Data mechanism combined driving method and system for direct current optimal power flow solution

Abstract

The invention provides a direct current optimal power flow solution-oriented data mechanism combined driving method and system, and is applied to the technical field of power data processing. Inputting the power demand into the DNN to obtain an initial solution; and if the initial solution does not meet the problem constraint, processing the problem to be solved and the initial solution by using a variable elimination method, and converting the problem to be solved and the initial solution into an optimization problem without equality constraint and a corresponding initial solution. Iteratively adjusting the position of the initial solution based on a constraint set only comprising the inequality constraint and the normal vector violating the inequality constraint until a non-feasible solution of the initial solution falls into a feasible region; and restoring the initial solution to obtain an intermediate feasible solution meeting problem constraints. And inputting the intermediate feasible solution into the FONET neural network to obtain adjustment information, and adjusting the intermediate feasible solution into a target feasible solution based on the adjustment information. The method is suitable for the DCOPF problem with a feasible solution under any power topology, the solving efficiency and the feasibility of the solving result can be guaranteed at the same time, and the obtained solving result has higher optimality.

Description

Technical Field

[0001] The present application relates to the field of power data processing technology, and in particular to a data mechanism joint driving method and system for solving DC optimal power flow. Background Art

[0002] DC Optimal Power Flow (DCOPF) is a simplified model used in power system optimization to solve the most economical unit output plan without violating the problem constraints at a low computational cost. The solution to the DCOPF problem must satisfy both equality and inequality constraints.

[0003] The neural network method is one of the general solutions to the DC optimal power flow problem. However, for neural networks solving DC optimal power flow problems, they lack the ability to generalize power grid topology. That is, they cannot output a DCOPF solution with high optimality based on a given power grid topology. However, after the power grid topology corresponding to the target DCOPF problem changes, the optimality and feasible probability of the solution given by the neural network often gradually decrease as the degree of topological change deepens.

[0004] Therefore, the existing technology has the problems of low efficiency in solving the DCOPF problem of the power system and low feasibility of the solution results when the power grid topology changes. Summary of the Invention

[0005] In view of this, the present application provides a data mechanism joint driving method and system for solving DC optimal power flow, which is applicable to the DCOPF problem with feasible solutions under any power topology. It can simultaneously ensure the solution efficiency and the feasibility of the solution results, and the obtained solution results have higher optimality.

[0006] To solve the above problems, the technical solutions provided by this application are as follows: In a first aspect, an embodiment of the present application provides a data mechanism joint driving method for solving a DC optimal power flow, the method comprising: Inputting the power demand into a deep neural network DNN to obtain a unit output plan output by the deep neural network DNN, wherein any unit output plan in the unit processing plan is an initial solution; If the initial solution satisfies the problem constraints, the initial solution is used as an intermediate feasible solution, wherein the problem constraints include equality constraints and inequality constraints; If the initial solution does not satisfy the problem constraints, the constraints and the initial solution of the problem to be solved are processed based on the equality constraints by a variable elimination method to obtain a transformed constraint set and an initial solution, wherein the constraint set consists only of inequality constraints; Iteratively adjusting the position of the initial solution based on the constraint set and the normal vector that violates the inequality constraint until an infeasible solution of the initial solution falls into a feasible region; Reducing the initial solution based on the equality constraint and the infeasible solution of the initial solution to obtain an intermediate feasible solution that satisfies the equality constraint and the inequality constraint; Inputting the intermediate feasible solution into a pre-built neural network FONET that guarantees feasibility and embeds optimality, and determining adjustment information of the intermediate feasible solution; The intermediate feasible solution is adjusted based on the adjustment information to obtain a target feasible solution.

[0007] In one possible implementation, iteratively adjusting the position of the initial solution based on the constraint set and the normal vector violating the inequality constraint until an infeasible solution of the initial solution falls into a feasible region includes: Calculate the inequality constraints violated by the infeasible solution of the initial solution according to the constraint set; Calculating a normal vector that violates the inequality constraint, and using the sum of the normal vectors as a moving direction of the infeasible solution of the initial solution; The distance between the infeasible solution of the initial solution and the initially violated inequality constraint is used as the single-iteration moving distance of the infeasible solution of the initial solution, wherein the initially violated inequality constraint is the first violated inequality constraint encountered in the moving direction; Calculating an intersection plane of a hyperplane where an infeasible solution of the initial solution is located after adjustment according to the moving direction and moving distance; Projecting the moving direction onto the intersection plane to generate a new moving direction; Until the intersection plane or the new moving direction meets a preset condition, it is determined that the infeasible solution of the initial solution falls into the feasible domain, and the preset condition is that the intersection plane is a point or the new moving direction is an all-zero vector.

[0008] In one possible implementation, the adjustment information includes an adjustment direction and an adjustment distance. Inputting the intermediate feasible solution into a pre-built neural network FONET that ensures feasibility and embeds optimality, and determining the adjustment information of the intermediate feasible solution include: Inputting the intermediate feasible solution into the FONET to obtain a first activation function, wherein the first activation function is used to determine the adjustment direction; determining a standby adjustment distance based on the adjustment direction and a first preset formula; Calculate the main adjustment distance according to the second preset formula; The adjustment distance is determined from the primary adjustment distance and the backup adjustment distance.

[0009] In one possible implementation, the FONET includes a neural network OAERN that is optimal and can ensure the satisfaction of equality constraints, and a neural network OAERN that is optimal and can ensure the satisfaction of inequality constraints; the OAERN is used to determine the adjustment direction, and the OAERN is used to determine the adjustment distance.

[0010] In a second aspect, an embodiment of the present application provides a data-mechanism combined drive system for solving DC optimal power flow, the system comprising: an acquisition module, configured to input the power demand into a deep neural network (DNN) to obtain a unit output plan output by the deep neural network (DNN), wherein any unit output plan in the unit processing plan is an initial solution; a judgment module, configured to use the initial solution as an intermediate feasible solution if the initial solution satisfies the problem constraints, wherein the problem constraints include equality constraints and inequality constraints; The judgment module is further configured to, if the initial solution does not satisfy the problem constraints, process the constraints and the initial solution of the problem to be solved based on the equality constraints by a variable elimination method to obtain a transformed constraint set and an initial solution, wherein the constraint set consists only of inequality constraints; an adjustment module, configured to iteratively adjust a position of the initial solution based on the constraint set and the normal vector violating the inequality constraint, until an infeasible solution of the initial solution falls into a feasible region; A restoration module, configured to restore the initial solution based on the equality constraint and the infeasible solution of the initial solution to obtain an intermediate feasible solution that satisfies the equality constraint and the inequality constraint; A determination module is used to input the intermediate feasible solution into a pre-built neural network FONET that ensures feasibility and embeds optimality, and determine adjustment information of the intermediate feasible solution; The adjustment module is further configured to adjust the intermediate feasible solution based on the adjustment information to obtain a target feasible solution.

[0011] In one possible implementation, the adjustment module is configured to iteratively adjust the position of the initial solution based on the constraint set and the normal vector violating the inequality constraint until the infeasible solution of the initial solution falls into the feasible region, including: a calculation submodule, configured to calculate the inequality constraints violated by the infeasible solution of the initial solution according to the constraint set; calculate the normal vectors violating the inequality constraints, and use the sum of the normal vectors as the moving direction of the infeasible solution of the initial solution; a determination submodule, configured to use the distance between the infeasible solution of the initial solution and the initially violated inequality constraint as the single-iteration moving distance of the infeasible solution of the initial solution, wherein the initially violated inequality constraint is the first violated inequality constraint encountered in the moving direction; The calculation submodule is further configured to calculate an intersection plane of a hyperplane where an infeasible solution of the initial solution is located after adjustment according to the moving direction and moving distance; A generating submodule, configured to project the moving direction onto the intersection plane to generate a new moving direction; The determination submodule is used to determine that the infeasible solution of the initial solution falls into the feasible domain until the intersection plane or the new moving direction meets a preset condition, and the preset condition is that the intersection plane is a point or the new moving direction is an all-zero vector.

[0012] In one possible implementation, the adjustment information includes an adjustment direction and an adjustment distance, and the determining module is specifically configured to: Inputting the intermediate feasible solution into the FONET to obtain a first activation function, wherein the first activation function is used to determine the adjustment direction; determining a standby adjustment distance based on the adjustment direction and a first preset formula; Calculate the main adjustment distance according to the second preset formula; The adjustment distance is determined from the primary adjustment distance and the backup adjustment distance.

[0013] In one possible implementation, the FONET includes a neural network OAERN that is optimal and can ensure the satisfaction of equality constraints, and a neural network OAERN that is optimal and can ensure the satisfaction of inequality constraints; the OAERN is used to determine the adjustment direction, and the OAERN is used to determine the adjustment distance.

[0014] In a third aspect, an embodiment of the present application provides a data mechanism joint drive device for solving DC optimal power flow, the device comprising: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, and the one or more programs include instructions. When the instructions are executed by the processor, the processor executes the data-mechanism joint driving method for solving DC optimal power flow described in the first aspect above.

[0015] In a fourth aspect, the present application provides a storage medium storing instructions, which, when executed by a computer, enable the computer to implement the data mechanism joint driving method for solving DC optimal power flow as described in the first aspect above.

[0016] It can be seen that this application has the following beneficial effects: The embodiment of the present application provides a data mechanism joint driving method for solving DC optimal power flow, which is applicable to the DCOPF problem with feasible solutions under any power topology. The power demand is input into the deep neural network DNN to obtain the unit output plan output by the DNN, and any unit output plan in the unit processing plan is the initial solution; if the initial solution satisfies the problem constraints, the initial solution is used as the intermediate feasible solution, wherein the problem constraints include equality constraints and inequality constraints; if the initial solution does not satisfy the problem constraints, the constraints and initial solution of the problem to be solved are processed based on the equality constraints by the variable elimination method to obtain the transformed constraint set and the initial solution, wherein the transformed constraint set only includes inequality constraints; based on the constraint set and the normal vector that violates the inequality constraint, the position of the initial solution is iteratively adjusted until the infeasible solution of the initial solution falls into the feasible domain; based on the equality constraints and the infeasible solution of the initial solution, the initial solution is restored to obtain the intermediate feasible solution that satisfies the equality constraints and the inequality constraints. The intermediate feasible solution is input into a pre-built neural network called FONET, which guarantees feasibility and embeds optimality. Adjustment information for the intermediate feasible solution is determined. Based on this adjustment information, the intermediate feasible solution is adjusted to obtain the target feasible solution. Determining the adjustment information and adjusting the initial solution using this information ensures both solution efficiency and the feasibility of the solution, resulting in a more optimal solution.

[0017] The embodiment of the present application also provides a system corresponding to the above method, which has the same beneficial effects as the above method. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flow chart of a data mechanism joint driving method for solving DC optimal power flow provided in an embodiment of the present application; Figure 2 An overall flow chart of solving DCOPF based on DNN preprocessing and then using neural network UDAP in an embodiment of the present application; Figure 3 A data flow diagram of a neural network OAERN that is optimal and can ensure equality constraint satisfaction, provided in an embodiment of the present application; Figure 4 A data flow diagram of a neural network OAIRN that is optimal and can ensure inequality constraint satisfaction, provided in an embodiment of the present application; Figure 5 A schematic diagram of an infeasible solution movement effect according to an embodiment of the present application; Figure 6 A schematic diagram of the structure of a data mechanism combined drive system for solving DC optimal power flow provided in an embodiment of the present application; Figure 7A schematic structural diagram of a data mechanism-based joint drive device for solving DC optimal power flow provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0020] In this application, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprises", "comprising" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further limitations, an element defined by the statement "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0021] In power system optimization, the DCOPF problem is used to solve the most economical unit output plan that does not violate the problem constraints at a low computational cost. The required data involves energy demand and power system parameters. The specific expression is: (1) (2) (3) (4) (5) in, For the unit combination, is the generator output matrix, The output of the generator set is The cost of electricity generation, It is The production cost of a generator, is the model parameter of power generation cost, for The values, and They are the minimum / maximum output matrix of the generator set, is the load demand matrix, is the matrix describing the power flow limit of the line, is the node input power-line power flow transfer function of the power system under the target topology, is the grid topology in the target state, is the sum of all elements in the matrix, Indicates that each element in the matrix takes the absolute value.

[0022] The constraints described above are the equations and inequalities that must be satisfied to solve the problem, including both equality and inequality constraints. The objective function is a metric used to evaluate the quality of the solution. For the DCOPF problem, this is typically to minimize the economic cost of the unit output scheme. A feasible solution indicates that the solution satisfies all constraints of the objective problem.

[0023] Currently, deep neural network (DNN) application methods for solving DCOPF problems, where the parameters of the DNN are not updated during the solution process, and where the output data is feasible, mainly focus on four technical directions: learning constraints on the target problem type from historical data (Learn From History, LFH), learning constraints on the target problem type from the problem model (Learn From Model, LFM), using DNN results as a warm start for optimization to achieve speedup (Predict As Warm Start, PAWS), and using DNN to obtain preliminary results and obtain the corresponding feasible solution through mapping (Predict And Projection, PAP).

[0024] in, LFH controls the internal structure of the DNN model or the preparation process of historical data, driving the DNN to fully learn the mapping relationship between energy demand and the feasible domain boundary in the process of learning the mapping law between energy demand and the optimal solution of the problem, so that the final output data meets the feasible domain boundary, which is jointly restricted by equality constraints and inequality constraints.

[0025] LFM adds the Lagrangian function of a class of problems to the loss function, and improves the probability that the output data is feasible by directly learning the mapping relationship between the input data and the boundary of the feasible region.

[0026] PAWS uses the output data of DNN as the initial solution or parameters of the optimization model, and uses traditional methods to solve the simplified optimization problem to ensure the feasibility of the results.

[0027] The PAP framework takes the infeasible data output by DNN as the starting point and pulls the starting point into the feasible domain through one or more constraint-based mappings.

[0028] Currently, the Gauge Mapping (GM) algorithm and the Deep Constraint Completion and Correction (DC3) algorithm are commonly used under the PAP framework.

[0029] The GM algorithm can only be used for problems with a constant feasible domain, restricting the DNN output data to a fixed complex high-dimensional space. The specific steps include: 1. Use simulation experiments or historical data to construct a sample set with power demand as sample input and power system scheduling plan as sample label.

[0030] 2. Use the tanh activation function to construct a DNN so that it outputs a vector with all elements between [-1, 1] based on the input power demand.

[0031] 3. Connect a differentiable canonical mapping GM function to the DNN. When the input of the GM function is a vector whose elements are all between [-1, 1], it outputs a vector with the same dimension and is within the spatial range corresponding to the GM function passing through the origin.

[0032] 4. Combine the DNN and GM functions to obtain a canonical mapping DNN, and use historical data to train the DNN.

[0033] 5. In actual application, the power demand is input into the DNN, which then outputs the power dispatch plan.

[0034] The GM algorithm can only map two convex sets that pass through the same points with fixed positions. The feasible domain of the DCOPF problem changes with the changes in the power system topology and energy demand. It is impossible to determine a fixed point that is always within the feasible domain of the DCOPF. Therefore, the GM algorithm cannot be directly applied to solving the DCOPF problem whose feasible domain changes with energy demand.

[0035] The DC3 algorithm consists of two parts: preparation and application. The specific steps of the preparation part include: 1. Divide the variables of the unit output plan into independent variable sets and the dependent variable , using the equality constraints in the target problem to construct the mapping function between the independent variable and the dependent variable .

[0036] 2. Use simulation experiments or historical data to construct a model with power demand as sample input. is the sample set of sample labels.

[0037] 3. The structure can be obtained by inputting power demand DNN.

[0038] 4. Use the gradient backpropagation algorithm to make DNN learn power demand and The mapping relationship between them.

[0039] The specific steps of the application part include: 1. By inputting the power demand into the trained DNN, we can obtain .

[0040] 2. Calculation , construct a set of variables that satisfies the equality constraints .

[0041] 3. Compute variable sets The inequality constraints that are not satisfied, calculate the corresponding degree of dissatisfaction , and calculate right The gradient of , and use this gradient to update , until All inequality constraints are satisfied.

[0042] The DC3 algorithm cannot guarantee convergence, and thus cannot guarantee its ability to obtain a feasible solution. In addition, step 3 of the application part contains an indefinite number of result adjustments based only on inequality constraints. Since the number and direction of mapping are only affected by the constraints and not guided by the DCOPF objective function, the DC3 algorithm is difficult to guarantee the optimality of the output data when the DCOPF problem changes greatly.

[0043] Furthermore, aside from the two aforementioned technologies, other major technical solutions are also unable to quickly obtain feasible DCOPF solutions for variable power topologies. Specifically, LFH and LFM lack data feasibility due to the black-box nature of the DNN used to output data. Therefore, the directly output scheduling solution cannot guarantee that it satisfies all problem constraints. Furthermore, PAWS, due to its unavoidable use of optimization methods, suffers from exponentially increasing solution time as the problem scale increases, thus lacking efficiency.

[0044] In summary, when the existing technology faces changes in the power grid topology, the initial solution obtained by DNN has low feasibility, and the post-processing method using mapping cannot ensure the computational efficiency of the acquisition process while ensuring the acquisition of a feasible solution. There are problems such as low efficiency in solving the DCOPF problem and low feasibility of the solution results.

[0045] In view of this, an embodiment of the present application provides a data mechanism joint driving method and system for solving DC optimal power flow, which is applicable to the DCOPF problem with feasible solutions under any power topology. The power demand is input into the deep neural network DNN to obtain the unit output plan output by the DNN, and any unit output plan in the unit processing plan is the initial solution; if the initial solution satisfies the problem constraints, the initial solution is used as the intermediate feasible solution, wherein the problem constraints include equality constraints and inequality constraints; if the initial solution does not satisfy the problem constraints, the constraints and initial solutions of the problem to be solved are processed based on the equality constraints by the variable elimination method to obtain the transformed constraint set and the initial solution, wherein the transformed constraint set only contains inequality constraints; based on the constraint set and the normal vector that violates the inequality constraint, the position of the initial solution is iteratively adjusted until the infeasible solution of the initial solution falls into the feasible domain; based on the equality constraints and the infeasible solution of the initial solution, the initial solution is restored to obtain the intermediate feasible solution that satisfies the equality constraints and the inequality constraints. The intermediate feasible solution is input into a pre-built neural network called FONET, which guarantees feasibility and embeds optimality. Adjustment information for the intermediate feasible solution is determined. Based on this adjustment information, the intermediate feasible solution is adjusted to obtain the target feasible solution. Determining the adjustment information and adjusting the initial solution using this information ensures both solution efficiency and the feasibility of the solution, resulting in a more optimal solution.

[0046] To facilitate understanding of the technical solution provided by the embodiment of the present application, a data mechanism joint driving method and system for solving DC optimal power flow provided by the embodiment of the present application are described below with reference to the accompanying drawings.

[0047] It should be noted that the embodiments of the present application can be applied to single-time unit scheduling in large-scale power systems to quickly obtain a scheduling plan that meets power balance, unit output constraints, and line flow constraints.

[0048] In this embodiment of the present application, power demand is input into an optimal solution prediction neural network, such as a deep neural network (DNN), to obtain an initial solution for the optimal power flow outputted by the neural network. If the initial solution does not meet the problem constraints, an infeasible solution correction algorithm is used to process the initial solution and output an intermediate feasible solution. If the initial solution meets the problem constraints, the initial solution is used as the intermediate feasible solution. Finally, in the third step, the intermediate feasible solution is input into FONET, and adjustment information for the intermediate feasible solution is output to guide adjustments to the intermediate feasible solution.

[0049] Among them, the infeasible solution correction algorithm first processes the constraints and initial solution of the problem to be solved based on the equality constraints through the variable elimination method to obtain the transformed constraint set and initial solution. The constraint set is composed only of inequality constraints; secondly, based on the constraint set and the normal vector that violates the inequality constraint, the position of the transformed initial solution is iteratively adjusted until the infeasible solution of the initial solution falls into the feasible domain; finally, based on the equality constraints and the infeasible solution of the initial solution, the initial solution is restored to obtain an intermediate feasible solution that satisfies the equality and inequality constraints.

[0050] FONET includes a feasible solution adjustment direction activation function and a feasible solution adjustment distance activation function. The former adjusts its parameters based on intermediate feasible solutions, the equality constraints of the DC optimal power flow problem, and the objective function, ensuring that a superior feasible solution exists in the adjustment direction output by the feasible solution adjustment neural network. The latter adjusts its parameters based on the adjustment direction, intermediate feasible solutions, the inequality constraints of the DC optimal power flow problem, and the objective function, ensuring that the adjustment displacement output by the feasible solution adjustment neural network is a non-suboptimal feasible solution. The feasible solution adjustment direction activation function is described in the OAERN section below; the feasible solution adjustment distance activation function is described in the OAIRN section below.

[0051] See also Figure 1 , Figure 1 This is a flow chart of a data mechanism joint driving method for solving DC optimal power flow provided by an embodiment of the present application. The data mechanism joint driving method for solving DC optimal power flow can be applied to a server or server cluster, etc. The method specifically includes S101-S107.

[0052] The embodiments of this application can be applied to a DNN application framework that uses DNN after preprocessing (UDAP), ensuring the feasibility and optimality of DNN output data. The methods and DNN application framework described in the embodiments of this application can achieve a balance between the computational efficiency of DNN methods, the feasibility of output data, and the optimality of output data.

[0053] See also Figure 2 , Figure 2The present invention provides an overall flow chart for solving DCOPF based on preprocessing of a DNN and then using a neural network UDAP. "Solving a DNN" is any neural network that can provide an initial solution to the DCOPF problem. UDAP does not require the feasibility or optimality of this neural network. In addition, all computational steps in UDAP, including the constraint-guided feasibility modification algorithm (FEMAⅡ) and the neural network with feasibility-ensured and optimality-embedded (FONET), which can be used within a finite number of steps, are key steps for ensuring the effectiveness of the proposed framework. The application of the FEMAⅡ algorithm includes the following steps S103-S105, and the application of the FONET neural network includes the following steps S106-S107.

[0054] S101: Inputting power demand into a deep neural network DNN to obtain a unit output plan output by the deep neural network DNN, wherein any unit output plan in the unit processing plan is an initial solution.

[0055] Step S101 is used to "solve the DNN." The power demand is input into multiple pre-trained deep neural network models (DNNs) to obtain different unit output plans output by the multiple DNNs. Any of the multiple unit output plans can be the initial solution. In one possible implementation, the unit output plan with the smallest violation of the inequality constraint is the initial solution, which is not limited in this application.

[0056] After determining the initial solution, the variable elimination method can be used to process the problem to be solved and the initial solution, converting it into an optimization problem without equality constraints and the corresponding initial solution.

[0057] Determine whether the initial solution satisfies the problem constraints. If the initial solution satisfies the problem constraints, execute step S102 and subsequent steps. If the initial solution does not satisfy the problem constraints, execute step S103 and subsequent steps. The problem constraints include equality constraints and inequality constraints.

[0058] S102: Take the initial solution as the intermediate feasible solution and jump to steps S106-S107.

[0059] S103: Processing the constraints and initial solution of the problem to be solved based on the equality constraints by a variable elimination method to obtain a transformed constraint set and an initial solution, wherein the constraint set consists only of inequality constraints.

[0060] The FEMAⅡ algorithm consists of three steps. Step A is the transformation of the constraint set and infeasible solution based on equality constraints. As a linear programming problem, all the inequality constraints (2) (4) and equality constraints (3) in DCOPF can be expressed as follows based on matrix multiplication: and , due to the particularity of the DCOPF problem, is a matrix with only one row, is a matrix with 1 row and 1 column. Middle The value of the first column in the row is divided by The first column value is , by and The Row Subtraction and , so that the inequality constraints are consistent with The variables corresponding to the first column in the constraint are decoupled, and then and The first column of is deleted, so that the constraint set of DCOPF problem is transformed into a constraint consisting only of inequality constraints. .

[0061] For infeasible solutions ,Will Delete the first variable in the list and construct the remaining variables in the original order. . Make conform to and , which is equivalent to conform to .

[0062] S104: Iteratively adjusting the position of the initial solution based on the constraint set and the normal vector that violates the inequality constraint until an infeasible solution of the initial solution falls into a feasible region.

[0063] Step B is an iterative algorithm for converting infeasible solutions into feasible solutions based on the spatial relationship between infeasible solutions and inequality constraints. Move iteratively so that Finally, the transformed constraint set is satisfied.

[0064] In one possible implementation, iteratively adjusting the position of the initial solution based on the constraint set and the normal vector violating the inequality constraint until the infeasible solution of the initial solution falls into the feasible region includes the following steps: 1. Calculate the infeasible solution of the initial solution based on the constraint set The inequality constraint that was violated.

[0065] 2. Calculate the normal vector that violates the inequality constraint and use the sum of the normal vectors as the infeasible solution of the initial solution Direction of movement .

[0066] 3. Calculate the infeasible solution of the initial solution In the direction of movement The first inequality constraint that is violated is encountered as Moving distance , which is the infeasible solution of the initial solution The distance from the initial violated inequality constraint is the moving distance of the infeasible solution of the initial solution in a single iteration. , the initially violated inequality constraint is in the moving direction The first violated inequality constraint encountered on .

[0067] 4. The infeasible solution of the initial solution Follow the above moving direction and moving distance After making the move, calculate the infeasible solution of the initial solution All hyperplanes where lies, and calculate the intersection plane of the hyperplanes.

[0068] 5. Change the current moving direction Project it onto the intersection plane to generate a new movement direction.

[0069] 6. Repeat steps 3-5 above until the intersection plane or the new movement direction meets the preset conditions, which are that the intersection plane is a point or the new movement direction is a zero vector.

[0070] 7. Repeat steps 1-6 above until an infeasible solution of the initial solution is determined Fall into the feasible region.

[0071] S105: Reducing the initial solution based on the equality constraint and the infeasible solution of the initial solution to obtain an intermediate feasible solution that satisfies the equality constraint and the inequality constraint.

[0072] Step C is to restore the feasible solution based on the equality constraint. and DCOPF equality constraint calculation The first variable in Perform splicing to obtain a solution that satisfies both the equality and inequality constraints of the DCOPF problem , which satisfies the equality and inequality constraints is an intermediate feasible solution.

[0073] S106: Input the intermediate feasible solution into a pre-built neural network FONET that guarantees feasibility and embeds optimality, and determine adjustment information of the intermediate feasible solution.

[0074] FONET can obtain the adjustment direction based on the existing feasible solution, ensuring that the adjustment result is an equally optimal or better feasible solution. In the embodiment of the present application, FONET is used to quickly iterate the feasible solution, thereby efficiently obtaining a feasible solution to the DCOPF problem considering optimality. Among them, the adjustment information includes the adjustment direction and the adjustment distance. FONET includes a neural network with optimality and the ability to ensure the satisfaction of equality constraints (optimality-aware-equality-constraint-remain neural network, OAERN) and a neural network with optimality and the ability to ensure the satisfaction of inequality constraints (optimality-aware-inequality-constraint-remain neural network, OAIRN). OAERN is used to determine the adjustment direction, and OAIRN is used to determine the adjustment distance.

[0075] In one possible implementation, inputting the intermediate feasible solution into a pre-built neural network (FONET) that ensures feasibility and embeds optimality, and determining adjustment information for the intermediate feasible solution, includes: inputting the intermediate feasible solution into FONET to obtain a first activation function (OAERN), where the first activation function is used to determine an adjustment direction; determining a backup adjustment distance based on the adjustment direction and a first preset formula; calculating a primary adjustment distance based on a second preset formula; and determining the adjustment distance from the primary adjustment distance and the backup adjustment distance.

[0076] In the embodiment of the present application, an activation function is designed to ensure that the moving direction of the DNN output does not change the satisfaction of the existing feasible solution to the equality constraint, and there is a better solution than the known feasible solution in this moving direction. The neural network that uses the above activation function as the activation function of the last layer of the neural network is called OAERN. The internal framework of OAERN is as follows: Figure 3 shown.

[0077] The steps to determine the activation function mainly include: 1. Calculate the average marginal cost matrix of all units under the current solution .

[0078] 2. Label units with marginal costs higher than the average cost with negative labels and units with marginal costs lower than the average cost with positive labels.

[0079] 3. Incorporate positive and negative labels into the activation function middle: (6) (7) (8) (9) in, It is a neural layer output that describes the characteristics of each unit. dimensional matrix, is the number of units and has no direct physical properties. and is a matrix used to extract the features of positive and negative labeled units, with dimensions of and .exist and In , each row represents a unit labeled as positive or negative, and each column represents a unit in the target problem. The unit is If there are units with positive marks, No. Row and The column is 1, while the other elements of the row are 0. represents the transpose operation, represents the diagonal matrix of the vector, Indicates taking the absolute value, It means to sum all the elements in the matrix.

[0080] After the adjustment direction is determined, the maximum distance that can be adjusted is in Substitute into formula (2) (4) As described in the following two formulas: (10) (11) By unifying these formulas into , the adjustment direction of the current solution in the OAERN output can be calculated. The first preset formula includes the following formulas (12) and (13). The maximum adjustable distance (Also the spare adjustment distance) is: (12) (13) In order to ensure that there is a better solution in the final moving direction, the minimum adjustment distance cannot be negative, so the minimum adjustment distance is 0. Based on the boundary value of the above distance interval, we can use the SoftMax function , limiting the output data of the neural network to between the minimum and maximum adjustment distances, The specific function is: (14) by The neural network with the activation function of the last layer can ensure the feasibility of the adjustment result, but cannot guarantee the optimality. Similar to the maximum adjustment distance, the optimal adjustment distance will also be determined as the adjustment direction is determined. The second preset formula is the following formula (15). The main adjustment distance can be expressed as: (15) in, It is composed of the marginal cost of each unit dimensional matrix, and yes dimensional unit output cost coefficient matrix, the value of each row element is the secondary or primary cost coefficient of the unit in the row; is element-wise multiplication.

[0081] comprehensive and It can simultaneously ensure the feasibility and optimality of the adjustment results. The specific principles are: When DCOPF optimality is improved, As the final adjustment distance, otherwise use As the final adjustment distance The above neural network embedded with inequality constraints and objective functions is called OAIRN, and it works with OAERN to obtain the flow chart of adjusting displacement as shown in the following figure: Figure 4 As shown in Figure 2, EC, IC, and OF are the equality constraints, inequality constraints, and objective function, respectively.

[0082] DNN application for DCOPF problem using DNN after preprocessing (UDAP), DNN preprocessing method with initial feasible solution as the core, FEMAⅡ algorithm that can obtain feasible solution within a finite number of steps, and DNN activation function that can ensure that the adjustment result is an equally good or better feasible solution. and .

[0083] S107: Adjust the intermediate feasible solution based on the target distance to obtain a target feasible solution.

[0084] The advantage of the FEMA II algorithm is that it is the first algorithm to require no gradient calculations and to guarantee a feasible solution within a finite number of calculations. Its equality constraints are guaranteed in step C above, while its satisfaction of inequality constraints is guaranteed by the following explanation: See also Figure 5 , Figure 5 The following is a schematic diagram of the effect of moving an infeasible solution according to an embodiment of the present application. The iterative feasibility of an infeasible solution is described as follows: I: For an infeasible solution and a constraint that it violates, while moving toward the hyperplane corresponding to the constraint along the normal of the hyperplane, and not reaching the hyperplane, the infeasible solution will continue to approach the feasible region.

[0085] See also Figure 5 As shown in part (a), once the infeasible solution and the hyperplane are given, the feasible region can be determined for each point. and Value. Obviously The distance between an infeasible solution and any feasible point in the process of approaching 0 It is bound to continue to shrink.

[0086] II: For an infeasible solution and any number of constraints violated by the infeasible solution, the hyperplane corresponding to the constraints is recorded as , and the normal vector of the hyperplane away from the infeasible solution is recorded as , then there must be the following conclusion: when the infeasible solution is along the direction obtained by any combination of these normal vectors ,in For any number, when moving, the infeasible solution always stays close to the feasible region before moving to any of the above hyperplanes.

[0087] Assume that the current point is along the direction Upper and The minimum distance between the hyperplanes in is , the displacement can be Decompose into Second-rate The sub-displacements are further decomposed into the sub-displacements in each sub-direction Under such subdivision, the starting point of each sub-displacement violates all the constraints violated by the starting point of the entire displacement. Therefore, no matter which hyperplane corresponding to the constraint violation is displaced, it will be one step closer to the feasible region as described in I above. For ease of understanding, please refer to Figure 5 The two parts (b) and (c) shown in the figure illustrate the There are two specific situations when .

[0088] III: After projecting the moving direction onto the hyperplane corresponding to the inequality constraint where the infeasible solution is located or the intersection of the hyperplanes, the infeasible solution continues to approach the feasible region before moving in the new moving direction until it encounters the hyperplane corresponding to the new inequality constraint.

[0089] The distance relationship between any three points in high-dimensional space will not change after dimensionality reduction. Therefore, if a direction in the original multi-dimensional space can make an infeasible solution close to the feasible domain, then after dimensionality reduction, the infeasible solution after dimensionality reduction will definitely move closer to the feasible domain after dimensionality reduction by moving in the direction after dimensionality reduction.

[0090] For easier understanding, see Figure 5 Part (d) shown, where the two moves correspond to and , the first move will change the infeasible solution from Move to , now in constraint On the corresponding hyperplane, the second moving direction is Normal vector In the constraints The projection of the hyperplane where it is located, therefore, it is an infeasible solution Will along Move towards the feasible region , thus obtaining a feasible solution.

[0091] IV: The number of correction steps required to make the infeasible solution satisfy any inequality constraint is no greater than the number of inequality constraints violated during the correction process , the total number of correction steps must be less than .

[0092] In the process of moving, when the infeasible solution moves to any inequality constraint After reaching the corresponding hyperplane or hyperplane intersection, the next move may lead to an infeasible solution to the inequality constraint only when moving to a vertex of a high-dimensional polyhedron consisting of multiple inequality constraints. , and since the infeasible solution is always approaching the feasible region, it will not move to this vertex repeatedly. The number of times an infeasible solution moves to the feasible region must be less than the number of vertices. A vertex is composed of inequality constraints that are not satisfied by multiple infeasible solutions, so the number of moves must be less than the number of infeasible solutions that have been violated during the movement.

[0093] Each iteration in OAIRN will bring the infeasible solution closer to the feasible region, so as the number of iterations increases, the infeasible solution will eventually enter the feasible region.

[0094] OAERN can ensure that all points in the moving direction do not change the satisfaction of the current solution to the equality constraints, and that there must be a solution that is better than the current solution in the moving direction.

[0095] According to the distributive law and associative law of matrix multiplication, the right side of formula (5) can be divided into two parts, namely: (16) According to the matrix and From the construction process of , we can infer that: and Each column of is a vector consisting of an element equal to 1 and multiple elements equal to 0, so the sum of a matrix multiplied on the left or before and after is unchanged. The above characteristics are expressed by the following equations (17) and (18): (17) (18) It can also be inferred that: If the diagonal matrix If the diagonal element of any column in is +1, then There must be one and only one element in the same row equal to 1, otherwise If the row is all zeros, If the diagonal element of any column is -1, then There is only one element in the same row that is equal to 1, otherwise This row of is all zero. It can be deduced that as well as Therefore, formula (16) can be expressed as: (19) Sum equation (19): (20) According to the definition of formula (7), and The sum of all elements of the matrix is ​​1, so the result of formula (20) must be equal to 0. equal When , there must be: , that is, the movement does not change the satisfaction of the current solution to the equality constraints.

[0096] Secondly, according to the record of formula (5), the objective function (1) can be rewritten as: (twenty one) Therefore, there exists: (twenty two) Among them, the satisfaction of formula (22) is equivalent to Compare According to the multiplication distributive law and formula (20), when the distance is adjusted When , the left side of inequality (22) can be equivalent to: (twenty three) in, It is composed of the marginal cost of each unit dimensional matrix. According to the labeling basis of positive and negative labels and The construction and use process, once , there is always the following phenomenon: in the marginal cost matrix, The value at each negative element in the marginal cost matrix must be greater than the average marginal cost, and vice versa. Therefore, if the minimum value greater than the average marginal cost and the maximum value less than the average marginal cost in the marginal cost matrix are recorded as and , then there must be: (twenty four) That is When , equation (24) is satisfied, so there is always a solution in the adjustment direction of the OAERN output that is better than the known feasible solution.

[0097] It's important to note that FONET's first neural layer receives the intermediate feasible solution, while the other neural layers receive relevant computational data, including power demand, unit generation cost model parameters, equality constraints, and inequality constraints. FONET processes this data to produce a more optimal target feasible solution.

[0098] The above embodiment of the present application provides a data mechanism joint driving method for solving the DC optimal power flow. Next, the present application embodiment also provides a data mechanism joint driving system for solving the DC optimal power flow, which is used to execute the above Figure 1 The function of the data mechanism joint drive system for solving the DC optimal power flow is explained. The structural diagram of the data mechanism joint drive system for solving the DC optimal power flow is shown in FIG. Figure 6 As shown, it includes an acquisition module 601 , a judgment module 602 , an adjustment module 603 , a restoration module 604 and a determination module 605 .

[0099] An acquisition module 601 is configured to input power demand into a deep neural network (DNN) to obtain a unit output solution output by the deep neural network (DNN), wherein any unit output solution in the unit processing solution is an initial solution. a judgment module, configured to use the initial solution as an intermediate feasible solution if the initial solution satisfies the problem constraints, wherein the problem constraints include equality constraints and inequality constraints; The judgment module 602 is further configured to process the constraints and the initial solution of the problem to be solved based on the equality constraints by using a variable elimination method if the initial solution does not satisfy the problem constraints, thereby obtaining a transformed constraint set and the initial solution, wherein the constraint set consists only of inequality constraints. An adjustment module 603 is configured to iteratively adjust the position of the initial solution based on the constraint set and the normal vector violating the inequality constraint until the infeasible solution of the initial solution falls into a feasible region; A restoration module 604 is configured to restore the initial solution based on the equality constraints and the infeasible solution of the initial solution to obtain an intermediate feasible solution that satisfies the equality constraints and the inequality constraints; A determination module 605 is configured to input the intermediate feasible solution into a pre-built neural network FONET that ensures feasibility and embeds optimality, and determine adjustment information of the intermediate feasible solution; The adjustment module 603 is further configured to adjust the intermediate feasible solution based on the adjustment information to obtain a target feasible solution.

[0100] In one possible implementation, the adjustment module 603 is configured to iteratively adjust the position of the initial solution based on the constraint set and the normal vector violating the inequality constraint until the infeasible solution of the initial solution falls into the feasible region, including: a calculation submodule, configured to calculate the inequality constraints violated by the infeasible solution of the initial solution according to the constraint set; calculate the normal vectors violating the inequality constraints, and use the sum of the normal vectors as the moving direction of the infeasible solution of the initial solution; a determination submodule, configured to use the distance between the infeasible solution of the initial solution and the initially violated inequality constraint as the single-iteration moving distance of the infeasible solution of the initial solution, wherein the initially violated inequality constraint is the first violated inequality constraint encountered in the moving direction; The calculation submodule is further configured to calculate an intersection plane of a hyperplane where an infeasible solution of the initial solution is located after adjustment according to the moving direction and moving distance; A generating submodule, configured to project the moving direction onto the intersection plane to generate a new moving direction; The determination submodule is used to determine that the infeasible solution of the initial solution falls into the feasible domain until the intersection plane or the new moving direction meets a preset condition, and the preset condition is that the intersection plane is a point or the new moving direction is an all-zero vector.

[0101] In a possible implementation, the adjustment information includes an adjustment direction and an adjustment distance, and the determining module 605 is specifically configured to: Inputting the intermediate feasible solution into the FONET to obtain a first activation function, wherein the first activation function is used to determine the adjustment direction; determining a standby adjustment distance based on the adjustment direction and a first preset formula; Calculate the main adjustment distance according to the second preset formula; The adjustment distance is determined from the primary adjustment distance and the backup adjustment distance.

[0102] In one possible implementation, the FONET includes a neural network OAERN that is optimal and can ensure the satisfaction of equality constraints, and a neural network OAERN that is optimal and can ensure the satisfaction of inequality constraints; the OAERN is used to determine the adjustment direction, and the OAERN is used to determine the adjustment distance.

[0103] An embodiment of the present application provides a data-mechanism joint drive system for solving DC optimal power flow, including an acquisition module, a judgment module, an adjustment module, a restoration module, and a determination module. The system is applicable to DCOPF problems with feasible solutions under any power topology. The acquisition module is used to input the power demand into a deep neural network (DNN) to obtain a unit output plan output by the DNN. Any unit output plan in the unit processing plan is the initial solution. The judgment module is used to use the initial solution as an intermediate feasible solution if the initial solution satisfies the problem constraints. The problem constraints include equality constraints and inequality constraints. If the initial solution does not satisfy the problem constraints, the constraints and initial solution of the problem to be solved are processed based on the equality constraints through variable elimination to obtain a converted constraint set and initial solution, wherein the converted constraint set only contains inequality constraints. The adjustment module is used to iteratively adjust the position of the initial solution based on the constraint set and the normal vector that violates the inequality constraint until the infeasible solution of the initial solution falls into the feasible domain. The restoration module is used to restore the initial solution based on the equality constraints and the infeasible solution of the initial solution to obtain an intermediate feasible solution that satisfies the equality constraints and inequality constraints. The determination module is configured to input the intermediate feasible solution into a pre-built neural network (FONET) that ensures feasibility and embeds optimality, and determine adjustment information for the intermediate feasible solution. The adjustment module is further configured to adjust the intermediate feasible solution based on the adjustment information to obtain a target feasible solution. In this way, determining the adjustment information and adjusting the initial solution using this adjustment information can simultaneously ensure solution efficiency and the feasibility of the solution result, resulting in a solution with higher optimality.

[0104] Based on the data mechanism joint driving method for solving DC optimal power flow provided by the above method embodiment, the embodiment of the present application provides a data mechanism joint driving device for solving DC optimal power flow, see Figure 7 , the device includes: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, and the one or more programs include instructions. When the instructions are executed by the processor, the processor executes the data-mechanism joint driving method for solving DC optimal power flow as described in any of the above embodiments.

[0105] Based on the data mechanism joint driving method for solving DC optimal power flow provided in the above method embodiment, an embodiment of the present application provides a storage medium, which stores instructions. When the instructions are executed by a computer, the computer implements the data mechanism joint driving method for solving DC optimal power flow as described in any of the above embodiments.

[0106] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple. For relevant parts, refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0107] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0108] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A data-mechanism joint driving method for solving DC optimal power flow, characterized by: The method comprises: Inputting the power demand into a deep neural network DNN to obtain a unit output plan output by the deep neural network DNN, wherein any unit output plan in the unit processing plan is an initial solution; If the initial solution satisfies the problem constraints, the initial solution is used as an intermediate feasible solution, wherein the problem constraints include equality constraints and inequality constraints; If the initial solution does not satisfy the problem constraints, the constraints and the initial solution of the problem to be solved are processed based on the equality constraints by a variable elimination method to obtain a transformed constraint set and an initial solution, wherein the constraint set consists only of inequality constraints; Iteratively adjusting the position of the initial solution based on the constraint set and the normal vector that violates the inequality constraint until an infeasible solution of the initial solution falls into a feasible region; Reducing the initial solution based on the equality constraint and the infeasible solution of the initial solution to obtain an intermediate feasible solution that satisfies the equality constraint and the inequality constraint; Inputting the intermediate feasible solution into a pre-built neural network FONET that guarantees feasibility and embeds optimality, and determining adjustment information of the intermediate feasible solution; The intermediate feasible solution is adjusted based on the adjustment information to obtain a target feasible solution.

2. The method according to claim 1, characterized in that The iterative adjustment of the position of the initial solution based on the constraint set and the normal vector violating the inequality constraint until the infeasible solution of the initial solution falls into the feasible domain includes: Calculate the inequality constraints violated by the infeasible solution of the initial solution according to the constraint set; Calculating a normal vector that violates the inequality constraint, and using the sum of the normal vectors as a moving direction of the infeasible solution of the initial solution; The distance between the infeasible solution of the initial solution and the initially violated inequality constraint is used as the single-iteration moving distance of the infeasible solution of the initial solution, wherein the initially violated inequality constraint is the first violated inequality constraint encountered in the moving direction; Calculating an intersection plane of a hyperplane where an infeasible solution of the initial solution is located after adjustment according to the moving direction and moving distance; Projecting the moving direction onto the intersection plane to generate a new moving direction; Until the intersection plane or the new moving direction meets a preset condition, it is determined that the infeasible solution of the initial solution falls into the feasible domain, and the preset condition is that the intersection plane is a point or the new moving direction is an all-zero vector.

3. The method according to claim 1, characterized in that The adjustment information includes an adjustment direction and an adjustment distance. The intermediate feasible solution is input into a pre-built neural network FONET that ensures feasibility and embeds optimality, and the adjustment information of the intermediate feasible solution is determined, including: Inputting the intermediate feasible solution into the FONET to obtain a first activation function, wherein the first activation function is used to determine the adjustment direction; determining a standby adjustment distance based on the adjustment direction and a first preset formula; Calculate the main adjustment distance according to the second preset formula; The adjustment distance is determined from the primary adjustment distance and the backup adjustment distance.

4. The method according to claim 3, characterized in that The FONET includes a neural network OAERN that is optimal and can ensure equality constraints are satisfied, and a neural network OAIRN that is optimal and can ensure inequality constraints are satisfied; the OAERN is used to determine the adjustment direction, and the OAIRN is used to determine the adjustment distance.

5. A data-mechanism combined drive system for solving DC optimal power flow, characterized by: The system comprises: an acquisition module, configured to input the power demand into a deep neural network (DNN) to obtain a unit output plan output by the deep neural network (DNN), wherein any unit output plan in the unit processing plan is an initial solution; a judgment module, configured to use the initial solution as an intermediate feasible solution if the initial solution satisfies the problem constraints, wherein the problem constraints include equality constraints and inequality constraints; The judgment module is further configured to, if the initial solution does not satisfy the problem constraints, process the constraints and the initial solution of the problem to be solved based on the equality constraints by a variable elimination method to obtain a transformed constraint set and an initial solution, wherein the constraint set consists only of inequality constraints; an adjustment module, configured to iteratively adjust a position of the initial solution based on the constraint set and the normal vector violating the inequality constraint, until an infeasible solution of the initial solution falls into a feasible region; A restoration module, configured to restore the initial solution based on the equality constraint and the infeasible solution of the initial solution to obtain an intermediate feasible solution that satisfies the equality constraint and the inequality constraint; A determination module is used to input the intermediate feasible solution into a pre-built neural network FONET that ensures feasibility and embeds optimality, and determine adjustment information of the intermediate feasible solution; The adjustment module is further configured to adjust the intermediate feasible solution based on the adjustment information to obtain a target feasible solution.

6. The system according to claim 5, characterized in that The adjustment module is configured to iteratively adjust the position of the initial solution based on the constraint set and the normal vector violating the inequality constraint until the infeasible solution of the initial solution falls into the feasible domain, comprising: a calculation submodule, configured to calculate the inequality constraints violated by the infeasible solution of the initial solution according to the constraint set; calculate the normal vectors violating the inequality constraints, and use the sum of the normal vectors as the moving direction of the infeasible solution of the initial solution; a determination submodule, configured to use the distance between the infeasible solution of the initial solution and the initially violated inequality constraint as the single-iteration moving distance of the infeasible solution of the initial solution, wherein the initially violated inequality constraint is the first violated inequality constraint encountered in the moving direction; The calculation submodule is further configured to calculate an intersection plane of a hyperplane where an infeasible solution of the initial solution is located after adjustment according to the moving direction and moving distance; A generating submodule, configured to project the moving direction onto the intersection plane to generate a new moving direction; The determination submodule is used to determine that the infeasible solution of the initial solution falls into the feasible domain until the intersection plane or the new moving direction meets a preset condition, and the preset condition is that the intersection plane is a point or the new moving direction is an all-zero vector.

7. The system according to claim 5, characterized in that The adjustment information includes an adjustment direction and an adjustment distance, and the determination module is specifically configured to: Inputting the intermediate feasible solution into the FONET to obtain a first activation function, wherein the first activation function is used to determine the adjustment direction; determining a standby adjustment distance based on the adjustment direction and a first preset formula; Calculate the main adjustment distance according to the second preset formula; The adjustment distance is determined from the primary adjustment distance and the backup adjustment distance.

8. The system according to claim 7, characterized in that The FONET includes a neural network OAERN that is optimal and can ensure the satisfaction of equality constraints and a neural network OAERN that is optimal and can ensure the satisfaction of inequality constraints; the OAERN is used to determine the adjustment direction and the OAERN is used to determine the adjustment distance.

9. A data mechanism combined drive device for solving DC optimal power flow, characterized in that: The device includes: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, and the one or more programs include instructions. When the instructions are executed by the processor, the processor executes the data-mechanism joint driving method for DC optimal power flow solution according to any one of claims 1 to 4.

10. A storage medium, characterized in that: The storage medium stores instructions, which, when executed by a computer, enable the computer to implement the data-mechanism joint driving method for solving DC optimal power flow according to any one of claims 1 to 4.

Images