Multi-objective optimization method and system for power grid carrying capacity considering source-load interaction

Abstract

The application discloses a power grid carrying capacity multi-objective optimization method and system considering source-load interaction, and relates to the technical field of power system and distribution network optimization control. The method comprises the following steps: acquiring power grid structure parameters, node net injection prediction baseline data, source-load controllable resource information and historical prediction error samples, constructing resource flexibility envelope and node aggregated control quantity feasible region, determining node voltage calibration parameters and branch current calibration parameters, establishing voltage opportunity constraints, branch current opportunity constraints and budget summary constraints, then solving a carrying capacity multi-objective optimization model to obtain node carrying capacity parameters and node aggregated control quantities, and decomposing to form resource-level control quantity sequences to execute adjustment. The application can perform more fine optimization analysis on the power grid carrying capacity under the condition of considering source-load interaction and operation uncertainty, and can also consider safety constraint satisfaction and resource execution feasibility.

Description

Technical Field

[0001] This invention relates to the field of power system and distribution network optimization control technology, and in particular to a multi-objective optimization method and system for power grid carrying capacity considering source-load interaction. Background Technology

[0002] With the integration of distributed power sources, energy storage devices, adjustable loads, and electric vehicle charging facilities into the distribution network, the operation of the distribution network is gradually shifting from the traditional one-way power supply to a mode of bidirectional interaction between power sources and loads, and dynamic changes in operating status. In this context, the grid's carrying capacity no longer depends solely on the rated capacity of the equipment, but is also closely related to changes in node injection, line operating status, load fluctuations, and the response capabilities of adjustable resources.

[0003] In existing technologies, grid carrying capacity assessment methods mostly employ static analysis or deterministic boundary checks, typically estimating the available capacity based on constraints such as node voltage, branch current, and transformer capacity. While these methods are relatively simple to implement, they do not adequately consider the coordinated adjustment capabilities of multiple source and load resources. Especially when there are constraints such as energy storage capacity, load timing constraints, or charging demand constraints, they are difficult to accurately reflect the actual adjustable space.

[0004] In addition, while some existing methods take into account prediction errors or operational uncertainties, they usually use uniform margins or fixed safety factors, which makes it difficult to reflect the differences in risk levels at different nodes, branches, and times. This can easily lead to a conservative carrying capacity assessment or insufficiently precise control of safety boundaries.

[0005] In addition, existing technologies also suffer from insufficient integration between optimization results and equipment execution processes. Even if adjustment results at the node level are obtained, there is a lack of effective mechanisms to further implement them into actual control sequences of various source load resources, which affects the engineering application effect of the solution.

[0006] Therefore, it is necessary to provide a grid carrying capacity optimization method that can take into account the characteristics of source-load interaction, the impact of uncertainty, grid security constraints, and the feasibility of resource execution. Summary of the Invention

[0007] To address the challenges of accurately assessing the carrying capacity of distribution networks, unifying risk constraints, and implementing optimization results at the resource execution layer under conditions of source-load interaction and uncertainty, this invention proposes a multi-objective optimization method and system for power grid carrying capacity that considers source-load interaction.

[0008] The present invention achieves the above objectives through the following technical solutions:

[0009] A multi-objective optimization method for grid carrying capacity considering source-load interaction is described, the method comprising:

[0010] Acquire power grid operation data, including power grid structure parameters, node net injection prediction baseline data, source and load controllable resource information, and historical prediction error samples;

[0011] Based on the source load controllable resource information, construct the resource flexibility envelope of each source load controllable resource in the time domain for optimization, establish cross-time coupling constraints for source load controllable resources with temporal coupling, and aggregate the resource flexibility envelope into the node aggregation flexibility envelope according to the mapping relationship between each source load controllable resource and the node, thereby limiting the feasible domain of the node aggregation control quantity.

[0012] Based on historical prediction error samples and power grid structure parameters, node voltage calibration parameters and branch current calibration parameters are determined, and probability budget parameters are set. Based on the node voltage calibration parameters, branch current calibration parameters and probability budget parameters, voltage opportunity constraints, branch current opportunity constraints and budget aggregation constraints are established.

[0013] Using node carrying capacity parameters, node aggregation control quantities, and probabilistic budget parameters as decision variables, the node net injection quantity is determined based on the node net injection prediction baseline data and the node aggregation control quantity. Based on the node carrying capacity parameters and the node net injection quantity formed by the smoothing and limiting injection function, a multi-objective optimization model for carrying capacity is constructed and solved under the feasible region of node aggregation control quantity, voltage opportunity constraints, branch current opportunity constraints, and budget aggregation constraints. The optimized values ​​of node carrying capacity parameters and node aggregation control quantities for each node in each discrete time period are determined. The optimized values ​​of node carrying capacity parameters are used to characterize the upper limit of the net injection that the corresponding node can carry in the corresponding discrete time period.

[0014] Based on the optimized value of the node aggregation control quantity, a sequence of resource-level control quantities that satisfies the node aggregation consistency constraint, resource flexibility envelope, and cross-time coupling constraint is solved in the optimization time domain, and then sent to the corresponding source-load controllable resources for adjustment.

[0015] As a preferred embodiment of the present invention, the source-load controllable resources include source-side controllable resources and load-side controllable resources; in the source-load controllable resource information, the constraint parameters of the source-side controllable resources include at least the upper and lower limits of distributed power output, ramping constraints, energy storage charging and discharging power constraints, and energy storage state of charge constraints; the constraint parameters of the load-side controllable resources include at least the upper and lower limits of adjustable load adjustment, cumulative adjustment constraints, charging facility power constraints, and energy demand constraints.

[0016] The cross-time coupling constraints include at least the energy storage state of charge balance constraint, the adjustable load cumulative adjustment constraint, and the charging facility time-period energy balance constraint.

[0017] As a preferred embodiment of the present invention, based on the mapping relationship between controllable resources and nodes of each source load, the resource flexibility envelope is aggregated into a node aggregation flexibility envelope, thereby defining the feasible domain of node aggregation control quantity, including:

[0018] set up For a node aggregation control space, there is a finite set of directions. For nodes During discrete time periods The node aggregation control quantity. For mapping to nodes The node feasible set is jointly induced by the resource flexibility envelope of each source load controllable resource and the cross-time coupling constraint; for any node Discrete time period and direction vector Calculate the support value, which satisfies:

[0019] ;

[0020] Based on the aforementioned support values, an outer approximation polyhedron of the feasible region of the node aggregation control quantity is constructed, satisfying:

[0021] ;

[0022] In the formula, For the finite set of directions The first in A directional vector, Represents a finite set of directions The total number of directions in; Represents a node During discrete time periods Along the first Support values ​​of each directional vector; Indicates transpose; Represents a node During discrete time periods The node aggregation control quantity is the outer approximation polyhedron of the feasible domain;

[0023] The external approximation polyhedron As a node During discrete time periods The feasible domain for node aggregation control.

[0024] As a preferred embodiment of the present invention, the step of determining the node voltage calibration parameters and branch current calibration parameters based on historical prediction error samples and power grid structure parameters includes:

[0025] For active power prediction error and reactive power prediction error, respectively, the active power calibration amplitude and reactive power calibration amplitude are calculated, and robust calibration is performed on the active power calibration amplitude and reactive power calibration amplitude based on coverage statistics and dispersion statistics; wherein, for any power type During discrete time periods Power calibration amplitude satisfy:

[0026] ;

[0027] In the formula, Indicates active power. Indicates reactive power; Indicates power type exist The prediction error sample sequence; express Partial operator Indicates quantile level; This represents the median absolute deviation operator; Indicates the coverage factor; Indicates the robustness coefficient;

[0028] Each node in discrete time intervals The active power calibration amplitude constitutes the active power calibration amplitude vector. Each node in discrete time intervals The reactive power calibration amplitude consists of the reactive power calibration amplitude vector. ;

[0029] A linearized power flow model of the distribution network is used to establish sensitivity matrices for active power disturbances, reactive power disturbances, and their effects on node voltages and branch currents. Based on these sensitivity matrices, the power calibration amplitude is determined. The parameters are propagated as node voltage calibration parameters and branch current calibration parameters. During the iterative solution process, the linearized running points are updated with weights according to the preset damping factor, and trust region constraints are applied to the node aggregation control quantities between adjacent iterations.

[0030] As a preferred embodiment of the present invention, the probability budget parameters include node voltage probability budget parameters. Branch current probability budget parameters ;

[0031] remember For nodes During discrete time periods Node voltage calibration parameters, branch road During discrete time periods The branch current calibration parameters then the voltage tightening amount and current tightening amount They respectively satisfy:

[0032] ;

[0033] ;

[0034] In the formula, Indicates the coverage factor; The operator representing the inverse cumulative distribution function of the standard normal distribution, and ;

[0035] Based on the voltage tightening amount and current tightening amount Establish voltage mechanism constraints, expressed as:

[0036] ;

[0037] Branch current opportunity constraints are expressed as follows: ;

[0038] Budget aggregation constraints are expressed as follows: ;

[0039] In the formula, Represents a node During discrete time periods Linearized prediction of node voltage, , They are nodes The upper and lower limits of the allowable voltage; Indicates a branch During discrete time periods Linearized prediction of branch current, Indicates a branch The maximum allowable current; Representing discrete time periods The total upper limit of the voltage budget; Representing discrete time periods The total upper limit of the current budget.

[0040] As a preferred embodiment of the present invention, the net injection amount of the node is expressed as: ;

[0041] The limited net injection amount is expressed as: ;

[0042] In the formula, Represents a node During discrete time periods The net injection amount of the node, Represents a node During discrete time periods The baseline value of the predicted net active power injection, Represents a node During discrete time periods The active component in the node aggregation control quantity; Represents a node During discrete time periods The limited net injection volume, Represents a node During discrete time periods The nodal bearing capacity parameters, Indicates the smoothing coefficient;

[0043] The node limiting ratio is determined based on the difference between the net injection amount at the node and the net injection amount for limiting the amplitude, expressed as follows:

[0044] ;

[0045] In the formula, Represents a node During discrete time periods The node issuance limit ratio, This represents the preset positive lower limit of the node bearing capacity parameter;

[0046] The node emission restriction ratio is used to characterize the relative emission restriction degree of a node due to carrying capacity limitations in the corresponding discrete time period, and is written into the objective function or constraint condition of the carrying capacity multi-objective optimization model.

[0047] As a preferred embodiment of the present invention, the objective function of the multi-objective optimization model for bearing capacity includes a node bearing capacity enhancement term, a node emission limitation and suppression term, and a budget distribution matching term, expressed as: ;

[0048] In the formula, Represent the objective function; , , These represent the objective function weight coefficients for the node carrying capacity enhancement term, the node emission restriction and suppression term, and the budget distribution matching term, respectively, and all are greater than 0; Represents a node The node importance coefficient; Representing discrete time periods The budget distribution matching item, and satisfies ;

[0049] in, ; ;

[0050] Represents a node During discrete time periods The actual voltage budget distribution ratio Represents a node During discrete time periods The proportion of the reference voltage budget distribution; Indicates a branch During discrete time periods The actual current budget distribution ratio Indicates a branch During discrete time periods The proportion of the reference current budget distribution; The total number of nodes. , Each of the following is a list of any node in the node set. Node voltage probability budget parameters, node voltage calibration parameters; The total number of branch roads, , Each is any one of the branches in the set of branches. The branch current probability budget parameters and branch current calibration parameters; Let be the regularization constant, and .

[0051] As a preferred embodiment of the present invention, the step of solving the resource-level control quantity sequence that satisfies the node aggregation consistency constraint, resource flexibility envelope, and cross-time coupling constraint in the optimization time domain includes:

[0052] For source load controllable resources mapped to the same node, a resource-level control quantity sequence decomposition model is established. Under the premise of satisfying node aggregation consistency constraints, resource flexibility envelopes corresponding to each source load controllable resource, and cross-time period coupling constraints, the resource-level control quantity sequence decomposition model reduces the deviation of the resource-level control quantity sequence of each source load controllable resource in the optimization time domain from the corresponding reference control quantity, and suppresses the jump of resource-level control quantity in adjacent discrete time periods.

[0053] The obtained resource-level control quantity sequence is converted into the active power setpoint, reactive power setpoint, charging and discharging power setpoint, load adjustment quantity or charging power setpoint of the corresponding source-load controllable resource, and sent to the corresponding source-load controllable resource for adjustment.

[0054] A multi-objective optimization system for power grid carrying capacity considering source-load interaction is provided, the system comprising:

[0055] The data acquisition module is used to acquire power grid operation data, including power grid structure parameters, node net injection prediction baseline data, source and load controllable resource information, and historical prediction error samples.

[0056] The resource modeling module is used to construct the resource flexibility envelope of each source load controllable resource in the time domain based on the source load controllable resource information, and to establish cross-time period coupling constraints for source load controllable resources with temporal coupling.

[0057] The node aggregation module is used to aggregate the resource flexibility envelope into a node aggregation flexibility envelope based on the mapping relationship between each source load controllable resource and the node, and thereby limit the feasible domain of the node aggregation control quantity.

[0058] The calibration and constraint construction module is used to determine the node voltage calibration parameters and branch current calibration parameters based on the historical prediction error samples and the power grid structure parameters, and to set the probability budget parameters to establish voltage opportunity constraints, branch current opportunity constraints and budget summary constraints.

[0059] The bearing capacity optimization module is used to take the node bearing capacity parameters, node aggregation control quantity and the probability budget parameters as decision variables, determine the node net injection quantity based on the node net injection prediction baseline data and the node aggregation control quantity, and construct and solve the bearing capacity multi-objective optimization model based on the node bearing capacity parameters and the node net injection quantity formed by the smoothing and limiting injection function, and determine the optimized values ​​of the node bearing capacity parameters and the node aggregation control quantity for each node in each discrete time period.

[0060] The resource decomposition and execution module is used to solve the resource-level control quantity sequence that satisfies the node aggregation consistency constraint, the resource flexibility envelope, and the cross-time coupling constraint in the optimization time domain based on the node aggregation control quantity optimization value, and send it to the corresponding source load controllable resource for execution adjustment.

[0061] The beneficial effects of this invention are as follows: Based on a unified consideration of controllable resource constraints on both the source and load sides, it forms a node-level adjustable capacity characterization, thereby more accurately reflecting the adjustable space of each node under different discrete time periods; it can combine historical error information to calibrate risks related to node voltage and branch current, and establish corresponding constraints on this basis, thereby improving the pertinence and effectiveness of carrying capacity optimization while ensuring the safe operation of the power grid; it can incorporate node carrying capacity parameters, node aggregated control quantities, and related risk allocation parameters into a unified optimization framework, comprehensively improving the net carrying capacity injection level of each node, while taking into account operational safety and control feasibility; it can further decompose the node-level optimization results into resource-level control sequences and send them for execution, thereby achieving a closed-loop connection from carrying capacity optimization to resource control implementation, and improving the engineering application value. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the modular structure of the system in an embodiment of the present invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0064] like Figure 1 The illustration shows an embodiment of the present invention, which provides a multi-objective optimization method for power grid carrying capacity considering source-load interaction. This method is applicable to distribution networks or active distribution networks containing distributed power sources, energy storage devices, adjustable loads, and charging facilities. The method mainly includes the following steps:

[0065] S1: Obtain power grid operation data, including power grid structure parameters, node net injection prediction baseline data, source and load controllable resource information, historical prediction error samples, etc.

[0066] The power grid structure parameters characterize the network topology and static parameters of each electrical component, and include at least: node set, branch set, node-branch connection relationship, branch resistance, branch reactance, transformer parameters, node rated voltage, branch allowable current upper limit, and node allowable upper and lower voltage limits. Preferably, the power grid structure parameters also include node type identifiers, such as balanced nodes, ordinary load nodes, and distributed generation access nodes. In this embodiment, node serial numbers are represented by letters. This indicates that the branch numbering uses letters. The discrete time period number is indicated by letters. Let the total number of nodes be... The total number of branch roads is Optimization of the time domain includes There are discrete time intervals, the length of which is denoted as . Preferably, 15 minutes, 30 minutes, or 60 minutes can be used; if used for intraday rolling assessment, 15 minutes or 30 minutes are preferred; if used for day-ahead coarse-grained assessment, 60 minutes can be used.

[0067] The node net injection prediction baseline data is used to characterize the basic net injection state of each node in the optimization time domain without additional adjustments. In this embodiment, the node... During discrete time periods The baseline net active power injection can be denoted as ,node During discrete time periods The baseline net reactive power injection can be denoted as The superscript "0" represents the baseline value, and the symbol "0" indicates the baseline value. "" indicates the predicted value. Net injection refers to the net value after subtracting the power absorbed by the node from the power injected into the grid by the node. If the node is a distributed generation access node, the net injection value is usually positive; if the node is a pure load node, the net injection value is usually negative; in scenarios involving source-load interaction, the net injection value may also switch between positive and negative over time.

[0068] Source-load controllable resource information is used to characterize the location, type, and controllability boundary of each participating regulating device. In this embodiment, the source-load controllable resource serial numbers are represented by letters. Indicates. Each resource Each node uniquely corresponds to one access node, which is denoted as . The source-load controllable resource information should include at least the resource number, resource type, access node number, rated power, output upper and lower limits, ramping capability, energy storage capacity, energy storage efficiency, initial state of charge, adjustable load adjustment upper and lower limits, target energy demand of charging facilities, and allowed adjustment period. To avoid confusion between constraints of different types of resources, it is preferable to classify resource types into two categories: source-side controllable resources and load-side controllable resources, and establish parameter tables for each category.

[0069] Historical prediction error samples are used to characterize the distribution of deviations between predicted and actual values. In this embodiment, the power type is represented by a letter. It means that among them , Indicates active power. This represents reactive power. Sample numbers are represented by letters. The total sample size is indicated by letters. Representation. Preferably, for each node Each discrete time period and each power type Save prediction error samples from multiple historical days, denoted as , Represents a node During discrete time periods The Power type The prediction error samples. To improve statistical stability, it is preferable to construct the sample database according to the same type of day, that is, separate databases for weekdays and holidays, and separate databases for summer and winter. Sample size. The sample size should preferably be no less than 20, and more preferably 30 to 90. When the sample size is insufficient, the backtracking time window can be expanded, but data with similar working conditions should be selected first.

[0070] S2: Based on the source load controllable resource information, construct the resource flexibility envelope of each source load controllable resource in the optimized time domain, establish cross-time coupling constraints for source load controllable resources with temporal coupling, and aggregate the resource flexibility envelope into the node aggregation flexibility envelope according to the mapping relationship between each source load controllable resource and the node, thereby limiting the feasible domain of the node aggregation control quantity.

[0071] The essence of resource flexibility envelope is: within a given optimization time domain, the set of all feasible control sequences of a resource, under the premise of satisfying its own physical characteristics, operational boundaries and business requirements.

[0072] To standardize the representation of different resources, this embodiment defines resources... During discrete time periods The resource-level control quantity is defined as When only active power regulation is considered, It is a scalar; when active and reactive power regulation are considered simultaneously, It is a two-dimensional vector. Preferably, in the source-load interaction carrying capacity assessment scenario, the resource-level control quantity is represented as... ;in, Representing resources During discrete time periods The active power regulation, Representing resources During discrete time periods The reactive power regulation quantity. If a certain type of resource does not have reactive power regulation capability, its reactive power component is fixed at zero.

[0073] Controllable resources on the source side include at least distributed power sources and energy storage devices.

[0074] For distributed power sources, it is preferable to establish upper and lower limits for output and ramping constraints. Let the distributed power source resources... During discrete time periods The active power output is The superscript DG indicates distributed generation; its upper and lower output limits are respectively and Then it satisfies ;

[0075] Assume distributed power resources The upper limit of the ascent and climb is The maximum descent ramp height is Then adjacent time periods satisfy The two types of constraints mentioned above together limit the adjustable range of distributed power sources in the optimization time domain.

[0076] For energy storage devices, it is preferable to establish charging power constraints, discharging power constraints, energy storage state of charge constraints, and state of charge balance constraints. Let energy storage resources... During discrete time periods The charging power is The discharge power is Its charging power limit is The upper limit of discharge power is Then it satisfies ;

[0077] Energy storage resources During discrete time periods The stored energy at the end is Its permissible lower limit for energy storage is The upper limit is Then it satisfies ;

[0078] Assume the charging efficiency is The discharge efficiency is Then the state-of-charge equilibrium constraint can be expressed as ;in, and The preferred value is 0.90 to 0.98, and more preferably 0.93 to 0.97 for lithium battery energy storage systems. The above energy storage balance equation reflects typical cross-time period coupling characteristics, that is, the adjustability of the current time period is affected by the state of charge of the previous time period, thus belonging to cross-time period coupling constraints.

[0079] Load-side controllable resources include at least adjustable loads and charging facilities. For adjustable loads, assume they exist in discrete time periods... The adjustment amount is The upper and lower limits are adjusted as follows: and Then it satisfies ;

[0080] If the overall adjustment intensity is limited by business requirements, then a cumulative adjustment constraint can be further established. Assume resources... The set of allowable adjustment periods is The cumulative adjustment limit is Then it satisfies ;in, The optimal method is to determine the value by multiplying the rated load power by the allowable adjustment duration, thereby balancing comfort, production process continuity, and business acceptability.

[0081] For charging facilities, let's assume they exist in discrete time periods. The charging power is Its charging power limit is The target energy requirement is The set of permitted charging periods is Then it satisfies as well as This energy balance constraint is also a cross-time period coupling constraint, because whether the target charging amount before departure is met depends on the cumulative charging results of multiple discrete time periods.

[0082] In terms of implementation, to facilitate subsequent unified aggregation, this embodiment will use each resource The control variables, constraints, and necessary equations for all discrete time periods throughout the entire optimization time domain are uniformly written as a convex constraint set, denoted as . , Representing resources The resource flexibility envelope is optimized within the time domain. In other words, It is not the upper and lower limits of a single time period, but a set of feasible control sequences encompassing the entire optimization time domain. After modeling the source-side controllable resources and the load-side controllable resources separately, the flexibility envelope of all resources can be used as the input for subsequent node aggregation.

[0083] In this embodiment, based on the mapping relationship between each source load controllable resource and node, the resource flexibility envelope is aggregated into a node aggregation flexibility envelope, and the feasible domain of node aggregation control quantity is defined accordingly. The purpose of this step is to uniformly convert multiple resources connected to the same node, of different types and with different constraints, and which may have cross-time coupling, into an aggregation adjustment space that can be provided to the power grid at the node level.

[0084] To clarify the mapping relationship between resources and nodes, this embodiment establishes a mapping table or mapping matrix. Let the mapping matrix be... If resources Access Node Then matrix elements ;otherwise For nodes During discrete time periods The node aggregation control quantity is denoted as When considering both active and reactive power regulation, it is preferable to define it as a two-dimensional vector: ,in This represents the aggregated active power regulation at the node. This represents the aggregated reactive power regulation at the node. If only active power regulation is considered, then... It can be simplified to a scalar.

[0085] Since resources on the same node often contain cross-time-period coupled devices such as energy storage, adjustable loads, and charging facilities, the feasible region of the node layer cannot be directly obtained by simply adding the upper and lower limits of resources in each time period. Therefore, this embodiment uses a support value + external approximation polyhedron approach to construct the feasible region of the node aggregation control quantity. Specifically, a finite set of directions is first selected in the control space, denoted as . ,in Indicates the first A directional vector, This represents the total number of directions. If the node aggregation control quantity is a two-dimensional vector, it is preferable to select at least the positive and negative directions of the coordinate axes and the diagonal directions, for example, (1,0), (-1,0), (0,1), (0,-1), (1,1), (1,-1), (-1,1), (-1,-1). When pursuing higher approximation accuracy, the above directions can be further subdivided and normalized, preferably making... Choose 8, 16, 24, or 32. Optimal choice for rapid evaluation scenarios. or Optimization in a detailed evaluation scenario or The higher the number of directions, the higher the accuracy of the external approximation, but the computational cost also increases accordingly.

[0086] set up Represents a node During discrete time periods The feasible set of nodes; it should be noted that... The formation process is not just about time periods. Instead of local variables, they are mapped to nodes. The control sequence of all resources throughout the entire optimization time domain is the optimization variable, satisfying the flexibility envelope of each resource. Under the premise of cross-time period coupling constraints, for time periods The induced feasible set is obtained by projecting the aggregated control quantity. Therefore, this set implicitly considers the impact of cross-time period coupling constraints on the achievable adjustment capability of the current time period.

[0087] For any node Any discrete time period and any direction vector Define support value For nodes During discrete time periods Along the first The maximum reachable projection value of each directional vector is calculated using the following formula: ;in This indicates transpose. In actual solution processing, the optimization variables are not limited to... It is not itself, but mapped to nodes. The entire resource control sequence over the entire optimization time domain; the objective function only applies to the first... The node aggregation control quantity in the time period is in the direction The projected value is maximized, while variables in other time periods are constrained through resource flexibility envelope and cross-time period coupling constraints. Using this method to calculate support values ​​allows the complex temporally coupled feasible set to be projected onto the node-level control space of the current discrete time period.

[0088] After obtaining the support values ​​in each direction, construct the nodes. During discrete time periods The outer approximation polyhedron of the feasible domain of the node aggregation control quantity Its expression is: ;

[0089] External approximation polyhedron That is, a node During discrete time periods The feasible region of the node aggregation control quantity is determined. The reason for choosing external approximation over exact projection is twofold: firstly, exact projection often has high computational complexity, especially when there are many nodes, many resource types, and significant cross-time coupling; secondly, the external approximation polyhedron can be directly written as a finite number of linear inequalities, making it easier to embed into the optimization model later. Those skilled in the art only need to solve the support value problem once for each node, each time period, and each direction to obtain all the inequality boundaries of the feasible region of the corresponding node aggregation control quantity.

[0090] In engineering implementation, to further improve computational efficiency, a parallel solution method is preferred for calculating support values ​​at different nodes and in different directions. If the node aggregate control quantity only considers the active component, the direction set degenerates into a one-dimensional direction set. In this case, the support value calculation and feasible region construction can be further simplified to a single-variable interval or a set of one-dimensional linear inequalities.

[0091] S3: Determine the node voltage calibration parameters and branch current calibration parameters based on historical prediction error samples and power grid structure parameters, and set the probability budget parameters. Based on the node voltage calibration parameters, branch current calibration parameters and probability budget parameters, establish voltage opportunity constraints, branch current opportunity constraints and budget summary constraints.

[0092] In this embodiment, the node voltage calibration parameters and branch current calibration parameters are determined based on historical prediction error samples and power grid structure parameters. This includes two consecutive sub-processes: the first is to calculate the power calibration amplitude based on historical prediction error samples; the second is to construct a linearized power flow sensitivity matrix of the distribution network based on the power grid structure parameters and propagate the power calibration amplitude into node voltage calibration parameters and branch current calibration parameters.

[0093] To ensure that those skilled in the art can directly implement the calculation of the power calibration amplitude, this embodiment specifically implements it as "calculating separately for each node component and then concatenating them into a vector". Specifically, for each node... Each discrete time period and each power type , sample sequence As a computational object, define a node. During discrete time periods Power type The power calibration range is The preferred method is to calculate using the following formula:

[0094] ;

[0095] In the formula, express Partial operator; Indicates quantile level, and The preferred value is 0.90 to 0.98. When more attention is paid to operational safety margin, 0.95 or 0.97 is preferred. When the sample size is small, it is not advisable to use 0.90 to 0.98. To avoid the statistical results being dominated by a very small number of tail samples; This represents the median absolute deviation operator; Represents the coverage factor, and The initial value is preferably 1.0. If the proportion of exceeding the limit after calibration is found to be too high in historical playback, then... If the system is too conservative and the load-bearing capacity loss is too large, it can be increased to 1.1 to 1.3. Lowered to 0.9 to 1.0; Represents the robustness coefficient, and The initial value is preferably 0.5. When the error sample fluctuations are small and the distribution is stable, it is preferably 0.1 to 0.5. When there are many abnormal fluctuations in the error sample or the meteorological uncertainty is large, it is preferably 0.5 to 1.0. If the number of samples... If the sample size is less than 10, it is preferable to first expand the sample time window before performing parameter estimation. If the value is greater than or equal to 30, then the above parameters have good statistical stability.

[0096] The meaning of this formula is: first use the absolute error samples... Quantile values ​​reflect the tail coverage level of the error, while the median absolute deviation reflects the degree of error dispersion, and robustness coefficients are used to further analyze this. Adjusting the contribution of the dispersion term, and finally using the coverage coefficient. The overall amplitude is uniformly scaled. The resulting power calibration amplitude takes into account both the tail risk of error samples and the discreteness of the error distribution, making it suitable for scenarios with skewed error distributions, heavy tails, or a small number of outliers.

[0097] After completing the node-by-node power calibration amplitude calculation, each node will be divided into discrete time periods. The active power calibration amplitude constitutes the active power calibration amplitude vector. Each node in discrete time intervals The reactive power calibration amplitude consists of the reactive power calibration amplitude vector. These two vectors serve as inputs for subsequent sensitivity propagation.

[0098] A linearized power flow model of the distribution network is established based on the power grid structure parameters. Let... Representing discrete time periods The sensitivity matrix from active power to node voltage. Representing discrete time periods The sensitivity matrix from reactive power to node voltage. Representing discrete time periods The sensitivity matrix from active power to branch current. Representing discrete time periods The sensitivity matrix from reactive power to branch current is given. This sensitivity matrix can be obtained by linearizing the power flow equations of the distribution network near the current operating point. To make the model easy to implement, it is preferable to use any of the following linearization methods: the Distribution Network DistFlow model, the Small Angle Approximation model, or the Jacobian matrix linearization model, as long as the resulting matrix can characterize the local impact of node injection changes on node voltage and branch current.

[0099] Based on the above sensitivity matrix, the power calibration amplitude is propagated into node voltage calibration parameters and branch current calibration parameters. Let... Representing discrete time periods The node voltage calibration parameter vector, Representing discrete time periods The branch current calibration parameter vector is preferably calculated using the following formula:

[0100] ;

[0101] ;

[0102] In the formula, This indicates taking the absolute value of each element. The purpose of using absolute values ​​is to focus on the maximum disturbance of the error magnitude to voltage and current, without distinguishing the direction of the disturbance. Vector The Each component is denoted as , representing a node During discrete time periods Node voltage calibration parameters; vector The Each component is denoted as , indicating a branch During discrete time periods The branch current calibration parameters.

[0103] Since the sensitivity matrix depends on the current running point, this embodiment preferably uses an iterative method to update the linearized running point. Let the linearized running point of the i-th iteration be . The new running point obtained based on this running point is The damping factor is denoted as The update rule can then be written as ;in, The preferred value range is 0.2 to 0.8, with an initial value preferably of 0.5. If oscillations occur during the iteration process, then decrease... If the iterative process converges and stabilizes, the value can be appropriately increased. To improve the convergence speed.

[0104] To prevent excessive changes in node aggregation control between adjacent iterations from causing linearization error out of control, this embodiment also preferably sets a trust region constraint for the node aggregation control. Let the... Node at the next iteration During discrete time periods The node aggregation control quantity is , No. During the next iteration The trust region radius is denoted as Then it satisfies ;in, Represents the L2 norm; trust region radius. The initial value is preferably 5% to 15% of the rated adjustable power of the corresponding node. If the linearization error in this round is small, the next round can be adjusted accordingly. If the linearization error is large in the current round, the value can be magnified to 1.2 times the original value in the next round. The linearization error is reduced to 0.5 times the original value. Linearization error can be measured using voltage prediction residuals or branch current prediction residuals. For example, when the maximum voltage prediction deviation for all nodes does not exceed 0.002 pu, and the maximum current prediction deviation for all branches does not exceed 1% of the rated current, the current linearization accuracy can be considered to meet the requirements.

[0105] After determining the node voltage calibration parameters and branch current calibration parameters, this embodiment further sets probabilistic budget parameters to transform the operational uncertainties caused by prediction errors into safety constraints that can be directly invoked by subsequent optimization models. Based on these probabilistic budget parameters, voltage opportunity constraints, branch current opportunity constraints, and budget aggregation constraints are constructed. The core idea of ​​this part is: first, the node voltage calibration parameters and branch current calibration parameters are used to characterize the fluctuation range of each node and branch under a given discrete time period affected by prediction errors; then, the probabilistic budget parameters are used to allocate the risk level that the system can bear to different nodes and branches; finally, the allocated local risk level is converted into voltage tightening and current tightening, thereby forming a safe operating boundary that can be directly embedded into the load-bearing capacity optimization model.

[0106] In this embodiment, the probability budget parameters are divided into node voltage probability budget parameters. (Indicates the discrete time period) Internally, assigned to nodes Local permissible over-limit risk level of voltage constraints and branch current probability budget parameters (Indicates the discrete time period) Internally, allocated to branch roads The local allowable risk level of current constraint exceedance); since the probability budget parameter represents the allowable risk proportion, it satisfies .

[0107] In engineering implementation, to avoid the inverse cumulative distribution function becoming excessively large or unstable when the budget parameter approaches zero, it is preferable to set a lower and upper limit for the local probability budget parameter. Let the lower limit of the node voltage probability budget parameter be... The upper limit of the value is Let the lower limit of the branch current probability budget parameter be... The upper limit of the value is Preferably, ,and .

[0108] The aforementioned upper and lower bounds are not mandatory limitations, but rather preferred implementation methods that facilitate numerical implementation. Those skilled in the art can adjust the specific values ​​according to actual security requirements and network scale, but as long as the local probability budget parameter remains within the open interval (0,1), the basic technical idea of ​​this invention can be achieved.

[0109] To translate probability budget parameters into specific safety margins, this embodiment introduces voltage tightening and current tightening. Definitions For nodes During discrete time periods The amount of voltage tightening, branch road During discrete time periods The current tightening amount has the following physical meanings: considering the uncertainty of prediction errors, it is the safety margin that is pre-deducted from the original allowable boundary to ensure that the node voltage and branch current meet the predetermined risk level. In other words, the larger the voltage tightening amount, the larger the safety margin reserved for the node voltage; the larger the current tightening amount, the larger the safety margin reserved for the branch current.

[0110] In this embodiment, the voltage tightening and current tightening amounts are jointly determined by the node voltage calibration parameters, branch current calibration parameters, and probability budget parameters. For nodes During discrete time periods Node voltage calibration parameters, branch road During discrete time periods The branch current calibration parameters then the voltage tightening amount and current tightening amount They respectively satisfy:

[0111] ;

[0112] ;

[0113] In the formula, This represents the coverage factor, which, as mentioned above, is preferably between 0.9 and 1.3, with an initial value preferably of 1.0. If historical operational data is used for playback verification and it is found that the proportion of exceeding the limit is too high, then... Adjust it to 1.1 to 1.3; if the system is found to be too conservative and its load-bearing capacity is underutilized, then... Lowered to 0.9 to 1.0; The operator representing the inverse cumulative distribution function of the standard normal distribution. This is used to convert a given tail risk level into an equivalent quantile coefficient under a standard normal distribution. Thus, when the local probability budget parameter is small, the corresponding quantile coefficient is large, and the resulting tightening is also large, indicating that the system adopts a more conservative safety margin for that node or branch; conversely, when the local probability budget parameter is large, the corresponding quantile coefficient is small, and the tightening is also small, indicating that the system allows that node or branch to bear a relatively higher local risk.

[0114] It should be noted that the inverse cumulative distribution function of the standard normal distribution is used in this embodiment. Converting local risk levels into quantile coefficients is a preferred implementation method. Its engineering implications are: treating node voltage disturbances and branch current disturbances obtained from historical error statistics and sensitivity propagation as equivalent standardized random fluctuations, and constructing a safety margin using the standard normal tail quantile values. This method is computationally simple, easy to embed into optimization models, and matches the aforementioned robust calibration methods based on quantile statistics and median absolute deviation.

[0115] Based on voltage tightening amount and current tightening amount Establish voltage mechanism constraints, expressed as:

[0116] ;

[0117] This inequality does not directly require prediction of node voltages. Falling within the original allowed range Instead of requiring it to fall within a narrower, tightened safety margin, the actual voltage can be controlled within the corresponding budget range, provided that the amplitude of the disturbance does not exceed the safety margin determined by the probability budget parameter and the calibration parameter.

[0118] Branch current opportunity constraints are expressed as follows: This is used to reserve a safety margin related to the branch current calibration parameters and probability budget parameters based on the linearized predicted branch current, so that the probability of the branch current exceeding the allowable upper limit is still controlled after considering the influence of prediction error. Since the branch current is usually expressed in the form of amplitude, and the focus of operation is whether the upper limit is exceeded, this embodiment preferably sets only the upper limit type chance constraint, and does not set a lower limit type chance constraint separately.

[0119] To avoid the overall risk from spiraling out of control due to all local probability budget parameters simultaneously taking large values ​​during the optimization process, this embodiment further sets a budget aggregation constraint, expressed as: ;

[0120] The purpose of budget aggregation constraints is to control the local risk budgets of each node and branch within a globally permissible range, thereby establishing a link between local adjustability and overall network risk controllability. In other words, if a node or branch wants to obtain a larger local risk budget, it must be at the expense of budget contraction for other nodes or branches, or allocated within the limits of the total budget cap, thus preventing the independent amplification of local risks from causing the overall operational risk to exceed the limit.

[0121] In the formula, Represents a node During discrete time periods Linearized prediction of node voltage, , They are nodes The upper and lower limits of the allowable voltage; Indicates a branch During discrete time periods Linearized prediction of branch current, Indicates a branch The maximum allowable current; Representing discrete time periods The total upper limit of the voltage budget; Representing discrete time periods The total upper limit of the current budget.

[0122] Regarding the setting of the total budget ceiling, this embodiment provides the following preferred rules. For the total budget ceiling for voltage... and the total upper limit of the current budget If the system has high safety requirements, such as in important industrial parks, hospital power supply scenarios, or high-risk scenarios near network operation boundaries, it is preferable to set its value to 0.01 to 0.05; if the system operates relatively smoothly and the focus is more on capacity utilization, it can be set to 0.05 to 0.10. Preferably, and The budget ceiling can be dynamically set based on the load level of different time periods, the intensity of renewable energy fluctuations, and the frequency of historical overrun events. For example, during periods of strong photovoltaic fluctuations at midday or heavy evening peak loads, the total budget ceiling can be appropriately lowered to increase conservatism; during periods of low load at night, the total budget ceiling can be appropriately relaxed to improve capacity utilization. If the dispatching department already has clear safety assessment indicators, these indicators can also be directly mapped to the total budget ceiling.

[0123] In subsequent optimization, the local probabilistic budget parameters can be used either as preset values ​​or as decision variables in joint optimization. In the basic implementation, if a more complex budget allocation strategy is not introduced, two simple initialization methods can be used. The first method is uniform initialization, which sets the same initial values ​​for all nodes and all branches, and then solves under budget aggregation constraints. The second method is initialization according to the calibration parameter ratio, where the initial local budget is proportional to the calibration parameter. or The relative sizes are proportional, thus allowing nodes and branches with greater fluctuations to obtain higher budget values ​​in the initial state. Even if any of the above basic initialization methods are used, as long as the subsequent optimization model satisfies the voltage opportunity constraint, branch current opportunity constraint, and budget summation constraint, the core technical effect of this invention can still be achieved.

[0124] S4: Using node carrying capacity parameters, node aggregation control quantities, and probabilistic budget parameters as decision variables, determine the node net injection quantity based on the node net injection prediction baseline data and node aggregation control quantities. Based on the node carrying capacity parameters and the node net injection quantity formed by the smoothed and limited injection function, construct and solve a multi-objective optimization model for carrying capacity under the feasible region of node aggregation control quantities, voltage opportunity constraints, branch current opportunity constraints, and budget aggregation constraints. Determine the optimized values ​​of node carrying capacity parameters and node aggregation control quantities for each node in each discrete time period. The optimized values ​​of node carrying capacity parameters are used to characterize the upper limit of the net injection capacity that the corresponding node can carry in the corresponding discrete time period.

[0125] In this embodiment, node During discrete time periods The nodal bearing capacity parameter is denoted as Its physical meaning is: under the current power grid topology, operating baseline, adjustable resource capacity, and uncertainty calibration level, the node During discrete time periods The net injection limit that the system can safely bear. This parameter is not a simple static equipment rating, but a dynamic upper limit obtained by an optimization model after comprehensively considering the local adjustable resources of the node, line constraints, node voltage constraints, and the influence of prediction errors. Therefore, the node carrying capacity parameter of the same node is usually different at different discrete time periods, and the node carrying capacity parameter of different nodes at the same discrete time period is also usually different.

[0126] In this embodiment, the set of decision variables for the multi-objective optimization model of bearing capacity preferably includes: When certain resources or nodes lack reactive power regulation capabilities, the corresponding nodes' It can be fixed at zero.

[0127] To facilitate the solution, it is preferable to set upper and lower bounds for the nodal bearing capacity parameters. Let the node... The upper limit of the node bearing capacity parameter is The lower limit is ,satisfy Preferably, A value of 0 is acceptable, indicating that the node's carrying capacity is allowed to degrade to the point where no new positive net injection is permitted under extreme circumstances. The upper limit of the engineering limit can be determined in any of the following ways: First, take the upper limit of the engineering limit calculated comprehensively at the equipment level, line level, and transformer level for that node; second, take 1.05 to 1.20 times the historical maximum safe injection value of that node as the initial upper limit of the engineering limit; third, if there is planned access capacity, take the smaller value between the planned capacity and the current safe access capacity as the upper limit. For most distribution network implementation scenarios, to avoid unbounded optimization, the second or third method is preferred.

[0128] The net injection amount at a node is determined jointly by the node net injection prediction baseline data and the node aggregation control quantity. Specifically, let... Represents a node During discrete time periods The baseline value of the predicted net active power injection, Represents a node During discrete time periods The active component in the node aggregation control quantity, then the node During discrete time periods Net injection volume of nodes The preferred method is determined by the following formula: ;in, This represents the net injection amount to a node after considering resource aggregation and adjustment, but before being constrained by the node's carrying capacity limit. If this value exceeds the safe carrying capacity level of the corresponding node, it needs to be mapped to a limited net injection amount using a smoothing and limiting injection function.

[0129] When constructing a multi-objective optimization model for bearing capacity, it is preferable to simultaneously input the following constraints: feasible region of node aggregation control quantity. The constraints include node voltage opportunity constraints, branch current opportunity constraints, budget aggregation constraints, upper and lower bound constraints on node carrying capacity parameters, and upper and lower bound constraints on local probabilistic budget parameters. This ensures that the optimization model, while expanding node carrying capacity, does not deviate from the aforementioned adjustable resource boundaries and grid security boundaries.

[0130] In the solution process, this embodiment preferably adopts the following order. First, read the node net injection prediction baseline data, the feasible region of the node aggregate control quantity, the node voltage calibration parameters, the branch current calibration parameters, and the total budget upper limit. Second, for the node bearing capacity parameters... Node aggregation control quantity and local probability budget parameters , Perform initialization. During initialization, it is preferable to initialize the node bearing capacity parameters. Take the sum of the historical secure access capacity or current baseline net injection volume of the corresponding node and the engineering margin; aggregate the node control volume. The zero vector or the optimal value of the previous rolling cycle is taken; the local probability budget parameters are initialized in a uniform manner or proportionally to the calibration parameters. Then, the objective function is iteratively solved under constraints. During the iteration process, the solution space is continuously limited by the opportunity constraints and budget aggregation constraints described above, while the sensitivity matrix and predicted running quantities are updated using a linearized running point update mechanism. Finally, the optimized values ​​of the node bearing capacity parameters for each node at each discrete time period are output. Optimization value of node aggregation control quantity Optimized values ​​of node voltage probability budget parameters and optimized values ​​of branch current probability budget parameters .

[0131] Since the optimization model in this embodiment includes a smooth nonlinear function, an inverse cumulative distribution function, and a relative entropy term, a solution method capable of handling smooth nonlinear constraint optimization problems is preferred, such as the interior-point method, sequential quadratic programming, or sequential convex approximation method. For scenarios with small network size or requiring rapid verification, the interior-point method can be used directly. For scenarios with large network size and requiring rolling optimization to update the sensitivity matrix, the sequential convex approximation method is preferred. In each round, the linearized running point is fixed, the approximate problem is solved, and then iterated according to the damping update rule defined above. To ensure a clear termination condition, both the relative change threshold of the objective function and the constraint residual threshold are preferred as stopping criteria. For example, when the relative change of the objective function between two adjacent iterations is no greater than... Furthermore, the maximum residual of all node voltage constraints, branch current constraints, budget aggregation constraints, and node aggregation feasible region constraints is no greater than [value missing]. When the model converges, it is determined that the model has converged.

[0132] In this embodiment, in order to measure the node bearing capacity parameters Net Injection Amount at Nodes To ensure continuous coupling and avoid the use of hard truncation functions that could lead to non-differentiable points in the optimization problem, a smooth, limited injection function is preferred to map the node net injection amount to a limited net injection amount. Let the node... During discrete time periods The limited net injection amount is The smoothing coefficient is The preferred expression is as follows: The technical principle of this function is as follows: when the net injection amount of the node... Significantly smaller than the node bearing capacity parameter hour, When the value is a large positive value, the exponent term is large, and the entire function approximately satisfies That is, it does not significantly compress the net injection volume of nodes; when the net injection volume of nodes Significantly greater than the node bearing capacity parameter hour, When the value is a large negative value, the exponent term approaches zero, and the entire function approximately satisfies... This means that the net injection volume at the node is smoothly limited to around the node's bearing capacity parameters. This is different from directly using... In comparison, the smoothing and limiting injection function described above is continuously differentiable throughout its entire domain, making it easier to solve in conjunction with multi-objective optimization models.

[0133] Smoothing coefficient Used to control the degree of smoothing approximation. If If the value is too small, the smoothing function will be very close to the hard cutoff function. Although the amplitude limit boundary is clear, it will worsen the numerical conditions, making the solver sensitive to gradient changes near the boundary; if... If the value is too large, the smoothing interval will be too wide, which will prevent the portion exceeding the bearing capacity boundary from being sufficiently suppressed, thus reducing the accuracy of the trigger limit judgment. Therefore, this embodiment preferably uses... Set to 1% to 5% of the node's rated injection capacity or historical maximum safe net injection volume; for low-voltage distribution areas or small-capacity nodes, preferably 2% to 3%; for medium-voltage feeder nodes or new energy access nodes with large capacity fluctuations, preferably 3% to 5%. If a uniform value is used, it can also be set as follows: ,in, This represents the average or median value of the upper bound of the bearing capacity parameters of all nodes in the system. Represents the smoothing ratio coefficient, preferably satisfying .

[0134] To quantitatively characterize the degree of node injection compression caused by amplitude limiting, this embodiment introduces a node injection limiting ratio. .set up If the preset positive lower limit of the node bearing capacity parameter is used, then the preferred definition is: ;in, Its purpose is to avoid when When the denominator is close to zero, it becomes too small, leading to numerical instability. Preferably, The value can be 1% to 5% of the rated injection capacity of the corresponding node, or a uniform lower limit value for the entire system, such as 0.01 pu. If the net injection volume of the node... The bearing capacity parameters of the nodes were not exceeded. ,but Connect to 0; if the net injection volume at a node significantly exceeds the node's bearing capacity parameter, then Take a positive value and increase it as the exceedance increases. This determines the node's transmission limit ratio. It can be directly used to quantify the balance between expanding carrying capacity and the actual level of emission restriction.

[0135] In this embodiment, the objective function of the multi-objective optimization model for bearing capacity preferably includes a node bearing capacity enhancement term, a node emission limitation and suppression term, and a budget distribution matching term. First, the node bearing capacity enhancement term is used to increase the upper limit of net injection capacity that each node can bear in each discrete time period. Second, the node emission limitation and suppression term is used to reduce the proportion of net injection capacity that is compressed due to bearing capacity constraints. Third, the budget distribution matching term is used to match the optimized local probability budget distribution with the local risk distribution characterized by the node voltage calibration parameters and the branch current calibration parameters, so as to avoid unreasonable concentration of probability budget in a few nodes or a few branches.

[0136] To construct a budget distribution matching term, this embodiment introduces a reference budget distribution and an actual budget distribution. For the node voltage side, a node is defined. During discrete time periods The reference voltage budget distribution ratio is The actual voltage budget distribution ratio is For the branch current side, define the branch. During discrete time periods The proportion of the reference current budget distribution is The actual current budget distribution ratio is To avoid the problem of the denominator being zero when all calibration parameters are close to zero during certain periods, a regularization constant is introduced. ,and Preferably, Pick to , better The reference distribution and the actual distribution are preferably determined by the following formula: ; ;in, , Each of the following is a list of any node in the node set. Node voltage probability budget parameters, node voltage calibration parameters; , Each is any one of the branches in the set of branches. The branch current probability budget parameters and branch current calibration parameters;

[0137] After obtaining the reference distribution and the actual distribution, define the discrete time period. Budget distribution matching items for:

[0138] ;

[0139] This formula is essentially the sum of two relative entropy terms: one for the voltage side and one for the current side. Its technical implication is: if the optimized actual budget distribution is close to the local risk distribution characterized by the calibration parameters, then... The budget is relatively small; however, if the actual budget is overly concentrated on a small number of nodes or branches, or deviates significantly from the local risk distribution, then... Increase this term. Introducing this term into the objective function can make the model more inclined to allocate the probability budget to locations with greater volatility and higher risk, rather than randomly concentrating it in local locations, while improving the node carrying capacity, thereby improving the physical rationality of the budget allocation.

[0140] In this embodiment, the node carrying capacity enhancement term, the node emission limitation and suppression term, and the budget distribution matching term together constitute the objective function. Let... Represents a node The node importance coefficient, , , Let represent the weight coefficients of the objective function for the node carrying capacity enhancement term, the node emission limitation and suppression term, and the budget distribution matching term, respectively. Then the objective function... The preferred option is written as:

[0141] ;

[0142] In the formula, , , All values ​​are positive and satisfy the following condition: .

[0143] Before the actual solution, it is preferable to normalize the three objectives first, and then set the weights. The preferred normalization method is as follows: divide the nodal bearing capacity improvement term by the sum of its upper engineering bounds, and divide the nodal load-bearing capacity suppression term by... Divide the budget distribution matching item by Alternatively, its historical reference value. After normalization, the following weighting rules can be used:

[0144] When the emphasis is on improving load-bearing capacity, the preferred option is... Up to 0.65, Up to 0.30, Up to 0.25;

[0145] When there is a greater emphasis on reducing restrictions, the preferred approach is to... Up to 0.50, Up to 0.45, Up to 0.25;

[0146] When greater emphasis is placed on the balance and physical rationality of risk budget allocation, the preferred option is... Up to 0.45, Up to 0.35, Up to 0.35.

[0147] In equilibrium scenarios without particular preferences, the recommended initial value is .

[0148] Node importance coefficient This is used to differentiate the priority of different nodes in the capacity enhancement process. Preferably, ordinary nodes are selected as... Important load nodes, key planned new energy access nodes, or feeder end constrained sensitive nodes can be selected. Up to version 2.0; if node differences are not required, then all nodes should be uniformly represented. To avoid arbitrary and unfounded value assignments, it is preferable to set values ​​according to the following rules. :

[0149] If node If the increase in the load capacity of the feeder has a direct impact on the planned access capacity, then it will be increased. ;

[0150] If node If the supply of electricity is connected to important public services or key industrial users and excessive restrictions are costly, then increase... ;

[0151] If node If it is a regular node in the entire network, then maintain .

[0152] In the specific solution process, the following steps are recommended. First, according to the aforementioned implementation methods, input the feasible region of the node aggregation control quantity, the node voltage opportunity constraint, the branch current opportunity constraint, and the budget summary constraint. Second, initialize... , , , And the objective function weights. Then, the net injection amount of the node is calculated based on the initial values. Net injection volume with limited amplitude Node issuance limit ratio Reference budget distribution, actual budget distribution, and budget distribution matching items Next, construct the overall objective function. The solution is then obtained by combining all constraints. If rolling optimization or sequential convex approximation methods are used, the sensitivity matrix, linearized predictions in the chance constraints, and budget distribution matching terms are updated after each iteration before proceeding to the next iteration. Finally, when the objective function and constraint residuals simultaneously satisfy the convergence criteria, the optimized values ​​of the nodal bearing capacity parameters, the nodal aggregation control variables, and the probabilistic budget parameters are output.

[0153] S5: Based on the optimized value of the node aggregation control quantity, solve the resource-level control quantity sequence that satisfies the node aggregation consistency constraint, resource flexibility envelope and cross-time coupling constraint in the optimization time domain, and send it to the corresponding source-load controllable resource for adjustment.

[0154] The purpose of this step is to accurately map the aggregation and adjustment results of the node layer to the resource layer, so as to realize the closed-loop implementation from node capacity optimization to equipment control execution, and avoid the problem that the node layer results cannot be directly applied to the actual equipment.

[0155] To ensure strict consistency between the resource-level control quantity sequence decomposition results and the node-level optimization results, this embodiment first establishes node aggregation consistency constraints. Let... Represents a node During discrete time periods The optimized value of the node aggregation control variable is then applied to any node. and any discrete time period The resource-level control quantity sequence should satisfy ;

[0156] Node aggregation consistency constraints alone are insufficient to obtain an executable resource-level control sequence, because different resources must also satisfy their respective resource flexibility envelopes and cross-time coupling constraints. Therefore, in this embodiment, the feasible region of the resource-level control sequence decomposition model is determined by three parts: the first part is the node aggregation consistency constraint; the second part is the resource flexibility envelope of each resource in the optimization time domain; and the third part is the cross-time coupling constraint corresponding to each resource. Specifically, for any resource... Its control sequence over all discrete time periods It must belong to the resource flexibility envelope defined above. If resources For an energy storage device, the above control sequence must also simultaneously satisfy the state-of-charge balance constraint and the energy storage boundary constraint; if resources For adjustable loads, cumulative adjustment constraints must also be met; if resources For charging facilities, energy demand constraints before departure must also be met. This ensures that the decomposition results are correct not only at the node level but also at the equipment level.

[0157] To obtain a resource-level control quantity sequence that better meets actual operational needs while satisfying the aforementioned feasible region constraints, this embodiment further sets an objective function for the resource-level control quantity sequence decomposition model. This objective function considers at least two technical requirements: first, reducing the deviation of resource-level control quantities from equipment reference control quantities to avoid excessive disturbance to the original equipment operation plan; and second, suppressing jumps in resource-level control quantities between adjacent discrete time periods to avoid frequent and large-scale changes in control commands. Therefore, the following is set... Indicates controllable resources of source load During discrete time periods The reference control quantity, whose physical meaning is: before the node bearing capacity optimization control described in this invention is implemented, the resources... The reference control value is the control value originally planned to be executed or the basic operation plan value with higher priority. The source of the reference control value may be different for different types of resources. For example, for distributed power sources, the reference control value can be the planned output corresponding to the maximum local output capacity; for energy storage devices, the reference control value can be the original charging and discharging plan; for adjustable loads, the reference control value can be the original load adjustment plan or the zero adjustment value; for charging facilities, the reference control value can be the user's default charging power plan.

[0158] In this embodiment, the following objective function for the sequential decomposition of resource-level control variables is preferably constructed:

[0159] ;

[0160] In the formula, This represents the objective function for the sequential decomposition of resource-level control variables; Representing resources The reference tracking weight coefficient is used to measure the cost of resource-level control quantities deviating from the reference control quantity; Representing resources The time-series smoothing weighting coefficient is used to measure the cost of control jumps between adjacent discrete time periods; This represents the L2 norm. When the resource-level control variable is a scalar, the L2 norm degenerates into the square of the absolute value. The first term of the objective function above is used to reduce the deviation of the decomposition result from the reference control plan, and the second term is used to suppress jump variables in the decomposition result between adjacent discrete time periods.

[0161] Reference tracking weight coefficient The value should be matched with the resource type and business sensitivity. Preferably, after normalizing the dimensions of each resource control variable, the value should be... The value should be set between 0.5 and 3.0. For energy storage devices that are flexible, responsive, and allow deviations from the original plan, a smaller value, such as 0.5 to 1.0, is preferred; for distributed power sources, 0.8 to 1.5 is preferred; for adjustable loads related to user comfort or production continuity, 1.5 to 3.0 is preferred; for charging facilities, if a certain degree of adjustment in charging power is allowed, 1.0 to 2.0 is preferred. If no fine classification is required, all resources can be considered... The initial value is 1, and then it is adjusted based on the running effect.

[0162] Time-series smoothing weight coefficient The value should be matched with the dynamic response characteristics of the resource and the sensitivity of equipment wear. Preferably, after normalization, the value should be... The value should be set between 0.1 and 2.0. For energy storage devices and adjustable switching equipment subject to frequent switching losses or lifespan limitations, a larger value is recommended, such as 0.8 to 2.0. For adjustable loads with continuous adjustment capabilities and low operating costs, a value of 0.1 to 0.8 is recommended. For inverter control of distributed power sources, a value of 0.3 to 1.2 is recommended. For charging facilities, a value of 0.2 to 1.0 is recommended. If network operation requires particularly smooth equipment operation, the overall efficiency of each resource can be increased. If a greater focus is placed on precisely meeting the node aggregation control parameters, then the parameters can be appropriately reduced. .

[0163] Before the actual solution, it is preferable to first normalize all types of resource control quantities to eliminate the impact of differences in the rated capacity of different resources on the weights of the objective function. The normalization method can be based on the rated power of the resources. Let the resources... The rated active capacity is Rated reactive power capacity is Then the active control component can be divided by The reactive power control component can be divided by For resources with only active power control, normalization based on active power capacity is sufficient. After normalization, then configure... and This can prevent certain large-capacity resources from being over-amplified in the objective function simply because of their large numerical scale.

[0164] In terms of the solution process, this embodiment preferably adopts the following steps. First, the optimized values ​​of the node aggregate control quantity for each node at each discrete time period are read from the multi-objective optimization model of node bearing capacity. Secondly, establish the resource set corresponding to each node based on the resource node mapping relationship. And read the resource flexibility envelope of each resource in the optimization time domain. And cross-time period coupling constraints. Then, based on the current state of the equipment, the execution result of the previous rolling cycle, or the original scheduling plan, reference control quantities for each resource are generated. After that, the construction was based on For the resource-level control variable sequence decomposition model of decision variables, node aggregation consistency constraints, resource flexibility envelope, and cross-time coupling constraints are all incorporated, and the above objective function is solved. Once the model is solved, it outputs the optimized sequence of resource-level control variables for each resource across all discrete time periods. .

[0165] For resource reference control quantity In this embodiment, further implementable rules are provided for the generation of the resource decomposition model. For distributed power sources, if the basic objective is to prioritize power generation, the reference control quantity can be set to the control value corresponding to the currently available power output; if the basic objective is to track existing day-ahead plans, the day-ahead plan value can be directly adopted. For energy storage devices, it is preferable to set the reference control quantity to a baseline control value that satisfies the original energy storage charging and discharging plan or maintains a stable state of charge. For adjustable loads, if it is desired to reduce the impact on user comfort or production processes, it is preferable to set the reference control quantity to zero adjustment or the original planned adjustment value. For charging facilities, it is preferable to set the reference control quantity to the original charging power plan under the premise of meeting the predetermined off-site energy demand. Through the above methods, the resource decomposition model can control the disturbance to the original operating plan of the equipment while satisfying the node-level objectives.

[0166] Obtain resource-level control sequence Then, it needs to be converted into execution instructions that can be recognized by the corresponding device. In this embodiment, different instruction mapping methods are used for different resource types. For distributed power sources, if they support bidirectional regulation of active and reactive power, then... Convert to active power output setpoint, Convert to reactive power setpoint or power factor command; if only active power limiting is supported, only active power setpoint is output. For energy storage devices, resource-level control variables are mapped to charging power setpoint or discharging power setpoint; when When, it can be understood as a discharge command, when At this time, it can be understood as a charging command, and the specific symbol convention can be uniformly specified in the equipment access protocol. For adjustable loads, resource-level control quantities are mapped to load reduction quantities, load transfer quantities, or equipment operation level adjustment commands. For charging facilities, resource-level control quantities are mapped to charging power setpoints or charging current setpoints. If the resource has reactive power regulation capabilities, reactive power control setpoints can be further output.

[0167] To reduce frequent communication switching and device jitter, this embodiment preferably sets local dead zones and tracking tolerances at the execution layer. Resources are... The control dead zone threshold is If the optimized value of the resource-level control variable obtained this time... The difference between the current set value and the previously issued value is less than This way, the current setting value can be kept unchanged, and it will not be issued repeatedly. Preferably, Take 0.5% to 2% of the resource's rated power. For devices with slow response times or high costs associated with frequent actions, a larger value can be used; for devices with fast response times and low action costs, a smaller value can be used. After execution, the device controller preferably sends back the current actual execution value and availability status in real time. If a resource fails to execute, communication is interrupted, or its availability status changes, the feasible domain of that resource is updated again in the next rolling optimization cycle. and reference control quantity Then, the resource-level control sequence decomposition model is solved again. This ensures that the entire system maintains closed-loop executability even when equipment states change.

[0168] Regarding the control cycle arrangement, it is preferable to use the same rolling cycle for the resource-level control quantity sequence decomposition and the aforementioned multi-objective optimization of node carrying capacity. Alternatively, a two-layer cycle structure with a slower upper layer and a faster lower layer can be adopted. For example, the multi-objective optimization of node carrying capacity can be rolled every 15 minutes, while the resource-level control quantity execution instructions can be refreshed every 1 minute or every 5 minutes. If a two-layer cycle structure is adopted, within the upper-layer optimization cycle, the resource-level control quantity sequence decomposition can be locally corrected according to the real-time status of the equipment, but the most recent node aggregate control quantity optimization value output by the upper layer will still be used. This serves as the overall boundary. This balances global optimality with execution flexibility.

[0169] like Figure 2 As shown, another embodiment of the present invention provides a multi-objective optimization system for power grid carrying capacity considering source-load interaction, comprising:

[0170] The data acquisition module is used to acquire power grid operation data, including power grid structure parameters, node net injection prediction baseline data, source and load controllable resource information, and historical prediction error samples.

[0171] The resource modeling module is used to construct the resource flexibility envelope of each source load controllable resource in the time domain based on the source load controllable resource information, and to establish cross-time period coupling constraints for source load controllable resources with temporal coupling.

[0172] The node aggregation module is used to aggregate the resource flexibility envelope into a node aggregation flexibility envelope based on the mapping relationship between each source load controllable resource and node, and thereby limit the feasible domain of node aggregation control quantity.

[0173] The calibration and constraint construction module is used to determine the node voltage calibration parameters and branch current calibration parameters based on historical prediction error samples and the power grid structure parameters, and to set probability budget parameters to establish voltage opportunity constraints, branch current opportunity constraints and budget summary constraints.

[0174] The bearing capacity optimization module is used to take the node bearing capacity parameters, node aggregation control quantity and probability budget parameters as decision variables, determine the node net injection quantity based on the node net injection prediction baseline data and the node aggregation control quantity, and construct and solve the bearing capacity multi-objective optimization model based on the node bearing capacity parameters and the node net injection quantity formed by the smoothing and limiting injection function, and determine the optimized values ​​of the node bearing capacity parameters and the node aggregation control quantity for each node in each discrete time period.

[0175] The resource decomposition and execution module is used to solve the resource-level control quantity sequence that satisfies the node aggregation consistency constraint, resource flexibility envelope and cross-time coupling constraint in the optimization time domain based on the node aggregation control quantity optimization value, and send it to the corresponding source load controllable resource for execution adjustment.

[0176] In summary, this invention enables more refined optimization analysis of power grid carrying capacity under the conditions of source-load interaction and operational uncertainties, while taking into account both safety constraint satisfaction and resource execution feasibility, thereby improving the distribution network's carrying capacity for new net injections and the actual operation control effect.

[0177] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A multi-objective optimization method for power grid carrying capacity considering source-load interaction, characterized in that, The method includes: Acquire power grid operation data, including power grid structure parameters, node net injection prediction baseline data, source and load controllable resource information, and historical prediction error samples; Based on the source load controllable resource information, a resource flexibility envelope is constructed for each source load controllable resource in the time domain to optimize the resource flexibility of each source load controllable resource. Cross-time period coupling constraints are established for source load controllable resources with temporal coupling. Furthermore, according to the mapping relationship between each source load controllable resource and nodes, the resource flexibility envelope is aggregated into a node aggregation flexibility envelope, thereby limiting the feasible domain of the node aggregation control quantity, including: set up For a node aggregation control space, there is a finite set of directions. For nodes During discrete time periods The node aggregation control quantity. For mapping to nodes The node feasible set is jointly induced by the resource flexibility envelope of each source load controllable resource and the cross-time coupling constraint; for any node Discrete time period and direction vector Calculate the support value, which satisfies: ； Based on the aforementioned support values, an outer approximation polyhedron of the feasible region of the node aggregation control quantity is constructed, satisfying: ； In the formula, For the finite set of directions The first in A directional vector, Represents a finite set of directions The total number of directions in; Represents a node During discrete time periods Along the first Support values ​​of each directional vector; Indicates transpose; Represents a node During discrete time periods The node aggregation control quantity is the outer approximation polyhedron of the feasible domain; The external approximation polyhedron As a node During discrete time periods The feasible domain of node aggregation control quantity; Based on historical prediction error samples and power grid structure parameters, node voltage calibration parameters and branch current calibration parameters are determined, and probability budget parameters are set. Based on these parameters, voltage opportunity constraints, branch current opportunity constraints, and budget aggregation constraints are established. The probability budget parameters include node voltage probability budget parameters. Branch current probability budget parameters ; remember For nodes During discrete time periods Node voltage calibration parameters, branch road During discrete time periods The branch current calibration parameters then the voltage tightening amount and current tightening amount They respectively satisfy: ； ； In the formula, Indicates the coverage factor; The operator representing the inverse cumulative distribution function of the standard normal distribution, and ; Based on the voltage tightening amount and current tightening amount Establish voltage mechanism constraints, expressed as: ； Branch current opportunity constraints are expressed as follows: ; Budget aggregation constraints are expressed as follows: ; In the formula, Represents a node During discrete time periods Linearized prediction of node voltage, , They are nodes The upper and lower limits of the allowable voltage; Indicates a branch During discrete time periods Linearized prediction of branch current, Indicates a branch The maximum allowable current; Representing discrete time periods The total upper limit of the voltage budget; Representing discrete time periods The total upper limit of the current budget; Using node carrying capacity parameters, node aggregation control quantities, and probabilistic budget parameters as decision variables, the node net injection quantity is determined based on the node net injection prediction baseline data and the node aggregation control quantity. Based on the node carrying capacity parameters and the node net injection quantity formed by the smoothing and limiting injection function, a multi-objective optimization model for carrying capacity is constructed and solved under the feasible region of node aggregation control quantity, voltage opportunity constraints, branch current opportunity constraints, and budget aggregation constraints. The optimized values ​​of node carrying capacity parameters and node aggregation control quantities for each node in each discrete time period are determined. The optimized values ​​of node carrying capacity parameters are used to characterize the upper limit of the net injection that the corresponding node can carry in the corresponding discrete time period. Based on the optimized value of the node aggregation control quantity, a sequence of resource-level control quantities that satisfies the node aggregation consistency constraint, resource flexibility envelope, and cross-time coupling constraint is solved in the optimization time domain, and then sent to the corresponding source-load controllable resources for adjustment.

2. The multi-objective optimization method for power grid carrying capacity considering source-load interaction as described in claim 1, characterized in that, Controllable resources include source-side controllable resources and load-side controllable resources; among the source-side controllable resource information, the constraint parameters of the source-side controllable resources include at least the upper and lower limits of distributed power output, ramping constraints, energy storage charging and discharging power constraints, and energy storage state of charge constraints; the constraint parameters of the load-side controllable resources include at least the upper and lower limits of adjustable load adjustment, cumulative adjustment constraints, charging facility power constraints, and energy demand constraints. The cross-time coupling constraints include at least the energy storage state of charge balance constraint, the adjustable load cumulative adjustment constraint, and the charging facility time-period energy balance constraint.

3. The multi-objective optimization method for power grid carrying capacity considering source-load interaction as described in claim 1, characterized in that, The determination of node voltage calibration parameters and branch current calibration parameters based on historical prediction error samples and power grid structure parameters includes: For active power prediction error and reactive power prediction error, respectively, the active power calibration amplitude and reactive power calibration amplitude are calculated, and robust calibration is performed on the active power calibration amplitude and reactive power calibration amplitude based on coverage statistics and dispersion statistics; wherein, for any power type During discrete time periods Power calibration amplitude satisfy: ； In the formula, Indicates active power. Indicates reactive power; Indicates power type exist The prediction error sample sequence; express Partial operator Indicates quantile level; This represents the median absolute deviation operator; Indicates the coverage factor; Indicates the robustness coefficient; Each node in discrete time intervals The active power calibration amplitude constitutes the active power calibration amplitude vector. Each node in discrete time intervals The reactive power calibration amplitude consists of the reactive power calibration amplitude vector. ; A linearized power flow model of the distribution network is used to establish sensitivity matrices for active power disturbances, reactive power disturbances, and their effects on node voltages and branch currents. Based on these sensitivity matrices, the power calibration amplitude is determined. The parameters are propagated as node voltage calibration parameters and branch current calibration parameters. During the iterative solution process, the linearized running points are updated with weights according to the preset damping factor, and trust region constraints are applied to the node aggregation control quantities between adjacent iterations.

4. The multi-objective optimization method for power grid carrying capacity considering source-load interaction as described in claim 1, characterized in that, The net injection amount at the node is expressed as: ; The limited net injection amount is expressed as: ; In the formula, Represents a node During discrete time periods The net injection amount of the node, Represents a node During discrete time periods The baseline value of the predicted net active power injection, Represents a node During discrete time periods The active component in the node aggregation control quantity; Represents a node During discrete time periods The limited net injection volume, Represents a node During discrete time periods The nodal bearing capacity parameters, Indicates the smoothing coefficient; The node limiting ratio is determined based on the difference between the net injection amount at the node and the net injection amount for limiting the amplitude, expressed as follows: ； In the formula, Represents a node During discrete time periods The node issuance limit ratio, This represents the preset positive lower limit of the node bearing capacity parameter; The node emission restriction ratio is used to characterize the relative emission restriction degree of a node due to carrying capacity limitations in the corresponding discrete time period, and is written into the objective function or constraint condition of the carrying capacity multi-objective optimization model.

5. The multi-objective optimization method for power grid carrying capacity considering source-load interaction as described in claim 4, characterized in that, The objective function of the multi-objective optimization model for bearing capacity includes a node bearing capacity enhancement term, a node emission limitation and suppression term, and a budget distribution matching term, expressed as follows: ; In the formula, Represent the objective function; , , These represent the objective function weight coefficients for the node carrying capacity enhancement term, the node emission restriction and suppression term, and the budget distribution matching term, respectively, and all are greater than 0; Represents a node The node importance coefficient; Representing discrete time periods The budget distribution matching item, and satisfies ; in, ; ; Represents a node During discrete time periods The actual voltage budget distribution ratio Represents a node During discrete time periods The proportion of the reference voltage budget distribution; Indicates a branch During discrete time periods The actual current budget distribution ratio Indicates a branch During discrete time periods The proportion of the reference current budget distribution; The total number of nodes. , Each of the following is a list of any node in the node set. Node voltage probability budget parameters, node voltage calibration parameters; The total number of branch roads, , Each is any one of the branches in the set of branches. The branch current probability budget parameters and branch current calibration parameters; Let be the regularization constant, and .

6. The multi-objective optimization method for power grid carrying capacity considering source-load interaction according to claim 1, characterized in that, The process of solving the resource-level control quantity sequence that satisfies node aggregation consistency constraints, resource flexibility envelope, and cross-time coupling constraints within the optimization time domain includes: For source load controllable resources mapped to the same node, a resource-level control quantity sequence decomposition model is established. Under the premise of satisfying node aggregation consistency constraints, resource flexibility envelopes corresponding to each source load controllable resource, and cross-time period coupling constraints, the resource-level control quantity sequence decomposition model reduces the deviation of the resource-level control quantity sequence of each source load controllable resource in the optimization time domain from the corresponding reference control quantity, and suppresses the jump of resource-level control quantity in adjacent discrete time periods. The obtained resource-level control quantity sequence is converted into the active power setpoint, reactive power setpoint, charging and discharging power setpoint, load adjustment quantity or charging power setpoint of the corresponding source-load controllable resource, and sent to the corresponding source-load controllable resource for adjustment.

7. A multi-objective optimization system for power grid carrying capacity considering source-load interaction, used to implement the multi-objective optimization method for power grid carrying capacity considering source-load interaction as described in any one of claims 1-6, characterized in that, The system includes: The data acquisition module is used to acquire power grid operation data, including power grid structure parameters, node net injection prediction baseline data, source and load controllable resource information, and historical prediction error samples. The resource modeling module is used to construct the resource flexibility envelope of each source load controllable resource in the time domain based on the source load controllable resource information, and to establish cross-time period coupling constraints for source load controllable resources with temporal coupling. The node aggregation module is used to aggregate the resource flexibility envelope into a node aggregation flexibility envelope based on the mapping relationship between each source load controllable resource and the node, and thereby limit the feasible domain of the node aggregation control quantity. The calibration and constraint construction module is used to determine the node voltage calibration parameters and branch current calibration parameters based on the historical prediction error samples and the power grid structure parameters, and to set the probability budget parameters to establish voltage opportunity constraints, branch current opportunity constraints and budget summary constraints. The bearing capacity optimization module is used to take the node bearing capacity parameters, node aggregation control quantity and the probability budget parameters as decision variables, determine the node net injection quantity based on the node net injection prediction baseline data and the node aggregation control quantity, and construct and solve the bearing capacity multi-objective optimization model based on the node bearing capacity parameters and the node net injection quantity formed by the smoothing and limiting injection function, and determine the optimized values ​​of the node bearing capacity parameters and the node aggregation control quantity for each node in each discrete time period. The resource decomposition and execution module is used to solve the resource-level control quantity sequence that satisfies the node aggregation consistency constraint, the resource flexibility envelope, and the cross-time coupling constraint in the optimization time domain based on the node aggregation control quantity optimization value, and send it to the corresponding source load controllable resource for execution adjustment.

Images