Method for optimal operation of microgrid considering multi-dimensional stability index balance and related product

Abstract

The application discloses a kind of microgrid optimization operation method and related product considering the balance of multidimensional stability index, and the application relates to microgrid technical field, method includes: based on input convex neural network constructs multidimensional stability proxy model, wherein, multidimensional stability proxy model exports analytic gradient information;Multidimensional stability proxy model is used to establish two-stage optimization scheduling mathematical model considering dynamic performance cost;Based on nested decomposition architecture, two-stage optimization scheduling model is solved, and the global optimal scheduling scheme of microgrid is obtained, wherein, the discrete start-stop state decision variable of generator set in microgrid is handled by Benders decomposition in outer layer of nested decomposition architecture, and analytic gradient information is handled by outer approximation algorithm in inner layer Nonlinear stability constraint and dynamic performance cost. The present application solves the problem of low operation efficiency of existing microgrid operation method from the aspects of calculation solving efficiency and operation economy.

Description

Technical Field

[0001] This invention relates to the field of microgrid technology, and in particular to a method for optimizing the operation of a microgrid that takes into account the balance of multidimensional stability indices, and related products. Background Technology

[0002] With the increasing penetration rate of new energy sources, microgrids, as an important carrier for the effective utilization of distributed energy, have attracted widespread attention for their operational safety. Due to the generally low inertia characteristics of new energy power generation equipment, microgrids have weak resistance to disturbances, making their stability issues increasingly prominent. As microgrid structures become increasingly complex, such as AC / DC hybrid microgrids, the dynamic coupling between their internal subdomains further increases the difficulty of stability assessment.

[0003] In the daily operation and scheduling of microgrids, stability constraints must always be considered to ensure that issued scheduling commands do not lead to microgrid system instability. A major current trend is to utilize surrogate modeling techniques to transform time-consuming physical simulations into analytical mappings, thereby meeting the timeliness requirements for stability assessment.

[0004] One existing technology utilizes a metaheuristic algorithm to perform a black-box search on optimization scheduling problems with stability constraints. This approach mainly includes steps such as stability constraint embedding, population iterative evolution, and feasible solution selection. However, this existing technology has the following shortcomings: First, the optimization efficiency is low when dealing with complex optimization problems involving discrete decisions on generator start-up and shutdown and nonlinear stability constraints; second, the computational results of this approach are random, making it difficult to guarantee the global optimality of the scheduling decision scheme; third, the computation time of this approach is long in large-scale systems, making it difficult to meet the timeliness requirements of scheduling scenarios.

[0005] Another existing technique employs the Big-M linearization method, embedding stability discrimination logic into the scheduling model and using binary variables to transform nonlinear stability constraints into linear inequalities for solution. This technique achieves model linearization by introducing a large number of discrete variables. However, this technique has the following shortcomings: First, the introduction of a large number of binary variables in the modeling process transforms the scheduling model into a complex mixed-integer linear programming problem, with the search space growing exponentially with the system size. Second, in scenarios involving large-scale decision variables, such as day-ahead unit combinations, it faces a severe curse of dimensionality, making it difficult to obtain the optimal scheduling solution within the specified time. Third, this approach treats stability as a binary hard constraint of pass / fail, ignoring the dynamic response quality of the system within the stable region. This leads the optimizer to tend to select extreme control parameters to obtain redundant stability, resulting in an overly conservative scheduling scheme and low operational economy. Summary of the Invention

[0006] This invention provides a microgrid optimization operation method and related products that take into account the balance of multi-dimensional stability indices, aiming to solve the problem of low operating efficiency in existing microgrid optimization operation methods.

[0007] In a first aspect, embodiments of the present invention provide a microgrid optimization operation method considering the balance of multidimensional stability indices, comprising: A multidimensional stability surrogate model is constructed based on an input convex neural network. The multidimensional stability surrogate model outputs analytical gradient information and includes a stability discrimination model, a stability index level regression model, and a stability time regression model. A two-stage optimization scheduling mathematical model considering dynamic performance costs is established using the aforementioned multidimensional stability proxy model; The two-stage optimization scheduling model is solved based on a nested decomposition architecture to obtain the global optimal scheduling scheme of the microgrid. The outer layer of the nested decomposition architecture uses Benders decomposition to process the discrete start-stop state decision variables of the generators in the microgrid, while the inner layer uses an external approximation algorithm to process the nonlinear stability constraints and the dynamic performance cost using the analytical gradient information.

[0008] Secondly, embodiments of the present invention also provide a microgrid optimized operation device that considers the balance of multi-dimensional stability indicators, comprising: A construction unit is used to construct a multidimensional stability surrogate model based on an input convex neural network. The multidimensional stability surrogate model outputs analytical gradient information and includes a stability discrimination model, a stability index level regression model, and a stable time regression model. A unit is established to utilize the multidimensional stability proxy model to establish a two-stage optimization scheduling mathematical model that takes into account dynamic performance costs. The solution operation unit is used to solve the two-stage optimization scheduling model based on the nested decomposition architecture to obtain the global optimal scheduling scheme of the microgrid. The outer layer of the nested decomposition architecture uses Benders decomposition to process the discrete start-stop state decision variables of the generators in the microgrid, and the inner layer uses the external approximation algorithm to process the nonlinear stability constraints and the dynamic performance cost using the analytical gradient information.

[0009] Thirdly, embodiments of the present invention also provide a computer device, the computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the above-described method.

[0010] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the above-described method.

[0011] This invention provides a microgrid optimization operation method and related products that consider the balance of multidimensional stability indices. The method includes: constructing a multidimensional stability surrogate model based on an input convex neural network, wherein the multidimensional stability surrogate model outputs analytical gradient information, and the multidimensional stability surrogate model includes a stability discrimination model, a stability index level regression model, and a stability time regression model; establishing a two-stage optimization scheduling mathematical model considering dynamic performance costs using the multidimensional stability surrogate model; and solving the two-stage optimization scheduling model based on a nested decomposition architecture to obtain the globally optimal scheduling scheme for the microgrid, wherein the outer layer of the nested decomposition architecture uses Benders decomposition to process the discrete start-stop state decision variables of the generator units in the microgrid, and the inner layer uses an external approximation algorithm to process nonlinear stability constraints and the dynamic performance costs using the analytical gradient information. The technical solution of this invention constructs a multidimensional stability surrogate model by inputting a convex neural network and outputting analytical gradient information, providing a precise search direction for optimization and overcoming the low optimization efficiency of metaheuristic algorithms. It employs a nested decomposition architecture, with the outer Benders decomposition processing discrete start-stop state decision variables to avoid the curse of dimensionality in the Big-M method, and the inner outer approximation algorithm utilizing analytical gradients to handle nonlinear stability constraints, significantly improving solution efficiency. Furthermore, it establishes a two-stage optimization scheduling mathematical model that considers dynamic performance costs, avoiding overly conservative scheduling schemes and improving operational economy. Therefore, this invention solves the problem of low operational efficiency in existing microgrid operation methods from both computational solution efficiency and operational economy perspectives. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating a microgrid optimization operation method considering multidimensional stability index balancing, as provided in an embodiment of the present invention. Figure 2 This is a topology diagram of an AC / DC hybrid microgrid provided in an embodiment of the present invention; Figure 3 A schematic diagram of a sub-process of a microgrid optimization operation method considering the balance of multidimensional stability indices provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of another sub-process of a microgrid optimization operation method that takes into account the balance of multi-dimensional stability indices, provided in an embodiment of the present invention. Figure 5 This is a schematic diagram of another sub-process of a microgrid optimization operation method that takes into account the balance of multi-dimensional stability indices, provided as an embodiment of the present invention. Figure 6 A schematic block diagram of a microgrid optimized operation device considering the balance of multidimensional stability indices, provided in an embodiment of the present invention; Figure 7 This is a schematic block diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0016] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0017] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0018] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0019] Currently, in the field of microgrid technology, existing microgrid optimization operation methods suffer from low operating efficiency. To address this issue, this invention proposes a microgrid optimization operation method that considers the balance of multidimensional stability indices. This method constructs a multidimensional stability surrogate model by inputting a convex neural network and outputs analytical gradient information, providing a precise search direction for optimization and overcoming the low efficiency of metaheuristic algorithms. It employs a nested decomposition architecture: the outer Benders decomposition process handles discrete start-stop state decision variables, avoiding the curse of dimensionality in the Big-M method; the inner outer approximation algorithm utilizes analytical gradients to handle nonlinear stability constraints, significantly improving solution efficiency. Furthermore, a two-stage optimization scheduling mathematical model considering dynamic performance costs is established to avoid overly conservative scheduling schemes and improve operational economy. Therefore, this invention solves the problem of low operating efficiency in existing microgrid operation methods from both computational efficiency and operational economy perspectives. The invention is described in detail below through specific embodiments.

[0020] Please see Figure 1 , Figure 1 This is a flowchart illustrating a microgrid optimization operation method considering multi-dimensional stability index balancing, provided by an embodiment of the present invention. Figure 1 As shown, the method includes the following steps S110-S130.

[0021] S110. Construct a multidimensional stability surrogate model based on an input convex neural network, wherein the multidimensional stability surrogate model outputs analytical gradient information, and the multidimensional stability surrogate model includes a stability discrimination model, a stability index level regression model, and a stable time regression model.

[0022] In this embodiment of the invention, a microgrid refers to an AC / DC hybrid microgrid, such as... Figure 2 As shown, the AC / DC hybrid microgrid includes an AC subgrid and a DC subgrid. The AC subgrid connects the diesel generator set and AC loads via an AC bus, while the DC subgrid connects photovoltaic units, energy storage systems, and DC loads via a DC bus. Power exchange between the AC and DC subgrids occurs through a bidirectional AC / DC converter. The photovoltaic units are connected to the DC bus via a unidirectional DC / DC converter, and the energy storage system is also connected to the DC bus via a bidirectional DC / DC converter. The rated voltage of the AC bus is [voltage value missing]. The rated voltage of the DC bus is .

[0023] Under islanded conditions, due to the lack of synchronous machine inertial support, microgrids exhibit low inertia and weak damping, resulting in significant stability issues. To ensure the safe and stable operation of microgrid systems, existing technologies often employ the following approaches: (1) In terms of computational solutions: Due to the complexity of transient stability criteria, it is difficult to directly embed them into the scheduling model, leading to low computational efficiency. (2) In terms of operational economy: To address the instability risks caused by low inertia and weak damping, existing technologies treat stability as a binary hard constraint, resulting in conservative scheduling schemes that sacrifice operational economy for safety redundancy, leading to low operational economy. Therefore, the operating methods for microgrids under islanded conditions suffer from low operational efficiency.

[0024] Among them, such as Figure 3 As shown, step S110 specifically includes steps S111-S114: S111. Multiple input feature vectors are generated within a preset parameter value range using a sampling method, wherein each input feature vector is generated based on the operating control parameters and power disturbance. S112. For each input feature vector, the microgrid transient simulation model is called to solve the problem to obtain the multidimensional stability response vector and the settling time. Based on the input feature vector, the multidimensional stability response vector and the settling time, a working condition sample library is constructed. S113. Based on the input convex neural network, an initial stability discrimination model, an initial stability index level regression model, and an initial stability time regression model are constructed using a convex activation function. S114. Based on the operating condition sample library, the initial stability discrimination model, the initial stability index level regression model, and the initial stable time regression model are independently trained using physical gradient consistency constraints to obtain the stability discrimination model, the stability index level regression model, and the stable time regression model. In the process of training the initial stability discrimination model, a conservative penalty term is constructed to iteratively optimize the initial stability discrimination model.

[0025] In this embodiment of the invention, multiple input feature vectors are generated within a preset parameter value range using the Latin hypercube sampling method. The microgrid's operating status data includes frequency deviation, voltage deviation, power disturbance, frequency change rate, voltage change rate, steady-state frequency, steady-state voltage, AC bus rated voltage, DC bus rated voltage, AC side inertia, DC side inertia, AC side damping, and DC side damping. The operating control parameters include DC side damping and DC side inertia. These operating control parameters are vectorized according to a predetermined dimension to obtain the input feature vector, as shown below:

[0026] in, This represents the input feature vector. Indicates DC-side damping. Indicates DC-side inertia. This indicates a power disturbance.

[0027] It should be noted that, for each input feature vector, a microgrid transient simulation model is called to solve the problem, in order to obtain the multidimensional stability response vector r and the settling time. The transient simulation model of the power grid is a set of differential algebraic equations describing the dynamic characteristics of the AC / DC hybrid microgrid. The model takes control parameters and disturbance intensity as inputs, obtains the time response trajectories of variables such as frequency and voltage through numerical integration, and outputs a multidimensional stability response vector and settling time to evaluate the transient stability of the microgrid system. Stability labels are generated based on the operating limits set for each stability index in the multidimensional stability response vector. The multidimensional stability response vector and settling time are normalized to generate stability index level labels and settling time labels.

[0028] The transient stability of a microgrid in islanded operation is determined by a multidimensional stability response vector. The following operating limits must be met for an assessment to determine stability: Wherein, for a given AC-side inertia and input feature vector The stability response exhibits an implicit nonlinear mapping. It should be noted that, Indicates frequency deviation. Indicates voltage deviation. Indicates the maximum value of voltage deviation Indicates the rate of change of frequency. This represents the minimum rate of change of frequency. This represents the maximum value of the rate of change of frequency. Indicates the rate of change of voltage. This represents the minimum rate of change of voltage. This represents the maximum value of the rate of change of voltage. Represents the steady-state frequency. This represents the minimum steady-state frequency. This represents the maximum value of the steady-state frequency. Represents steady-state voltage. This represents the minimum value of the steady-state voltage. This represents the maximum value of the steady-state voltage. Understandably, stability indicators include frequency deviation, voltage deviation, rate of change of frequency, rate of change of voltage, steady-state frequency, and steady-state voltage. An operating condition is considered stable only when all stability indicators are within safe limits, and the corresponding stability label is [value missing]. =1, otherwise it is judged as an unstable state, and the corresponding stability label is =0, The input feature vector is paired and associated with the corresponding stability label, stability index level label and stability time label, and stored and managed using a key-value pair structure to form a working condition sample library.

[0029] It should also be noted that the Input Convex Neural Network (ICNN) is a multi-layer feedforward network structure, which includes an input layer, an output layer, and multiple hidden layers. Skip connections are established between the input layer and each hidden layer to directly pass the input feature vector received by the input layer to each hidden layer. Non-negativity constraints are applied to the connection weights between hidden layers to ensure that the mapping from the input feature vector to the output of the multi-dimensional stability surrogate model satisfies the mathematical convexity requirement. Based on the Input Convex Neural Network, a convex activation function that simultaneously satisfies convexity and non-increasing properties is selected as the activation function for each hidden layer to construct the initial stability discrimination model, the initial stability index level regression model, and the initial stable time regression model. The convex activation function is the Softplus activation function.

[0030] It should be further explained that the physical gradient consistency constraint includes hard gradient direction constraint and soft gradient numerical constraint. The hard gradient direction constraint means that the partial derivatives of the outputs of the initial stability discrimination model, the initial stability index level regression model, and the initial stable time regression model with respect to each variable in the input feature vector must strictly follow the physical monotonicity law. The soft gradient numerical constraint means that the gradient magnitude of the outputs of the initial stability discrimination model, the initial stability index level regression model, and the initial stable time regression model with respect to each input variable is guided to approximate the global gradient reference value obtained by fitting the training sample set. To balance differentiability during training and convexity during optimization, a wrapper function is added after the output layer of the initial stability discriminant model, initial stability index level regression model, and initial stable time regression model during training. This wrapper function maps the outputs of the initial stability discriminant model, initial stability index level regression model, and initial stable time regression model to stable probabilities. After the initial stability discriminant model, initial stability index level regression model, and initial stable time regression model are trained to obtain the stable discriminant model, stability index level regression model, and stable time regression model, the wrapper function is removed from these models. The wrapper function is a mirrored hard sigmoid wrapper function.

[0031] It should also be noted that the operating condition sample library is divided into a training sample set, a test sample set, and a validation sample set. During the training of the initial stability discrimination model, active sampling is performed based on the false positive probability to update the training sample set and obtain the target training sample set. The false positive probability is calculated by the initial stability discrimination model to obtain the probability of each operating condition sample in the operating condition sample library. Understandably, a higher false positive probability means a greater risk that the initial stability discrimination model will misclassify an unstable operating condition sample as a stable state. The initial stability discrimination model is obtained by training the initial stability discrimination model using the target training sample set. The training sample set is used when training the initial stability index level regression model and the initial stability time regression model. Furthermore, during the training of the initial stability discrimination model, a conservative penalty term is constructed to iteratively optimize the model. This conservative penalty term involves selecting operating condition samples labeled as unstable from the target training sample set to obtain a first sample set; selecting operating condition samples predicted as stable by the initial stability discrimination model from the target training sample set to obtain a second sample set; determining the intersection of the first and second sample sets to obtain a false positive sample set; and determining the conservative penalty term based on the probability that each operating condition sample in the false positive sample set is predicted as stable. The calculation formula is: , in, Describes the first sample set Number of samples under medium operating conditions; Used for conservative penalties Normalization is performed; i is the sample index in the false positive sample set; The probability value for the initial stability discrimination model to predict the i-th working condition sample as a stable state is denoted by 0, and its value ranges from 0 to 1.

[0032] It should be further noted that the initial stability discrimination model uses a total loss consisting of binary cross-entropy loss and a conservative penalty term during training, while the initial stability index level regression model and the initial stable time regression model use mean squared error loss during training.

[0033] S120. A two-stage optimization scheduling mathematical model considering dynamic performance costs is established using the multidimensional stability proxy model.

[0034] In this embodiment of the invention, the two-stage optimized scheduling mathematical model includes a first-stage unit combination decision model and a second-stage operation decision model; such as Figure 4 As shown, step S120 specifically includes steps S121-S122: S121. Using the set of binary variables representing the start-up and shutdown states of the generator set as decision variables, and the sum of the start-up cost and shutdown cost of the generator set as the optimization objective, establish the first-stage unit combination decision model. S122. Based on the generator set start-up and shutdown status determined in the first stage, and with the weighted sum of operating cost and dynamic performance cost as the optimization objective, the output of the stability discrimination model is constrained to a positive value to establish the second stage operation decision model.

[0035] In this embodiment of the invention, the dynamic performance cost includes a stability index level cost and a stabilization time cost. The stability index level cost is obtained by normalizing the output of the stability index level regression model, and the stabilization time cost is obtained by normalizing the output of the stabilization time regression model.

[0036] It should be noted that the unit combination decision model for the first stage is as follows: in, The set of binary variables representing the start-up and shutdown states of the generator set is used as the decision variable. and These are indicators for starting and stopping the generator set, respectively. and These are the start-up cost and shutdown cost of the generator set, respectively.

[0037] Given the set of binary variables for the start-stop state of the generator set in the first stage. Under the premise that the second phase of operation decision-making model is as follows:

[0038] in, This represents the cost of the stability index level. This represents the cost of stabilizing time. The weighting coefficients represent the cost of the stability index level. The weighting coefficients represent the time cost of stabilization. express Constantly monitor the load shearing capacity of the busbar. express The load sheared by the DC bus at any given time; The weighting factor represents the load shedding amount. express Time generator set Actual output power The weighting coefficient represents the actual output power. This represents the second-stage decision variables in scenario s; Indicates a given and The second phase of the feasible region, The first-stage decision variable vector represents the set of binary variables representing the generator set's start-up and shutdown states. Representing a scene The random parameter vector below, This indicates operating costs.

[0039] S130. Solve the two-stage optimization scheduling model based on the nested decomposition architecture to obtain the global optimal scheduling scheme of the microgrid. The outer layer of the nested decomposition architecture uses Benders decomposition to process the discrete start-stop state decision variables of the generator units in the microgrid, and the inner layer uses the external approximation algorithm to process the nonlinear stability constraints and the dynamic performance cost using the analytical gradient information.

[0040] In embodiments of the present invention, such as Figure 5 As shown, step S130 specifically includes steps S131-S133: S131. Use the Benders master problem decomposed by Benders to process the discrete start-stop state decision variables of the generator set, and solve the mixed integer linear programming problem based on the Benders cutting plane constraints obtained in the current iteration step to obtain the optimal generator set start-stop scheme. S132. Using the Benders subproblem of Benders decomposition to process continuous operating variables, after receiving the optimal generator set start-up and shutdown scheme, based on the stability feasibility prediction result, select to execute a unified solution path or a collaborative decomposition path, and use the external approximation algorithm to linearize the nonlinear stability constraints and the dynamic performance cost to obtain the optimization results of continuous operating variables. S133. Generate the Benders cutting plane based on the optimization results of the continuous running variables and feed it back to the Benders main problem. Repeat the solution of the Benders main problem and the solution of the Benders subproblems until the convergence condition is met, and obtain the global optimal scheduling scheme of the microgrid.

[0041] In this embodiment of the invention, Benders decomposition is a classic decomposition algorithm for solving mixed-integer programming problems. The problem is decomposed into Benders decomposition variables that handle the start-up and shutdown states of generator sets. The Benders Master Problem (MP) and its specific implementation scenarios. The Benders subproblem (SP) involves power allocation and control parameter optimization. The Benders main problem is based on the current iteration step. Given the Benders cutting plane, solve the following mixed-integer linear programming problem: in, This represents the first-stage decision variable vector. This is the cost coefficient vector for the first stage. For scene collection, For the second-phase cost estimation variable in scenario s, the lower bound estimate of the objective value of the Benders subproblem is given. This represents the probability of scenario s occurring. Let be the coefficient matrix of the generator set start-stop constraints, and b be the constraint limit value. This represents the lower bound of the cost for scenario s.

[0042] It should be noted that the step of selecting the unified solution path or the collaborative decomposition path based on the stability feasibility prediction result includes: calculating the stability discrimination value of the microgrid under the limiting control parameters; if the stability discrimination value is greater than a preset discrimination value, then the unified solution path is selected; if the stability discrimination value is not greater than the preset discrimination value, then the collaborative decomposition path is selected. Specifically, the optimal generator start-up and shutdown scheme is obtained by solving the problem. And pass it to the Benders subproblem, upon receiving the Benders main problem's pass. back Stability discriminant value : 1) If This indicates that the microgrid system is in a critical stability state under nominal disturbances and enters the unified solution path (Full Path). 2) If This indicates that stability requirements are not the performance bottleneck for load shedding in microgrid systems, and the system has entered the Coordinated Decomposition path.

[0043] in, This represents the upper limit of the DC-side damping value. This represents the upper limit of the virtual inertia value. The disturbance intensity under nominal operating conditions (zero load shedding) is preset to a value of 0.

[0044] It should also be noted that the step of using the external approximation algorithm to linearize the nonlinear stability constraints and the dynamic performance cost to obtain the continuous running variable optimization result includes: obtaining candidate solutions; forming linear cutting plane constraints based on the analytical gradient information output by the multidimensional stability surrogate model at the running point of the candidate solutions; and obtaining the continuous running variable optimization result by alternately solving the external approximation master problem and the external approximation subproblem until convergence. It should also be further noted that in the unified solution path, scheduling instructions and control parameters participate as variables in the external approximation algorithm iteration, and the continuous running variable optimization result is obtained through global collaborative optimization; in the collaborative decomposition path, the optimal perturbation intensity is determined by independently solving the linear optimization model that does not include stability constraints and the dynamic performance cost, the optimal perturbation intensity is fixed, and the external approximation algorithm is executed on the control parameters to obtain the continuous running variable optimization result.

[0045] For ease of understanding, the unified solution path and collaborative decomposition path are described in detail below: Unified solution path: 1) Solve the main problem of the external approximation algorithm to obtain candidate scheduling schemes. ; 2) Fixed and the determined first Perturbation strength in the next iteration Solve the external approximation algorithm subproblem. If the external approximation algorithm subproblem has no solution, then solve for minimizing the constraint violation. Feasibility recovery subproblem: in, Indicates the DC-side damping parameters. Indicates the DC-side inertia parameter; 3) In the current solution A cutting plane constraint is formed at the location: in, The outer approximation algorithm is represented by the first... The linearized expansion points obtained in the next iteration This represents the auxiliary variable characterizing the upper bound of dynamic performance and cost in the main problem of the external approximation algorithm. It should be noted that, in the current solution... In the formation of cutting plane constraints Represent decision variables; 4) By alternately solving the main problem and subproblems of the external approximation algorithm, the linear approximation of the stability domain boundary is continuously refined until the solution is completed.

[0046] In each iteration of the external approximation algorithm, the unified solution path involves scheduling instructions and control parameters as variables participating in the optimization, achieving global collaborative optimization.

[0047] Collaborative decomposition path: 1) Solve the linear model independently without a multidimensional stability surrogate model, with the optimal perturbation strength fixed. ; 2) In a fixed position The following describes an independent iterative process for the external approximation algorithm, performed for the control parameters:

[0048] This stage optimizes the nonlinear performance cost through rapid iteration without altering the already determined linear output scheme. That is, under the cooperative decomposition path, the generator output allocation and disturbance intensity determined by the first step of independently solving the linear optimization model remain unchanged in subsequent external approximation algorithm iterations, and only the control parameters are optimized.

[0049] After the Benders subproblem completes nonlinear optimization based on the external approximation algorithm, the solution results are fed back to the Benders main problem. Since the microgrid system introduces load shedding decisions, it ensures that the Benders subproblem has a feasible solution for any first-stage decision, i.e., it satisfies the Complete Recourse property; therefore, only the optimal cut needs to be considered. Because the subproblem has integer variables, a Lagrange cut is obtained through Lagrange relaxation and added to the Benders main problem as the optimal cut. The above iterative process of the main and subproblems is repeated until the difference between the first-stage objective function value and the lower bound of the second-stage feedback satisfies the convergence threshold, at which point the globally optimal scheduling decision is output.

[0050] It should be noted that in this embodiment, the horizontal axis represents the scenario number and the vertical axis represents the stabilization time. Experimental results show that, in all scenarios, the stabilization time of the scheduling scheme of the two-stage optimized scheduling mathematical model that takes into account the dynamic performance cost is significantly lower than that of the conventional scheduling scheme that does not take into account the dynamic performance cost. This indicates that by establishing a two-stage optimized scheduling mathematical model that takes into account the dynamic performance cost and coordinating the optimization of multi-dimensional stability indicators, the stabilization time of the microgrid system can be effectively shortened and the dynamic response performance of the microgrid after being disturbed can be improved.

[0051] In summary, the microgrid optimization operation method considering the balance of multidimensional stability indices in this invention constructs a multidimensional stability surrogate model based on an input convex neural network, outputting analytical gradient information to provide a precise search direction for optimization, overcoming the shortcomings of existing metaheuristic algorithms that lack analytical gradient guidance, have low optimization efficiency, and are difficult to guarantee global optimality. By establishing a two-stage optimization scheduling mathematical model considering dynamic performance costs, the stability index level and settling time are transformed into a continuous dynamic performance cost function, quantitatively evaluating the dynamic response quality of the system within the stable region, avoiding the problem of overly conservative scheduling schemes caused by binary hard constraints in existing technologies, and achieving synergistic optimization of multidimensional stability indices. By adopting a nested decomposition architecture, the outer layer uses Benders decomposition to process the discrete start-stop state decision variables of generator units, while the inner layer uses an external approximation algorithm to process nonlinear stability constraints and dynamic performance costs using analytical gradient information, avoiding the dimensionality curse caused by the introduction of a large number of binary variables in the Big-M linearization method, and significantly improving the computational efficiency. By adopting a dual-path solution mechanism based on stability feasibility prediction, the system adaptively selects to execute a unified solution path or a collaborative decomposition path. Under critical stability conditions, the unified solution path is used to achieve global collaborative optimization of scheduling instructions and control parameters. Under non-bottleneck stability conditions, the collaborative decomposition path is used to fix the determined linear output scheme and only perform external approximation algorithm iteration on the control parameters. This significantly reduces computational redundancy and improves solution efficiency without sacrificing global optimality.

[0052] Figure 6 This is a schematic block diagram of a microgrid optimized operation device 200 considering multi-dimensional stability index balancing provided in an embodiment of the present invention. Figure 6 As shown, corresponding to the above-described microgrid optimization operation method considering multi-dimensional stability index balance, the present invention also provides a microgrid optimization operation device 200 considering multi-dimensional stability index balance. This microgrid optimization operation device 200 includes a unit for executing the above-described microgrid optimization operation method considering multi-dimensional stability index balance, and the device can be configured in a computer device. Specifically, please refer to... Figure 6The microgrid optimization operation device 200, which considers the balance of multi-dimensional stability indices, includes a construction unit 201, an establishment unit 202, and a solution operation unit 203. The functional modules are described in detail below: The construction unit 201 is used to construct a multidimensional stability surrogate model based on the input convex neural network. The multidimensional stability surrogate model outputs analytical gradient information and includes a stability discrimination model, a stability index level regression model, and a stable time regression model. Establishment unit 202 is used to establish a two-stage optimization scheduling mathematical model that takes into account dynamic performance costs using the multidimensional stability proxy model; The solution execution unit 203 is used to solve the two-stage optimization scheduling model based on the nested decomposition architecture to obtain the global optimal scheduling scheme of the microgrid. The outer layer of the nested decomposition architecture uses Benders decomposition to process the discrete start-stop state decision variables of the generators in the microgrid, and the inner layer uses the external approximation algorithm to process the nonlinear stability constraints and the dynamic performance cost using the analytical gradient information.

[0053] In some embodiments, such as this one, the construction unit 201 is specifically used for: Multiple input feature vectors are generated within a preset parameter value range using a sampling method, wherein each input feature vector is generated based on the operating control parameters and power disturbance. For each input feature vector, a microgrid transient simulation model is called to solve the problem, so as to obtain a multidimensional stability response vector and a settling time. Based on the input feature vector, the multidimensional stability response vector and the settling time, a working condition sample library is constructed. Based on the input convex neural network, an initial stability discrimination model, an initial stability index level regression model, and an initial stability time regression model are constructed using convex activation functions. Based on the aforementioned working condition sample library, the initial stability discrimination model, the initial stability index level regression model, and the initial stable time regression model are independently trained using physical gradient consistency constraints to obtain the stability discrimination model, the stability index level regression model, and the stable time regression model. In the process of training the initial stability discrimination model, a conservative penalty term is constructed to iteratively optimize the initial stability discrimination model.

[0054] In some embodiments, such as this one, the establishment unit 202 is specifically used for: Using the set of binary variables representing the start-up and shutdown states of the generator sets as decision variables, and the sum of the start-up cost and shutdown cost of the generator sets as the optimization objective, a first-stage unit combination decision model is established. Based on the generator set start-up and shutdown status determined in the first stage, and with the weighted sum of operating cost and dynamic performance cost as the optimization objective, the output of the stability discrimination model is constrained to a positive value to establish the second stage operation decision model. The dynamic performance cost includes a stability index level cost and a stabilization time cost. The stability index level cost is obtained by normalizing the output of the stability index level regression model, and the stabilization time cost is obtained by normalizing the output of the stabilization time regression model.

[0055] In some embodiments, such as this one, the solution execution unit 203 is specifically used for: The discrete start-stop state decision variables of the generator set are processed using the Benders master problem of Benders decomposition. Based on the Benders cutting plane constraints obtained in the current iteration step, the mixed integer linear programming problem is solved to obtain the optimal generator set start-stop scheme. The Benders subproblem of Benders decomposition is used to process continuous operating variables. After receiving the optimal generator set start-up and shutdown scheme, a unified solution path or a collaborative decomposition path is selected based on the stability feasibility prediction result. The external approximation algorithm is used to linearize the nonlinear stability constraints and the dynamic performance cost to obtain the optimization results of continuous operating variables. The Benders cut plane is generated based on the optimization results of the continuous running variables and fed back to the Benders master problem. The solution of the Benders master problem and the solution of the Benders subproblems are repeated until the convergence condition is met, and the global optimal scheduling scheme of the microgrid is obtained.

[0056] In some embodiments, such as this one, the solution execution unit 203 is further configured to: Calculate the stability discrimination value of the microgrid under the limiting control parameters; If the stability discrimination value is greater than the preset discrimination value, then the unified solution path is selected to be executed; If the stability discrimination value is not greater than the preset discrimination value, then the collaborative decomposition path is selected to be executed.

[0057] In some embodiments, such as this one, the solution execution unit 203 is further configured to: Obtain candidate solutions; At the running point of the candidate solution, a linear cutting plane constraint is formed based on the analytical gradient information output by the multidimensional stability surrogate model; The optimization results of the continuous running variables are obtained by alternately solving the external approximation master problem and the external approximation subproblems until convergence.

[0058] In some embodiments, such as this one, the solution execution unit 203 is further configured to: In the unified solution path, scheduling instructions and control parameters are both used as variables in the iteration of the external approximation algorithm. The optimization results of the continuous running variables are obtained through global collaborative optimization. In the collaborative decomposition path, the optimal disturbance intensity is determined by independently solving the linear optimization model that does not include stability constraints and the dynamic performance cost. The optimal disturbance intensity is fixed, and the external approximation algorithm is executed on the control parameters to obtain the optimization results of the continuous running variables.

[0059] The aforementioned microgrid optimization operation device, which considers the balance of multidimensional stability indices, can be implemented as a computer program. This computer program can be used in various ways, such as... Figure 7 It runs on the computer device shown.

[0060] Please see Figure 7 , Figure 7 This is a schematic block diagram of a computer device provided in an embodiment of the present invention. The computer device 300 is a device capable of optimizing the operation of a microgrid, taking into account the balancing of multi-dimensional stability indicators.

[0061] See Figure 7 The computer device 300 includes a processor 302, a memory, and a network interface 305 connected via a system bus 301. The memory may include a non-volatile storage medium 303 and internal memory 304.

[0062] The non-volatile storage medium 303 can store an operating system 3031 and a computer program 3032. When the computer program 3032 is executed, it enables the processor 302 to execute a microgrid optimization operation method that takes into account the balance of multi-dimensional stability indicators.

[0063] The processor 302 provides computing and control capabilities to support the operation of the entire computer device 300.

[0064] The internal memory 304 provides an environment for the operation of the computer program 3032 in the non-volatile storage medium 303. When the computer program 3032 is executed by the processor 302, the processor 302 can execute a microgrid optimization operation method that takes into account the balance of multi-dimensional stability indicators.

[0065] This network interface 305 is used for network communication with other devices. Those skilled in the art will understand that... Figure 7The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device 300 to which the present invention is applied. The specific computer device 300 may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0066] The processor 302 is used to run a computer program 3032 stored in a memory to implement any embodiment of the microgrid optimization operation method that takes into account the balance of multidimensional stability indicators.

[0067] It should be understood that, in this embodiment of the invention, the processor 302 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0068] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by a processor in the computer system to implement the process steps of the embodiments of the above methods.

[0069] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program. When executed by a processor, the computer program causes the processor to perform any embodiment of the microgrid optimization operation method considering the equilibrium of multidimensional stability indices described above.

[0070] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0071] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0072] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0073] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0074] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0075] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0076] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Since these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

[0077] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A microgrid optimization operation method considering the equilibrium of multidimensional stability indices, characterized in that, include: A multidimensional stability surrogate model is constructed based on an input convex neural network. The multidimensional stability surrogate model outputs analytical gradient information and includes a stability discrimination model, a stability index level regression model, and a stability time regression model. A two-stage optimization scheduling mathematical model considering dynamic performance costs is established using the aforementioned multidimensional stability proxy model; The two-stage optimization scheduling model is solved based on a nested decomposition architecture to obtain the global optimal scheduling scheme of the microgrid. The outer layer of the nested decomposition architecture uses Benders decomposition to process the discrete start-stop state decision variables of the generators in the microgrid, while the inner layer uses an external approximation algorithm to process the nonlinear stability constraints and the dynamic performance cost using the analytical gradient information.

2. The method according to claim 1, characterized in that, The multidimensional stability surrogate model includes a stability discrimination model, a stability index level regression model, and a stable time regression model; the step of constructing the multidimensional stability surrogate model based on the input convex neural network includes: Multiple input feature vectors are generated within a preset parameter value range using a sampling method, wherein each input feature vector is generated based on the operating control parameters and power disturbance. For each input feature vector, a microgrid transient simulation model is called to solve the problem, so as to obtain a multidimensional stability response vector and a settling time. Based on the input feature vector, the multidimensional stability response vector and the settling time, a working condition sample library is constructed. Based on the input convex neural network, an initial stability discrimination model, an initial stability index level regression model, and an initial stability time regression model are constructed using convex activation functions. Based on the aforementioned working condition sample library, the initial stability discrimination model, the initial stability index level regression model, and the initial stable time regression model are independently trained using physical gradient consistency constraints to obtain the stability discrimination model, the stability index level regression model, and the stable time regression model. In the process of training the initial stability discrimination model, a conservative penalty term is constructed to iteratively optimize the initial stability discrimination model.

3. The method according to claim 2, characterized in that, The two-stage optimized scheduling mathematical model includes a first-stage unit combination decision model and a second-stage operation decision model; the step of establishing a two-stage optimized scheduling mathematical model considering dynamic performance costs using the multidimensional stability proxy model includes: Using the set of binary variables representing the start-up and shutdown states of the generator sets as decision variables, and the sum of the start-up cost and shutdown cost of the generator sets as the optimization objective, a first-stage unit combination decision model is established. Based on the generator set start-up and shutdown status determined in the first stage, and with the weighted sum of operating cost and dynamic performance cost as the optimization objective, the output of the stability discrimination model is constrained to a positive value to establish the second stage operation decision model. The dynamic performance cost includes a stability index level cost and a stabilization time cost. The stability index level cost is obtained by normalizing the output of the stability index level regression model, and the stabilization time cost is obtained by normalizing the output of the stabilization time regression model.

4. The method according to claim 1, characterized in that, The steps for solving the two-stage optimization scheduling model based on a nested decomposition architecture to obtain the globally optimal scheduling scheme for the microgrid include: The discrete start-stop state decision variables of the generator set are processed using the Benders master problem of Benders decomposition. Based on the Benders cutting plane constraints obtained in the current iteration step, the mixed integer linear programming problem is solved to obtain the optimal generator set start-stop scheme. The Benders subproblem of Benders decomposition is used to process continuous operating variables. After receiving the optimal generator set start-up and shutdown scheme, a unified solution path or a collaborative decomposition path is selected based on the stability feasibility prediction result. The external approximation algorithm is used to linearize the nonlinear stability constraints and the dynamic performance cost to obtain the optimization results of continuous operating variables. The Benders cut plane is generated based on the optimization results of the continuous running variables and fed back to the Benders master problem. The solution of the Benders master problem and the solution of the Benders subproblems are repeated until the convergence condition is met, and the global optimal scheduling scheme of the microgrid is obtained.

5. The method according to claim 4, characterized in that, The step of selecting either a unified solution path or a collaborative decomposition path based on the stability and feasibility prediction results includes: Calculate the stability discrimination value of the microgrid under the limiting control parameters; If the stability discrimination value is greater than the preset discrimination value, then the unified solution path is selected to be executed; If the stability discrimination value is not greater than the preset discrimination value, then the collaborative decomposition path is selected to be executed.

6. The method according to claim 4, characterized in that, The step of using the external approximation algorithm to linearize the nonlinear stability constraints and the dynamic performance cost to obtain the optimization results of the continuous running variables includes: Obtain candidate solutions; At the running point of the candidate solution, a linear cutting plane constraint is formed based on the analytical gradient information output by the multidimensional stability surrogate model; The optimization results of the continuous running variables are obtained by alternately solving the external approximation master problem and the external approximation subproblems until convergence.

7. The method according to any one of claims 4-6, characterized in that, In the unified solution path, scheduling instructions and control parameters are both used as variables in the iteration of the external approximation algorithm. The optimization results of the continuous running variables are obtained through global collaborative optimization. In the collaborative decomposition path, the optimal disturbance intensity is determined by independently solving the linear optimization model that does not include stability constraints and the dynamic performance cost. The optimal disturbance intensity is fixed, and the external approximation algorithm is executed on the control parameters to obtain the optimization results of the continuous running variables.

8. A microgrid optimized operation device considering the balance of multidimensional stability indices, characterized in that, include: A construction unit is used to construct a multidimensional stability surrogate model based on an input convex neural network. The multidimensional stability surrogate model outputs analytical gradient information and includes a stability discrimination model, a stability index level regression model, and a stable time regression model. A unit is established to utilize the multidimensional stability proxy model to establish a two-stage optimization scheduling mathematical model that takes into account dynamic performance costs. The solution operation unit is used to solve the two-stage optimization scheduling model based on the nested decomposition architecture to obtain the global optimal scheduling scheme of the microgrid. The outer layer of the nested decomposition architecture uses Benders decomposition to process the discrete start-stop state decision variables of the generators in the microgrid, and the inner layer uses the external approximation algorithm to process the nonlinear stability constraints and the dynamic performance cost using the analytical gradient information.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, can implement the method as described in any one of claims 1-7.

Images