric car drives to is taken as the exploration cell, and the time series data of the traffic flow on the exploration cell and the neighboring cells before the time t is extracted, and the extracted time series data is divided according to the time series order to obtain multiple time series data, and the time series data of the preset time step is taken as an input sample; the input sample is input into the input layer of the LSTM, and the memory unit in the hidden layer of the LSTM captures the characteristics of the traffic flow data on the exploration cell; At each time step, the hidden layer performs calculations based on the current input sample and the hidden state of the previous moment, and controls the flow and update of information through the input gate, forget gate, and output gate. The output layer predicts the vehicle change data on the cell at the next time step t+1; finally, the probability k that there is traffic flow in the exploration cell at time t+1 is calculated.

8. The electric vehicle charging and replacement decision-making method considering the personalized needs of users according to claim 7 is characterized by: Step 4 specifically includes: S41: With the goal of minimizing travel cost, the improved A* algorithm is used to obtain the optimal travel path. In the A* algorithm, each searched cell is first evaluated to obtain the best cell, and then the search is continued from this cell until the target is reached. When the exploration target reaches the nth cell, the cost required to travel from the initial point to the nth cell and the cost required to travel from the nth cell to the end point are recorded, and it is determined whether there is a vehicle on the nth cell. If there is, the congestion waiting cost caused by the vehicle is added. The specific dynamic measurement heuristic of the improved A* search algorithm is as follows: In the above formula: g represents the distance of the transportation network contained in the corresponding exploration cell; v n represents the speed at the nth cell, following the road driving rules in formula (1); h n Represents the Euclidean distance between the nth cell and the target cell; represents the preferred speed of different users in different functional areas; w is the evaluation The weight coefficient is larger the closer to the target unit space, and smaller on the contrary; k represents the possibility that a vehicle exists in the nth exploration cell when driving to the cell; σ represents the time lost by a vehicle; f(n) represents the value of the nth node; S42: Construct an improved prospect theory decision model; in this model, the value function is constructed by using the charging and swapping budget of electric vehicles and the travel cost under decision simulation. The value function of this decision model is shown as follows: In the above formula, α and β represent the sensitivity of electric vehicles to the risks of charging and swapping, and their values ​​are preset; λ represents the loss aversion coefficient, and its value is also preset; C s is the travel cost under decision simulation, is the actual simulated travel cost from the starting point to the end point; C i is the travel budget cost of the electric vehicle, and is the travel cost obtained by the straight-line distance from the starting point to the end point; S43: Due to the uncertainty of the travel conditions of each electric vehicle, there is a random slowing-down rule. Therefore, for an electric vehicle making a decision, the specific time it will travel to the destination is uncertain. The CA-LSTM integrated model is used to simulate the optimal travel plan obtained by the decision-making electric vehicle multiple times to obtain the travel time and its probability, that is, the risk probability of the plan; Assuming that the electric vehicle has m+n+1 possible arrival times, C s(-m) <···<C i <···<C s(n) , and their occurrence probabilities are p -m ···p n , using C s =(C s(-m) ···C s(n) ) and P = (p -m ···p n ) represents; then the cumulative decision weight function is as follows: In the above formula: and are the cumulative decision weights when facing gains and losses, respectively, and their calculation methods are respectively determined by the function ω + (·) and ω - (·) is represented as follows: In the above formula, γ and δ are the probability weight coefficients when the car owner faces gains or losses, and their preset values ​​are: γ﹦0.61, δ﹦0.69; Then the cumulative prospect value of different electric vehicles for charging decisions is calculated as shown in the following formula: In the above formula, CPV j + represents the positive cumulative prospect value under scenario j; CPV j － represents the negative cumulative prospect value under scenario j; CPV j represents the actual comprehensive cumulative prospect value under scenario j, j = 1, 2, 3…; When making a decision on charging or battery replacement, electric vehicles choose the charging and battery replacement solution with the largest comprehensive cumulative prospect value as the optimal charging and battery replacement option, as shown in the following formula: CPV i =MAX(CPV i,1 ,CPV i,2 ,...,CPV i,j ),j∈{1,2,3,...} (20) In the above formula, CPV stands for Cumulative Prospect Value.

Images