Electric vehicle charging and discharging planning method under large-scale V2G scene based on constraint evolution

Abstract

The invention designs a constrained evolution-based electric vehicle charging and discharging planning method under a large-scale V2G scene aiming at the charging and discharging behavior planning problem of an electric vehicle under the large-scale V2G scene. According to the method, a population restart strategy comprising a multi-level restart opportunity estimation method and a multi-scene restart population generation method is used, and meanwhile, a guide type initialization strategy is added. Wherein the guiding type initialization strategy accelerates the population to enter a feasible region for searching, the population restart strategy enables the algorithm to accurately estimate the restart time at a low cost, and pertinently processes different types of evolutionary dilemma, so that the algorithm solving efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the field of energy management, and in particular relates to a method for planning electric vehicle charging and discharging in a large-scale V2G scenario based on constrained evolution. Technical Background

[0002] With the rapid rise in electricity demand, various renewable energy technologies are being widely adopted. However, new energy technologies such as photovoltaics, wind power, and tidal power are limited by natural resources, resulting in high energy supply volatility and low flexibility. They also lack inherent energy storage capabilities, making them difficult to meet the high reliability and stability requirements of smart grids. Thanks to the popularity of electric vehicles, vehicle-to-grid (V2G) technology has received considerable attention and research. V2G technology essentially transforms large-scale electric vehicles (EVs) from "load terminals" into distributed "energy storage units," supporting bidirectional energy exchange between EVs and the grid, thereby achieving goals such as peak load shaving and valley filling and enhancing the grid's regulatory capabilities.

[0003] With the increasing number of electric vehicles connected to the power grid, V2G systems are becoming increasingly large-scale, high-dimensional, and highly constrained. Solving the overall charging and discharging planning problem for electric vehicles presents challenges such as shrinking and discrete feasible domains, significant conflicting objectives, and high computational complexity. Evolutionary algorithms, due to their excellent global search capabilities and adaptability to complex optimization problems, have been widely used in optimizing charging and discharging planning for electric vehicles in V2G scenarios. However, when implementing charging and discharging planning for large-scale electric vehicle deployments, such as large logistics parks, stations, and ports, evolutionary algorithms are prone to falling into infeasible regions during the search process or encountering an "evolutionary dilemma"—premature convergence and search stagnation within the feasible domain—severely impacting solution efficiency and planning solution quality. Existing research has paid insufficient attention to the algorithmic mechanisms themselves, primarily relying on traditional optimization algorithms for solution resolution, making it difficult to rapidly respond to and address evolutionary dilemmas. Therefore, an optimization algorithm that can effectively address evolutionary dilemmas and improve search efficiency is urgently needed to address the needs of electric vehicle charging and discharging planning in large-scale V2G scenarios. Summary of the Invention

[0004] To achieve the above objectives, the present invention adopts a large-scale electric vehicle charging and discharging planning method (EPR-COE) based on constrained evolution in V2G scenarios, which includes the following steps:

[0005] Step 1: Build an optimization model for the electric vehicle charging and discharging planning problem in a large-scale V2G scenario;

[0006] Step 2: Use the constrained evolutionary algorithm based on population restart to solve the problem model and calculate the starting time and duration of charging and discharging for each electric vehicle.

[0007] The optimization model in step 1 specifically includes:

[0008] The objective function of the optimization model is:

[0009]

[0010] Among them, f1 represents the overall charging cost of the electric vehicle, f2 represents the mean square error of the total charging and discharging load of the power grid in each period, ω represents the weight coefficient; K represents the number of electric vehicles, T represents the number of time periods; Pc k and Pd k represents the charging power and discharging power of the kth tram, ct k,t and dt k,t represents the charging and discharging time of the kth tram in the tth period, p t represents the electricity price in the tth period; P t Indicates the load after total charge and discharge in the tth period, PB t represents the base load before summing in the tth period, represents the mean value of the total afterload in all periods, ηd k represents the discharge efficiency of the kth tram.

[0011] Constraints for the optimization model:

[0012]

[0013] Among them, SOC k,min and SOC k,max Indicates the lower and upper limits of the battery capacity of the kth electric vehicle, SOC k,cs and SOC k,ds represents the amount of electricity before and after charging and discharging of the kth tram; ct k and dt k represents the total charging and discharging time of the kth tram, ηc k represents the charging efficiency of the kth electric vehicle; SOC k,d represents the amount of electricity when the kth tram leaves the charging area, l k and u k represents the mileage and power consumption per kilometer of the kth tram; ta k and td k represents the time when the kth tram arrives and leaves the charging area, cs k and ds k Indicates the time when the kth tram starts charging and discharging; state k represents the charging state of the kth tram, and state k ∈{0,1}, where the value 0 indicates that the kth trolley is not charging, and the value 1 indicates that it is charging; P max Indicates the maximum charging load allowed in the charging area, N maxIndicates the number of charging piles in the charging area.

[0014] The step 2 specifically includes the following steps:

[0015] Step 2.1: Input the basic parameters of the problem model and the basic parameters of the constrained evolution algorithm;

[0016] Step 2.2: Encode the solution variables;

[0017] Step 2.3: Generate the initial population P = (x1, ..., x NP )';

[0018] Step 2.4: Calculate the fitness value f(x i ) and default degree G(x i );

[0019] Step 2.5, perform mutation and crossover operations to generate the offspring population P'=(y1,...,y NP )', calculate the fitness value f(y i ) and default degree G(y i );

[0020] Step 2.6, perform the selection operation to obtain the next generation population and update the optimal plan for charging and discharging the electric vehicle x best ;

[0021] Step 2.7, record the population search traces and calculate the feasible edge probability, elite edge probability, and unexplored edge probability of each region for each decision variable;

[0022] Step 2.8, executing a multi-level restart timing estimation method;

[0023] Step 2.9: Based on the type of evolutionary dilemma determined in step 2.8, perform corresponding operations in the multi-scenario restart population generation method;

[0024] Step 2.10: Determine whether the maximum number of iterations has been reached. If so, terminate the algorithm and output the optimal plan for charging and discharging the electric vehicle. Otherwise, execute 2.4.

[0025] The step 2.7 specifically includes the following steps:

[0026] Step 2.7.1, divide the search range of each decision variable into s regions;

[0027] Step 2.7.2, record the number of times each region of each decision variable is explored, the number of times feasible solutions appear, and the number of times elite solutions appear;

[0028] In step 2.7.3, calculate the feasible edge probability, elite edge probability, and unexplored edge probability for each region of each decision variable.

[0029] The multi-level restart timing estimation method in step 2.8 specifically includes the following steps:

[0030] Step 2.8.1, calculate the current number of iterations g and the dividing point G b The size relationship between them; if g<G b , then the algorithm executes 2.8.2, otherwise executes step 2.10;

[0031] Step 2.8.2, calculate the expected value ρ of whether the population has been trapped in the feasible region or the infeasible region for a long time. The calculation formula is:

[0032]

[0033] Where FR∈{0,1}, FR=0 means all planning schemes are infeasible, and FR=1 means all planning schemes are feasible; N1 represents the number of consecutive generations with FR value 0, and N2 represents the number of consecutive generations with FR value 1; if the restart expectation is met and FR is 0, the algorithm executes the restart population generation method for the population trapped in the infeasible region; if the restart expectation is met and FR is 1, the algorithm executes 2.8.3; if the restart expectation is not met or the value of FR is neither 0 nor 1, execute step 2.10;

[0034] Step 2.8.3, calculate the evolution rate OR(g) of the optimal solution of the population, which is calculated as follows:

[0035]

[0036] Compare OR(g) with the evolution rate of the previous generation optimal solution OR(g-1), and use N3 to record the number of consecutive generations of evolution rate decline; when N3 reaches the threshold If , the algorithm executes step 2.8.4, otherwise executes step 2.10;

[0037] Step 2.8.4, calculate the population density PD, which is calculated as follows:

[0038]

[0039]

[0040] in, Represents the mean of all individuals of the jth decision variable; when PD is less than the threshold λ, the algorithm executes a restart population generation method for the population falling into the dilemma of premature convergence in the feasible domain; otherwise, it executes a restart population generation method for the population falling into the dilemma of search stagnation in the feasible domain.

[0041] The beneficial effects of the present invention are reflected in the following: Based on the differential evolution algorithm and multi-objective constraint processing technology, a population restart strategy is adopted, which includes a multi-level restart timing estimation method and a multi-scenario restart population generation method, and a guided initialization strategy is also used. The multi-level restart timing estimation method balances the accuracy of restart timing estimation with computational overhead, and the multi-scenario restart population generation method uses population search traces to provide targeted population restart measures for three evolutionary dilemmas, thereby improving the algorithm's efficiency in solving the electric vehicle charging and discharging planning problem in large-scale V2G systems. Compared with random initialization, the initial population generated by the guided initialization strategy is more consistent with the characteristics of the electric vehicle charging and discharging planning problem in V2G systems and accelerates the population's search into the feasible domain. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a schematic diagram of a scenario according to an embodiment of the present invention;

[0043] Figure 2 This is a diagram of the EPR-COE algorithm architecture of an embodiment of the present invention;

[0044] Figure 3 is a graph showing the convergence curve of the ablation experiment default rate of the guided initialization strategy according to an embodiment of the present invention;

[0045] Figure 4 This is a graph showing the fitness value convergence of the comparison algorithm according to an embodiment of the present invention;

[0046] Figure 5 This is a comparison chart of load optimization results of comparison algorithms in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] To provide a clearer understanding of the technical features, objectives, and beneficial effects of this invention, the following detailed description of the technical solutions of this invention is provided. Obviously, the described implementation examples represent only a portion of the present invention, not all of it, and should not be construed as limiting the scope of the invention. Based on the embodiments of this invention, all other implementations derived by persons of ordinary skill in the art without inventive effort fall within the scope of protection of this invention.

[0048] Aiming at the problem that when using existing evolutionary algorithms to solve the electric vehicle charging and discharging planning in V2G scenarios, the algorithm is easily trapped in an evolutionary dilemma due to the large number of electric vehicles and multiple constraints, this paper proposes an electric vehicle charging and discharging planning method EPR-COE in large-scale V2G scenarios based on constrained evolution.

[0049] Schematic diagram of the research problem scenario Figure 1 As shown in the diagram, the algorithm architecture is as follows Figure 2The specific implementation is divided into three parts: the first part is the description of the simulation experiment scene and the setting of parameters, the second part is the specific application process of the method, and the third part is the experimental results.

[0050] The first part is: simulation experiment scene description and parameter setting.

[0051] The experimental scenario of the present invention is a logistics park charging area equipped with 100 V2G charging piles, which allows a maximum charging power of 2000kW at the same time. It needs to provide charging and discharging services for 300 electric vehicles from 8:00 to 19:00 and for 200 electric trucks from 19:00 to 8:00 the next day.

[0052] In the experimental scenario of the present invention, the performance parameters of electric vehicles and electric trucks are shown in the following table:

[0053] The second part is: the specific application process of the method. It includes the following steps:

[0054] Step 1: Build an optimization model for the electric vehicle charging and discharging planning problem in a large-scale V2G scenario, including the objective function and constraints:

[0055] The objective function of the optimization model is:

[0056]

[0057] Wherein, f1 represents the overall charging cost of the electric vehicle, f2 represents the mean square error of the total charging and discharging load of the power grid in each period, ω represents the weight coefficient, and ω∈(0,1); K represents the number of electric vehicles, T represents the number of time periods divided into a day, which is set to 24 in the present invention; Pc k and Pd k represents the charging power and discharging power of the kth tram, ct k,t and dt k,t represents the charging and discharging time of the kth tram in the tth period, p t represents the electricity price in the tth period; P t Indicates the load after total charge and discharge in the tth period, PB t represents the base load before summing in the tth period, represents the mean value of the total afterload in all periods, ηd k represents the discharge efficiency of the kth tram.

[0058] Constraints for the optimization model:

[0059]

[0060] Among them, SOC k,minand SOC k,max Indicates the lower and upper limits of the battery capacity of the kth electric vehicle, SOC k,cs and SOC k,ds represents the amount of electricity before and after charging and discharging of the kth tram; ct k and dt k represents the total charging and discharging time of the kth tram, ηc k represents the charging efficiency of the kth electric vehicle; SOC k,d represents the amount of electricity when the kth tram leaves the charging area, l k and u k represents the mileage and power consumption per kilometer of the kth tram; ta k and td k represents the time when the kth tram arrives and leaves the charging area, cs k and ds k Indicates the time when the kth tram starts charging and discharging; state k represents the charging state of the kth tram, and state k ∈{0,1}, where the value 0 indicates that the kth trolley is not charging, and the value 1 indicates that it is charging; P max Indicates the maximum charging load allowed in the charging area, N max Indicates the number of charging piles in the charging area.

[0061] Step 2: Use the constrained evolutionary algorithm based on population restart to solve the optimization model and calculate the starting time and duration of charging and discharging for each electric vehicle. Figure 2 The provided algorithm architecture diagram includes the following steps:

[0062] Step 2.1: Input the basic parameters of the problem model and the basic parameters of the constrained evolution algorithm;

[0063] Step 2.2: Encode the solution variables;

[0064] Step 2.3: Generate the initial population P = (x1, ..., x NP )';

[0065] Step 2.4: Calculate the fitness value f(x i ) and default degree G(x i );

[0066] Step 2.5: Use the differential evolution algorithm to perform mutation and crossover operations to generate the offspring population P'=(y1,...,y NP )', calculate the fitness value f(y i ) and default degree G(y i );

[0067] Step 2.6: Use multi-objective constraint processing technology to perform selection operations, obtain the next generation population, and update the optimal electric vehicle charging and discharging plan x best ;

[0068] Step 2.7: Record the population search traces and calculate the feasible edge probability, elite edge probability, and unexplored edge probability of each region for each decision variable;

[0069] Step 2.8: Execute a multi-level restart timing estimation method;

[0070] Step 2.9: Based on the type of evolutionary dilemma determined in 2.8, perform corresponding operations in the multi-scenario restart population generation method;

[0071] Step 2.10: Determine whether the maximum number of iterations has been reached. If so, terminate the algorithm and output the optimal electric vehicle charging and discharging plan. Otherwise, execute 2.4.

[0072] In 2.1, enter all the parameters required for the problem model; enter the population size NP, the maximum number of iterations g max , the maximum number of evaluations MaxFEs. Among them, the population size NP = 100, the maximum number of iterations g max =20000, maximum number of evaluations MaxFEs=2000000.

[0073] In 2.2, the solution variable is encoded, x i It can be expressed as x i =(cs i,1 ,ct i,1 ,ds i,1 ,dt i,1 ,...cs i,K ,ct i,K ,ds i,K ,dt i,K ), represents a plan for electric vehicle charging and discharging. i,k ,ct i,k ,ds i,k ,dt i,k They represent the starting charging time, total charging time, discharging starting time, and total discharging time of the kth electric vehicle in the i-th individual, and their value ranges are cs i,k ∈[ta k ,td k ), ct i,k ∈[0,td k -ta k ), ds i,k ∈[ta k ,td k ), dt i,k ∈[0,tdk -ta k ).

[0074] In 2.3, set the current iteration number g = 1, and use the guided initialization strategy to generate the initial population P = (x1, ..., x NP )'. In this strategy, the order of charging and discharging of the trams is first determined. If the tram arrives during off-peak or off-peak periods, it is charged first and then discharged. If it arrives during peak periods, it is discharged first and then charged. If the kth tram in the i-th individual is charged first and then discharged: first, two moments are randomly selected in this section and the subsequent off-peak period as the charging start time cs i,k The duration between the time when charging ends is the total charging time ct i,k , and then randomly select two moments in the longest peak segment as the discharge start time ds i,k and the end time of discharge, the duration between them is the total discharge time dt i,k For the case of discharging first and then charging: randomly select two moments in the peak period as the discharge start time ds i,k and the end time of discharge, the duration between them is the total discharge time dt i,k , and then randomly select two moments in the longest continuous off-peak period as the charging start time cs i,k The duration between the time when charging ends is the total charging time ct i,k .

[0075] In 2.4, calculate the population of each individual x i The fitness value f(x i ) and default degree G(x i ).

[0076] In 2.5, for each individual x i , randomly generate a number rand between 0 and 1, if rand<g / G max , using DE / rand-to-best / 1 / bin operator to perform population mutation and crossover operations to generate the offspring population P'=(y1,...,y NP )'; otherwise, DE / current-to-rand / 1 operator is used to perform population mutation and crossover operations to generate the offspring population P'=(y1,...,y NP )'. Calculate y for each individual in the offspring population i The fitness value f(y i ) and default degree G(y i ).

[0077] In 2.6, for each individual x in the parent populationi and the corresponding individual y in the offspring population i , through the formula g ws (x|λ)=λf(x)+(1-λ)G(x) calculate g respectively ws (x i |λ i ) and g ws (y i |λ i ), if g ws (y i |λ i )≤g ws (x i |λ i ), then use y i Replace x i ,f(y i ) instead of f(x i ), G(y i ) instead of G(x i ). where λ i =i×η / NP, The values of Γ and α are 30 and 0.75 respectively. If g ws (y i |λ i )≤g ws (x best |λ i ), then use y i Replace x best ,f(y i ) instead of f(x best ), G(y i ) instead of G(x best ).

[0078] In 2.7, the marginal probability calculation for each region of each decision variable includes the following three steps:

[0079] In 2.7.1, the search range of each decision variable is first divided into s regions. l max and u max They represent the lower and upper limits of the search corresponding to the decision variable with the widest search range.

[0080] In 2.7.2, the variables RF(j,k) and RI(j,k) are used to represent the number of times the feasible and infeasible variables of the j-th dimension appear in the k-th region of that dimension up to the present time; RE(j,k) represents the number of times the j-th dimension variable appears in the k-th region, RE(j,k)=RI(j,k)+RF(j,k); RB(j,k) represents the number of times the j-th dimension variable appears in the k-th region among the top 25% of elite individuals in the contemporary population.

[0081] In 2.7.3, the feasible edge probability FP(j,k), elite edge probability BP(j,k), and unexplored edge probability UP(j,k) of the k-th region in the j-th dimension are calculated according to the following methods:

[0082]

[0083]

[0084]

[0085]

[0086] In 2.8, the multi-level restart timing estimation method includes the following four steps:

[0087] In 2.8.1, perform the first level estimation and calculate the current number of iterations g and the cutoff point G b The size relationship between them, where G b =0.8×g max When g<G b When , the algorithm executes 2.8.2, otherwise executes 2.10.

[0088] In 2.8.2, we perform the second-level estimation. First, we calculate the expected value ρ of whether the population has been trapped in the feasible region or the infeasible region for a long time. The calculation formula is:

[0089]

[0090] Where FR∈{0,1}, FR=0 means that all planning schemes are infeasible, and FR=1 means that all planning schemes are feasible; N1 represents the number of consecutive generations with FR value of 0, N2 represents the number of consecutive generations with FR value of 1, and α=0.05; a random number rand between 0 and 1 is generated. If rand<ρ and FR is 0, the algorithm immediately executes the restart population generation method for the population trapped in the infeasible region dilemma; if rand<ρ and FR is 1, the algorithm executes 2.8.3; if the restart expectation is not met or the value of FR is neither 0 nor 1, execute 2.10.

[0091] In 2.8.3, the third level of estimation is performed. First, the evolution rate of the optimal solution of the population, OR(g), is calculated. The calculation formula is:

[0092]

[0093] Compare OR(g) with the evolution rate of the previous generation optimal solution OR(g-1), and use N3 to record the number of consecutive generations of evolution rate decline. When N3 reaches the threshold When , the algorithm executes 2.8.4, where The value is equal to the population size NP; if N3 does not reach the threshold, execute 2.10.

[0094] In 2.8.4, the fourth level of estimation is performed to calculate the population density PD, which is calculated as follows:

[0095]

[0096]

[0097] in, represents the mean of the individuals in the jth dimension. When PD is less than the threshold λ, the algorithm immediately executes the restart population generation method to prevent the population from falling into the dilemma of premature convergence in the feasible domain; otherwise, the algorithm immediately executes the restart population generation method to prevent the population from falling into the dilemma of search stagnation in the feasible domain. Where λ = 0.5.

[0098] In 2.9, the type of dilemma the algorithm is in is determined according to 2.8, and the corresponding restart population generation method is executed. There are three specific methods:

[0099] When the population falls into the infeasible region dilemma, the probability p(j,k) of the restarted individual's j-th dimension variable appearing in the k-th region is calculated as p(j,k)=UP(j,k);

[0100] For the population that falls into the premature convergence dilemma in the feasible region, the probability p(j,k) of the restarted individual j-th dimension variable appearing in the k-th region is calculated as

[0101] For the population stuck in the search stagnation dilemma in the feasible region, the probability p(j,k) that the j-th dimension variable of the restarted individual appears in the k-th region is calculated as

[0102] In 2.10, if the current iteration number g = g max , then exit the algorithm and enter the optimal planning solution x best Otherwise, add 1 to the value of g and execute step 2.4.

[0103] The third part is: presentation of simulation experiment results and analysis of results.

[0104] Figure 1: Simulation Experiment Results Figure 3 , Figure 4 , Figure 5 .

[0105] exist Figure 3In the initial population generated by the EPR-COE of the present invention, the optimal default value is much lower than the optimal default value of the EPR-COE-rand using random initialization. In addition, the EPR-COE in the 15000th generation (FEs is 1.5×10 6 ) has found a feasible solution, and EPR-COE-rand has found a feasible solution in the 25000th generation (FEs is 2.5×10 6 ) before a feasible solution is found. Compared with random initialization, the guided initialization strategy proposed in this paper can accelerate the population to enter the feasible region for search.

[0106] exist Figure 4 The EPR-COE algorithm of the present invention outperforms other compared algorithms in terms of optimal fitness. Specifically, it improves by 58.7% compared to IUDE, 69.2% compared to ECO-HCT, and 55.5% compared to DeCODE.

[0107] exist Figure 5 The EPR-COE algorithm of the present invention outperformed the other compared algorithms in terms of mean square error of load in each time period. Specifically, it improved by 68.9% compared to the base load, 69.0% compared to IUDE, 68.7% compared to ECO-HCT, and 44.2% compared to DeCODE.

[0108] This paper proposes a method called EPR-COE and successfully applies it to the electric vehicle charging and discharging planning problem in large-scale V2G scenarios. By comparing it with several advanced algorithms, it proves that the EPR-COE algorithm has good ability to handle this type of problem.

[0109] The above illustrates and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the foregoing embodiments. The foregoing embodiments and descriptions merely illustrate the principles of the present invention. Various modifications and improvements are possible without departing from the spirit and scope of the present invention. These modifications and improvements are intended to fall within the scope of the claimed invention. The scope of the claimed invention is defined by the appended claims and their equivalents.

Claims

1. A method for planning electric vehicle charging and discharging in large-scale V2G scenarios based on constrained evolution, characterized by: It includes the following 2 steps: Step 1: Build an optimization model for the electric vehicle charging and discharging planning problem in a large-scale V2G scenario; Step 2: Use the constrained evolutionary algorithm based on population restart to solve the problem model and calculate the starting time and duration of charging and discharging for each electric vehicle.

2. The method for planning electric vehicle charging and discharging in a large-scale V2G scenario based on constrained evolution according to claim 1 is characterized in that: The objective function of the optimization model in step 1 includes: Among them, f1 represents the overall charging cost of the electric vehicle, f2 represents the mean square error of the total charging and discharging load of the power grid in each period, ω represents the weight coefficient; K represents the number of electric vehicles, T represents the number of time periods; Pc k and Pd k represents the charging power and discharging power of the kth tram, ct k,t and dt k,t represents the charging and discharging time of the kth tram in the tth period, p t represents the electricity price in the tth period; P t Indicates the load after total charge and discharge in the tth period, PB t represents the base load before summing in the tth period, represents the mean value of the total afterload in all periods, ηd k represents the discharge efficiency of the kth tram.

3. The method for planning electric vehicle charging and discharging in a large-scale V2G scenario based on constrained evolution according to claim 1 is characterized in that: The constraints of the optimization model in step 1 include: Among them, SOC k,min and SOC k,max Indicates the lower and upper limits of the battery capacity of the kth electric vehicle, SOC k,cs and SOC k,ds represents the amount of electricity before and after charging and discharging of the kth tram; ct k and dt k represents the total charging and discharging time of the kth tram, ηc k represents the charging efficiency of the kth electric vehicle; SOC k,d represents the amount of electricity when the kth tram leaves the charging area, l k and u k represents the mileage and power consumption per kilometer of the kth tram; ta k and td k represents the time when the kth tram arrives and leaves the charging area, cs k and ds k Indicates the time when the kth tram starts charging and discharging; state k represents the charging state of the kth tram, and state k ∈{0,1}, where the value 0 indicates that the kth trolley is not charging, and the value 1 indicates that it is charging; P max Indicates the maximum charging load allowed in the charging area, N max Indicates the number of charging piles in the charging area.

4. The method for planning electric vehicle charging and discharging in a large-scale V2G scenario based on constrained evolution according to claim 1 is characterized in that: The step 2 specifically includes the following sub-steps: Step 2.1: Input the basic parameters of the problem model and the basic parameters of the constrained evolution algorithm; Step 2.2, encode the solution variables; Step 2.3, use the guided initialization strategy to generate the initial population P = (x1,...,x NP )'; Step 2.4, calculate the fitness value f(x i ) and default degree G(x i ); Step 2.5, perform mutation and crossover operations to generate the offspring population P'=(y1,...,y NP )', calculate the fitness value f(y i ) and default degree G(y i ); Step 2.6, perform the selection operation to obtain the next generation population and update the optimal plan for charging and discharging the electric vehicle x best ; Step 2.7, record the population search traces and calculate the feasible edge probability, elite edge probability, and unexplored edge probability of each region for each decision variable; Step 2.8, executing a multi-level restart timing estimation method; Step 2.9: Based on the type of evolutionary dilemma determined in step 2.8, perform corresponding operations in the multi-scenario restart population generation method; Step 2.10: Determine whether the maximum number of iterations has been reached. If so, terminate the algorithm and output the optimal plan for charging and discharging the electric vehicle. Otherwise, execute 2.

4.

5. The method for planning electric vehicle charging and discharging in a large-scale V2G scenario based on constrained evolution according to claim 4 is characterized in that: The step 2.7 specifically includes the following sub-steps: Step 2.7.1, divide the search range of each decision variable into s regions; Step 2.7.2, record the number of times each region of each decision variable is explored, the number of times feasible solutions appear, and the number of times elite solutions appear; In step 2.7.3, calculate the feasible edge probability, elite edge probability, and unexplored edge probability for each region of each decision variable.

6. The method for planning electric vehicle charging and discharging in a large-scale V2G scenario based on constrained evolution according to claim 4 is characterized in that: The multi-level restart timing estimation method in step 2.8 specifically includes the following sub-steps: Step 2.8.1, calculate the current number of iterations g and the dividing point G b The size relationship between them; if g<G b , then the algorithm executes 2.8.2, otherwise executes step 2.10; Step 2.8.2, calculate the expected value ρ of whether the population has been trapped in the feasible region or the infeasible region for a long time. The calculation formula is: Where FR∈{0,1}, FR=0 means all planning schemes are infeasible, and FR=1 means all planning schemes are feasible; N1 represents the number of consecutive generations with FR value 0, and N2 represents the number of consecutive generations with FR value 1; if the restart expectation is met and FR is 0, the algorithm executes the restart population generation method for the population trapped in the infeasible region; if the restart expectation is met and FR is 1, the algorithm executes 2.8.3; if the restart expectation is not met or the value of FR is neither 0 nor 1, execute step 2.10; Step 2.8.3, calculate the evolution rate OR(g) of the optimal solution of the population, which is calculated as follows: Compare OR(g) with the evolution rate of the previous generation optimal solution OR(g-1), and use N3 to record the number of consecutive generations of evolution rate decline; when N3 reaches the threshold If , the algorithm executes step 2.8.4, otherwise executes step 2.10; Step 2.8.4, calculate the population density PD, which is calculated as follows: in, Represents the mean of all individuals of the jth decision variable; when PD is less than the threshold λ, the algorithm executes a restart population generation method for the population falling into the dilemma of premature convergence in the feasible domain; otherwise, it executes a restart population generation method for the population falling into the dilemma of search stagnation in the feasible domain.

Images