Improved NSGA-II algorithm-based layered and partitioned cluster optimization and regulation and control method for vehicle network interaction

Abstract

The invention relates to a vehicle network interaction layering and partitioning cluster optimization and regulation and control method based on an NSGA-II algorithm. The method comprises the following steps that 1, a power grid layer collects historical data and basic load data of renewable energy output; 2, predicting conditions of power over-generation, over-use and insufficient supply possibly occurring in the region; step 3, the load aggregation layer constructs an electric vehicle and power grid interaction layering and partitioning cluster multi-objective optimization and scheduling model; and step 4, solving the electric vehicle and power grid interaction layered and partitioned cluster multi-target optimization and scheduling model by using an NSGA-II algorithm improved by Levy flight, and issuing a charging and discharging instruction to a distribution network layer. According to the method, peak regulation and renewable energy consumption regulation and control targets can be comprehensively considered, and the layered and partitioned interaction optimization and regulation and control strategy of the electric vehicle and the power grid is determined.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system optimal dispatching, and relates to a hierarchical and partitioned cluster optimization and control method for vehicle-grid interaction, in particular to a hierarchical and partitioned cluster optimization and control method for vehicle-grid interaction based on an improved NSGA-II algorithm. Background Art

[0002] Electric vehicles have the ability to respond quickly, accurately, and stably to grid regulation and control, and are an important resource for grid safety support. After aggregation, they can provide flexible regulation capabilities such as frequency modulation and reserve for the main grid, improve the grid restoration ability under extreme operating conditions, and effectively support the safe operation of the grid.

[0003] The access of a large number of distributed photovoltaics and electric vehicle charging and discharging equipment to the low-voltage distribution network has increased sharply. Facing conventional and extreme operating scenarios considering grid safety, the low-voltage distribution network layer needs to solve its transient safety problems, which have high-frequency and instantaneous characteristics; the main grid level needs to solve the supply-demand balance safety problems, and higher requirements are put forward for the flexibility of resources.

[0004] Existing technical means still pose great challenges in the hierarchical and partitioned collaborative optimization of multi-energy systems, and it is difficult to achieve efficient processing and accurate analysis of massive, real-time, and high-frequency data, thereby affecting the timeliness and accuracy of dispatching decisions. In the grid operation scenario, there is a lack of collaborative control technology for system stratification and control partition that comprehensively considers the safety requirements of the low-voltage distribution network and the main grid, and it is difficult to achieve efficient processing and accurate analysis of massive, real-time, and high-frequency data, thereby affecting the timeliness and accuracy of dispatching decisions.

[0005] Therefore, the present invention proposes a hierarchical and partitioned cluster optimization and control method for vehicle-grid interaction based on an improved NSGA-II algorithm.

[0006] After retrieval, no public literature of prior art identical or similar to the present invention has been found. Summary of the Invention

[0007] The purpose of the present invention is to overcome the deficiencies of the prior art and propose a hierarchical and partitioned cluster optimization and control method for vehicle-grid interaction based on an improved NSGA-II algorithm, which can comprehensively consider the regulation objectives of peak shaving and renewable energy consumption, and determine the optimal interaction and control strategy between electric vehicles and the grid at different levels and partitions.

[0008] The present invention solves its practical problems by adopting the following technical solutions:

[0009] A hierarchical and partitioned cluster optimization and control method for vehicle-grid interaction based on the NSGA-II algorithm, comprising the following steps:

[0010] Step 1: The grid layer collects historical data of renewable energy output and basic load data, and provides electricity price and grid load information to the load aggregation layer;

[0011] Step 2: Based on the renewable energy generation data and base load data collected in Step 1, and combined with the load conditions of the distribution network layer, the load aggregation layer predicts possible situations of over-generation, over-usage, and insufficient supply of electricity within the region.

[0012] Step 3: Based on the prediction results of Step 1, the load aggregation layer constructs a multi-objective optimization and scheduling model for the interaction between electric vehicles and the power grid with hierarchical and zonal clustering.

[0013] Step 4: Use the NSGA-II algorithm improved by Levy flight to solve the multi-objective optimization and scheduling model for the interaction between electric vehicles and the power grid with hierarchical and zonal clustering, and issue the charging and discharging instructions to the distribution network layer for guidance.

[0014] Advantages and beneficial effects of the present invention:

[0015] 1. Based on the hierarchical and zonal vehicle-grid interaction optimization and control strategy, the present invention can take into account the requirements of both the main grid layer and the distribution network layer at the same time. By intelligently managing the charging and discharging behaviors of electric vehicles, the adaptability of the power grid to different load scenarios is effectively improved.

[0016] 2. The improved NSGA-II algorithm proposed by the present invention enhances the global search ability by introducing Levy flight, avoids falling into local optima, and helps to find more efficient solutions in a more complex power grid control environment.

[0017] 3. By optimizing the charging and discharging behaviors of electric vehicles, the present invention can better track the power generation fluctuations of renewable energy, reduce power interaction, improve the utilization rate of renewable energy, and reduce the dependence on traditional power grid resources.

[0018] 4. The present invention effectively improves the flexibility and response speed of the power grid in multi-level zonal control. Especially when facing high-frequency and instantaneous power demands, it can achieve precise scheduling and real-time control of the charging and discharging behaviors of electric vehicles, reduce scheduling decision delays, and improve the overall control efficiency of the system. Brief Description of the Drawings

[0019] Figure 1 is a flowchart of a vehicle-grid interaction hierarchical and zonal clustering optimization and control method based on the NSGA-II algorithm of the present invention;

[0020] Figure 2 is a framework diagram of a vehicle-grid interaction hierarchical and zonal clustering optimization and control method based on the NSGA-II algorithm of the present invention.

[0021] Figure 3 is a flowchart of the NSGA-II algorithm improved by Levy of the present invention. Detailed Embodiments

[0022] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings:

[0023] A hierarchical and zonal clustering optimization and regulation method for vehicle-grid interaction based on the NSGA-II algorithm, as Figure 1 and Figure 2 shown, the method specifically includes the following steps:

[0024] Step 1: The power grid layer collects historical data of renewable energy output and basic load data, and provides electricity prices and grid load information to the load aggregation layer;

[0025] The specific steps of Step 1 include:

[0026] (1) The historical data of renewable energy output collected by the power grid layer includes: meteorological data (such as wind speed, temperature, humidity, light intensity, etc.), renewable power generation data, and regional electricity consumption historical data, such as the basic load on the power consumption side, reserve capacity, charging and discharging power of electric vehicles, etc.

[0027] (2) The power grid layer provides electricity prices and load information to the load aggregation layer by collecting and analyzing the basic load data of the power system.

[0028] In this embodiment, the power grid layer is responsible for collecting various basic data related to the operation of the regional power system, including historical data of renewable energy output, such as power generation records of new energy sources such as photovoltaic and wind power. In addition, the power grid layer systematically collects the historical electricity consumption data in the region, covering key parameters such as the basic load situation on the power consumption side, reserve power capacity, and charging and discharging power of electric vehicles. By continuously collecting and deeply analyzing these data, the power grid layer can construct the operation situation of the entire power system, and on this basis, calculate time-of-use electricity prices and load information, and timely provide necessary support information to the load aggregation layer to achieve subsequent power dispatch optimization.

[0029] Step 2: The load aggregation layer predicts possible situations of over-generation, over-consumption, and insufficient supply within the region based on the renewable energy generation data and basic load data collected in Step 1, in combination with the load situation of the distribution network layer.

[0030] The specific method of Step 2:

[0031] (1) Data cleaning and preprocessing:

[0032] Apply the quartile-based outlier detection method to screen the historical data of renewable energy output collected in Step 1, calculate the first quartile (Q1) and the third quartile (Q3) of the data, and then determine the upper and lower boundaries of the data according to the interquartile range (IQR = Q3 - Q1);

[0033] Any value exceeding the range of Q1 - 1.5×IQR or Q3 + 1.5×IQR is considered an outlier; data points outside these boundary ranges are regarded as outliers and these outlier points are deleted;

[0034] Using the linear interpolation method, construct a straight line through the data points before and after the missing point, and insert the missing value according to this straight line.

[0035] (2) Feature extraction:

[0036] Perform Min - Max normalization on the data; transform each data point to ensure that the values of all variables are within the range of (0, 1), unify the scales of different data sources; and use principal component analysis for feature extraction.

[0037]

[0038] (3) Divide the training set and the test set:

[0039] Divide the cleaned data set into a training set and a test set, generally with a ratio of 80% for training and 20% for testing; ensure that the time continuity or other relevant characteristics of the data are preserved during the division process.

[0040] (4) Predict the output of renewable energy based on LSTM:

[0041] The specific steps of step (4) in step 2 include:

[0042] ① Forget stage: The forget gate is responsible for removing irrelevant or outdated information in the historical data, ensuring that only the data valuable for predicting the current output of renewable energy is retained;

[0043] ② Update stage: The input gate determines which new output data of renewable energy should be updated into the memory unit, helping the model capture the changing trend of the output of renewable energy;

[0044] ③ Output stage: The output gate controls which information will be used for the final prediction result.

[0045] The relevant calculation formulas for each layer are shown in formulas (2) - (7):

[0046] f t =σ(W f [h t-1 ,x t +b f ) (2)

[0047] i t =σ(W i [x t ,h t-1 +b i ) (3)

[0048] C t = tanh(W c [x t ,h t-1 +b c ) (4)

[0049] U t = i t C t + f t C t-1 (5)

[0050] O t = σ(W o [x t ,h t-1 +b o ) (6)

[0051] h t = O t tanh(U t ) (7)

[0052] Wherein, f t , i t , O t , C t respectively represent the forgetting gate, input gate, output gate and control gate; W f , W i , W c , W o and b f , b i , b c , b o respectively represent the weight matrix and bias of each gate; U t , h t and σ respectively represent the cell state, hidden layer output and Sigmoid function; tanh is the activation function.

[0053] In this embodiment, the load aggregation layer serves as the core decision-making layer of power dispatching. After receiving the renewable energy generation data and basic load data provided by the power grid layer, it combines the current load conditions of the distribution network layer to perform power supply and demand balance prediction. Through the analysis and modeling of the collected data, the load aggregation layer can identify potential power supply and demand imbalance problems within the region and predict risks such as power over-generation, over-consumption, and insufficient supply that may occur.

[0054] Step 3: Based on the prediction results of Step 1, the load aggregation layer constructs a hierarchical and partitioned cluster multi-objective optimization and dispatching model for the interaction between electric vehicles and the power grid.

[0055] The specific method of Step 3 is as follows:

[0056] Based on the prediction results of Step 1, the load aggregation layer considers the selection and decision-making of electric vehicle charging methods,

[0057] constructs an electric vehicle charging model based on the travel chain, and then establishes a multi-objective optimization and scheduling model for the interaction between electric vehicles and the power grid with hierarchical zoning clusters.

[0058] The specific steps of Step 3 include:

[0059] (1) Based on the prediction results of Step 1, the load aggregation layer considers the selection and decision-making of electric vehicle charging methods, and constructs an electric vehicle charging model based on the travel chain;

[0060] The specific steps of Step (1) in Step 3 include:

[0061] 1) Input basic parameters. Input the basic parameters of the electric vehicle, including the travel time Initial grid-connected state of charge Parking duration Off-grid target state of charge Energy consumption during the trip Minimum standard of state of charge Maximum standard of state of charge Conventional AC pile charging power P i AC and fast DC pile charging power P i DC and the rated capacity of the power battery of the i-th electric vehicle

[0062] 2) Calculate the battery charge and driving energy consumption of the electric vehicle. The driving energy consumption of the i-th electric vehicle during the k-th trip is as shown in Equation (8):

[0063]

[0064] In the formula: v i,k and are respectively the average driving speed and driving duration of the i-th electric vehicle during the k-th trip, is the rated capacity of the power battery of the i-th electric vehicle.

[0065] Calculate the state of charge of the electric vehicle when it arrives at the destination

[0066]

[0067] 3) Judge whether charging is required. According to the state of charge at arrival and the off-grid target state of charge Determine whether charging is required. If charging is required, select a charging method. Based on the parking duration and the required charging duration, determine whether to choose conventional charging or fast charging.

[0068]

[0069] In the formula: is the user's charging duration at the conventional AC charging pile power P i AC under the standard. If the initial state of charge is higher than the target state of charge when leaving then no charging behavior occurs during the stay in this area.

[0070] If the user charges using the conventional charging method; if the user charges using the fast charging method. Assume that the electric vehicle starts charging immediately after parking in the charging area until at least is satisfied or reaches the highest standard of state of charge and then stops charging.

[0071] 4) Calculate the charging load and form a charging load curve. According to the charging method and charging duration, calculate the charging load curve of the electric vehicle during the stay at the destination. Simulate the charging loads of all electric vehicles to obtain the total charging load curve of the electric vehicles in the area.

[0072] (2) Construct a multi-objective optimization and scheduling model for the interaction between electric vehicles and the power grid with hierarchical partitioning and clustering;

[0073] The specific steps of step (2) in step 3 are as follows:

[0074] 1) Considering the following two factors comprehensively, construct the objective function of the multi-objective optimization and scheduling model for the interaction between electric vehicles and the power grid:

[0075] ① Minimize the response deviation of the total guiding charge and discharge power command P t ev _ sum as shown in Equation (11).

[0076]

[0077] In the formula: is the guiding charge and discharge power command of the i-th electric vehicle cluster in the t-th power grid scheduling cycle.

[0078] ② Minimize the operating economic cost of the electric vehicle cluster, as shown in Equation (12).

[0079] F2 = F t 1 -Ft 2 +F t 3 -F t 4 +F t 5 (12)

[0080] where: F t 1 represents the cost of all electric vehicle clusters in the region purchasing electricity from the superior power grid during the t-th power grid dispatching cycle, F t 2 represents the income obtained by each electric vehicle cluster in the region during the t-th power grid dispatching cycle when charging occurs at its location, F t 3 represents the cost of all electric vehicles participating in V2G incentives in the region during the t-th power grid dispatching cycle, F t 4 represents the income from V2G participation in the region during the t-th power grid dispatching cycle, F t 5 represents the economic penalty for the deviation of the total charge and discharge guidance power command during the t-th power grid dispatching cycle.

[0081] The specific calculation methods for each part of the cost are as follows:

[0082] a) Cost of purchasing electricity from the superior power grid

[0083]

[0084] where: is the electricity purchase price for the i-th electric vehicle cluster purchasing electricity from the superior power grid during the t-th power grid dispatching cycle.

[0085] b) User charging income

[0086]

[0087] where: is the charging power of the j-th electric vehicle in the i-th electric vehicle cluster during the t-th power grid dispatching cycle, N i is the number of electric vehicles in the i-th electric vehicle cluster during the t-th power grid dispatching cycle, is the real-time electricity price for the electric vehicle charging of the i-th electric vehicle cluster during the t-th power grid dispatching cycle. Among them, the real-time electricity price consists of two parts: a fixed electricity price and a floating electricity price.

[0088] Real-time electricity price The expression is as shown in Equation (17):

[0089]

[0090] In the formula: is the fixed electricity price of the i-th electric vehicle cluster in the t-th power grid dispatching period, is the floating electricity price of the i-th electric vehicle cluster in the t-th power grid dispatching period.

[0091] c) Incentive cost for users to participate in V2G:

[0092]

[0093] In the formula: The incentive electricity price for the operator of the i-th electric vehicle cluster in the t-th power grid dispatching period to promote users' participation in V2G. The specific expression is shown in Equation (21).

[0094]

[0095] In the formula: δ is the incentive response coefficient, and δ > 1.

[0096] d) Revenue obtained by the operator for participating in V2G:

[0097]

[0098] In the formula: is the electricity selling price of the i-th electric vehicle cluster responding to the power grid V2G in the t-th power grid dispatching period. To ensure the actual profit of the operator, it should be slightly higher than the incentive electricity price for electric vehicles to participate in V2G.

[0099]

[0100] In the formula: θ is the operator's electricity selling coefficient, and θ > 1.

[0101] e) Deviation economic penalty cost:

[0102]

[0103] In the formula: is the deviation penalty electricity price (yuan / kWh).

[0104] 2) Construct the constraint conditions of the multi-objective optimization and dispatching model for the interaction between electric vehicles and the power grid with hierarchical and zonal clustering;

[0105] The charging and discharging power instruction of the i-th electric vehicle cluster in the t-th power grid dispatching period shall not exceed the upper and lower boundaries of the adjustable capacity power of this cluster during this period, that is:

[0106]

[0107] In the formula: The guiding charging and discharging power command expected to be issued to the \(i\)-th electric vehicle cluster by the coordination layer in the \(t\)-th grid dispatching cycle.

[0108] The total guiding charging and discharging power command \(P\) of the electric vehicle cluster formulated in the \(t\)-th grid dispatching cycle t ev_sum shall not exceed the sum of the upper and lower bounds of the adjustable capacity power of each cluster during this period, as shown in Eqs. (26)-(28).

[0109] \(P\) t DCH_sum \(\leq P\) t ev_sum \(\leq P\) t CH_sum (24)

[0110]

[0111]

[0112] Where: is the lower limit of the maximum discharging capacity of the \(i\)-th electric vehicle cluster in the \(t\)-th grid dispatching cycle, and \(P\) t DCH _sum is the sum of the lower limits of the maximum discharging capacities of all \(n\) electric vehicle clusters in the region, is the upper limit of the maximum charging capacity of the \(i\)-th electric vehicle cluster in the \(t\)-th grid dispatching cycle, and \(P\) t CH_sum is the sum of the upper limits of the maximum charging capacities of all \(n\) electric vehicle clusters in the region.

[0113] All loads in the region cannot exceed the load upper limit \(P\) net set by the grid during any control period within the optimization interval, as shown in Eq. (27).

[0114] \(\vert P\) t ev_sum +\(P\) t non_ev \(\vert\leq P\) net (27)

[0115] Where: The non-electric vehicle load data \(P\) t non_ev predicted in the \(t\)-th grid dispatching cycle and the total guiding charging and discharging power command \(P\) t ev_sum of the electric vehicle cluster in the \(t\)-th grid dispatching cycle net .

[0116] Step 4: Solve the hierarchical and zonal clustering multi-objective optimization and scheduling model for the interaction between electric vehicles and the power grid using the Levy flight improved NSGA-II algorithm, and issue the charging and discharging instruction guidance to the distribution network layer.

[0117] Solve the hierarchical and zonal clustering multi-objective optimization and scheduling model for the interaction between electric vehicles and the power grid using the Levy flight improved NSGA-II algorithm, and issue the charging and discharging instruction guidance to the distribution network layer.

[0118] 1) Initialize the population. Randomly generate the initial population according to the charging and discharging demands of electric vehicles, the prediction of renewable energy output, and the power grid load demand. Each individual represents a charging and discharging scheduling plan for an electric vehicle cluster, including the charging power and discharging power at each time period.

[0119] 2) Calculate the objective functions. For each individual, calculate its corresponding two objective functions.

[0120] Objective 1: Minimize the charging and discharging power deviation. Calculate the deviation between the actual charging and discharging power and the charging and discharging power instruction required by the power grid. The goal is to minimize this deviation as much as possible to ensure that the electric vehicle cluster can effectively assist the power grid in balancing the load.

[0121] Objective 2: Minimize the economic cost. Calculate the economic cost of the electric vehicle cluster when performing charging and discharging operations, including the cost of purchasing electricity from the power grid, the incentive cost for participating in V2G, and the penalty cost caused by the power deviation.

[0122] 3) Introduce Levy flight. On the basis of the traditional NSGA-II algorithm, introduce the Levy flight mechanism. Levy flight is an optimization strategy based on random walk, which has the characteristics of short-distance search and occasional large-jump search, and can avoid the algorithm falling into local optimum. In the mutation operation, new solutions are generated through Levy flight to increase the diversity of the population and improve the global search ability, ensuring that the algorithm can find the global optimum solution. 4) Perform non-dominated sorting and calculate the crowding degree. Perform non-dominated sorting on the individuals in the population to identify the individuals with better performance in the two objective functions. Non-dominated sorting divides the individuals into different levels, and gives priority to those individuals that achieve a good balance between the minimum charging and discharging deviation and the minimum economic cost. For the individuals in the non-dominated solution set, calculate the crowding distance of each individual (i.e., the distribution density of the individual in the solution space), and give priority to retaining the individuals with a larger crowding degree to ensure the population diversity and avoid premature convergence.

[0123] 5) Cross - variation and population combination. Perform crossover and mutation operations on the individuals after non - dominated sorting to generate a new offspring population. The crossover operation generates offspring by combining the characteristics of two parent individuals, and the mutation operation introduces new solutions by randomly changing a part of the characteristics of individuals, further improving the diversity of the population. Combine the parent population and the offspring population to form a new population, and then repeat the non - dominated sorting and crowding degree calculation steps to retain the individuals with the best performance.

[0124] 6) Issue charging and discharging instructions. According to the current operating requirements and constraints of the power grid, select the most suitable charging and discharging scheduling scheme from the Pareto optimal solution set. This scheme should ensure that the electric vehicle cluster can flexibly respond to the load regulation requirements of the power grid at different times and meet the economic requirements at the same time. Convert the finally selected optimal scheduling scheme into specific charging and discharging instructions and transmit them to the distribution network layer. The distribution network layer manages the charging and discharging behaviors of the electric vehicle group according to this instruction, ensuring that electric vehicles perform corresponding charging or discharging operations at different time periods to support the real - time load balancing and optimal operation of the power grid.

[0125] In this embodiment, the working principle of step 3 is as follows:

[0126] 1) Electric vehicle charging method selection and decision - making

[0127] During a day's travel, an electric vehicle user may make three decisions: charging (including fast charging and slow charging), driving, and being idle (neither charging nor driving). Based on the travel chain model, it is possible to determine the driving time period and possible charging time periods of an electric vehicle user in a day.

[0128] The charging method selection of electric vehicle user i before the k - th trip may depend on the following factors: travel time Initial grid - connected state of charge of the battery Parking duration Off - grid target state of charge of the battery Energy consumption during the trip Minimum standard of state of charge Maximum standard of state of charge Conventional AC pile charging power P i AC and fast DC pile charging power P i DC and the rated capacity of the power battery of the i - th electric vehicle Among them, the initial grid - connected state of charge of the battery and the parking duration are determined by the (k - 1) - th trip of the i - th electric vehicle in the travel chain model.

[0129] Energy consumption during the k - th trip of the i - th electric vehicle while driving on the road As shown in formula (28).

[0130]

[0131] Where: v i,k and are respectively the average driving speed and driving duration of the i-th electric vehicle during the k-th trip, is the rated capacity of the power battery of the i-th electric vehicle.

[0132] Since the long-term low state of charge of the battery will affect the state of charge of the battery, it is necessary to set the minimum standard of the state of charge of the battery

[0133]

[0134] The charging decision-making judgment process of electric vehicles is as follows:

[0135] 1. Calculate the user's charging duration under the standard of the charging power P i AC of the conventional AC charging pile

[0136]

[0137] Where: is the user's charging duration under the standard of the charging power P i AC of the conventional AC charging pile. If the initial state of charge is higher than the departure target no charging behavior will occur during the stay in this area.

[0138] 2. Judge the relationship between the user's charging duration and the parking duration :

[0139] If the user charges in the conventional charging mode; if the user charges in the fast charging mode. It is assumed that the electric vehicle starts charging immediately after parking in the charging area until at least is satisfied or the highest standard of the state of charge is reached and charging stops. It should be noted that when using the fast charging mode and still unable to meet before the departure time , judge the relationship between before leaving and the energy consumption during driving on the road :

[0140] If that is, charge continuously in this parking area until the departure time If it can meet the driving energy consumption, fast charging is always used for charging;

[0141] If the sampling result of this travel chain is unreasonable, resample a new travel chain for subsequent calculations.

[0142] 2) Electric vehicle charging model based on travel chain

[0143] Based on the travel chain model and the selection and decision-making of electric vehicle charging methods, an electric vehicle charging model within the region is constructed. The specific steps are as follows.

[0144] 1. Input basic parameters, including the total number of electric vehicles N within the region, the total simulation time period T, the basic parameters of electric vehicles (conventional AC pile charging power P i AC , fast DC pile charging power P i DC and the rated capacity of the power battery of the i-th electric vehicle the minimum standard of state of charge the maximum standard of state of charge );

[0145] 2. Extract the travel chains of N electric vehicles, record the first travel time of each electric vehicle Determine the cumulative travel times num trip of each electric vehicle, and record the travel time driving duration arrival time and the parking duration at the destination of each electric vehicle during each travel, and establish an electric vehicle travel matrix Trip;

[0146] 3. Initialize the variable i, let i = 0, and record the electric vehicle number;

[0147] 4. Initialize the variable j, let j = 0, and record that the electric vehicle is about to make the j-th travel, that is, the travel number;

[0148] 5. Let i = i + 1, j = j + 1, and the simulation starts;

[0149] 6. If j = 1, that is, the i-th electric vehicle is about to complete its first travel, extract the first travel time then the state of charge at the travel time of the first travel of the i-th electric vehicle

[0150] 7. If j > 1, calculate the state of charge at the travel time of the j-th travel of the i-th electric vehicle:

[0151]

[0152] 8. Extract driving duration Calculate the energy consumption during driving on the road according to Equation (31)

[0153] 9. Calculate the entry at the destination

[0154]

[0155] 10. Extract the parking duration of the j-th trip at the destination Extract the departure time from the destination Extract the driving duration of the (j + 1)-th trip Calculate the energy consumption during driving on the road for the (j + 1)-th trip according to Equation (28) Calculate the minimum state of charge requirement for the i-th electric vehicle to meet the (j + 1)-th trip according to Equation (29)

[0156] 11. Determine the charging load curve of the electric vehicle during the parking period at the destination for the j-th trip according to the user's charging selection and decision-making scheme;

[0157] 12. Let j = j + 1. If j < num trip Go to Step 6 to calculate the charging load to ensure charging at the original parking location for the next trip; if j > num trip , that is, the charging load of the i-th electric vehicle has been simulated. Let i = i + 1 and j = 1 to calculate the charging load of the next electric vehicle;

[0158] 13. If i > N, that is, the charging loads of all N electric vehicles have been simulated, couple and compare with the electric vehicle travel matrix Trip and superimpose on the time axis to obtain the charging load curve after superposition of electric vehicles in each region.

[0159] 3) Hierarchical, Zonal, and Cluster Multi-objective Optimization and Scheduling Model for Electric Vehicle and Grid Interaction

[0160] In the hierarchical and zonal framework, direct communication between individual electric vehicle clusters is not possible, and the formulation of charging and discharging power commands for a certain electric vehicle cluster is restricted by the charging and discharging behaviors of other electric vehicle clusters. The load aggregator needs to formulate specific charging and discharging plans for each electric vehicle cluster on the basis of considering the interests of each cluster. On the basis of meeting the travel needs of electric vehicle users, considering the economic interests of each operator, and considering the response to the total charging and discharging power command P t ev _ sum to achieve coordinated cooperation between individual electric vehicle clusters.

[0161] a Objective function

[0162] One of the main objectives of the model is to minimize the response deviation from the total guiding charging and discharging power command P t ev _ sum and the other is to ensure the optimal economy of each electric vehicle cluster during the scheduling process. Therefore, considering the two factors comprehensively, a multi-objective coordination layer optimization model is constructed.

[0163] ① Minimize the response deviation from the total guiding charging and discharging power command P t ev _ sum as shown in Equation (33).

[0164]

[0165] In the formula: is the guiding charging and discharging power command for the i-th electric vehicle cluster in the t-th power grid scheduling cycle.

[0166] ② Minimize the operating economic cost of the electric vehicle cluster.

[0167] The composition of the economic cost mainly includes: the cost of purchasing electricity from the superior power grid, the incentive cost for users to participate in V2G, the income obtained by users from charging, the income of operators participating in V2G, and the deviation penalty cost, as shown in Equation (34).

[0168] F2 = F t 1 - F t 2 + F t 3 - F t 4 + F t 5 (34)

[0169] In the formula: F t 1 represents the cost of purchasing electricity from the superior power grid for all electric vehicle clusters in the region during the t-th power grid scheduling cycle, F t 2 represents the income obtained by each electric vehicle cluster in the region from charging during the t-th power grid scheduling cycle, F t 3 represents the incentive cost for all electric vehicles in the region to participate in V2G during the t-th power grid scheduling cycle, F t 4 represents the income from participating in V2G in the region during the t-th power grid scheduling cycle, F t 5 represents the economic penalty for the deviation from the total charging and discharging guiding power command during the t-th power grid scheduling cycle. The specific calculation methods for each part of the cost are as follows:

[0170] a) Cost of purchasing electricity from the superior power grid

[0171]

[0172] Wherein: is the electricity purchase price for the i-th electric vehicle cluster purchasing electricity from the superior power grid in the t-th power grid dispatching period.

[0173] b) User charging revenue

[0174]

[0175] Wherein: is the charging power of the j-th electric vehicle in the i-th electric vehicle cluster in the t-th power grid dispatching period, and N i is the number of electric vehicles in the i-th electric vehicle cluster in the t-th power grid dispatching period, is the real-time electricity price for electric vehicle charging in the i-th electric vehicle cluster in the t-th power grid dispatching period. Among them, the real-time electricity price consists of two parts: a fixed electricity price and a floating electricity price.

[0176] To ensure meeting the travel needs of electric vehicle users and the revenue of charging stations. The fixed electricity price adopts the electricity cost price for electric vehicle charging, and the floating electricity price ratio coefficient is determined by the proportion of the sum of non-electric vehicle load and electric vehicle charging load in the power grid to the capacity of the regional power grid transformer. The electric vehicle charging electricity price changes with the change of the real-time load. The sum of non-electric vehicle load and electric vehicle charging load in the power grid reflects the total load at the current moment. When the non-electric vehicle load is relatively high, as the electric vehicle charging load increases, the charging electricity price will increase, thereby suppressing the further increase of the load; on the contrary, when the non-electric vehicle load is relatively low, as the electric vehicle charging load decreases, the charging electricity price will decrease, thereby suppressing the further decrease of the load. The real-time electricity price The expression is as shown in Equation (39).

[0177]

[0178] Wherein: is the fixed electricity price of the i-th electric vehicle cluster in the t-th power grid dispatching period, is the floating electricity price of the i-th electric vehicle cluster in the t-th power grid dispatching period.

[0179] c) Incentive cost for users to participate in V2G:

[0180]

[0181] Wherein: The incentive electricity price for the operator of the i-th electric vehicle cluster in the t-th power grid dispatching period to promote users to participate in V2G. The specific expression is as shown in Equation (43).

[0182]

[0183] In the formula: δ is the excitation response coefficient, and δ > 1.

[0184] d) Revenue obtained by the operator participating in V2G:

[0185]

[0186] In the formula: is the electricity selling price for the i-th electric vehicle cluster responding to the grid V2G in the t-th grid scheduling period. To ensure the actual profit of the operator, it should be slightly higher than the incentive electricity price for electric vehicles participating in V2G.

[0187]

[0188] In the formula: θ is the electricity selling coefficient of the operator, and θ > 1.

[0189] e) Deviation economic penalty cost:

[0190]

[0191] In the formula: is the deviation penalty electricity price (yuan / kWh).

[0192] 2. Constraints

[0193] It provides a theoretical basis for formulating the specific guiding charging and discharging power instructions for each electric vehicle cluster in the response control layer in the t-th grid scheduling period. The guiding charging and discharging power instruction of the i-th electric vehicle cluster in the t-th grid scheduling period shall not exceed the upper and lower boundaries of the adjustable capacity power of the cluster during this period, that is:

[0194]

[0195] In the formula: is the guiding charging and discharging power instruction expected to be issued to the i-th electric vehicle cluster by the coordination layer in the t-th grid scheduling period.

[0196] The total guiding charging and discharging power instruction P t ev _ sum issued for the electric vehicle cluster in the t-th grid scheduling period shall not exceed the sum of the upper and lower boundaries of the adjustable capacity power of each cluster during this period, that is, as shown in Equations (46)-(48).

[0197] P t DCH_sum ≤P tev_sum ≤P t CH_sum (46)

[0198]

[0199] In the formula: is the lower limit of the maximum discharge capacity of the i-th electric vehicle cluster in the t-th power grid dispatching cycle, and P t DCH _sum is the sum of the lower limits of the maximum discharge capacities of all n electric vehicle clusters in the region, is the upper limit of the maximum charging capacity of the i-th electric vehicle cluster in the t-th power grid dispatching cycle, and P t CH_sum is the sum of the upper limits of the maximum charging capacities of all n electric vehicle clusters in the region.

[0200] All loads in the region cannot exceed the load upper limit P set by the power grid during any control period within the optimization interval net , as shown in formula (49).

[0201] |P t ev_sum +P t non_ev |≤P net (49)

[0202] In the formula: The non-electric vehicle load data P predicted in the t-th power grid dispatching cycle t non_ev and the total guiding charging and discharging power command P of the electric vehicle cluster in the t-th power grid dispatching cycle t ev_sum The absolute value of the sum cannot exceed the load upper limit P set by the power grid net .

[0203] 3. NSGA-II Algorithm Improved by Levy Flight

[0204] The Non-dominated Sorting Genetic Algorithms-II (NSGA-II) is a multi-objective genetic algorithm with a large influence and a wide range of applications. Because of its good global convergence and high generality, it is often used to solve optimization problems with multiple objective functions. However, as the number of iterations increases, the problem of insufficient population characteristics will appear, and it is easy to fall into local optima.

[0205] Levy flight is a random walk model, characterized by the flight step lengths following a heavy-tailed distribution, which means that compared with traditional random walks, the step lengths of Levy flights are smaller in most cases but occasionally produce large jumps. This search mechanism is very suitable for the global search of optimization algorithms and can help the algorithm quickly jump out of local optima. The randomness and jump characteristics of Levy flights endow it with stronger exploration ability in the search space, thus improving the global convergence performance of the algorithm. Introducing Levy flight into multi-objective optimization algorithms can enhance their global search ability and effectively solve the local convergence problem of NSGA-II in high-dimensional and complex environments. The Levy flight position update formula is shown in Eqs. (50)-(54).

[0206]

[0207] where is the position of x t at the t-th generation; α is the step size control parameter; Levy(λ) satisfies Eq. (53).

[0208] Levy~u=t -λ 1<λ≤3 (51)

[0209] The Mantegna algorithm is used to simulate the Levy distribution, and its step size s is:[[]]

[0210]

[0211] where: μ, ν are direction vectors, satisfying the normal distribution as shown in Eq. (55), and the standard deviation is as shown in (56):

[0212]

[0213] The improved NSGA-II algorithm generates a new sub-population according to the characteristics of Levy flight and merges it with the offspring population of the traditional NSGA-II. In the process of finding the global optimal solution, this is used to enhance the population diversity, strengthen the exploration ability, and minimize the probability of falling into local optima, thereby improving the standard NSGA-II algorithm in terms of calculation accuracy and convergence speed. Its algorithm flow is as follows:

[0214] 1. Initialize the population. According to the predicted parameters such as daily load, wind power generation, etc. input by the model, randomly generate an initial population of size N, and each individual in the population represents a possible solution.

[0215] 2. Generate the offspring population. Use the selection, crossover, and mutation operations of the genetic algorithm to generate the first-generation offspring population from the initial population. At the same time, use Levy flight to generate a new offspring population Q', and the individuals randomly generated by Levy flight help to enhance the population diversity and jump out of local optimal solutions.

[0216] 3. Merge the populations. Combine the parent population with the two offspring populations to form a population that encompasses a broader solution space.

[0217] 4. Non-dominated sorting and crowding distance calculation. Based on the objective functions and constraints of the multi-objective optimization problem, perform non-dominated sorting on the merged population and calculate the crowding distance of each individual. Select individuals with a lower non-dominated rank and a larger crowding distance, and retain the suitable individuals to form a new parent population.

[0218] 5. Iteration and termination conditions. If the maximum number of iterations is reached, output the Pareto optimal solution set that satisfies the constraints. If the maximum number of iterations is not reached, return to step 2, continue to iterate to generate new offspring populations and optimize the solution set until the output conditions are met.

[0219] Step 4: After receiving the guiding charging and discharging instructions from the load aggregation layer, the distribution network layer performs real-time management and control of the electric vehicles within the region based on the actual load conditions of the substation area. Ensure that resources within the region are optimally allocated and managed in the case of excessive load or insufficient power supply. The distribution network layer feeds back the key data during the execution process to the load aggregation layer to further enhance the overall coordination and adaptability of the system.

[0220] Step 5: The electric vehicles in the user layer adjust their own charging and discharging behaviors according to the control signals issued by the distribution network layer. During the peak period of the power grid load, they feed back power to the power grid to reduce the system load; while during the low load period, the electric vehicles perform charging operations to optimize the power utilization efficiency, and ensure the efficient execution of the charging and discharging behaviors through real-time monitoring. The charging and discharging data of the electric vehicles, including power, capacity utilization rate, response delay, etc., are fed back to the distribution network layer in real time to form an efficient control closed-loop.

[0221] The present invention will be further described below through specific examples:

[0222] Taking the electric vehicle charging and discharging pilot in a certain area as an example, simulate and analyze a hierarchical and zonal clustering optimization and control method for vehicle-grid interaction based on the improved NSGA-II algorithm.

[0223] (1) The power grid layer is responsible for collecting various basic data related to the operation of the regional power system, including historical data on the output of renewable energy sources, such as the power generation records of new energy sources like photovoltaic and wind power. In addition, the power grid layer systematically collects the historical electricity consumption data within the region, covering key parameters such as the basic load conditions on the electricity consumption side, the standby power capacity, and the charging and discharging power of electric vehicles. By continuously collecting and deeply analyzing this data, the power grid layer can construct the operating situation of the entire power system, and on this basis, calculate the time-of-use electricity price and load information, and timely provide the necessary support information to the load aggregation layer to achieve subsequent optimization of power dispatching. In the present invention, the dispatching period is divided into 24 hours, and the time-of-use electricity price data is shown in Table 1.

[0224] Table 1 Time-of-use electricity price

[0225]

[0226]

[0227] Statistically analyze the output of renewable energy sources and the basic load.

[0228] For simplicity of calculation, in the present invention, the number of vehicles refers to 0.1% of the vehicle ownership of 2.505 million in a certain city, and after rounding, it is approximately 2,500 vehicles as the electric vehicle ownership data. Taking a typical vehicle model for simulation verification, the battery capacity of this vehicle model is 56.4 kWh, the fast charging power is 35 kW, the conventional charging power is 7 kW, the power consumption per 100 kilometers is 20.5 kWh, the average driving speed on the road is 40 km / h, the initial state of charge at the start of the trip follows a uniform random distribution of U(0.5, 0.9), the minimum standard of the state of charge is set to 0.5, the simulation period is 1 day, the simulation interval is 1 h (60 min), and the upper limit of the distribution network load is 15 MW. The NSGA-II algorithm is set as follows: the population size pop is 300, the number of iterations gen is 50 times, the crossover rate is 0.8, the mutation rate is 1 / V (V is the number of control variables), the simulation period is 1 day, and the simulation interval is 1 h.

[0229] (2) As the core decision-making layer of power dispatching, the load aggregation layer, after receiving the renewable energy generation data and basic load data provided by the power grid layer, combines the current load situation of the distribution network layer to predict the power supply and demand balance. Through the analysis and modeling of the collected data, the load aggregation layer can identify potential power supply and demand imbalance problems within the region and predict risks such as power over-generation, over-consumption, and insufficient supply that may occur.

[0230] (3) The load aggregation layer further considers the selection and decision-making of electric vehicle charging methods. Based on the characteristics of the electric vehicle travel chain, it constructs a hierarchical, zonal, and clustered multi-objective optimization and scheduling model for the interaction between electric vehicles and the power grid. This model integrates factors such as the travel needs of electric vehicle users, the selection of charging periods, and the power grid load, aiming to optimize the impact of electric vehicle charging and discharging on the power grid load while meeting the travel needs of users. To solve the complex multi-objective optimization problem, the load aggregation layer adopts the NSGA-II algorithm, which can seek the optimal solution set among multiple objectives, finally generate a set of feasible charging and discharging guidance instructions, and issue them to the distribution network layer.

[0231] Under the strategy of the present invention, the total load in the region can track the output of renewable energy under the total guidance of the charging and discharging power instructions of the scheduling layer, so as to reduce the power interaction with the superior power grid and achieve the coordination of the output of renewable energy and the load in the region. At the same time, during the periods of 10:00 and 20:00 - 22:00, due to the limitation of the adjustable capacity boundary range of the electric vehicle cluster in the region, it is unable to completely track the output of renewable energy. During the 10:00 period, the renewable energy that cannot be absorbed in the region will be fed back to the superior power grid. During the 20:00 - 22:00 period, power purchase behavior needs to occur with the superior power grid to meet the load supply and demand balance in the region. Compared with unordered charging, the model of the present invention has better optimization effects. The analysis of the renewable energy consumption effect is carried out.

[0232] (4) After receiving the guidance charging and discharging instructions from the load aggregation layer, the distribution network layer conducts real-time management and control of the electric vehicles in the region based on the actual load conditions of the transformer substations. Ensure that the resources in the region are optimally allocated and managed in the case of excessive load or insufficient power supply. The distribution network layer feeds back the key data during the execution process to the load aggregation layer to further improve the overall coordination and adaptability of the system.

[0233] (5) The electric vehicles in the user layer adjust their own charging and discharging behaviors according to the control signals issued by the distribution network layer. During the peak period of the power grid load, they feed back power to support the power grid and reduce the system load; while during the low load period, the electric vehicles perform charging operations to optimize the power utilization efficiency and ensure the efficient execution of the charging and discharging behaviors through real-time monitoring. The charging and discharging data of electric vehicles, including power, capacity utilization rate, response delay, etc., are fed back to the distribution network layer in real time to form an efficient control closed-loop.

[0234] Under the action of the optimization model, the guidance charging and discharging power instructions of each cluster can take values within the upper and lower boundaries of the adjustable capacity, and specific guidance charging and discharging power instructions for each cluster are formulated under the dual objectives of economic optimality and minimum deviation. Moreover, the actual charging and discharging power of each cluster can well track the guidance charging and discharging power instructions of its own cluster, and the total guidance charging and discharging power instructions and the actual charging and discharging power are analyzed.

[0236] After retaining 4 decimal places for the deviation power and 2 decimal places for the economic cost in the operation results of the NSGA-II algorithm, the specific operation results are shown in Table 2. By analyzing the data in Table 2 and comparing the strategy in this section with unordered charging, it can be seen that an economic benefit of 43,069.90 yuan is obtained under the strategy in this section, while under the condition of unordered charging, the economic cost is 279,149.72 yuan, indicating that the strategy of the present invention has more advantages in terms of economy.

[0237] Table 2 Economic comparison between the strategy of the present invention and unordered charging

[0238]

[0239] It should be emphasized that the embodiments described in the present invention are illustrative rather than restrictive. Therefore, the present invention includes but is not limited to the embodiments described in the specific implementation manners. Any other implementation manners obtained by those skilled in the art based on the technical solutions of the present invention also belong to the scope protected by the present invention.

Claims

1. A vehicle-grid interactive hierarchical partition cluster optimization and control method based on NSGA-Ⅱ algorithm, characterized by: The following steps are involved: Step 1: The grid layer collects historical data on renewable energy output and basic load data, and provides electricity prices and grid load information to the load aggregation layer; Step 2: The load aggregation layer predicts the possible over-generation, over-usage and under-supply of electricity in the region based on the renewable energy generation data and basic load data collected in step 1 and the load conditions of the distribution network layer; Step 3: Based on the prediction results of step 1, the load aggregation layer constructs a multi-objective optimization and scheduling model for the interaction between electric vehicles and power grids by hierarchical partitioning and clustering; Step 4: Use the NSGA-Ⅱ algorithm improved by Levy flight to solve the multi-objective optimization and scheduling model of the interactive hierarchical and partitioned clusters of electric vehicles and power grids, and issue the charging and discharging instructions to the distribution network layer.

2. According to claim 1, a vehicle-grid interactive hierarchical partition cluster optimization and control method based on NSGA-Ⅱ algorithm is characterized by: The specific steps of step 1 include: (1) The power grid layer collects historical data on renewable energy output, including meteorological data, renewable power data, and regional electricity consumption data; (2) The grid layer collects and analyzes the basic load data of the power system and provides electricity price and load information to the load aggregation layer.

3. The vehicle-grid interactive hierarchical partition cluster optimization and control method based on NSGA-Ⅱ algorithm according to claim 1 is characterized by: The specific method of step 2 is: (1) Data cleaning and preprocessing: Apply the quartile-based outlier detection method to screen the historical data of renewable energy output collected in step 1, calculate the first quartile Q1 and the third quartile Q3 of the data, and then determine the upper and lower boundaries of the data according to the interquartile range IQR = Q3-Q1; Any value beyond Q1-1.5×IQR or Q3+1.5×IQR is considered an outlier; data points outside these boundary ranges are considered outliers and deleted; Use linear interpolation to construct a straight line through the data points before and after the missing point, and insert the missing value based on this straight line; (2) Feature extraction: right The data is processed by Min-Max standardization; each data point is transformed to ensure that the values ​​of all variables are in the range of (0,1) and to unify the scales of different data sources; and feature extraction is performed using principal component analysis: (3) Divide the training set and test set: Divide the cleaned data set into training and test sets, with a general ratio of 80% for training and 20% for testing; ensure that the temporal continuity or other relevant characteristics of the data are preserved during the division process; (4) Predict renewable energy output based on LSTM.

4. The vehicle-grid interactive hierarchical partition cluster optimization and control method based on NSGA-Ⅱ algorithm according to claim 3 is characterized by: The specific steps of step (4) of step 2 include: ① Forget stage: The forget gate is responsible for removing irrelevant or outdated information in historical data to ensure that only data that is valuable for current renewable energy output forecasts is retained; ② Update phase: The input gate determines which new renewable energy output data should be updated to the memory unit, helping the model capture the changing trend of renewable energy output; ③ Output stage: The output gate controls which information will be used for the final prediction result; The relevant calculation formulas for each layer are shown in formulas (2)-(7): f t =σ(W f [h t-1 ,x t ]+b f ) (2) i t =σ(W i [x t ,h t-1 ]+b i ) (3) C t =tanh(W c [x t ,h t-1 ]+b c ) (4) U t =i t C t +f t C t-1 (5) The t =σ(W o [x] t ,h t-1 ]+b o ) (6) h t =O t fishy t ) (7) In the formula, f t 、i t , O t , C t Respectively represent the forget gate, input gate, output gate and control gate; W f , W i , W c , W o and b f 、b i 、b c 、b o Represents the weight matrix and bias of each gate respectively; U t 、h t and σ represent the unit state, hidden layer output and Sigmoid function respectively; tanh is the activation function.

5. The vehicle-grid interactive hierarchical partition cluster optimization and control method based on NSGA-Ⅱ algorithm according to claim 1 is characterized by: The specific steps of step 3 include: (1) Based on the prediction results of step 1, the load aggregation layer considers the selection and decision of electric vehicle charging methods and constructs an electric vehicle charging model based on the travel chain; (2) Construct a multi-objective optimization and scheduling model for the interaction between electric vehicles and power grids through hierarchical and partitioned clustering.

6. The vehicle-grid interactive hierarchical partition cluster optimization and control method based on NSGA-Ⅱ algorithm according to claim 5 is characterized by: The specific steps of step (1) of step 3 include: 1) Input basic parameters: Input basic parameters of electric vehicles, including travel time Initial battery charge level Parking duration Off-grid target battery charge level Travel energy consumption Minimum state of charge Highest state of charge Conventional AC charging pile power P i AC 、Fast DC charging power P i DC and the rated capacity of the power battery of the i-th electric vehicle 2) Calculate the battery charge and driving energy consumption of electric vehicles; the driving energy consumption of the i-th electric vehicle on the k-th trip As shown in formula (8): Where: v i,k and are the average speed and driving time of the kth trip of the i-th electric vehicle, is the rated capacity of the power battery of the i-th electric vehicle; Calculate the charge of an electric vehicle when it arrives at its destination 3) Determine whether charging is needed; based on the battery charge at arrival and off-grid target charge level Determine whether charging is needed; if charging is needed, select the charging method; determine whether to choose regular charging or fast charging based on the parking time and the required charging time; Where: is the charging power P at a conventional AC pile i AC Standard user charging time, if the entry charge Above exit target No charging will occur during the stay in this area; like If the user uses conventional charging methods to charge; The user uses fast charging; assuming that the electric vehicle starts charging after entering the charging area and stops, until at least or reach the highest state of charge Stop charging; 4) Calculate the charging load and form a charging load curve; calculate the charging load curve of the electric vehicle during the parking period at the destination according to the charging method and charging time; simulate the charging load of all electric vehicles to obtain the total charging load curve of electric vehicles in the area.

7. The vehicle-grid interactive hierarchical partition cluster optimization and control method based on NSGA-Ⅱ algorithm according to claim 5 is characterized by: The specific steps of step (2) of step 3 include: 1) Considering the following two factors comprehensively, the objective function of the multi-objective optimization and dispatching model of the interactive hierarchical partition cluster of electric vehicles and power grid is constructed: ①Total charge and discharge power instruction P t ev_sum The response deviation is the smallest, as shown in formula (11); Where: is the guiding charging and discharging power instruction of the ith electric vehicle cluster in the tth grid dispatch cycle; ② The economic cost of electric vehicle cluster operation is the lowest, as shown in formula (12); F2=F t 1 -F t 2 +F t 3 -F t 4 +F t 5 (12) Where: F t 1 represents the cost of all electric vehicle clusters purchasing electricity from the upper grid in the tth grid dispatch cycle, F t 2 represents the revenue gained by each electric vehicle cluster from charging in the tth grid dispatching cycle, F t 3 represents the incentive cost for all electric vehicles to participate in V2G in the tth grid dispatch cycle, F t 4 represents the benefits of V2G participation in the tth grid dispatching period, F t 5 represents the economic penalty for the deviation of the total charge and discharge guidance power instruction in the tth grid dispatch cycle; The specific calculation method of each part’s cost is as follows: a) Cost of purchasing electricity from the upper-level power grid Where: is the electricity purchase price of the ith electric vehicle cluster from the superior power grid in the tth power grid dispatch cycle; b) User charging revenue Where: is the charging power of the jth electric vehicle in the i-th electric vehicle cluster in the t-th grid dispatch period, N i is the number of electric vehicles in the ith electric vehicle cluster in the tth grid dispatch period, is the real-time electricity price of electric vehicle charging in the ith electric vehicle cluster in the tth grid dispatch cycle; among them, the real-time electricity price It consists of two parts: fixed electricity price and floating electricity price; Real-time electricity prices The expression is shown in formula (17): Where: is the fixed electricity price of the ith electric vehicle cluster in the tth grid dispatch period, is the floating electricity price of the ith electric vehicle cluster in the tth grid dispatch period; c) Incentive cost for users to participate in V2G: Where: The incentive electricity price set by the ith electric vehicle cluster operator in the tth grid dispatch cycle to promote user participation in V2G; The specific expression is shown in formula (21); Where: δ is the excitation response coefficient, and δ>1; d) Benefits of operators participating in V2G: Where: is the electricity price of the ith electric vehicle cluster responding to the V2G of the grid in the tth grid dispatch cycle; to ensure the actual profit of the operator, it should be slightly higher than the incentive price for electric vehicles to participate in V2G; Where: θ is the operator's electricity sales coefficient, and θ>1; e) Deviation economic penalty cost: Where: is the penalty electricity price for deviation (yuan / kWh); 2) Construct the constraints of the multi-objective optimization and dispatching model for the interaction between electric vehicles and power grids; Guiding charging and discharging power instructions for the ith electric vehicle cluster in the tth grid dispatch cycle The upper and lower limits of the adjustable power capacity of the cluster during the period must not be exceeded, that is: Where: The coordination layer estimates the charging and discharging power instructions issued to the i-th electric vehicle cluster in the t-th grid dispatch cycle; The total guiding charging and discharging power instruction P of the electric vehicle cluster formulated in the tth grid dispatch cycle t ev_sum It shall not exceed the sum of the upper and lower limits of the controllable power of each cluster during the period, as shown in equations (26)-(28); P t DCH_sum ≤P t ev_sum ≤P t CH_sum (24) Where: is the lower limit of the maximum discharge capacity of the ith electric vehicle cluster in the tth grid dispatch cycle, P t DCH_sum is the sum of the lower limits of the maximum discharge capacity of all n electric vehicle clusters in the region, is the upper limit of the maximum charging capacity of the ith electric vehicle cluster in the tth grid dispatch cycle, P t CH_sum The sum of the maximum charging capacity limits of all n electric vehicle clusters in the region; All loads in the area cannot exceed the load upper limit P set by the power grid during any control period in the optimization interval. net , which is shown in formula (27); |P t ev_sum +P t non_ev |≤P net (27) Where: The non-electric vehicle load data P predicted in the tth grid dispatch cycle t non_ev The total charging and discharging power instruction P of the electric vehicle cluster in the tth grid dispatch cycle t ev_sum The absolute value of the sum cannot exceed the load upper limit P set by the power grid. net .

8. The vehicle-grid interactive hierarchical partition cluster optimization and control method based on NSGA-Ⅱ algorithm according to claim 1 is characterized by: The specific steps of step 4 include: 1) Initialize the population; randomly generate the initial population according to the charging and discharging requirements of electric vehicles, the output forecast of renewable energy, and the load demand of the power grid; each individual represents a charging and discharging scheduling plan for an electric vehicle cluster, including the charging power and discharging power in each period; 2) Objective function calculation: For each individual, calculate the two corresponding objective functions; Goal 1: Minimize the charge and discharge power deviation; calculate the deviation between the actual charge and discharge power and the charge and discharge power instructions required by the power grid, with the goal of minimizing this deviation; Objective 2: Minimize economic costs; calculate the economic costs of the electric vehicle cluster when performing charging and discharging operations, including the cost of purchasing electricity from the grid, the incentive fee for participating in V2G, and the penalty cost caused by power deviation; 3) Introducing Levy flight: Based on the traditional NSGA-II algorithm, the Levy flight mechanism is introduced; 4) Non-dominated sorting and crowding calculation: Perform non-dominated sorting on the individuals in the population and identify the individuals that perform better on the two objective functions; Non-dominated sorting divides individuals into different levels and gives priority to those individuals that achieve a good balance between the minimum charge-discharge deviation and the lowest economic cost; For the individuals in the non-dominated solution set, calculate the crowding distance of each individual and give priority to retaining individuals with a larger crowding degree; 5) Crossover, mutation and merging populations: Perform crossover and mutation operations on the individuals after non-dominated sorting to generate a new offspring population; then repeat the non-dominated sorting and crowding calculation steps to retain the individuals with the best performance; 6) Issue charging and discharging instructions; select the most suitable charging and discharging scheduling scheme from the Pareto optimal solution set according to the current operation requirements and constraints of the power grid; convert the final selected optimal scheduling scheme into specific charging and discharging instructions and pass them to the distribution network layer; the distribution network layer manages the charging and discharging behavior of the electric vehicle group according to the instructions.

Images