Electricity-hydrogen coupling active power distribution network distributed optimization method considering privacy protection

Abstract

The invention provides an electricity-hydrogen coupling active power distribution network distributed optimization method considering privacy protection, and the method comprises the steps: building a mathematic model based on a FedADMM algorithm, and carrying out the improvement to obtain a mathematic model of an improved FedADMM algorithm; an active power distribution network optimization model is constructed for partitioning, sub-regions and boundary coupling constraints are obtained, based on the sub-region and boundary coupling constraints, a sub-region optimization model is established for convex processing, and a convex processing sub-region optimization model is output; and constructing a three-layer distributed optimization framework based on the improved and convex processing model, solving the convex processing model in each sub-region by using a bilevel programming method, generating an optimization strategy, uploading the optimization strategy to the constructed framework for iterative calculation, and obtaining a global optimal solution, namely an optimal operation result of the distributed power supply. By flexibly adjusting hydrogen energy conversion and storage, the regulation and control capability of the power distribution network and the new energy consumption level are effectively improved; through the improved FedADMM algorithm, the calculation complexity is reduced, the iteration convergence speed is accelerated, and the calculation efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of active distribution network operation optimization technology, and specifically relates to a distributed optimization method for an electric-hydrogen coupled active distribution network that takes privacy protection into account. Background Technology

[0002] With the large-scale integration of distributed energy resources, the structure and operating characteristics of traditional distribution networks are becoming increasingly complex. Against this backdrop, active distribution networks, through advanced information and communication technologies, have transformed from passive dispatch to active control, providing strong support for the construction of new power systems. However, distributed power sources are highly random and volatile. While the large-scale grid connection of distributed energy resources improves the utilization rate of clean energy, it also presents numerous challenges to the safe operation of distribution networks, such as voltage exceeding limits and power flow reversal.

[0003] Furthermore, for the optimization problem of active distribution networks, traditional centralized optimization methods suffer from limitations such as low computational efficiency and insufficient data privacy protection in large-scale, distributed energy environments, making it difficult to meet the demands of modern power systems for efficient, real-time, and secure dispatch. Therefore, distributed optimization algorithms have gradually become a research focus. However, while distributed algorithms provide practical solutions to the complex operation and optimization problems of active distribution networks, their iterative mechanisms increase the computational burden to some extent. Moreover, distributed optimization inevitably requires information exchange between sub-regions, and with the continuous expansion of the power grid, data privacy becomes an unavoidable consideration.

[0004] In summary, with the continuous increase in the penetration rate of distributed energy resources in active distribution networks and the increasing diversification of operational demands, the operation optimization problem of active distribution networks has become more complex. How to quickly solve complex optimization problems while ensuring real-time performance, high scalability, and strong data privacy protection has become one of the urgent research challenges.

[0005] A search revealed no published patent documents that are identical or similar to this invention. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a distributed optimization method for an active distribution network with electro-hydrogen coupling that takes privacy protection into account, aiming to solve the problems in the background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a distributed optimization method for an active distribution network with electro-hydrogen coupling, considering privacy protection, comprising: Step S1: Establish the mathematical model of the FedADMM algorithm, obtain the global augmented Lagrange equation and the iterative update equation of each variable based on the mathematical model of the FedADMM algorithm, introduce auxiliary variables into the iterative update equation of each variable, find approximate solutions for the auxiliary variables, and when the approximate solution meets the conditions for satisfying the approximate solution, the improved mathematical model of the FedADMM algorithm is obtained. Step S2: Establish an active distribution network operation optimization mathematical model including EHCS, PV and EES, and divide the distribution network into reasonable regions according to the regional characteristics to obtain each sub-region of the distribution network and the corresponding boundary coupling constraints. Step S3: Based on the boundary coupling constraints in Step S2, establish a mathematical model for the optimization of the sub-region of the distribution network. Then, perform convex processing on the mathematical model for the optimization of the sub-region of the distribution network using the second-order cone programming method, and output the convex-processed mathematical model for the optimization of the sub-region of the distribution network. Step S4: Based on the improved FedADMM algorithm mathematical model and the optimized mathematical model of the sub-region of the distribution network after convex processing, a three-layer distributed optimization solution framework for the active distribution network is constructed. Each sub-region uses the bi-layer programming method to solve the optimized mathematical model of the sub-region of the distribution network after convex processing, generates optimization strategies, and uploads the optimization strategies to the three-layer distributed optimization solution framework for the active distribution network for iterative calculation to obtain the global optimal solution, which is the optimal operating result of the distributed power source of the distribution network.

[0008] Furthermore, the specific process of step S1 is as follows: Suppose there are m clients locally, and each client has a local dataset denoted as m. By introducing local variables from the i-th client. ,express: ; In the formula, It is the minimum value; It is a global variable; Let be a vector belonging to the d-dimensional real space; m is the total number of clients; Let be the weight coefficient for the i-th client; Let be the loss function for the i-th client; Will introduce Rewritten as the client's global total loss function, it means: ; ; In the formula, This is the global total loss function for the client. These are the parameter values ​​that allow the global total loss function to reach its minimum. To construct the mathematical model of the FedADMM algorithm, the gradient of the global total loss function for all clients should satisfy the Lipshitz continuity condition. Based on satisfying the Lipshitz continuity condition, the corresponding global augmented Lagrange equation is defined as follows: ; ; In the formula, For the corresponding globally augmented Lagrange equation; For local variables; As dual variables; Let be the global augmented Lagrange equation for the i-th client; Let be the dual variable of the i-th client; Let be the penalty factor for the i-th client; The corresponding iterative update process of each variable in the globally augmented Lagrange equation for round t+1 is represented as follows: ; ; ; In the formula, This is a global variable during the (t+1)th iteration. In a given and In the case of finding the corresponding globally augmented Lagrange equation minimization value; and These are the local and dual variables at the t-th iteration, respectively; This is a local variable for the i-th client during the (t+1)-th iteration. In a given and In the case of finding the global augmented Lagrange equation minimized for the i-th client. value; Let be the dual variable for the i-th client in the t-th iteration; Let be the dual variable for the i-th client in the (t+1)-th iteration; Among them, the introduction ,make , express: ; In the formula, This is an intermediate auxiliary variable for the i-th client during the (t+1)-th iteration. based on Obtain an approximate solution The condition for satisfying the condition is: ; In the formula, For the first Loss function for each client The gradient; For the first Approximate values ​​of local variables for each client in the (t+1)th iteration; For the first The accuracy of the approximate value obtained by each client in the (t+1)th iteration; Among them, solving for approximate solutions The process is as follows: Set initial value ; Solve for an approximate solution in the i-th client. Solving for the initial values ​​of auxiliary variables; for Perform the (k+1)th iteration to obtain express: ; In the formula, Solve for an approximate solution in the i-th client. Solve for the auxiliary variables in the (k+1)th iteration; Solve for an approximate solution in the i-th client. Solve for the auxiliary variables during the k-th iteration; Let be the Lipschitz constant for the i-th client; ; For updates The maximum number of iterations preset during the process; Let the loss function of the i-th client be... The gradients all satisfy the Lipschitz continuity condition and the Hessian matrix satisfaction condition. If the condition is met, it means: ; In the formula, Let be the constraint residual for the i-th client; It is a constant greater than 1; For the i-th client, by updating At most Within each iteration, the Lipschitz and Hessian matrices are obtained. Approximate solution to the conditions ,Right now ; Solve for an approximate solution in the i-th client. The Solve for auxiliary variables in the +1st iteration; based on approximate solution. The conditions for obtaining the final result Make a judgment, when the result is Satisfying approximate solution The condition that is satisfied is the optimal solution, which is the improved mathematical model of the FedADMM algorithm. When the obtained... Not satisfying the approximate solution If the conditions are met, then re-apply... Perform the (k+1)th iteration until the optimal solution is obtained.

[0009] Furthermore, an active distribution network operation optimization mathematical model including EHCS, PV, and EES is established, which represents: ; In the formula, The objective function is... This is the line loss weighting coefficient; The set of all branches in the distribution network; Let t be the current flowing through branch l at time t; Let L be the resistance of branch l; This is the voltage deviation weighting coefficient; The set of all nodes in the distribution network; When t, node i exceeds the voltage optimization range. The voltage amplitude; and These represent the maximum and minimum values ​​within the voltage optimization range, respectively. Weighting coefficients for the operating costs of PV, EES, and EHCS; The weighting coefficient for the PV power term; For the PV collection in the distribution network; Let be the active power reduction of the i-th PV at time t; The weighting coefficient for the EES operating cost item; For the EES set in the distribution network; and Let $t$ be the operating costs of the $i$-th EES and $i$-th EHCS at time $t$. The weighting coefficient for the EHCS operating cost item; For the EHCS set in the distribution network; Weighting factor for the revenue from the sale of excess hydrogen energy; The revenue obtained from selling excess hydrogen energy at time t; PV stands for photovoltaic system; EES stands for energy storage system; EHCS stands for hydrogen energy storage system; Using the DisFlow line power flow model, the constraints on the DisFlow power flow equations are expressed as follows: ; ; ; In the formula, , Each node represents the set of the first and last nodes of a branch with node j as its end and first node, respectively. , Let be the active power and reactive power flowing from node i to node j in the branches of node i and node j; , and These represent the resistance, reactance, and current from node i to node j in the branch; , These represent the active and reactive power injected into node j, respectively. Let be the active power from node j to node k; Let be the reactive power from node j to node k; For node j exceeding the voltage optimization range The voltage amplitude; For node i exceeding the voltage optimization range The voltage amplitude; Let be the branch for all nodes i and j; Safe operation constraints, representing: ; In the formula, , These are the maximum and minimum values ​​of the safe operating range of the voltage at the distribution network nodes, respectively. This represents the maximum allowable current from node i to node j in the branch.

[0010] Furthermore, the EHCS mainly consists of an alkaline electrolyzer, a hydrogen fuel cell, and a hydrogen storage tank; the alkaline electrolyzer, the hydrogen fuel cell, and the hydrogen storage tank are constrained respectively; at the same time, the PV and EES are also constrained respectively.

[0011] Furthermore, based on the DisFlow power flow equations, EHCS, PV, and EES constraints, the branch node i to node j is taken as the boundary coupling line. The branch node i to node j and the two end nodes i and j of the boundary coupling line are copied to each sub-region of the adjacent distribution network. Based on the DisFlow power flow equations, EHCS, PV, and EES constraints in the characteristics of the distribution network sub-region, reasonable partitioning is performed to obtain each sub-region of the distribution network and the corresponding boundary coupling constraints.

[0012] Furthermore, the specific process of step S3 is as follows: An auxiliary variable, the square of the voltage, is introduced using a second-order cone programming method. and The constraints of the DisFlow power flow equations are rewritten to obtain the rewritten active distribution network operation optimization mathematical model. Based on the rewritten active distribution network operation optimization mathematical model, a distribution network sub-region optimization mathematical model is established. Then, the non-convex problem of the distribution network sub-region optimization mathematical model is transformed into a convex problem through the second-order cone programming method, and the convex-processed distribution network sub-region optimization mathematical model is output. and These are auxiliary variables representing the squared voltages at nodes i and j, respectively. Establish a mathematical model for the optimization of a sub-region of the distribution network, which represents: ; ; In the formula, N is the total number of sub-regions; The global augmented Lagrange equation for the nth subregion; This is a local variable for the nth sub-region; For the dual variable of the nth subregion; Let the objective function be the local variable of the nth subregion; The penalty parameter for the nth sub-region; and These are the equality and inequality constraints for the nth subregion.

[0013] Furthermore, based on the improved FedADMM algorithm mathematical model and the convex-processed distribution network sub-region optimization mathematical model, a three-layer distributed optimization solution framework for active distribution networks is constructed. The three-layer distributed optimization solution framework for active distribution networks includes a lower layer, a middle layer, and an upper layer; the lower layer is the distribution network sub-region layer; the middle layer is the client layer; and the upper layer is the central server layer. Will Perform initialization, Perform l iterations to obtain ; This is a global variable during the l-th iteration; The original global variables and original local variables are passed into the distribution network sub-region layer for processing to obtain the results. ; This is a local variable for the nth sub-region during the lth iteration; Will Uploaded to the client layer for processing, resulting in ; This is an auxiliary variable for the nth subregion during the lth iteration. The process of passing the original global variables and original local variables into the distribution network sub-region layer for the (l+1)th iteration update is represented as follows: ; ; ; In the formula, This is a global variable for the (l+1)th iteration; and These are the local and dual variables during the l-th iteration, respectively; This is a local variable for the i-th client during the (l+1)-th iteration; Let be the dual variable for the i-th client during the l-th iteration; Let be the dual variable for the i-th client during the (l+1)-th iteration; Combining the mathematical model of the distribution network sub-region optimization after output convex processing, the update process of the local variable of the i-th client is rewritten as the local variable of the n-th sub-region, represented as: ; ; ; ; In the formula, This is a local variable for the nth sub-region during the (l+1)th iteration; In a given and In the case of finding the minimized global augmented Lagrange equation for the nth subregion value; Let be the dual variable of the nth subregion during the lth iteration; Let be the dual variable of the nth subregion during the (l+1)th iteration; This is the auxiliary variable for the nth subregion during the (l+1)th iteration.

[0014] Furthermore, the solution is obtained based on the local variables of the nth sub-region. Suppose the original optimization problem is transformed into a mixed-integer nonlinear programming problem. The mixed-integer nonlinear programming problem is solved using a bi-level programming method, with the following specific steps: The outer iteration of the mixed-integer nonlinear programming problem first determines the integer variables of the running states of EES and EHCS. Then, the inner iteration of the mixed-integer nonlinear programming problem calculates the coupling variables in the (e+1)th iteration based on the integer variables. ; The coupling variables are obtained for the nth sub-region in the outer iteration of round e; Will Given known quantities, the mixed-integer nonlinear programming problem for the subregion is transformed into a mixed-integer linear programming problem. Solving the mixed-integer linear programming problem yields integer variables. ,express: ; ; In the formula, Let the objective function be the minimum value of the nth subregion; Let n be an integer variable representing the nth subregion. For the nth sub-region, in the (e+1)th iteration, there is an integer variable. This is a mixed-integer linear programming problem. In the inner iteration process of the mixed-integer linear programming problem, the inner layer fixes integer variables. Transform the mixed-integer linear programming problem into a nonlinear programming problem and solve it quickly. ,express: ; ; In the formula, This represents the minimum value of the globally augmented Lagrange equation for the nth subregion; This is a nonlinear programming problem. The coupling variables are obtained for the nth sub-region in the (e+1)th round of outer iteration; when When the bi-level programming method for solving mixed-integer nonlinear programming problems converges, the optimization strategy obtained by the inner iteration is the optimal solution for sub-region n, i.e. ;Will Upload to the client layer for encrypted calculation to obtain ,based on Calculate the central server layer .

[0015] Furthermore, based on the central server layer The original residuals were calculated. and dual residuals ; When the original residual and dual residuals The maximum values ​​in all cases are less than or equal to the residual threshold. The improved FedADMM algorithm mathematical model is determined to be globally converged. The optimization strategies for each sub-region are output, and the globally optimal solution is obtained, which is the optimal operating result of the distributed generation in the distribution network. When the original residual... and dual residuals All are greater than the residual threshold The improved FedADMM algorithm mathematical model was determined not to have global convergence, and the output of the central server layer was determined. Continue with the (l+1)th iteration until the improved FedADMM algorithm mathematical model converges globally.

[0016] Compared with existing technologies, this invention has the following advantages: It addresses the technical problems of insufficient renewable energy absorption, high risk of data privacy leakage, and low computational efficiency in active distribution networks with a high proportion of distributed power sources; it fully considers the operating characteristics of hydrogen-electric coupling devices, and through flexible adjustment of hydrogen energy conversion and storage, it can effectively improve the self-regulation capability of the distribution network and the level of renewable energy absorption; it can effectively reduce computational complexity by avoiding the solution of closed-form solutions, thus significantly improving the efficiency of distributed optimization strategy solutions; and it ensures that real data remains within the designated area, guaranteeing the privacy and security of critical information and effectively reducing the risk of data leakage during distributed computing. Attached Figure Description

[0017] Figure 1 This is a flowchart of the present invention.

[0018] Figure 2 This is a topology diagram of the IEEE 33-node system of the present invention.

[0019] Figure 3 This is a diagram showing the voltage distribution of the power distribution network under adverse conditions according to the present invention.

[0020] Figure 4 This is a comparison chart showing the computation time of different algorithms in this invention. Detailed Implementation

[0021] like Figure 1 As shown, the present invention provides a technical solution: a distributed optimization method for an active distribution network with electro-hydrogen coupling that considers privacy protection, comprising: Step S1: Establish the mathematical model of the FedADMM algorithm, obtain the global augmented Lagrange equation and the iterative update equation of each variable based on the mathematical model of the FedADMM algorithm, introduce auxiliary variables into the iterative update equation of each variable, find approximate solutions for the auxiliary variables, and when the approximate solution meets the conditions for satisfying the approximate solution, the improved mathematical model of the FedADMM algorithm is obtained. Step S2: Establish an active distribution network operation optimization mathematical model including EHCS (electricity-hydrogen coupled system), PV and EES (electrochemical energy storage), and divide the distribution network into reasonable regions according to the regional characteristics to obtain each sub-region of the distribution network and the corresponding boundary coupling constraints. Step S3: Based on the boundary coupling constraints in Step S2, establish a mathematical model for the optimization of the sub-region of the distribution network. Then, perform convex processing on the mathematical model for the optimization of the sub-region of the distribution network using the second-order cone programming method, and output the convex-processed mathematical model for the optimization of the sub-region of the distribution network. Step S4: Based on the improved FedADMM algorithm mathematical model and the optimized mathematical model of the sub-region of the distribution network after convex processing, a three-layer distributed optimization solution framework for the active distribution network is constructed. Each sub-region uses the bi-layer programming method to solve the optimized mathematical model of the sub-region of the distribution network after convex processing, generates optimization strategies, and uploads the optimization strategies to the three-layer distributed optimization solution framework for the active distribution network for iterative calculation to obtain the global optimal solution, which is the optimal operating result of the distributed power source of the distribution network.

[0022] The improved FedADMM algorithm mathematical model includes the federated learning algorithm model, the ADMM algorithm model, and the FedADMM algorithm mathematical model. Suppose there are m clients locally, and each client has a local dataset denoted as m. The client's loss function is defined as: (1); In the formula, Let be the loss function for the i-th client; Let be the total number of data in the dataset of the i-th client; For a single sample in the dataset; The dataset owned by the i-th client; Let be the continuous loss function for the i-th client; It is a global variable; The global total loss function for all clients is expressed as follows: (2); In the formula, Let be the weight coefficient for the i-th client; Let be the global total loss function for all clients, and satisfy . ; The training objective of a federated learning algorithm model is a global loss minimization problem, expressed as: (3); In the formula, These are the parameter values ​​that allow the global total loss function to reach its minimum. It is the minimum value; Let be a vector belonging to the d-dimensional real space; Suppose the original optimization problem is decomposed into several subproblems. The ADMM algorithm model solves these subproblems through alternating iterations, as shown below: (4); In the formula, and These are the objective functions within the two subproblems, respectively; z and z are the original variables. and the original variable z; for The constraint matrix; Let z be the constraint matrix; It is a constant vector; We introduce the globally augmented Lagrange equation for solving, which is expressed as: (5); In the formula, The equation is the globally augmented Lagrange equation; y is the dual variable; The penalty factor is T; T represents the transpose operation. The iterative equation obtained from equation (5) above is as follows: (6); (7); (8); In the formula, These are the original variables during the (k+1)th iteration; In a given and Finding the minimizer of the globally augmented Lagrange equation in the case of value; These are the original variables at the k-th iteration; These are the dual variables in the k-th iteration; These are the original variables during the (k+1)th iteration; In a given and In the case of finding the z-value that minimizes the globally augmented Lagrange equation; For the (k+1)th iteration, this is the dual variable; for The constraint matrix; for The constraint matrix.

[0023] Specifically, the FedADMM algorithm model and the ADMM algorithm model are integrated through the FedADMM algorithm mathematical model; Taking client i as an example, by introducing local variables of the i-th client... ,express: (9); Rewrite equation (9) as follows: (10); (11); To construct the mathematical model of the FedADMM algorithm, the gradient of the global total loss function for all clients must satisfy the Lipshitz continuity condition, i.e.: (12); In the formula, The global total loss function f for all clients is... gradient at a point; The global total loss function f for all clients is... gradient at a point; It is Lipschitz's constant, which is The maximum rate of change, ; Based on the Lipschitz continuity condition, the corresponding globally augmented Lagrange equation is defined as follows: (13); (14); In the formula, For the corresponding globally augmented Lagrange equation; For local variables; Let be the global augmented Lagrange equation for the i-th client; Let be the dual variable of the i-th client; Let be the penalty factor for the i-th client; Taking the (t+1)th iteration as an example, the update process of each variable in the corresponding global augmented Lagrange equation is as follows: (15); (16); (17); In the formula, This is a global variable during the (t+1)th iteration. In a given and In the case of finding the corresponding globally augmented Lagrange equation minimization value; and These are the local and dual variables at the t-th iteration, respectively; This is a local variable for the i-th client during the (t+1)-th iteration. In a given and In the case of finding the global augmented Lagrange equation minimized for the i-th client. value; Let be the dual variable for the i-th client in the t-th iteration; Let be the dual variable for the i-th client in the (t+1)-th iteration; Among them, the introduction ,make , express: (18); In the formula, This is an intermediate auxiliary variable for the i-th client during the (t+1)-th iteration. based on Obtain an approximate solution The condition for satisfying the condition is: (19); In the formula, For the first Loss function for each client The gradient; For the first Approximate values ​​of local variables for each client in the (t+1)th iteration; For the first The accuracy of the approximate value obtained by each client in the (t+1)th iteration; Let equation (19) be in Defined as feasible, let equation optimal solution The following optimality condition is satisfied, indicating that: (20); In the formula, The optimal solution is found in the i-th client; Among them, the approximate solution is obtained based on the condition of satisfying formula (19). The process is as follows: Set initial value ; Solve for an approximate solution in the i-th client. The initial values ​​of the auxiliary variables are determined; taking the (k+1)th iteration as an example. The expression is as follows: (twenty one); In the formula, Solve for an approximate solution in the i-th client. Solve for the auxiliary variables in the (k+1)th iteration; Solve for an approximate solution in the i-th client. Solve for the auxiliary variables during the k-th iteration; Let be the Lipschitz constant for the i-th client; ; For updates The maximum number of iterations preset during the process; Let the loss function of the i-th client be... The gradients all satisfy the Lipschitz continuity condition, and The loss function for the i-th client The Hessian matrix satisfies the following condition, indicating that: (twenty two); In the formula, The loss function for the i-th client The Hessian matrix; This represents the positive semi-definite order relation of the Hessian matrix; It is the loss function with respect to the i-th client. The corresponding positive numbers; I is the Lipschitz constant, and I is the identity matrix; The following conditions must be met: (twenty three); In the formula, Let be the constraint residual for the i-th client; It is a constant greater than 1; For the i-th client, by updating At most Within each iteration, an approximate solution satisfying the conditions of formulas (22) and (23) is obtained. ,Right now ; Solve for an approximate solution in the i-th client. The Solve for auxiliary variables in the +1st iteration; based on approximate solution. The conditions for obtaining the final result Make a judgment, when the result is Satisfying approximate solution The condition that is satisfied is the optimal solution, which is the improved mathematical model of the FedADMM algorithm. When the obtained... Not satisfying the approximate solution If the conditions are met, then re-apply... Perform the (k+1)th iteration until the optimal solution is obtained; Among them, the most The formula for calculating the next iteration is: (twenty four); In the formula, It is a logarithmic operation with base σ.

[0024] The specific process of step S2 is as follows: A mathematical model for optimizing the operation of an active distribution network, including EHCS, PV, and EES, is established, representing: (25); In the formula, The objective function is... This is the line loss weighting coefficient; The set of all branches in the distribution network; Let t be the current flowing through branch l at time t; Let L be the resistance of branch l; This is the voltage deviation weighting coefficient; The set of all nodes in the distribution network; When t, node i exceeds the voltage optimization range. The voltage amplitude; and These represent the maximum and minimum values ​​within the voltage optimization range, respectively. Weighting coefficients for the operating costs of PV, EES, and EHCS; The weighting coefficient for the PV power term; For the PV collection in the distribution network; Let be the active power reduction of the i-th PV at time t; The weighting coefficient for the EES operating cost item; For the EES set in the distribution network; and Let $t$ be the operating costs of the $i$-th EES and $i$-th EHCS at time $t$. The weighting coefficient for the EHCS operating cost item; For the EHCS set in the distribution network; Weighting factor for the revenue from the sale of excess hydrogen energy; The revenue obtained from selling excess hydrogen energy at time t; PV stands for photovoltaic system; EES stands for energy storage system; EHCS stands for hydrogen energy storage system; The established constraints are described as follows: Using the DisFlow line power flow model, the constraints on the DisFlow power flow equations are as follows: (26); (27); (28); In the formula, , Each node represents the set of the first and last nodes of a branch with node j as its end and first node, respectively. , Let be the active power and reactive power flowing from node i to node j in the branches of node i and node j; , and These represent the resistance, reactance, and current from node i to node j in the branch; , These represent the active and reactive power injected into node j, respectively. Let be the active power from node j to node k; Let be the reactive power from node j to node k; For node j exceeding the voltage optimization range The voltage amplitude; For node i exceeding the voltage optimization range The voltage amplitude; Let be the branch for all nodes i and j; Safe operation constraints, representing: (29); In the formula, , These are the maximum and minimum values ​​of the safe operating range of the voltage at the distribution network nodes, respectively. This represents the maximum allowable current from node i to node j in the branch.

[0025] The EHCS mainly consists of an alkaline electrolyzer, a hydrogen fuel cell, and a hydrogen storage tank; the alkaline electrolyzer, hydrogen fuel cell, and hydrogen storage tank are constrained respectively; and the PV and EES are also constrained respectively. The constraints of an alkaline electrolyzer are expressed as follows: (30); (31); (32); (33); In the formula, , , The input power, hydrogen production efficiency, and hydrogen production volume of the alkaline electrolyzer during time period t are respectively. , , These are the rated power, maximum input power, and upper limit of ramp rate for the alkaline electrolyzer, respectively. , , They are respectively The coefficients of the quadratic, linear, and constant terms; This represents the input power of the alkaline electrolyzer during the time period t−1; The constraints of hydrogen fuel cells are expressed as follows: (34); (35); (36); In the formula, , These represent the output power and hydrogen consumption volume of the hydrogen fuel cell at time t, respectively. It is Faraday's constant; , These are the stack voltage and power generation efficiency of the hydrogen fuel cell, respectively. and These are the maximum input power and the upper limit of ramp rate for hydrogen fuel cells, respectively. The hydrogen consumption volume of the hydrogen fuel cell at time t−1; The constraints of the hydrogen storage tank are as follows: (37); (38); (39); (40); (41); In the formula, Let be the hydrogen storage volume of the hydrogen storage tank at the end of time period t; The hydrogen storage volume of the hydrogen storage tank at the end of time period t−1; and These represent the amount of hydrogen added to and the amount of hydrogen released from the hydrogen storage tank at time t, respectively. , , These are the hydrogen storage tank's self-consumption rate, hydrogen filling efficiency, and hydrogen release efficiency, respectively. and These represent the maximum and minimum capacities of the hydrogen storage tank, respectively. and These represent the hydrogen filling and hydrogen discharging states of the hydrogen storage tank at time t, respectively. This represents the maximum hydrogen capacity of the hydrogen storage tank. This represents the maximum hydrogen release capacity of the hydrogen storage tank. The constraints of PV are expressed as follows: (42); In the formula, and These are the active power and reactive power of the i-th PV at time t, respectively; Let be the rated apparent power capacity of the i-th PV; θ is the power factor angle of PV; tan is the tangent function in trigonometric functions; The constraints of EES are expressed as follows: (43); (44); (45); (46); (47); (48); In the formula, The stored energy of EES at the end of time period t; The stored energy of EES at the end of time period t−1; and These are the charging and discharging power of the EES at time t, respectively; , , These are the energy self-consumption rate, charging efficiency, and discharging efficiency of the EES, respectively. and These are the maximum and minimum values ​​of the EES capacity, respectively. and These are integer variables, representing the charging and discharging states of EES at time t, respectively. This represents the maximum permissible charging power of the EES during charging. This represents the maximum permissible discharge power of the EES under discharge conditions.

[0026] Specifically, based on the constraints of the DisFlow power flow equations, EHCS constraints, PV constraints, and EES constraints, branch node i to node j is treated as a boundary coupling line. The branch node i to node j and the two endpoints i and j of this boundary coupling line are copied to various sub-regions of the adjacent distribution network. Based on the characteristics of the distribution network sub-regions, the DisFlow power flow equation constraints, EHCS constraints, PV constraints, and EES constraints are selected and rationally partitioned to obtain the various sub-regions of the distribution network and their corresponding boundary coupling constraints. Specifically: By region For example, the boundary coupling variable is { }; Let A be the set of boundary coupling variables for region A′; and For the region The active and reactive power from branch node i to node j in the boundary coupling variable set. and For the region The voltage magnitudes at nodes i and j at both ends of the boundary coupling variable set; Similarly, we can obtain the region The boundary coupling variables are ={ }; Let B′ be the set of boundary coupling variables of region B′; and Let the active and reactive power of the branch node i to node j in the boundary coupling variable set of region B′ be . and Let be the voltage magnitudes of nodes i and j at both ends of the boundary coupling variable set of region B′; to ensure the consistency and accuracy of the distributed algorithm solution, region and region The constraints also need to satisfy the boundary coupling constraints, as shown in the following equation: (49).

[0027] The specific process of step S3 is as follows: An auxiliary variable, the square of the voltage, is introduced using a second-order cone programming method. and The constraints of the DisFlow power flow equations are rewritten to obtain a revised active distribution network operation optimization mathematical model. Based on this revised model, a sub-regional optimization mathematical model for the distribution network is established. Then, the non-convex problem of the sub-regional optimization mathematical model is transformed into a convex problem using a second-order cone programming method, outputting a convex-processed sub-regional optimization mathematical model for the distribution network. Specifically: Introducing an auxiliary variable of voltage square and The constraints of the DisFlow power flow equations are rewritten as follows: (50); (51); (52); In the formula, and These are auxiliary variables representing the squared voltages at nodes i and j, respectively. The squares of the current magnitudes at nodes i and j; The relaxation of equation (52) above is: (53); Establish a mathematical model for the optimization of a sub-region of the distribution network, which represents: (54); (55); In the formula, N is the total number of sub-regions; The global augmented Lagrange equation for the nth subregion; This is a local variable for the nth sub-region; For the dual variable of the nth subregion; Let the objective function be the local variable of the nth subregion; The penalty parameter for the nth sub-region; and These are the equality and inequality constraints for the nth subregion.

[0028] The specific process of step S4 is as follows: Based on the improved FedADMM algorithm mathematical model and the distribution network sub-region optimization mathematical model after convex processing, a three-layer distributed optimization solution framework for active distribution networks is constructed. The three-layer distributed optimization solution framework for active distribution networks includes a lower layer, a middle layer, and an upper layer; the lower layer is the distribution network sub-region layer; the middle layer is the client layer; and the upper layer is the central server layer. Will Perform initialization, Perform l iterations to obtain ; This is a global variable during the l-th iteration; The original global variables and original local variables are passed into the distribution network sub-region layer for processing to obtain the results. ; This is a local variable for the nth sub-region during the lth iteration; Will Uploaded to the client layer for processing, resulting in ; This is an auxiliary variable for the nth subregion during the lth iteration. The process of passing the original global variables and original local variables into the distribution network sub-region layer for the (l+1)th iteration update is represented as follows: (56); (57); (58); In the formula, This is a global variable for the (l+1)th iteration; and These are the local and dual variables during the l-th iteration, respectively; This is a local variable for the i-th client during the (l+1)-th iteration; Let be the dual variable for the i-th client during the l-th iteration; Let be the dual variable for the i-th client during the (l+1)-th iteration; Combining the mathematical model of the distribution network sub-region optimization after output convex processing, the update process of the local variable of the i-th client is rewritten as the local variable of the n-th sub-region, represented as: (59); (60); (61); (62); In the formula, This is a local variable for the nth sub-region during the (l+1)th iteration; In a given and In the case of finding the minimized global augmented Lagrange equation for the nth subregion value; Let be the dual variable of the nth subregion during the lth iteration; Let be the dual variable of the nth subregion during the (l+1)th iteration; This is the auxiliary variable for the nth subregion during the (l+1)th iteration.

[0029] The solution is based on the local variables of the nth sub-region. Let the original optimization problem be transformed into a mixed-integer nonlinear programming problem. The mixed-integer nonlinear programming problem is solved using the bi-level programming method. Taking the (e+1)th iteration as an example, the specific steps are as follows: The outer iteration of the mixed-integer nonlinear programming problem first determines the integer variables of the running states of EES and EHCS. Then, the inner iteration of the mixed-integer nonlinear programming problem calculates the coupling variables in the (e+1)th iteration based on the integer variables. ; The coupling variables are obtained for the nth sub-region in the outer iteration of round e; Will Given known quantities, the mixed-integer nonlinear programming problem for the subregion is transformed into a mixed-integer linear programming problem. Solving the mixed-integer linear programming problem yields integer variables. ,express: (63); (64); In the formula, Let the objective function be the minimum value of the nth subregion; Let n be an integer variable representing the nth subregion. For the nth sub-region, in the (e+1)th iteration, there is an integer variable. This is a mixed-integer linear programming problem. In the inner iteration process of the mixed-integer linear programming problem, the inner layer fixes integer variables. Transform the mixed-integer linear programming problem into a nonlinear programming problem and solve it quickly. ,express: (65); (66); In the formula, This represents the minimum value of the globally augmented Lagrange equation for the nth subregion; This is a nonlinear programming problem. The coupling variables are obtained for the nth sub-region in the (e+1)th round of outer iteration; when When the bi-level programming method for solving mixed-integer nonlinear programming problems converges, the optimization strategy obtained by the inner iteration is the optimal solution for sub-region n, i.e. ;Will Upload to the client layer for encrypted calculation to obtain ,based on The central server layer in formula (59) is calculated. .

[0030] Among them, based on the central server layer The calculation process for the original residual and the dual residual is shown below: (67); In the formula, This represents the original residual of the nth sub-region during the (l+1)th iteration. This is the dual residual at the (l+1)th iteration; The determination condition is: (68); (69); In the formula, For the first The original residual at +1 iteration; Infinitely large; The threshold value for the residual; when and The maximum values ​​in all are less than or equal to The improved FedADMM algorithm mathematical model is determined to be globally converged. The optimization strategies for each sub-region are output, and the globally optimal solution is obtained, which represents the optimal operating result of distributed generation in the distribution network. and Both are greater than The improved FedADMM algorithm mathematical model was determined not to have global convergence, and the output of the central server layer was determined. Continue with the (l+1)th iteration until the improved FedADMM algorithm mathematical model converges globally.

[0031] Verification Example The accuracy and effectiveness of this invention were verified using the IEEE 33-node system as a test case.

[0032] The example of this invention is an IEEE 33-bus AC power distribution system, the topology of which is shown below. Figure 2 As shown; Busbar 1 is the reference node, with a voltage reference value set to 1.0 pu; the safe voltage range for the distribution network nodes is [0.95, 1.05] pu, and the voltage optimization range is [0.985, 1.015] pu. When the node voltage exceeds the safe voltage range, the IEEE 33-node system begins optimization control; the IEEE 33-node system is divided into two regions, region... The set of nodes in the region is {1,2,3,…,9,19,20,21,…,33}. The set of nodes in the array is {10,11,…,18}; pu (per unit) represents the relative unit or per-unit value; Among them, in the region and region In the process, some nodes have installed distributed photovoltaic (PV) systems to support voltage optimization, specifically including: PV1 is installed at nodes 5, 7, 10, 12, 14, 20, 22 and 32, with an installed capacity of 0.6 MVA; PV3 is installed at nodes 2, 24, and 28, with an installed capacity of 1.2 MVA; EES is installed on node 29; PV2 is installed at nodes 26 and 30, with an installed capacity of 0.8 MVA; Node 30 also has EHCS installed; Node 18 is connected to a variable load; The power factor of the inverters in the photovoltaic system in the IEEE 33-node system varies in the range of [-0.95, 0.95]; the EES is installed at nodes 14 and 29, and the EHCS is installed at node 30.

[0033] Table 1. Photovoltaic System Installation Locations and Capacities .

[0034] Two scenarios were set up to verify the optimization effect of the method proposed in this invention, as follows: Scenario 1: No control measures were taken for the calculation examples; Scenario 2: Distribution network optimization strategies considering multiple control methods such as PV, EES and EHCS; Optimized simulations were performed on scenarios 1 and 2 under harsh conditions, and the resulting distribution network voltage distribution is as follows: Figure 3As shown, before optimized control, the voltage limit exceeding problem in the IEEE 33-node system was severe; after adopting the control strategy of Scenario 2, the voltage limit exceeding problem was effectively solved, and the voltage level of the IEEE 33-node system was significantly improved; among them, in Figure 3 (a) t=12:00: In scenario 1, the voltage of node 14 reached 1.0646 pu, which significantly exceeded the voltage safety limit of 1.05 pu; while in scenario 2, by optimizing the control strategy, the voltage of node 14 was reduced to 1.0200 pu, which effectively alleviated the voltage over-limit problem. exist Figure 3 (b) t=20:00: In scenario 1, the voltage of node 18 dropped to 0.9354 pu, which is below the safe voltage limit of 0.95 pu. In scenario 2, after optimization measures were taken, the voltage of node 18 was increased to 0.9504 pu, which solved the voltage over-limit problem and improved the power supply quality.

[0035] The above results show that the present invention can effectively improve the voltage level and power supply quality of the power distribution network.

[0036] Furthermore, the traditional ADMM algorithm is compared with the method proposed in this invention to verify the effectiveness of the proposed method. Figure 4 As shown in the figure, the computation time of different algorithms is compared. The FedADMM algorithm proposed in this invention significantly outperforms the traditional ADMM algorithm in terms of computation time. Specifically, the computation time of a single iteration of the FedADMM algorithm is significantly shorter than that of the ADMM algorithm. For example, in the first iteration, the computation time of the ADMM algorithm is 32.1 seconds, while that of the FedADMM algorithm is only 14.2 seconds. As the number of iterations increases, the cumulative iteration time of the FedADMM algorithm increases slowly. By the 10th iteration, the cumulative computation time is only 66.1 seconds, a reduction of 112.86 seconds compared to the 172.9 seconds of the ADMM algorithm, representing only 38.23% of the computation time of the ADMM algorithm. This is because the FedADMM algorithm decomposes the optimization problem into multiple subproblems and computes them in parallel on different clients through distributed optimization and privacy protection mechanisms, reducing data transmission and processing time. Furthermore, it accelerates the iteration convergence speed by introducing auxiliary variables and improving the iterative update equation.

[0037] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0038] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A distributed optimization method for an electro-hydrogen coupled active distribution network that considers privacy protection, characterized in that, include: Step S1: Establish the mathematical model of the FedADMM algorithm, obtain the global augmented Lagrange equation and the iterative update equation of each variable based on the mathematical model of the FedADMM algorithm, introduce auxiliary variables into the iterative update equation of each variable, find approximate solutions for the auxiliary variables, and when the approximate solution meets the conditions for satisfying the approximate solution, the improved mathematical model of the FedADMM algorithm is obtained. Step S2: Establish an active distribution network operation optimization mathematical model including EHCS, PV and EES, and divide the distribution network into reasonable regions according to the regional characteristics to obtain each sub-region of the distribution network and the corresponding boundary coupling constraints. Step S3: Based on the boundary coupling constraints in Step S2, establish a mathematical model for the optimization of the sub-region of the distribution network. Then, perform convex processing on the mathematical model for the optimization of the sub-region of the distribution network using the second-order cone programming method, and output the convex-processed mathematical model for the optimization of the sub-region of the distribution network. Step S4: Based on the improved FedADMM algorithm mathematical model and the optimized mathematical model of the sub-region of the distribution network after convex processing, a three-layer distributed optimization solution framework for the active distribution network is constructed. Each sub-region uses the bi-layer programming method to solve the optimized mathematical model of the sub-region of the distribution network after convex processing, generates optimization strategies, and uploads the optimization strategies to the three-layer distributed optimization solution framework for the active distribution network for iterative calculation to obtain the global optimal solution, which is the optimal operating result of the distributed power source of the distribution network.

2. The distributed optimization method for an active distribution network with electro-hydrogen coupling considering privacy protection as described in claim 1, characterized in that: The specific process of step S1 is as follows: Suppose there are m clients locally, and each client has a local dataset denoted as m. By introducing local variables from the i-th client. ,express: ； In the formula, It is the minimum value; It is a global variable; Let be a vector belonging to the d-dimensional real space; m is the total number of clients; Let be the weight coefficient for the i-th client; Let be the loss function for the i-th client; Will introduce Rewritten as the client's global total loss function, it means: ； ； In the formula, This is the global total loss function for the client. These are the parameter values ​​that allow the global total loss function to reach its minimum. To construct the mathematical model of the FedADMM algorithm, the gradient of the global total loss function for all clients should satisfy the Lipshitz continuity condition. Based on satisfying the Lipshitz continuity condition, the corresponding global augmented Lagrange equation is defined as follows: ； ； In the formula, For the corresponding globally augmented Lagrange equation; For local variables; As dual variables; Let be the global augmented Lagrange equation for the i-th client; Let be the dual variable of the i-th client; Let be the penalty factor for the i-th client; The corresponding iterative update process of each variable in the globally augmented Lagrange equation for round t+1 is represented as follows: ； ； ； In the formula, This is a global variable during the (t+1)th iteration. In a given and In the case of finding the corresponding globally augmented Lagrange equation minimization value; and These are the local and dual variables at the t-th iteration, respectively; This is a local variable for the i-th client during the (t+1)-th iteration. In a given and In the case of finding the global augmented Lagrange equation minimized for the i-th client. value; Let be the dual variable for the i-th client in the t-th iteration; Let be the dual variable for the i-th client in the (t+1)-th iteration; Among them, the introduction ,make , express: ； In the formula, This is an intermediate auxiliary variable for the i-th client during the (t+1)-th iteration. based on Obtain an approximate solution The condition for satisfying the condition is: ； In the formula, For the first Loss function for each client The gradient; For the first Approximate values ​​of local variables for each client in the (t+1)th iteration; For the first The accuracy of the approximate value obtained by each client in the (t+1)th iteration; Among them, solving for approximate solutions The process is as follows: Set initial value ; Solve for an approximate solution in the i-th client. Solving for the initial values ​​of auxiliary variables; for Perform the (k+1)th iteration to obtain express: ； In the formula, Solve for an approximate solution in the i-th client. Solve for the auxiliary variables in the (k+1)th iteration; Solve for an approximate solution in the i-th client. Solve for the auxiliary variables during the k-th iteration; Let be the Lipschitz constant for the i-th client; ; For updates The maximum number of iterations preset during the process; Let the loss function of the i-th client be... The gradients all satisfy the Lipschitz continuity condition and the Hessian matrix satisfaction condition. If the condition is met, it means: ； In the formula, Let be the constraint residual for the i-th client; It is a constant greater than 1; For the i-th client, by updating At most Within each iteration, the Lipschitz and Hessian matrices are obtained. Approximate solution to the conditions ,Right now ; Solve for an approximate solution in the i-th client. The Solve for auxiliary variables in the +1st iteration; based on approximate solution. The conditions for obtaining the final result Make a judgment, when the result is... Satisfying approximate solution The condition that is met is the optimal solution, which is the improved mathematical model of the FedADMM algorithm. When the obtained... Not satisfying the approximate solution If the conditions are met, then re-apply... Perform the (k+1)th iteration until the optimal solution is obtained.

3. The distributed optimization method for an active distribution network with electro-hydrogen coupling, considering privacy protection, as described in claim 2, is characterized in that: A mathematical model for optimizing the operation of an active distribution network, including EHCS, PV, and EES, is established, representing: ； In the formula, The objective function is... This is the line loss weighting coefficient; The set of all branches in the distribution network; Let t be the current flowing through branch l at time t; Let L be the resistance of branch l; This is the voltage deviation weighting coefficient; The set of all nodes in the distribution network; When t, node i exceeds the voltage optimization range. The voltage amplitude; and These represent the maximum and minimum values ​​within the voltage optimization range, respectively. Weighting coefficients for the operating costs of PV, EES, and EHCS; The weighting coefficient for the PV power term; For the PV collection in the distribution network; Let be the active power reduction of the i-th PV at time t; The weighting coefficient for the EES operating cost item; For the EES set in the distribution network; and Let $t$ be the operating costs of the $i$-th EES and $i$-th EHCS at time $t$. The weighting coefficient for the EHCS operating cost item; For the EHCS set in the distribution network; Weighting factor for the revenue from the sale of excess hydrogen energy; The revenue obtained from selling excess hydrogen energy at time t; PV stands for photovoltaic system; EES stands for energy storage system; EHCS stands for hydrogen energy storage system; Using the DisFlow line power flow model, the constraints on the DisFlow power flow equations are expressed as follows: ； ； ； In the formula, , Each node represents the set of the first and last nodes of a branch with node j as its end and first node, respectively. , Let be the active power and reactive power flowing from node i to node j in the branches of node i and node j; , and These represent the resistance, reactance, and current from node i to node j in the branch; , These represent the active and reactive power injected into node j, respectively. Let be the active power from node j to node k; Let be the reactive power from node j to node k; For node j exceeding the voltage optimization range The voltage amplitude; For node i exceeding the voltage optimization range The voltage amplitude; Let be the branch for all nodes i and j; Safe operation constraints, representing: ； In the formula, , These are the maximum and minimum values ​​of the safe operating range of the voltage at the distribution network nodes, respectively. This represents the maximum allowable current from node i to node j in the branch.

4. The distributed optimization method for an active distribution network with hydrogen-electric coupling, considering privacy protection, as described in claim 3, is characterized in that: EHCS mainly consists of an alkaline electrolyzer, a hydrogen fuel cell, and a hydrogen storage tank; the alkaline electrolyzer, hydrogen fuel cell, and hydrogen storage tank are constrained respectively; at the same time, PV and EES are also constrained respectively.

5. A distributed optimization method for an active distribution network with electro-hydrogen coupling, considering privacy protection, as described in claim 4, characterized in that: Based on the DisFlow power flow equations, EHCS, PV and EES constraints, the branch node i to node j is taken as the boundary coupling line. The branch node i to node j and the two end nodes i and j of the boundary coupling line are copied to each sub-region of the adjacent distribution network. Based on the DisFlow power flow equations, EHCS, PV and EES constraints in the characteristics of the distribution network sub-region, reasonable partitioning is carried out to obtain each sub-region of the distribution network and the corresponding boundary coupling constraints.

6. A distributed optimization method for an active distribution network with electro-hydrogen coupling, considering privacy protection, as described in claim 5, characterized in that: The specific process of step S3 is as follows: An auxiliary variable, the square of the voltage, is introduced using a second-order cone programming method. and The constraints of the DisFlow power flow equations are rewritten to obtain the rewritten active distribution network operation optimization mathematical model. Based on the rewritten active distribution network operation optimization mathematical model, a distribution network sub-region optimization mathematical model is established. Then, the non-convex problem of the distribution network sub-region optimization mathematical model is transformed into a convex problem through the second-order cone programming method, and the convex-processed distribution network sub-region optimization mathematical model is output. and These are auxiliary variables representing the squared voltages at nodes i and j, respectively. Establish a mathematical model for the optimization of a sub-region of the distribution network, which represents: ； ； In the formula, N is the total number of sub-regions; The global augmented Lagrange equation for the nth subregion; This is a local variable for the nth sub-region; For the dual variable of the nth subregion; Let the objective function be the local variable of the nth subregion; The penalty parameter for the nth sub-region; and These are the equality and inequality constraints for the nth subregion.

7. A distributed optimization method for an active distribution network with electro-hydrogen coupling, considering privacy protection, as described in claim 6, characterized in that: The improved FedADMM algorithm mathematical model and the convex-processed distribution network sub-region optimization mathematical model are used to construct a three-layer distributed optimization solution framework for active distribution networks. The three-layer distributed optimization solution framework for active distribution networks includes a lower layer, a middle layer, and an upper layer; the lower layer is the distribution network sub-region layer; the middle layer is the client layer; and the upper layer is the central server layer. Will Perform initialization, Perform l iterations to obtain ; This is a global variable during the l-th iteration; The original global variables and original local variables are passed into the distribution network sub-region layer for processing to obtain the results. ; This is a local variable for the nth sub-region during the lth iteration; Will Uploaded to the client layer for processing, resulting in ; This is an auxiliary variable for the nth subregion during the lth iteration. The process of passing the original global variables and original local variables into the distribution network sub-region layer for the (l+1)th iteration update is represented as follows: ； ； ； In the formula, This is a global variable for the (l+1)th iteration; and These are the local and dual variables during the l-th iteration, respectively; This is a local variable for the i-th client during the (l+1)-th iteration; Let be the dual variable for the i-th client during the l-th iteration; Let be the dual variable for the i-th client during the (l+1)-th iteration; Combining the mathematical model of the distribution network sub-region optimization after output convex processing, the update process of the local variable of the i-th client is rewritten as the local variable of the n-th sub-region, represented as: ； ； ； ； In the formula, This is a local variable for the nth sub-region during the (l+1)th iteration; In a given and In the case of finding the minimized global augmented Lagrange equation for the nth subregion value; Let be the dual variable of the nth subregion during the lth iteration; Let be the dual variable of the nth subregion during the (l+1)th iteration; This is the auxiliary variable for the nth subregion during the (l+1)th iteration.

8. A distributed optimization method for an active distribution network with electro-hydrogen coupling, considering privacy protection, as described in claim 7, characterized in that: Solving based on the local variables of the nth subregion Suppose the original optimization problem is transformed into a mixed-integer nonlinear programming problem. The mixed-integer nonlinear programming problem is solved using a bi-level programming method, with the following specific steps: The outer iteration of the mixed-integer nonlinear programming problem first determines the integer variables of the running states of EES and EHCS. Then, the inner iteration of the mixed-integer nonlinear programming problem calculates the coupling variables in the (e+1)th iteration based on the integer variables. ; The coupling variables are obtained for the nth sub-region in the outer iteration of round e; Will Given known quantities, the mixed-integer nonlinear programming problem for the subregion is transformed into a mixed-integer linear programming problem. Solving the mixed-integer linear programming problem yields integer variables. ,express: ； ； In the formula, Let the objective function be the minimum value of the nth subregion; Let n be an integer variable representing the nth subregion. For the nth sub-region, in the (e+1)th iteration, there is an integer variable. This is a mixed-integer linear programming problem. In the inner iteration process of the mixed-integer linear programming problem, the inner layer fixes integer variables. Transform the mixed-integer linear programming problem into a nonlinear programming problem and solve it quickly. ,express: ； ； In the formula, This represents the minimum value of the globally augmented Lagrange equation for the nth subregion; This is a nonlinear programming problem. The coupling variables are obtained for the nth sub-region in the (e+1)th round of outer iteration; when When the bi-level programming method for solving mixed-integer nonlinear programming problems converges, the optimization strategy obtained by the inner iteration is the optimal solution for sub-region n, i.e. ;Will Upload to the client layer for encrypted calculation to obtain ,based on Calculate the central server layer .

9. A distributed optimization method for an active distribution network with electro-hydrogen coupling, considering privacy protection, as described in claim 8, characterized in that: Based on the central server layer The original residuals were calculated. and dual residuals ; When the original residual and dual residuals The maximum values ​​in all cases are less than or equal to the residual threshold. The improved FedADMM algorithm mathematical model is determined to be globally converged. The optimization strategies for each sub-region are output, and the globally optimal solution is obtained, which is the optimal operating result of distributed generation in the distribution network. When the original residual... and dual residuals All are greater than the residual threshold The improved FedADMM algorithm mathematical model was determined not to have global convergence, and the output of the central server layer was determined. Continue with the (l+1)th iteration calculation until the improved FedADMM algorithm mathematical model converges globally.

Images