Power and transportation multi-layer network coupling model-based power transmission and distribution network collaborative optimization method and system

Abstract

The application relates to the technical field of power transmission and distribution network collaborative optimization, in particular to a power transmission and distribution network collaborative optimization method and system considering a power and traffic multi-layer network coupling model, which aims at the problems of convergence difficulty and time-consuming solution of the power transmission and distribution network collaborative optimization model considering power-traffic network coupling, adopts a McCormick envelope relaxation method to relax non-convex constraints, and adopts a large M convex optimization balance method to determine the upper and lower bounds of variables. Since the operation feasible region of the power distribution network at a single time is influenced by current state parameters of the system, the application adopts a deep learning method to realize the prediction of the operation feasible region of the power distribution network at a single time, and simultaneously, in order to improve the prediction accuracy, the lower boundary of the operation feasible region of the power distribution network is approximately processed.

Description

Technical Field

[0001] The present invention relates to the technical field of coordinated optimization of power transmission and distribution networks, and in particular to a coordinated optimization method and system for power transmission and distribution networks considering a multi-layer network coupling model of power and transportation. Background Art

[0002] For electric vehicle users, when driving in the transportation network, they will have a demand for charging at some point, and thus connect to the power grid through the charging facilities of a charging station at a certain node. In other words, the charging behavior of electric vehicles is a key factor affecting the operational reliability of the power-transportation network coupling, which creates certain requirements for collaborative optimization methods.

[0003] The centralized solution method for the coordinated optimization of the transmission and distribution network considering the coupling of the power and transportation networks has many variables, is difficult to solve, and requires the collection of parameter information of all devices, which makes privacy protection difficult; while the distributed optimization model requires multiple iterations to complete the solution, resulting in difficulty in model convergence and time-consuming solution.

[0004] For example, the Chinese invention patent application with publication number CN117649082A discloses a method and system for optimizing the operation reliability of a traffic-power coupling network. The method first constructs a traffic-power coupling network model; obtains real-time operation data of the traffic-power coupling network based on the traffic-power coupling network model; calculates a traffic network reliability index and a power network reliability index based on the real-time operation data of the traffic-power coupling network; collects facial image information and calculates a user's acceptable scheduling rationality value based on the facial image information; generates and executes an optimal scheduling plan based on the traffic network reliability index, the power network reliability index and the user's acceptable scheduling rationality value; as disclosed in the Chinese invention patent application with publication number CN11672068. The Chinese invention patent application for 9A discloses a two-layer game model and equilibrium solution method for an electric power-transportation coupling network that considers the elastic demand behavior of electric vehicles. It establishes a hybrid traffic flow model that includes electric vehicle demand elasticity and charging behavior, realizing the coupling of urban electric power and transportation networks. Considering the optimization objectives and operational constraints of the coupled network equilibrium problem, a quasi-variational inequality framework is used to characterize the equilibrium state of elastic mixed traffic users, and a mathematical optimization model for the urban electric power and transportation coupling network is established. In response to the non-convexity of the problem, an outer layer layered fixed point iteration is used to decouple the problem into a two-layer convex optimization problem. The inner layer equilibrium problem is solved based on the viscous projection approximation algorithm, forming a complete method for solving the two-layer game equilibrium state. However, none of the above considers the processing method of the three-layer model and does not involve the relevant processing and calculation of collaborative optimization. Summary of the Invention

[0005] Purpose of the invention: In order to overcome the shortcomings of the existing technology, the present invention provides a transmission and distribution network collaborative optimization method considering the multi-layer network coupling model of electricity and transportation. This method solves the problems of privacy protection difficulties and time-consuming solutions in traditional optimization methods. The present invention also provides a transmission and distribution network collaborative optimization system considering the multi-layer network coupling model of electricity and transportation.

[0006] Technical solution: According to a first aspect of the present invention, a method for collaborative optimization of a transmission and distribution network considering a multi-layer network coupling model of power and transportation is provided, the method comprising:

[0007] A transmission and distribution network collaborative optimization model considering the coupling of power and transportation networks is constructed. The model is a three-layer distributed structure consisting of the transmission network layer, the distribution network layer, and the transportation network layer.

[0008] Construct the KKT conditions corresponding to the transportation network layer and project the KKT conditions into the distribution network layer, thereby reducing the dimension of the three-layer optimization model to a two-layer optimization model;

[0009] The distribution network layer after adding KKT conditions is set as a convex model using the large-M convex optimization balance method, and the upper and lower bounds of Lagrange multipliers and traffic flow are determined;

[0010] The vertex search method is used to solve the feasible domain of the above convex model, and the number of lower boundary points of the feasible domain obtained at different times is controlled by linear interpolation and Douglas-Peucker algorithm to obtain training data of uniform dimension.

[0011] The deep neural network is trained using the data set obtained from the training data, and then the transmission and distribution network is optimized collaboratively based on the prediction results output by the neural network.

[0012] Further, including:

[0013] The objective function of the transmission and distribution network collaborative optimization model considering the coupling of power and transportation networks is set to minimize the total cost of the transmission and distribution networks, specifically: the generator cost of the transmission and distribution networks is minimized, the abandoned photovoltaics are minimized, and the soft switch SOP loss in the distribution network is minimized;

[0014] The constraints considered at the transmission network layer include: node power balance constraints, DC power flow constraints, generator output upper and lower limit constraints, unit ramp constraints, unit minimum start and stop time constraints, and photovoltaic constraints;

[0015] The constraints considered at the distribution network layer include: node power balance constraints, voltage drop constraints, current constraints, generator output constraints, photovoltaic constraints, voltage constraints and transmission and distribution network boundary node constraints;

[0016] The traffic network layer includes a charging power model and a traffic flow distribution model based on each path.

[0017] Further, including:

[0018] The traffic flow distribution model is expressed as: The objective function of this model is to minimize the total driving time, charging time and the cost of charging electric vehicles, where: is the set of all paths, It is the collection of all charging stations. is the set of all OD pairs, It is an OD pair, where the OD pair is used to describe the traffic demand between the starting point and the end point. It's OD The set of all paths of Is the path Traffic on It's a charging station Traffic on Is the path The driving time on It's at the charging station The waiting time required for charging, and The paths are Basic driving time and charging stations Basic charging waiting time, and The road sections and charging stations capacity, It's OD path The cost of It's OD path The travel time on the road segment, It's OD path traffic flow, For OD travel needs, It's OD The demand function of It's OD The greatest travel demand, and OD pairs path Whether it passes through the road section and charging stations The flag variable, It's OD path of electric vehicles at charging stations The charging capacity, It's a charging station charging load.

[0019] Further, including:

[0020] The step of constructing the KKT conditions corresponding to the transportation network layer and projecting the KKT conditions to the distribution network layer includes:

[0021] Set the KKT condition corresponding to the traffic flow assignment model, expressed as:

[0022] in, It's OD path The sum of the total driving time, charging time and cost, and They are respectively and The corresponding Lagrange multiplier;

[0023] Therefore, After offset, it is expressed as: Formulas (16)-(18) are added as constraints to the constraints considered at the distribution network layer.

[0024] Further, including:

[0025] The method of using the large-M convex optimization balance method to set the distribution network layer after adding the KKT condition as a convex model includes:

[0026] According to the complementary relaxation constraint It can be seen that the traffic flow distribution model is a bilinear term, which leads to the distribution network model after adding the KKT condition to be a non-convex model; the definition of the bilinear term involves two variables x and y , their product is x × y This product term often appears in optimization models, especially when it is necessary to deal with the interdependence between variables. Bilinear terms can be used to describe complex interactions.

[0027] The McCormick envelope relaxation method is used to transform the bilinear term into a linear form, thereby transforming the above non-convex model into a convex model, which can be expressed as follows: ; ; ;

[0028] in, and They are the Lagrange multipliers The lower and upper bounds of and Traffic flow The lower and upper bounds of .

[0029] Further, including:

[0030] The determination of the upper and lower bounds of the Lagrange multiplier and the traffic flow includes:

[0031] Using the Big M method to transform non-convex constraints into linear constraints, we get a mixed integer programming problem as follows: ;

[0032] in, is a very large positive number, is an auxiliary binary variable, by adjusting the auxiliary binary variable The value of , decomposes the problem into convex optimization sub-problems;

[0033] By changing the variable parameters in the traffic flow distribution model, and recording the range of the results obtained using the big M method and , in each record and Keep binary variables within the interval The value remains unchanged. At this time, all and All in and between, thus determining and The upper and lower bounds of .

[0034] Further, including:

[0035] The method of using the vertex search method to solve the operational feasible domain of the convex model includes:

[0036] Determine the vertex search model according to the constraints of the distribution network itself, determine the initial vertex according to the dimension of the coupling variable, thereby forming an initial convex hull, and store the initial vertex in the initialization vertex set;

[0037] Search for new vertices along the outer normal direction of each edge of the convex hull through an inner loop, and record the corresponding improvement ratio, which gradually decreases with the number of iterations;

[0038] Until the last iteration, the error between the current convex hull and the true projection is determined through the outer loop is smaller than the allowable error, and the loop is terminated, thereby obtaining the feasible domain of operation corresponding to the current moment.

[0039] Further, including:

[0040] The method of determining a vertex search model according to the constraints of the distribution network itself, determining an initial vertex according to the dimension of the coupling variable, thereby forming an initial convex hull, and storing the initial vertex in an initialization vertex set includes:

[0041] Determine the vertex search model

[0042] in, is the objective function of vertex search, The value of the vector represents the direction of vertex search. and The coupling variables are and internal variables The coefficient matrix of is the constant coefficient vector on the right side;

[0043] The number of initial vertices is at least the number of coupling variables The corresponding dimension is increased by one, and the initial vertex is determined along the axis.

[0044] Further, including:

[0045] The inner loop searches for new vertices along the outer normal direction of each edge of the convex hull and records the corresponding improvement ratio, which decreases gradually with the number of iterations, including:

[0046] Jordi The convex hull composed of existing vertices generated by the second cycle is , assuming Depend on planes, the new vertex is along Search in the direction of the external normal vector of the plane;

[0047] No. The outward normal vector of a plane is , then the newly identified vertex For the order The optimal solution found later;

[0048] Indicates new vertex The improvement ratio of the current convex hull is:

[0049] in, It is The external normal vector of a plane, is the constant term on the right side of the corresponding hyperplane equation; if ,show exist Outside, will Add vertex set , making Closer to the actual feasible domain, and record the current improvement ratio. As the number of iterations increases, the improvement ratio gradually decreases. Further, including:

[0050] Until the last iteration, the error between the current convex hull and the true projection is determined by the outer loop Less than the allowable error, and terminate the loop to obtain the feasible domain of the operation corresponding to the current moment, including: using Hausdorff distance to evaluate the error between the current convex hull and the true projection , expressed as:

[0051] in, The convex hull of the approximate feasible region composed of existing nodes The set of all planes of ;

[0052] When the error When the error is less than the allowed error, the vertex search method terminates the iteration and outputs the convex hull of the existing nodes. As the final result, otherwise, enter the next outer loop.

[0053] Further, including:

[0054] The number of lower boundary points of the operational feasible region obtained at different times is controlled by using linear interpolation and Douglas-Peucker algorithm, including:

[0055] The number of lower boundary points of the distribution network's operational feasible domain at the current moment is determined. When the current number of points is less than the specified number, new lower boundary points are added between the existing points using linear interpolation to carefully depict the feasible domain without changing its shape. When the number of points exceeds the specified number, the Douglas-Peucker algorithm is applied to simplify the number of lower boundary points to reduce the number of lower boundary points while minimizing the impact on the shape of the feasible domain.

[0056] Further, including:

[0057] The Douglas-Peucker algorithm is applied to simplify the number of lower boundary points to reduce the number of lower boundary points while minimizing the impact on the shape of the feasible region, including:

[0058] Connect the first and last points of the current lower boundary point to form the initial line segment, and mark the first and last points as reserved points;

[0059] Find the point farthest from the initial line segment among all internal points ,if In tolerance If the tolerance is not within the range, the simplification is completed and all internal points are discarded. Inside, then must be retained, and new parts are processed recursively; ultimately, the output curve is the set of all retained points where for The maximum distance to a line segment.

[0060] On the other hand, the present invention also provides a transmission and distribution network collaborative optimization system considering a multi-layer network coupling model of power and transportation, the system comprising:

[0061] The model building module is used to construct a transmission and distribution network collaborative optimization model that considers the coupling of power and transportation networks. The model is a three-layer distributed structure: the transmission network layer, the distribution network layer, and the transportation network layer.

[0062] The model dimensionality reduction module is used to construct the KKT conditions corresponding to the transportation network layer and project the KKT conditions into the distribution network layer, thereby reducing the dimensionality of the three-layer optimization model into a two-layer optimization model;

[0063] The model conversion module is used to set the distribution network layer after adding the KKT condition as a convex model using the large-M convex optimization balance method, and to determine the upper and lower bounds of the Lagrange multiplier and traffic flow;

[0064] The feasible domain calculation module is used to solve the operational feasible domain of the above-mentioned convex model using the vertex search method, and to control the number of lower boundary points of the operational feasible domain obtained at different times using linear interpolation and the Douglas-Peucker algorithm to obtain training data of uniform dimension;

[0065] A neural network training module is used to train a deep neural network using a data set obtained from the training data, and then achieve collaborative optimization of the transmission and distribution network based on the prediction results output by the neural network.

[0066] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0067] (1) In order to solve the problems of privacy protection difficulties and time-consuming solutions in traditional optimization methods, the present invention adopts the equivalent projection method and KKT conditions to map the operating feasible domain of the distribution network from a high-dimensional space to a low-dimensional space containing only coupling variables. While protecting privacy, it avoids iterative solutions, thereby achieving accelerated solution.

[0068] (2) The present invention adopts the McCormick envelope relaxation method to relax non-convex constraints, and adopts the large-M convex optimization balance method to determine the better upper and lower bounds of variables. This makes it possible to find a balance between feasibility and accuracy, and selects upper and lower bounds that can both ensure the feasibility of the model and reduce the error. The corresponding model is also more accurate.

[0069] (3) In order to improve the accuracy of prediction, the present invention adopts the vertex search method to collect the data set of the feasible domain of distribution network operation, and approximates the lower boundary of the feasible domain. On this basis, in order to solve the problem of inconsistent neural network input dimensions caused by the change in the number of lower boundary points in the distribution network operation analysis, an optimization method is proposed. This method combines linear interpolation and Douglas-Peucker algorithm to improve data consistency, optimize the training and prediction performance of the neural network model, and verifies its effectiveness in maintaining model accuracy and generalization ability through experiments.

[0070] (4) Since the operational feasible domain of the distribution network at a single moment is affected by the current state parameters of the system, the present invention adopts a deep learning method to predict the operational feasible domain of the distribution network at a single moment.

[0071] (5) The deep learning method of this application can accurately predict the shape of the lower boundary of the feasible region of distribution network operation, while ensuring the optimality of the solution and the speed of solution. Case studies show that the collaborative optimization prediction results obtained in this application are highly consistent with the actual values ​​in terms of overall trend, especially in the 0-2MW range. Although there are slight errors, the area enclosed by the predicted lower boundary is almost completely consistent with the actual feasible region, indicating that the model can effectively capture key features and has strong robustness. The stable performance in different power ranges demonstrates its good predictive ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 This is a topology diagram of the T57D33U12 test system according to an embodiment of the present invention;

[0073] Figure 2 is a parameter curve diagram of the test system according to an embodiment of the present invention;

[0074] Figure 3 is a comparison diagram of the solution results described in an embodiment of the present invention;

[0075] Figure 4 is the binary variable described in the embodiment of the present invention With the trend of traffic load proportional coefficient;

[0076] Figure 5 The maximum value of electric vehicle travel and the maximum value of Lagrange multiplier are change trends according to the embodiment of the present invention;

[0077] Figure 6 This is a comparison of the solution results of the three solutions described in the embodiments of the present invention;

[0078] Figure 7 This is the process of solving the feasible domain using the vertex search method described in the embodiment of the present invention;

[0079] Figure 8 is an improved ratio graph of the vertex search method according to an embodiment of the present invention;

[0080] Figure 9 This is a schematic diagram of approximating the lower boundary of the feasible region according to an embodiment of the present invention;

[0081] Figure 10 is a graph of training error and test error according to an embodiment of the present invention;

[0082] Figure 11 The comparison between the feasible domain prediction result and the actual value described in the embodiment of the present invention;

[0083] Figure 12 This is a comparison of feasible regions of distribution networks at different times according to an embodiment of the present invention;

[0084] Figure 13 This is a flow chart of a non-iterative solution method for collaborative optimization of transmission and distribution networks considering power-transportation network coupling according to an embodiment of the present invention. DETAILED DESCRIPTION

[0085] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention and not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0086] Example 1: Figure 13 As shown, the embodiment of the present invention considers the non-iterative solution method for the coordinated optimization of the power transmission and distribution network with the coupling of the power and transportation network. First, a coordinated optimization model of the power transmission and distribution network with the coupling of the power and transportation network is formed. The original model is time-consuming to solve and has difficulty in convergence. Secondly, the McCormick envelope relaxation method is used to process the non-convex constraints and convert it into a convex optimization model. Thirdly, the vertex search method is used to collect the data set of the feasible region of the distribution network operation, and the lower boundary of the feasible region is approximated. Finally, the training data set is used to train the deep neural network, and the prediction results output by the neural network are used to realize the rapid solution of the coordinated optimization of the power transmission and distribution network. The method mainly includes the following steps:

[0087] The first step is to form a transmission and distribution network collaborative optimization model that considers the coupling of power and transportation networks, and add the KKT conditions of the transportation network to the distribution network. This model is huge in scale, has many variables, and has complex constraints, making it difficult to solve.

[0088] In general, the objective function of the coordinated optimization of the transmission and distribution network is set to minimize the total cost of the transmission and distribution network, which mainly includes minimizing the cost of the transmission and distribution network generators, minimizing the abandoned photovoltaic power generation, and minimizing the soft switch (Soft Open Points, SOP) loss in the distribution network, such as: ; ;

[0089] in, and are the operating costs of the transmission and distribution networks, is the set of all time scales, and are the collection of all the group nodes in the transmission network and the distribution network, and are the collection of all photovoltaic nodes in the transmission grid and distribution grid respectively, is the collection of all SOP nodes in the distribution network, and are the unit power generation costs of generators in the transmission and distribution networks, and are the unit penalty costs of abandoning PV in the transmission grid and distribution grid, is the loss conversion cost coefficient of SOP in the distribution network, and are the active power injected into the node by the transmission network and distribution network generators, and are the photovoltaic power abandoned by the transmission grid and the distribution grid, is the loss of SOP in the distribution network.

[0090] The constraints considered in the transmission network are: node power balance constraints, DC power flow constraints, generator output upper and lower limit constraints, unit ramp constraints, unit minimum start and stop time constraints, and photovoltaic constraints, as shown below:

[0091] in, is the set of all nodes in the transmission network, and Node The pre-order node set and post-order node set of It is a branch road The active power on is the active power injected into the node by the photovoltaic system, is the active load of the node in the transmission network, is the set of all branches in the transmission network, A branch line in the transmission network The reactance, is the phase angle of the node, It is the start and stop sign of the unit. and are the lower and upper limits of the generator output, and are the minimum and maximum rates of the unit climbing, and They are the minimum start-up time and the minimum shutdown time of the unit, It is the upper limit of the active power output of the PV system.

[0092] The constraints considered in the distribution network include node power balance constraints, voltage drop constraints, current constraints, generator output constraints, photovoltaic constraints, voltage constraints, and transmission and distribution network boundary node constraints, such as: ; ; ; ; ; ;

[0093] in, is the set of all nodes in the distribution network, and They are the pre-order node set and post-order node set of the node, and are the active power and reactive power injected into the node by the distribution network generator, and Distribution network branches The active power and reactive power on It is a distribution network branch The square of the current amplitude flowing through it, and Distribution network branches The resistance and reactance, and are the active power and reactive power injected into the node by the photovoltaic system, and are the active load and reactive load of the distribution network node, is the charging power of the charging station in the distribution network, is the square of the voltage at the distribution network node. Assuming that all photovoltaic systems operate in constant power factor mode, It is the maximum active output of the photovoltaic system in the distribution network. is the power factor of photovoltaic operation, and are the minimum and maximum values ​​of the distribution network node voltage, respectively.

[0094] The boundary node constraints of the transmission and distribution network are:

[0095] in, It is the active power output of the balancing node in the distribution network.

[0096] The losses in each section of the SOP in the distribution network are as follows: ; ;

[0097] in, , and They are the losses of AC-DC rectifier 1, AC-DC rectifier 2 and DC-DC module, and is the loss factor of the AC-DC rectifier, and is the loss factor of the DC-DC module, , , , , , is an auxiliary variable for calculating loss and is defined as follows:

[0098] in, and are the capacities of the AC-DC rectifier and DC-DC module respectively, and SOP Active power and reactive power of the port.

[0099] The power balance constraints of SOP are as follows: ; ; ; For the urban transportation network, we first enumerate the offline routes of all OD pairs and calculate the charging power required for each route.

[0100] The model for calculating the charging capacity of each path is: ; ; ; ; ; ; ; ;

[0101] in, It's OD Path At the charging station The charging capacity, is the collection of all charging stations in all transportation networks, is the set of all road segments in the traffic network. Is electric vehicle in OD Node The remaining power, It is a section of the transportation network distance, is the unit mileage charge attenuation rate of electric vehicles, It's OD Whether it passes through the road section A binary flag variable, is a very large positive number. is the set of all nodes in the transportation network. A node in the transportation network Flag variable indicating whether charging or not, is the maximum charge of an electric vehicle. It is the range anxiety coefficient of electric vehicle users.

[0102] The traffic flow distribution model is: ; ; ; ; ; ; ; ; ;

[0103] The objective function is to minimize the total driving time, charging time and the cost of charging electric vehicles, where is the set of all paths, It is the collection of all charging stations. is the set of all OD pairs, It's OD The set of all paths of Is the path Traffic on It's a charging station Traffic on Is the path The driving time on It's at the charging station The waiting time required for charging, and The paths are Basic driving time and charging stations Basic charging waiting time, and The road sections and charging stations capacity, It's OD path The cost of It's OD path The travel time on the road segment, It's OD path of traffic, For OD travel needs, It's OD The demand function of It's OD The greatest travel demand, and OD pairs path Whether it passes through the road section and charging stations The flag variable, It's OD path of electric vehicles at charging stations The charging capacity, It's a charging station charging load.

[0104] The coordinated optimization model for the transmission and distribution network, which considers the coupling of the power and transportation networks, is a three-layer distributed model that is difficult to solve directly. Therefore, the KKT condition is used to transform the transportation network into the distribution network, reducing the three-layer optimization model to a two-layer one. The KKT condition (Karush-Kuhn-Tucker condition) is a set of necessary conditions in optimization theory that are applicable to solving nonlinear programming problems with equality and inequality constraints. When the objective function and the constraint condition are convex, the KKT condition is also a sufficient condition for finding the optimal solution. Specifically, this embodiment converts the traffic flow distribution model (including minimization of travel time, charging waiting time and cost, as well as constraints such as flow conservation and non-negativity) into equivalent KKT conditions and then projects them to the distribution network layer. These conditions eliminate the original traffic layer flow variables and Lagrange multipliers and replace them with the aggregated charging load variables and corresponding linear complementary constraints in the distribution network. This simplifies the original optimization model consisting of three coupled layers of transmission network, distribution network and transportation network into a two-layer model containing only the transmission network layer and the distribution network, with the KKT constraints of the transportation layer embedded in the distribution network model, thus achieving model dimensionality reduction, integrated solution and privacy protection.

[0105] In this embodiment, the KKT conditions of the optimization model of the urban transportation network are as follows: ; ; ; ;

[0106] in, It's OD path The sum of the total driving time, charging time and cost, and are the Lagrange multipliers of equations (1-46) and (1-47) respectively. Eliminate the Lagrange multipliers , which can be further written as: ; ;

[0107] In summary, the objective function of the centralized algorithm for coordinated optimization of the transmission and distribution network considering the coupling of the power and transportation networks is (1-1), and the constraints are the transmission network-related constraints (1-4)-(1-10), the distribution network-related constraints (1-11)-(1-17), the transmission and distribution network boundary node constraints (1-18), the SOP-related constraints (1-19)-(1-30), and the KKT conditions of the transportation network (1-54)-(1-56). The decision variables include a large number of binary variables and continuous variables in the transmission network, distribution network, and transportation network.

[0108] Centralized algorithms for collaborative optimization of transmission and distribution networks are mixed-integer nonlinear programming (MINLP) problems with numerous variables and complex constraints. Their solution time increases significantly with the size of the problem, sometimes even becoming unsolvable using existing commercial solvers. Furthermore, since centralized programs require the collection of parameter information from all devices in the transmission and distribution network, this raises privacy concerns.

[0109] exist Figure 1 The test was conducted in the example system shown. The transmission network uses an IEEE57-node network. The access to renewable energy sources such as wind power and photovoltaic power is not considered in the transmission network. There are seven units in the transmission network. The minimum start and stop time of the units is set to 3 hours, and the ramp rate is set to 20%. Among them, node 50 is connected to the IEEE33-node distribution network.

[0110] The distribution network has seven distributed photovoltaic systems, with a photovoltaic penetration rate of 0.5. A SOP is connected between nodes 7 and 8, which are connected to DC loads. Nodes 6 and 11 are connected to distributed generators. The transportation network is a 12-node urban transportation network with 20 routes and four OD pairs. The OD pair information is shown in Table 1-1. The battery capacity of electric vehicles is set to 100 kWh, the power attenuation coefficient is set to 0.2 kWh / km, the starting power of the electric vehicle is set to 30 kWh, and the user's range anxiety coefficient is set to 0.2.

[0111] Table 1-1 Node traffic network OD pair information

[0112]

[0113] First, the more likely paths for each OD pair are listed, and the charging power of the electric vehicle on each path is calculated.

[0114] Secondly, the electric vehicle charging capacity is used as a known value and entered into the traffic flow assignment program. Finally, the KKT conditions of the traffic flow assignment model are written and included in the transmission and distribution network coordinated optimization scheduling model to complete the solution.

[0115] The purpose of selecting possible paths is to reduce computational complexity. Enumerating all offline paths may take up a lot of time and computing power. In addition, in practice, electric vehicle users will only choose a few paths with the best overall cost, including time, distance, and required charging power. Therefore, it is unnecessary to enumerate all paths offline.

[0116] When selecting a route, the shortest route where electric vehicles do not need to be charged is first determined. Due to the traffic congestion effect, too many electric vehicles choosing the same route will result in longer travel time for electric vehicles. Therefore, the traffic flow allocation model will allocate traffic flow to other routes. However, electric vehicles need to be charged after passing through a longer route, otherwise they cannot complete the journey. Therefore, it is necessary to choose a route with a charging station.

[0117] The parameters of the test system are as follows: Figure 2 As shown in the figure, the research scenario is the day-ahead market. In order to adapt to the high volatility of renewable energy in the power system and improve market flexibility, the time interval is selected as 15 minutes. Figure 2 In a, the system load is at its peak between 16:00 and 20:00. Figure 2 In b, the photovoltaic system output reaches its peak between 11:00 and 12:00. Figure 2 As shown in c in Figure 2, the DC load simulates the rail transit load and reaches its peak at around 8:00 am and 20:00 pm. Figure 2As shown in d in Figure 3, the electric vehicle travel demand curve reaches its peak at 7:00 and 17:00.

[0118] The results obtained by the centralized and distributed algorithms are shown in Table 1-2. The objective function values ​​and transmission and distribution network costs are very similar, with the distributed algorithm achieving slightly higher results. The distributed algorithm also takes approximately 2.3 times longer to solve than the centralized algorithm, requiring 322.70 seconds to complete. Each distribution network optimization takes 161.25 seconds to complete. This is because the distribution network considers both the KKT conditions of the transportation network and constraints related to the SOP. It also includes integer variables, resulting in complex variables and numerous constraints, making it difficult to solve. The distributed algorithm reduces the size of each subproblem, facilitating faster solutions. However, it requires multiple iterations to converge, resulting in a long total solution time.

[0119] Table 1-2 Comparison of solution results between centralized and distributed methods

[0120]

[0121] The cost curves at each moment calculated by the two algorithms are as follows: Figure 3 As shown in a, b, c, and d in Figure 1, the transmission and distribution network costs calculated using the distributed algorithm are slightly higher at each moment than those calculated using the centralized algorithm. The transportation network costs are completely identical, and the distributed algorithm exhibits higher SOP losses, demonstrating the centralized algorithm's ability to calculate the global optimal solution. The costs calculated using both algorithms exhibit similar patterns. For example, the transmission network cost curve is similar in shape to the load curve, while the distribution network cost is influenced by multiple factors, including load and photovoltaics. Distribution network costs are higher during periods of high load but low photovoltaic output (8:00 AM - 12:00 AM).

[0122] Step 2: Use the vertex search method to solve the feasible domain. The model is required to be a convex optimization model. Therefore, the McCormick envelope relaxation method is used to deal with non-convex constraints, and the large M convex optimization balance method is used to determine the better upper and lower bounds of the variables.

[0123] Specifically, the vertex search method, when solving the feasible region for distribution network operation, requires that the optimization model of the distribution network be convex. This is because the feasible region of a convex model is a convex set, and its local optimal solution is also the global optimal solution. Furthermore, vertices can accurately describe the boundaries of the feasible region, simplifying the solution process and improving computational efficiency. However, the feasible region of a non-convex model may contain multiple local optimal solutions or discontinuous regions, making it difficult for the vertex search method to effectively find the global optimal solution and even causing it to become trapped in a local optimum.

[0124] Since the KKT condition from the urban transportation network optimization model is considered in the distribution network optimization model, the complementary relaxation constraint The bilinear term makes the distribution network optimization model non-convex, so it needs to be relaxed. The commonly used method to deal with bilinear terms is the McCormick Envelope Relaxation (MCE) method, which converts the bilinear term into a linear form, such as: ; ; ;

[0125] in, and They are the Lagrange multipliers The lower and upper bounds of and Traffic flow The lower and upper bounds of .

[0126] The MCE method can be used to relax nonconvex bilinear terms into linear constraints, transforming the distribution network optimization model into a convex model. The vertex search method can then be used to solve the distribution network's operational feasible region. However, the upper and lower bounds of the variables selected when using the MCE method have a significant impact on the solution. Loose upper and lower bounds ensure model feasibility but may result in large relaxation errors and inaccurate results. Tighter upper and lower bounds make the solution more accurate but may make the model infeasible. Therefore, it is necessary to strike a balance between feasibility and accuracy, selecting upper and lower bounds that both ensure model feasibility and minimize errors.

[0127] This example proposes the Big-M Convex Optimization Balancing Method (BCOBM) to determine precise upper and lower bounds on variables. It uses the Big-M method to handle non-convex constraints and decomposes the problem into convex optimization subproblems by adjusting the values ​​of binary variables. The core of this method is to transform a non-convex problem into a convex one by introducing auxiliary variables and the Big-M method. Furthermore, the upper and lower bounds on the variables are determined by recording and analyzing the solution results, thereby ensuring model feasibility while reducing error.

[0128] Table 1-3 BCOBM steps to determine the upper and lower bounds of variables

[0129]

[0130] First, the Big M method is used to transform the non-convex constraints into linear constraints, resulting in a Mixed Integer Programming (MIP) problem, as shown below: ;

[0131] in, is a very large positive number, is an auxiliary binary variable. By adjusting the binary variable The problem is decomposed into convex optimization subproblems.

[0132] By varying the variable parameters in the transportation network, such as the travel demand of electric vehicles, the range of results obtained using the Big M method is recorded. and , in each record and In the interval, binary variables unchanged, the model is a convex optimization model, so all and All in and between, so that we can determine and The specific steps are shown in Table 1-3.

[0133] Considering the heavier traffic load, the travel demands of the four OD pairs are 1850, 1450, 1400 and 1945 respectively. By changing the proportional coefficient of the traffic load, the binary variable obtained is The image of traffic load proportional coefficient changes is as follows Figure 4 shown.

[0134] according to Figure 4 As shown in a and b in the figure, when the travel demand coefficient increases from 0.20 to 0.29, the OD value of the binary variable corresponding to the 4th path 2 is The value changes from 1 to 0, affected by the The corresponding Lagrange multiplier is Becomes 0, traffic flow It changes from 0 to a positive number, indicating that as the traffic demand increases, the travel demand of OD pair 4 begins to allocate traffic flow to path 2.

[0135] When the travel demand coefficient is between 0.20 and 0.29, the binary variable The value of is fixed, and the model is relaxed to a convex optimization model. The value of the variable at this time is recorded. and , then all variables with travel demand between 0.20 and 0.29 and All in and In order to determine the upper and lower bounds of all variables within the interval.

[0136] because Figure 4In the figure, gray represents a traffic flow of 0 and a Lagrange multiplier of non-zero, and white represents a Lagrange multiplier of 0 and a traffic flow of non-zero. This figure illustrates that as the travel demand coefficient changes within a certain range, the upper and lower bounds of the Lagrange multiplier and traffic flow in that range can be determined. This example uses the range of 0.2-0.29 as an example. Figure 4 Analyses c, d, e, and f are similar to those above and will not be repeated here.

[0137] As the traffic load increases, the maximum value of electric vehicles and the maximum value of Lagrange multipliers for all OD paths change as follows: Figure 5 As shown in the figure, the maximum value of the Lagrange multiplier gradually decreases with the increase of traffic load, and the number of electric vehicles shows a trend of first increasing and then decreasing. This is because as the traffic load ratio increases, the total travel demand of electric vehicles increases, resulting in an increase in the maximum value of electric vehicles. When it increases to a certain extent, affected by the traffic congestion effect, the cost corresponding to the travel time of the path increases, causing the traffic flow allocation model to start allocating to other paths, resulting in a decrease in the maximum value of electric vehicles allocated to a single path.

[0138] Compare the results of three schemes: (1) MCE1: using the Big M method to solve; (2) MCE2: using the BCOBM method to determine the upper and lower bounds and then using MCE to relax and solve; (3) MCE3: using looser upper and lower bounds, that is, the traffic load and Lagrange multipliers are always selected Figure 3-4 The maximum and minimum values ​​that appear in .

[0139] The results of the three solutions are as follows: Figure 6 As shown. Figure 6 As shown in a in Figure , in terms of the comprehensive cost of OD pairs, the cost of MCE3 on all OD pairs is close to that of MCE1; Figure 6 As shown in (b), in the comparison of charging station loads, the traffic loads of MCE3 and MCE1 are concentrated on charging stations 5 and 6, while MCE2 has charging loads at all charging stations, indicating that the upper and lower bounds are too loose. Figure 6 As shown in Figure (c), in the comparison of path traffic flow, MCE3's flow on all paths is relatively close to that of MCE1. However, MCE2 also exhibits traffic flow on paths 4 and 6, indicating that the calculation error obtained using MCE2 is relatively large. Taking all indicators into consideration, the MCE3 solution has relatively small differences from MCE1 in all aspects. Therefore, it can be concluded that MCE3 performs better in approximating MCE1 and demonstrates higher accuracy.

[0140] Step 3: Use the vertex search method to collect the data set of the distribution network operation feasible region, and approximate the lower boundary of the operation feasible region to improve the accuracy of the prediction.

[0141] Common methods for solving the feasible domain include vertex search and multi-parameter planning. Vertex search characterizes the feasible domain by traversing its vertices from the inside out, while multi-parameter planning solves the problem by analyzing how the objective function and constraints vary with parameters. This embodiment uses the vertex search method to solve the VPP feasible domain. The process, combined with the previous embodiment, includes the following: The feasible domain prediction process: large-M convex optimization model → vertex search method to solve the feasible domain → obtain boundary points → linear interpolation / Douglas-Peucker algorithm simplification → construct a neural network dataset → network prediction. The large-M method handles the upper and lower bounds of the optimization variables to ensure stable operation of the vertex search method. The vertex search method collects data from the boundary of the feasible domain to obtain the lower boundary point set. Linear interpolation and Douglas-Peucker algorithm further regularize the boundary point set to form a unified dimensional data input to the neural network. Finally, deep learning is used to achieve non-iterative and efficient collaborative optimization prediction.

[0142] (1) Determine the vertex search model

[0143]

[0144] Where, is the objective function of vertex search, The value of the vector represents the direction of vertex search. and The coupling variables are and internal variables The coefficient matrix of is the constant coefficient vector on the right side.

[0145] The feasible domain of the distribution network is solved, and the constraints are the constraints of the distribution network itself and the constraints related to the distribution network cost (1-64). is a very large positive number. In this embodiment, a very large positive number can be represented by "positive infinity", that is, "+∞".

[0146] (2) Initialization: At least you need to determine The initial convex hull can be formed by initial vertices, where is the dimension of the coupling variable. In this embodiment, the dimension of the coupling variable is 2, and initialization requires at least 3 initial vertices to be determined. These initial vertices can be determined along the axis, that is, when solving , It is The standard basis vector with the first element being 1 and the other elements being 0. The initial vertex is stored in the initialization vertex set. middle.

[0147] In this embodiment, variables related to distribution network operation and scheduling are defined as two categories:

[0148] The first category: coupled variables (coordination variables) ;

[0149] in, and The time periods Active power and operating costs are exchanged between the distribution network boundary nodes. The coupled variables are submitted to the distribution network for collaborative optimization and clearing. The optimization results are used as boundary conditions for the distribution network optimization scheduling to fix the variables.

[0150] The boundary nodes in this embodiment are nodes shared by the transmission and distribution networks and are located in both the topology of the transmission network and the topology of the distribution network. Therefore, the coupling variables are variables used to describe the interaction between the distribution network and the transmission network and are variables that belong to both networks.

[0151] Category 2: Internal variables ,in, Contains all decision variables of the distribution network except coupling variables. For fixed coupling variables , internal variables It can be determined by the self-dispatching optimization of the distribution network, that is, given , by solving the distribution network optimization subproblem to obtain the internal variables .

[0152] Assume that the distribution network period The feasible domain of the variable operation of the distribution network constraint condition is , the distribution network operation feasible region is defined as: For the operation feasible region Any coupled variable There exists a set of feasible internal variables that do not violate the distribution network operation constraints. , which is the following formula: in, For the period The set of all feasible solutions for the distribution network.

[0153] (3) Internal circulation: Figure 8 As shown in , the inner loop of the vertex search method constructs the convex hull of existing vertices and identifies new vertices outside the convex hull. The convex hull composed of existing vertices generated by the second cycle is , assuming Depend on planes, the new vertex is along Search in the direction of the outer normal vector of the plane. The outward normal vector of a plane is , then the newly identified vertex The optimal solution found later. Formula (1-65) calculates the vertex arrive The distance from the plane is always non-negative. For the new vertex The improvement ratio to the current convex hull, It is The external normal vector of a plane, is the constant term on the right side of the corresponding hyperplane equation. If ,show exist In addition, add it to the vertex set , making Get closer to the practical feasible region and record the improvement rate.

[0154] (4) Outer loop: The outer loop evaluates the error of the current approximation and compares it with the allowed error to decide whether to terminate the algorithm. The error between the current convex hull and the true projection is evaluated using the Hausdorff distance. : Where, The convex hull of the approximate feasible region composed of existing nodes The set of all planes.

[0155] when When the error is less than the allowed error, the vertex search method terminates and outputs the convex hull of the existing nodes. As the final result, otherwise enter the next outer loop.

[0156] The process of solving the feasible region of distribution network operation using vertex search method and the improvement ratio IR are as follows: Figure 7 shown.

[0157] The vertex search method first generates three initial vertices by initialization to form an initial convex hull; then searches for new vertices along the outer normal direction of each edge of the convex hull through an inner loop, and the improvement rate gradually decreases with the number of iterations; until the last iteration, the outer loop determines If the error is less than the allowable error, the loop is terminated. Figure 7 a to i in is the entire search process, Figure 7 In the feasible domain of distribution network operation in (i), any point in the feasible domain contains a set of feasible decision variable values, while any point outside the feasible domain cannot find a set of decision variable values ​​that meet the distribution network operation constraints.

[0158] The distribution network's operational feasible region at each moment is affected by the network's current system state parameters. Inconsistencies in the load level, photovoltaic output, DC load, and traffic load parameters at each moment can cause the distribution network's operational feasible region to change, necessitating a recalculation of the operational feasible region at each moment. Therefore, this embodiment employs a deep learning approach, using a neural network to predict the distribution network's operational feasible region at each moment, avoiding the need for repeated computations. Furthermore, the lower boundary of the distribution network's operational feasible region already contains all information about the network's operational feasible region. Therefore, only the coordinates of the lower boundary point can be predicted.

[0159] During actual data collection, we noticed that the number of points at the lower boundary of the distribution network's operational feasible region varied at different time points. To address this issue, a common approach is to use zero-padding, which involves adding zero values ​​to ensure consistent data dimensions. While zero-padding can achieve dimensional consistency, it does not provide additional useful information to the model; it simply formally extends the data length. This practice can cause the model to overfit to the zero-padding, affecting the accuracy of model predictions.

[0160] Furthermore, we found that only a certain number of points is needed to roughly outline the feasible region of the distribution network. However, too many points often only describe some local details of the feasible region, which is redundant in many cases. Therefore, in order to improve model performance and avoid unnecessary computational overhead, we need to find a more efficient method to handle this data dimensionality inconsistency problem, rather than simply relying on zero-padding.

[0161] In order to solve the problem of inconsistent neural network input dimensions caused by the change in the number of lower boundary points in the distribution network operation analysis, this embodiment proposes an optimization method, such as Figure 9 As shown in Figure 1, this method combines linear interpolation and the Douglas-Peucker algorithm. When the number of points is less than a specified number, linear interpolation is used to add new points between existing points to carefully delineate the feasible region without changing its shape. When the number of points exceeds the specified number, the Douglas-Peucker algorithm is applied to simplify the points, reducing the number of points while minimizing the impact on the shape of the feasible region. This approach aims to improve data consistency and optimize the training and prediction performance of the neural network model. Experiments have verified its effectiveness in maintaining model accuracy and generalization ability.

[0162] The Douglas-Peucker algorithm is a method for simplifying polylines that smoothes a polyline consisting of line segments by reducing the number of points. The degree of simplification of the algorithm is determined by the parameter control, Defines the maximum distance between the original point and the simplified curve. The algorithm first connects the first and last points to form an initial approximation and marks the first and last points as retained points. Then, the algorithm finds the point farthest from the line segment among all internal points. ,if (Right now Maximum distance to the line segment) within tolerance If it is within the tolerance, the simplification is completed and all internal points can be discarded; if it is not within the tolerance, then must be retained, and the algorithm recursively processes the new part. Ultimately, the output curve is the set of all retained points.

[0163] Specifically, in this embodiment, given the starting curve As an orderly A collection of points, and the distance dimension , the algorithm first calculates the first approximation to the line segment ;

[0164] Next, the algorithm checks all Internal points, find the distance from the line segment The farthest point , whose distance is ;

[0165] if , then the simplification is complete and all interior points can be discarded. Otherwise, must be retained, and the algorithm recursively processes the new part and .

[0166] Step 4: Use the training data set to train the deep neural network, and use the prediction results output by the neural network to quickly solve the coordinated optimization of the transmission and distribution network.

[0167] A neural network is used to predict the feasible region of the distribution network. In this embodiment, a fully connected neural network is used. The model parameters and training parameters of the neural network are shown in Table 1-4. There are 10,000 data sets in total. The data sets are divided into 70% training set, 15% validation set and 15% test set. The training error is as follows: Figure 10 As shown:

[0168] Table 1-4 Fully connected neural network model parameters and training parameters

[0169]

[0170] according to Figure 11The distribution network operational feasible region prediction results shown in the figure show that the neural network model performs well in terms of prediction accuracy, generalization ability, and practical application value. The overall trend of the prediction results is highly consistent with the actual values, especially in the 0-2MW range. Despite slight errors, the area enclosed by the predicted lower boundary is almost completely consistent with the actual feasible region, indicating that the model can effectively capture key features and has strong robustness. Its stable performance across different power ranges demonstrates its good predictive ability.

[0171] The distribution network operation feasible domain obtained at different times is as follows: Figure 12 As shown in the figure, the transmission power reaches its minimum around 4:00 AM, indicating that the lower boundary of the feasible region is at the far left. At this time, the system load level is low, and the active power deficit is small. The distribution network cost reaches its maximum around 8:00 PM, when the PV output is zero and the distribution network load, DC load, and traffic load are all high, resulting in the maximum system cost. The distribution network cost is relatively low around 0:00 AM and 12:00 PM. This is because the system load level is low at 0:00 AM, the PV output is maximum at 12:00 PM, and the active power deficit is small, resulting in lower transmission power and lower distribution network cost.

[0172] The solution results of the coordinated optimization of the transmission and distribution network using the three methods are shown in Tables 1-5. Thanks to the accuracy of the neural network prediction, the equivalent projection method used in this application is almost the same as the centralized method in terms of both the objective function value and the cost of the transmission and distribution network. At the same time, its solution time is only 95.12s, of which the transmission network optimization takes 0.18s and the distribution network solution takes 94.94s. Compared with the centralized method, the acceleration effect can reach 31.72%, and compared with the distributed method, the acceleration effect can reach 70.52%.

[0173] Table 1-5 Comparison of solution results

[0174]

[0175] Embodiment 2: The present invention further provides a transmission and distribution network collaborative optimization system considering a multi-layer network coupling model of power and transportation, the system comprising:

[0176] The model building module is used to construct a transmission and distribution network collaborative optimization model that considers the coupling of power and transportation networks. The model is a three-layer distributed structure: the transmission network layer, the distribution network layer, and the transportation network layer.

[0177] The model dimensionality reduction module is used to construct the KKT conditions corresponding to the transportation network layer and project the KKT conditions into the distribution network layer, thereby reducing the dimensionality of the three-layer optimization model into a two-layer optimization model;

[0178] The model conversion module is used to set the distribution network layer after adding the KKT condition as a convex model using the large-M convex optimization balance method, and to determine the upper and lower bounds of the Lagrange multiplier and traffic flow;

[0179] The feasible domain calculation module is used to solve the operational feasible domain of the above-mentioned convex model using the vertex search method, and to control the number of lower boundary points of the operational feasible domain obtained at different times using linear interpolation and the Douglas-Peucker algorithm to obtain training data of uniform dimension;

[0180] A neural network training module is used to train a deep neural network using a data set obtained from the training data, and then achieve collaborative optimization of the transmission and distribution network based on the prediction results output by the neural network.

[0181] Other technical features of the transmission and distribution network collaborative optimization system considering the power and transportation multi-layer network coupling model described in this embodiment are similar to the corresponding transmission and distribution network collaborative optimization method considering the power and transportation multi-layer network coupling model, and will not be repeated here.

[0182] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. "Multiple" means two or more, unless otherwise specifically defined.

[0183] In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they may refer to fixed connection, detachable connection, or integration; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; internal communication between two components or interaction between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0184] In the present invention, unless otherwise expressly specified or limited, when a first feature is "above" or "below" a second feature, it may mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediary. Furthermore, when a first feature is "above," "above," or "above" a second feature, it may mean that the first feature is directly above or diagonally above the second feature, or simply means that the first feature is at a higher level than the second feature. When a first feature is "below," "below," or "below" a second feature, it may mean that the first feature is directly below or diagonally below the second feature, or simply means that the first feature is at a lower level than the second feature.

[0185] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0186] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0187] The logic and / or steps represented in a flowchart or otherwise described herein, for example, can be considered a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (not exhaustive) of computer-readable media include: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0188] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0189] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0190] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0191] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A transmission and distribution network collaborative optimization method considering a multi-layer network coupling model of power and transportation, characterized by: The method includes: A transmission and distribution network collaborative optimization model considering the coupling of power and transportation networks is constructed. The model is a three-layer distributed structure consisting of the transmission network layer, the distribution network layer, and the transportation network layer. Construct the KKT conditions corresponding to the transportation network layer and project the KKT conditions into the distribution network layer, thereby reducing the dimension of the three-layer optimization model to a two-layer optimization model; The distribution network layer after adding KKT conditions is set as a convex model using the large-M convex optimization balance method, and the upper and lower bounds of Lagrange multipliers and traffic flow are determined; The vertex search method is used to solve the feasible domain of the above convex model, and the number of lower boundary points of the feasible domain obtained at different times is controlled by linear interpolation and Douglas-Peucker algorithm to obtain training data of uniform dimension. The deep neural network is trained using the dataset obtained from the training data, and then the transmission and distribution network is optimized collaboratively based on the prediction results output by the neural network. The traffic network layer includes a charging power model and a traffic flow distribution model based on each path. The traffic flow distribution model is expressed as: The objective function of this model is to minimize the total driving time, charging time and the cost of charging electric vehicles, where: is the set of all paths, It is the collection of all charging stations. is the set of all OD pairs, Represents a certain OD pair, It's OD The set of all paths of Is the path Traffic on It's a charging station Traffic on Is the path The driving time on It's at the charging station The waiting time required for charging, and The paths are Basic driving time and charging stations Basic charging waiting time, and The road sections and charging stations capacity, It's OD path The cost of It's OD path The travel time on the road segment, It's OD path traffic flow, For OD travel needs, It's OD The demand function of It's OD The greatest travel demand, and OD pairs path Whether it passes through the road section and charging stations The flag variable, It's OD path of electric vehicles at charging stations The charging capacity, It's a charging station charging load.

2. The transmission and distribution network collaborative optimization method considering the power and transportation multi-layer network coupling model according to claim 1 is characterized in that: The objective function of the transmission and distribution network collaborative optimization model considering the coupling of power and transportation networks is set to minimize the total cost of the transmission and distribution networks, specifically: the generator cost of the transmission and distribution networks is minimized, the abandoned photovoltaics are minimized, and the soft switch SOP loss in the distribution network is minimized; The constraints considered at the transmission network layer include: node power balance constraints, DC power flow constraints, generator output upper and lower limit constraints, unit ramp constraints, unit minimum start and stop time constraints, and photovoltaic constraints; The constraints considered at the distribution network layer include: node power balance constraints, voltage drop constraints, current constraints, generator output constraints, photovoltaic constraints, voltage constraints and transmission and distribution network boundary node constraints.

3. The method for collaborative optimization of transmission and distribution networks considering a multi-layer network coupling model of power and transportation according to claim 1 is characterized in that: The step of constructing the KKT conditions corresponding to the transportation network layer and projecting the KKT conditions to the distribution network layer includes: Set the KKT condition corresponding to the traffic flow assignment model, expressed as: in, It's OD path The sum of the total driving time, charging time and cost, and They are respectively and The corresponding Lagrange multiplier; Therefore, After offset, it is expressed as: Formulas (16)-(18) are added as constraints to the constraints considered at the distribution network layer.

4. The method for collaborative optimization of transmission and distribution networks considering a multi-layer network coupling model of power and transportation according to claim 3 is characterized in that: The method of using the large-M convex optimization balance method to set the distribution network layer after adding the KKT condition as a convex model includes: According to the complementary relaxation constraint It can be seen that the traffic flow distribution model is a bilinear term, which leads to the distribution network model after adding the KKT condition being a non-convex model; The McCormick envelope relaxation method is used to transform the bilinear term into a linear form, thereby transforming the above non-convex model into a convex model, which can be expressed as follows: in, and They are the Lagrange multipliers The lower and upper bounds of and Traffic flow The lower and upper bounds of .

5. The method for collaborative optimization of transmission and distribution networks considering a multi-layer network coupling model of power and transportation according to claim 4 is characterized in that: The determination of the upper and lower bounds of the Lagrange multiplier and the traffic flow includes: Using the Big M method to transform non-convex constraints into linear constraints, we get a mixed integer programming problem as follows: in, is a very large positive number, is an auxiliary binary variable, by adjusting the auxiliary binary variable The value of , decomposes the problem into convex optimization sub-problems; By changing the variable parameters in the traffic flow distribution model, and recording the range of the results obtained using the big M method and , in each record and Keep binary variables within the interval The value remains unchanged. At this time, all and All in and between, thus determining and The upper and lower bounds of .

6. The method for collaborative optimization of transmission and distribution networks considering a multi-layer network coupling model of power and transportation according to claim 2, characterized in that: The method of using the vertex search method to solve the operational feasible domain of the convex model includes: The vertex search model is determined based on the constraints of the distribution network itself, and the initial vertex is determined based on the dimension of the coupling variable, thereby forming an initial convex hull and storing the initial vertex in the initialization vertex set. The coupling variable is a variable used to describe the interaction between the distribution network and the transmission network, and is a variable that belongs to both networks. Search for new vertices along the outer normal direction of each edge of the convex hull through an inner loop, and record the corresponding improvement ratio, which gradually decreases with the number of iterations; Until the last iteration, the error between the current convex hull and the true projection is determined through the outer loop is smaller than the allowable error, and the loop is terminated, thereby obtaining the feasible domain of operation corresponding to the current moment.

7. The method for collaborative optimization of transmission and distribution networks considering a multi-layer network coupling model of power and transportation according to claim 6 is characterized in that: The method of determining a vertex search model according to the constraints of the distribution network itself, determining an initial vertex according to the dimension of the coupling variable, thereby forming an initial convex hull, and storing the initial vertex in an initialization vertex set includes: Determine the vertex search model in, is the objective function of vertex search, The value of the vector represents the direction of vertex search. and The coupling variables are and internal variables The coefficient matrix of is the constant coefficient vector on the right side; The number of initial vertices is at least the number of coupling variables The corresponding dimension is increased by one, and the initial vertex is determined along the axis.

8. The method for collaborative optimization of transmission and distribution networks considering a multi-layer network coupling model of power and transportation according to claim 7 is characterized in that: The inner loop searches for new vertices along the outer normal direction of each edge of the convex hull and records the corresponding improvement ratio, which decreases gradually with the number of iterations, including: Jordi The convex hull composed of existing vertices generated by the second cycle is , assuming Depend on planes, the new vertex is along Search in the direction of the external normal vector of the plane; No. The outward normal vector of a plane is , then the newly identified vertex For the order The optimal solution found later; Indicates new vertex The improvement ratio of the current convex hull is: in, It is The external normal vector of a plane, is the constant term on the right side of the corresponding hyperplane equation; if ,show exist Outside, will Add vertex set , making It is closer to the actual feasible domain and records the current improvement ratio. As the number of iterations increases, the improvement ratio gradually decreases.

9. The method for collaborative optimization of transmission and distribution networks considering a multi-layer network coupling model of power and transportation according to claim 8, characterized in that: Until the last iteration, the error between the current convex hull and the true projection is determined by the outer loop Less than the allowable error, and the loop is terminated, thus obtaining the feasible domain of operation corresponding to the current moment, including: Use Hausdorff distance to evaluate the error between the current convex hull and the true projection , expressed as: in, The convex hull of the approximate feasible region composed of existing nodes The set of all planes of ; When the error When the error is less than the allowed error, the vertex search method terminates the iteration and outputs the convex hull of the existing nodes. As the final result, otherwise, enter the next outer loop.

10. The transmission and distribution network collaborative optimization method considering the power and transportation multi-layer network coupling model according to claim 6 is characterized in that: The number of lower boundary points of the operational feasible region obtained at different times is controlled by using linear interpolation and Douglas-Peucker algorithm, including: The number of lower boundary points of the distribution network's operational feasible domain at the current moment is determined. When the current number of points is less than the specified number, new lower boundary points are added between the existing points using linear interpolation to depict the feasible domain without changing its shape. When the number of points exceeds the specified number, the Douglas-Peucker algorithm is applied to simplify the number of lower boundary points to reduce the number of lower boundary points while minimizing the impact on the shape of the feasible domain.

11. The method for collaborative optimization of transmission and distribution networks considering a multi-layer network coupling model of power and transportation according to claim 10, characterized in that: The Douglas-Peucker algorithm is applied to simplify the number of lower boundary points to reduce the number of lower boundary points while minimizing the impact on the shape of the feasible region, including: Connect the first and last points of the current lower boundary point to form the initial line segment, and mark the first and last points as reserved points; Find the point farthest from the line segment among all internal points ,if In tolerance If the tolerance is not within the range, the simplification is completed and all internal points are discarded. Inside, then must be retained, and new parts are processed recursively; ultimately, the output curve is the set of all retained points where for The maximum distance to a line segment.

12. A transmission and distribution network collaborative optimization system considering the multi-layer network coupling model of power and transportation, characterized by: The system includes: The model building module is used to construct a transmission and distribution network collaborative optimization model that considers the coupling of power and transportation networks. The model is a three-layer distributed structure: the transmission network layer, the distribution network layer, and the transportation network layer. The model dimensionality reduction module is used to construct the KKT conditions corresponding to the transportation network layer and project the KKT conditions into the distribution network layer, thereby reducing the dimensionality of the three-layer optimization model into a two-layer optimization model; The model conversion module is used to set the distribution network layer after adding the KKT condition as a convex model using the large-M convex optimization balance method, and to determine the upper and lower bounds of the Lagrange multiplier and traffic flow; The feasible domain calculation module is used to solve the operational feasible domain of the above-mentioned convex model using the vertex search method, and to control the number of lower boundary points of the operational feasible domain obtained at different times using linear interpolation and the Douglas-Peucker algorithm to obtain training data of uniform dimension; A neural network training module is used to train a deep neural network using a data set obtained from the training data, thereby achieving coordinated optimization of the transmission and distribution network based on the prediction results output by the neural network; the traffic network layer includes a charging power model and a traffic flow distribution model based on each path; the traffic flow distribution model is expressed as: The objective function of this model is to minimize the total driving time, charging time and the cost of charging electric vehicles, where: is the set of all paths, It is the collection of all charging stations. is the set of all OD pairs, Represents a certain OD pair, It's OD The set of all paths of Is the path Traffic on It's a charging station Traffic on Is the path The driving time on It's at the charging station The waiting time required for charging, and The paths are Basic driving time and charging stations Basic charging waiting time, and The road sections and charging stations capacity, It's OD path The cost of It's OD path The travel time on the road segment, It's OD path traffic flow, For OD travel needs, It's OD The demand function of It's OD The greatest travel demand, and OD pairs path Whether it passes through the road section and charging stations The flag variable, It's OD path of electric vehicles at charging stations The charging capacity, It's a charging station charging load.

Images