Power grid dispatching method and device considering wind power uncertainty and medium

Abstract

The invention discloses a power grid dispatching method and device considering wind power uncertainty and a medium, and the method comprises the steps: building an AC power flow-based chance constraint model according to the wind power uncertainty; analyzing the opportunity constraint model by adopting a sensitivity factor and first-order Taylor expansion, and converting opportunity constraint in the opportunity constraint model into linear constraint; according to the linear constraint, an improved double description method is adopted, and an original feasible region of the power grid is obtained through calculation; and constructing a local nonlinear adjustment optimization model based on the original feasible region, and solving the local nonlinear adjustment optimization model by adopting a deep learning method to obtain a final feasible region so as to perform power grid dispatching according to the feasible region. According to the method, the system feasible region can be quickly and accurately calculated under the wind power high permeability, and early warning information is provided for real-time dispatching of a power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid dispatching, and particularly relates to a power grid dispatching method, device and medium considering the uncertainty of wind power. Background Art

[0002] With the development of new energy technologies, the installed capacity of wind power and photovoltaic power has been increasing year by year, and wind power has gradually become the main body of power supply. However, the randomness and volatility of wind power make the power grid may face the problem of insufficient peak regulation and frequency modulation capabilities, and there is even a contradiction between wind power curtailment and the alternating occurrence of power and electricity shortages in the power grid under extreme weather conditions. Therefore, it is necessary to study a fast solution method for the feasible region of the power grid considering the uncertainty of wind power to determine whether the operating point of the future power system is within the safe range and perform power grid dispatching in a timely manner.

[0003] The feasible region of the power system can provide important support for research such as power markets, voltage control, and planning. In terms of time scale, the feasible region of the power system is divided into two aspects: steady-state feasible region and transient feasible region. However, the current research on the operating feasible region of the power system under high penetration of wind power is still in its infancy, and an efficient solution method for analyzing the operating feasible region has not been established. Among the feasible region analysis methods, the point domain analysis method can be used to describe the feasible region space under new energy, but there are problems such as difficulty in obtaining its analytical form, and with the increase of the power grid scale, its calculation time consumption increases sharply. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a power grid dispatching method, device and medium considering the uncertainty of wind power, which can quickly and accurately calculate the system feasible region under high wind power penetration and provide early warning information for power grid real-time dispatching.

[0005] An embodiment of the present invention provides a power grid dispatching method considering the uncertainty of wind power, including:

[0006] Construct an opportunity-constrained model based on AC power flow according to the uncertainty of wind power;

[0007] Analyze the opportunity-constrained model by using the sensitivity factor and the first-order Taylor expansion, and transform the opportunity constraint in the opportunity-constrained model into a linear constraint;

[0008] According to the linear constraint, use an improved double-description method to calculate the original feasible region of the power grid;

[0009] Based on the original feasible region, construct a local non-linear adjustment and optimization model, and use a deep learning method to solve the local non-linear adjustment and optimization model to obtain the final feasible region, so as to perform power grid dispatching according to the feasible region.

[0010] As an improvement of the above solution, constructing an opportunity-constrained model based on AC power flow according to the wind power uncertainty includes:

[0011] Modeling the new energy power as the sum of the ultra-short-term point prediction value of wind power and the wind power uncertainty deviation;

[0012] Based on the wind power uncertainty deviation, constructing the generator power balance constraint, the node voltage constraint, and the branch power flow constraint;

[0013] Expressing the generator power balance constraint, the node voltage constraint, and the branch power flow constraint as opportunity constraints to obtain the opportunity-constrained model.

[0014] As an improvement of the above solution, analyzing the opportunity-constrained model by using the sensitivity factor and the first-order Taylor expansion, and transforming the opportunity constraints in the opportunity-constrained model into linear constraints, including:

[0015] Determining the sensitivity factor with respect to the wind power uncertainty deviation according to the generator power, voltage, and branch power flow;

[0016] According to the sensitivity factor, re-expressing the opportunity-constrained model to obtain the opportunity constraint containing the sensitivity factor;

[0017] Using the first-order Taylor expansion to analyze the opportunity constraint containing the sensitivity factor to obtain the linear constraint.

[0018] As an improvement of the above solution, the sensitivity factor specifically includes:

[0019]

[0020]

[0021] Among them, and respectively represent the sensitivity factors of the active power injection and the reactive power injection of the generator; Γ v is the sensitivity factor of the voltage; and respectively represent the sensitivity factors of the active power flow and the reactive power flow of the branch. ω is the wind power uncertainty deviation, p G is the active power injection of the generator, q G is the reactive power injection of the generator, v is the node voltage, p L is the active power flow of the branch, and q L is the reactive power flow of the branch.

[0022] As an improvement of the above solution, according to the linear constraint, using the improved double-description method to calculate and obtain the original feasible region of the power grid, including:

[0023] For the problem of vertex description of a polyhedron with the given linear constraints and the feasible region to be solved, transform the problem of vertex description of the polyhedron into a corresponding polyhedral cone problem;

[0024] Initialize the vertex subset, extreme ray set, and adjacency pair set;

[0025] By gradually adding the inequalities in the linear constraints, update the adjacency pairs and vertices; the update of the adjacency pairs includes: traversing each extreme ray in the geometry, performing an adjacency test on each extreme ray and vertex to screen and update the adjacency pairs;

[0026] Delete the pairs in the adjacency pairs that no longer generate new rays to obtain the final set of polyhedron vertices and determine the original feasible region of the power grid.

[0027] As an improvement to the above solution, construct a local non - linear adjustment and optimization model based on the original feasible region, including:

[0028] Based on the original feasible region, calculate the geometric center of the polyhedron corresponding to the original feasible region; assume the polyhedron is n - dimensional and has m vertices, where the coordinate of each vertex u i is an n - dimensional vector u i =(u i1 ,u i2 ,…,u in ), then the coordinate of the geometric center C=(C 1 ,C 2 ,…,C n ) is:

[0029]

[0030] Take N points u old on each edge to construct a local non - linear adjustment and optimization model, specifically:

[0031] min‖u new -u old ‖

[0032] s.t. the linear constraint (21)

[0033] where, u new represents the vertex to be solved.

[0034] As an improvement to the above solution, use a deep learning method to solve the local non - linear adjustment and optimization model to obtain the final feasible region, including:

[0035] Use a deep learning method based on the optimized neural network ONN to batch - solve the local non - linear adjustment and optimization model to obtain new vertices of the feasible region, which form a new feasible region; the ONN model is defined as follows:

[0036] y(t; w) = y 0 + (1 - e -t ) N(t; w) (22)

[0037] where is the initial solution, is the time range, t ∈ [0, T] is the time variable; 1 - e -t is used to ensure that the ONN model always satisfies the initial condition y(0; w) = y 0 ;

[0038] The endpoints of the state solution of the ONN model are approximations of the optimal solution of the local non - linear adjustment optimization model, i.e.:

[0039]

[0040] Combining equations (22) - (23) gives:

[0041] P(y(T; w)) ≈ y * (24)

[0042] where P(·) is the projection function that projects the endpoints onto the feasible set, and equation (24) means that the endpoints of the ONN model together with the projection function are approximate solutions of the optimal solution of the local non - linear adjustment optimization model;

[0043] The loss function of the ONN model is:

[0044]

[0045] where φ(·) is the ordinary differential equation corresponding to the local non - linear adjustment optimization model, e -γ*t is used to re - distribute the weights in the loss and assign higher weights to times closer to the origin, and γ is the weighting hyperparameter;

[0046] The objective function of the ONN model is:

[0047]

[0048] where the objective function E(w) is the integral of the loss function over the time range [0, T], and the loss function L(T, w) represents the error of the ONN model at time t;

[0049] The model is trained by minimizing the batch loss:

[0050]

[0051] where is a set of times t randomly sampled from [0, T], represents the set size;

[0052] After obtaining the new feasible region, calculate its geometric center again, and perform iterative loops based on the local non-linear adjustment optimization model and the ONN model until the preset accuracy requirement is met, and the final feasible region is obtained.

[0053] As an improvement of the above solution, after obtaining the final feasible region, the method further includes:

[0054] Use the Hausdorff distance, Minkowski sum and volume ratio, Jaccard index, and mean square error of the support function to evaluate the final feasible region respectively, and obtain the evaluation result.

[0055] An embodiment of the present invention also provides a power grid scheduling device considering wind power uncertainty, including:

[0056] An opportunity constraint construction module, configured to construct an opportunity constraint model based on AC power flow according to wind power uncertainty;

[0057] A linear constraint conversion module, configured to analyze the opportunity constraint model by using a sensitivity factor and first-order Taylor expansion, and convert the opportunity constraint in the opportunity constraint model into a linear constraint;

[0058] A feasible region calculation module, configured to calculate the original feasible region of the power grid by using an improved double description method according to the linear constraint;

[0059] A feasible region optimization model, configured to construct a local non-linear adjustment optimization model based on the original feasible region, and use a deep learning method to solve the local non-linear adjustment optimization model to obtain the final feasible region, so as to perform power grid scheduling according to the feasible region.

[0060] An embodiment of the present invention also provides a computer-readable storage medium, which stores a computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the power grid scheduling method considering wind power uncertainty described in any one of the above.

[0061] Compared with the prior art, the beneficial effects of a power grid scheduling method, device and medium considering wind power uncertainty provided by the embodiments of the present invention are as follows: Through the chance-constraint analytical method based on the sensitivity factor and the first-order Taylor expansion, the chance constraint can be maximally analyzed while considering the wind power uncertainty and ensuring no over-limit, and the chance constraint is transformed into a directly processable manner; By using the improved double-description method to solve the vertices of the feasible region, the vertex calculation problem in the case of large-scale high dimensions can be solved, and the overall complexity of the algorithm is reduced, thereby improving the calculation efficiency; By using the ONN model, the large-scale non-linear optimization model can be quickly and accurately solved, further improving the accuracy and calculation efficiency. The embodiments of the present invention can quickly calculate the vertices of the feasible region on the basis of ensuring the safety of the operating points within the feasible region, and can meet the requirements of real-time power grid scheduling. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 FIG. is a schematic flow chart of a power grid scheduling method considering wind power uncertainty provided by an embodiment of the present invention;

[0063] Figure 2 FIG. is a schematic structural diagram of an improved IEEE case5 system provided by an embodiment of the present invention;

[0064] Figure 3 FIG. is a schematic diagram of the calculation results of the improved double-description method provided by an embodiment of the present invention;

[0065] Figure 4 FIG. is a schematic structural diagram of a power grid scheduling device considering wind power uncertainty provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0066] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0067] Please refer to Figure 1 , Figure 1 FIG. is a schematic flow chart of a power grid scheduling method considering wind power uncertainty provided by an embodiment of the present invention. The power grid scheduling method considering wind power uncertainty includes:

[0068] S1: Construct an opportunity-constraint model based on the AC power flow according to the wind power uncertainty;

[0069] As an optional embodiment, the constructing an opportunity-constraint model based on the AC power flow according to the wind power uncertainty includes:

[0070] Model the new - energy power as the sum of the ultra - short - term point prediction value of wind power and the deviation of wind - power uncertainty;

[0071] Based on the wind - power uncertainty deviation, construct the generator power balance constraint, the node voltage constraint, and the branch power flow constraint;

[0072] Express the generator power balance constraint, the node voltage constraint, and the branch power flow constraint as chance constraints to obtain a chance - constraint model.

[0073] Specifically, without loss of generality, assume that each node has a conventional generator, a new - energy power plant, and a load. The uncertainty comes from the fluctuation of new - energy output. The conventional generator and the new - energy power plant are collectively referred to as the power - generation side, and the node where they are located is denoted as The node with uncertainty is denoted as The node set and branch set of this power system are respectively

[0074] Since the nodes are divided into three types: PQ nodes, PV nodes, and balance nodes (i.e., θV nodes), the variables related to PQ nodes in the following text are all represented by the subscript PQ, and the same applies to the others. The active power P and reactive power Q of PQ nodes are given, and their node voltages and phases (V, δ) are the quantities to be solved; for PV nodes, the node active power injection and voltage amplitude are given, and their reactive power injection and voltage phase angles are the quantities to be solved; for balance nodes, the voltage amplitude and phase are given, and their injected active power and reactive power are the quantities to be solved.

[0075] Model the new - energy power as the sum of the deterministic prediction value and the deviation:

[0076]

[0077] where, p wind is the ultra - short - term point prediction value of wind power, ω is the wind - power uncertainty deviation, which is the source of uncertainty, represents the wind power considering wind - power uncertainty.

[0078] The covariance matrix of wind - power uncertainty is Do not assume that wind - power uncertainties are independent, and allow the non - diagonal elements of ∑ to be non - zero values. The total deviation caused by uncertainty is

[0079]

[0080] The standard deviation of the total deviation Ω is

[0081] Define the participation - factor vector α represents the contribution that each generator should make to cope with uncertainties, and it is required that to ensure that any given uncertainty deviation Ω can obtain the same

[0082] (1) Generator power balance constraint:

[0083] Assume that the change in active power caused by wind power uncertainty is unknown and is represented by and is balanced by the generators at the θV nodes. Then, for active power:

[0084]

[0085] Assume that after all generators connected to the PV and θV nodes are affected by wind power uncertainty, the uncertainty is The output of the generators connected to the PQ nodes remains unchanged and is deterministic. Then, for reactive power:

[0086]

[0087] (2) Node voltage constraint:

[0088] The voltage magnitude is fixed at the PV and θV nodes and changes with the uncertainty at the PQ nodes:

[0089]

[0090] (3) Branch power flow constraint:

[0091]

[0092] From the above analysis, the chance constraints that make up the feasible region can be expressed as:

[0093]

[0094]

[0095] θ θV = 0(11 - 12)

[0096]

[0097] where γ is called the power ratio, and ε represents the quantile. In constraint (11 - 1), f(·) is the source-load balance function, is the quantity to be solved for the uncertainty of the node voltage phase angle, is the quantity to be solved for the uncertainty of the node voltage magnitude, is the quantity to be solved for the uncertainty of the branch active power flow, is the substitution variable for the uncertainty of branch reactive power flow; Constraint (11-1) is the system source-load power balance constraint; Equations (11-2) and (11-3) are active power constraints, and Equations (11-4) and (11-5) are reactive power constraints; Equations (11-6) and (11-7) are voltage constraints; Equations (11-8) and (11-9) are branch active power flow constraints; Equations (11-10) and (11-11) are branch reactive power flow constraints; Equation (11-12) indicates that the voltage phase angle of the slack bus is 0; Equations (11-13) and (11-14) are the active power and reactive power injected into the node, q G is the reactive power injected by the node generator, q D is the reactive power injected by the node load. The current forms of the above chance constraints cannot be directly processed because the wind power uncertainty set is not enumerable. Therefore, it is impossible to perform the calculations of Equations (11-1) to (11-12) for all ω. Therefore, the chance constraints need to be reformulated into tractable constraints.

[0098] S2: Analyze the chance constraint model using the sensitivity factor and the first-order Taylor expansion, and transform the chance constraints in the chance constraint model into linear constraints;

[0099] As one of the optional embodiments, the analyzing the chance constraint model using the sensitivity factor and the first-order Taylor expansion and transforming the chance constraints in the chance constraint model into linear constraints includes:

[0100] Determine the sensitivity factor with respect to the wind power uncertainty deviation according to the generator power, voltage, and branch power flow;

[0101] According to the sensitivity factor, re-represent the chance constraint model to obtain a chance constraint containing the sensitivity factor;

[0102] Use the first-order Taylor expansion to analyze the chance constraint containing the sensitivity factor to obtain a linear constraint.

[0103] Furthermore, the sensitivity factor specifically includes:

[0104]

[0105] Among them, and respectively represent the sensitivity factors of the active power and reactive power injected by the generator; Γ v is the sensitivity factor of the voltage; and respectively represent the sensitivity factors of the branch active power flow and reactive power flow, ω is the wind power uncertainty deviation, p G is the active power injected by the generator, qG Inject reactive power into the generator, v is the node voltage, p L is the active power flow of the branch, q L is the reactive power flow of the branch.

[0106] Specifically, the impact of wind power uncertainty ω is explained using a linearization method based on the first-order Taylor expansion of the day-ahead operating point. Based on this, the sensitivity factor with respect to uncertainty ω is defined. The sensitivity factor depends non-linearly on the day-ahead operating point (θ, v, p, q) and linearly on α, γ. At this time, the generator power balance constraint, the node voltage constraint, and the branch power flow constraint can be approximately expressed as:

[0107]

[0108] From equations (13) - (17), the chance constraint in equation (11) can be rewritten as:

[0109]

[0110] Among them, the active power chance constraint of the θV node depends on The reactive power chance constraints of the PV and θV nodes depend on The reactive power chance constraint of the PQ node is determined because the injected power is fixed. Similarly, the chance constraint of the PQ node voltage depends on Γ v , while those of the PV and θV nodes are determined.

[0111] At this time, according to the first-order Taylor expansion, the chance constraint in equation (18) can be analytically expressed as the following linear constraints:

[0112]

[0113]

[0114] S3: According to the linear constraints, use the improved double-description method to calculate the original feasible region of the power grid;

[0115] As one of the optional embodiments, the step of calculating the original feasible region of the power grid according to the linear constraints by using the improved double-description method includes:

[0116] The linear constraints and the problem of describing the vertices of the feasible region to be solved are polyhedra. Transform the problem of describing the vertices of the polyhedron into a corresponding polyhedral cone problem;

[0117] Initialize the vertex subset, extreme ray set, and adjacency pair set;

[0118] Update the adjacent pairs and vertices by gradually adding the inequalities in the linear constraints; the update of the adjacent pairs includes: traversing each extreme ray in the geometry, and performing an adjacency test on each extreme ray and the vertices to filter and update the adjacent pairs.

[0119] Delete the pairs in the adjacent pairs that no longer generate new rays to obtain the final set of polyhedron vertices and determine the original feasible region of the power grid.

[0120] Specifically, the polyhedron composed of the above linear inequalities, that is, Equation (19), can be uniformly expressed as: where x represents the polyhedron vertices. Ax ≤ b is the unified mathematical expression of all inequalities (19), A can represent the coefficients of all variables to be solved in Equation (19), x represents all variables to be solved, and b represents all constant terms.

[0121] Since the designed algorithm is for polyhedral cones, and the inequality constraints and the feasible region to be solved are construction problems described by the vertices of the polyhedron, it is necessary to transform the vertex description problem of the polyhedron into the corresponding polyhedral cone problem. Specifically, the vertices of a polyhedron can be found by transforming it into a polyhedral cone problem and then set x d+1 = 1.

[0122] In the double description method, the vertex set needs to be updated in each iteration. A key step is to check the adjacency of the extreme rays on the cone, which involves determining whether two extreme rays share a face of the polyhedron. The adjacency test includes an algebraic test and a combinatorial test. The algebraic test means that when and only when the rank of the joint tensor of two extreme rays u 1 , u 2 is , u 1 , u 2 are adjacent in the set U(C) of extreme rays of the polyhedral cone; the combinatorial test means that for the extreme rays when and only when is 1 , u 2 are adjacent in the set U(C) of extreme rays of the polyhedral cone.

[0123] Condition 1: If {u 1 , u 2} is an edge in G, then |Z(u 1 ) ∩ Z(u 2 )| ≥ r - 2.

[0124] Let G be an undirected graph formed by the polyhedral cone C, and its vertex set is the set U of extreme rays of C. When and only when |Z(u 1 ) ∩ Z(u 2)|≥r - 2, {u 1 , u 2} forms an edge. The edge set of G is denoted as E(G). Condition 1 is a necessary condition for the adjacency of extreme rays. This is a simplified adjacency screening condition used for a quick preliminary screening to significantly reduce the pairs of extreme rays that need to be verified and avoid redundant calculations.

[0125] Condition 2: For vertex u in the adjacency graph G of the extreme rays of the polyhedral cone 1 , u 2 , if and only if there does not exist another vertex in G such that {u 1 , w} and {u 2 , w} are both edges of G, and when, {u 1 , u 2} is an edge in G.

[0126] Condition 2 is a sufficient condition for the adjacency of extreme rays. After the preliminary screening, it determines the final adjacency in a more precise way. According to Condition 2, if two rays are adjacent in G, then they are truly adjacent if and only if there does not exist another ray w such that that is, no other ray can "be sandwiched" between these two rays.

[0127] Furthermore, an improved double description method is adopted to determine the vertex set of the polyhedron. The specific steps include:

[0128] S31: Take the inequality coefficient matrix of the polyhedron and the constant term of the inequality as the input;

[0129] S32: According to the input, construct the matrix A′ = [A, -b] and the variable x′ = [x T , y] T , and transform the vertex description problem of the polyhedron into a corresponding polyhedral cone problem;

[0130] S33: Set the basis matrix U = I d+！ , select the initial subset such that |K| = d (rank(A) = d) and det(A k ) ≠ 0;

[0131] S34: Construct the set of extreme rays of the initialized subset Bx ≥ 0 (B is a sub - matrix of A determined by K);

[0132] S35: Initialize the adjacency pair set E = {};

[0133] S36: Gradually add inequalities, update the extreme points and adjacency pairs until K ≠ {1, 2, …, m}:

[0134] S361: Select a new inequality \(i\in K\) and construct the set

[0135] S362: Initialize the set of extreme rays \(U\) +- =\(\{\}\);

[0136] S363: Traverse each extreme ray \(u\) in \(U^+\) 1 \(\in U\) + , and for each extreme ray \(u\) in \(U^-\) 2 \(\in U\) - and \(u\) 1 perform an adjacency test: Screen using the above Conditions 1 and 2 to calculate a new extreme ray and then update \(U\) +- =\(U\) +- \(\cup\{w\}\), update the set of adjacent pairs \(E = E\cup\{\{u\) 1 ,w\},\{\{u\) 2 ,w\}\};

[0137] S364: Update the set of vertices \(U = U\) + \(\cup U\) +- \(\cup U\) 0

[0138] S365: Delete the pairs in the adjacent pairs that no longer generate new rays and remove the processed inequalities;

[0139] S37: Output the set of polyhedron vertices \(V\), which is the original feasible region of the power grid.

[0140] S4: Based on the original feasible region, construct a local non - linear adjustment optimization model and use a deep learning method to solve the local non - linear adjustment optimization model to obtain the final feasible region for power grid scheduling according to the feasible region.

[0141] Specifically, first calculate the geometric center of the original high - dimensional polyhedron feasible region.

[0142] The constructing the local non - linear adjustment optimization model based on the original feasible region includes:

[0143] Based on the original feasible region, calculate the geometric center of the polyhedron corresponding to the original feasible region; Suppose the polyhedron is \(n\) - dimensional and has \(m\) vertices, where the coordinates of each vertex \(u\) i is an \(n\) - dimensional vector \(u\) i =\((u\) i1 ,u\) i2 ,\(\cdots,u\) in ), then the coordinates of the geometric center \(C=(C\) 1 ,C\) 2 ,\(\cdots,C\)n ) is as follows:

[0144]

[0145] Take N points u on each edge old , and construct a local non-linear adjustment and optimization model, specifically as follows:

[0146] min‖u new -u old ‖ (21)

[0147] s.t. The linear constraints of Equation (19)

[0148] where u new represents the vertex to be solved.

[0149] Furthermore, the use of the deep learning method to solve the local non-linear adjustment and optimization model to obtain the final feasible region includes:

[0150] Use the deep learning method based on the Optimization-Neural Networks (ONN) to batch solve the local non-linear adjustment and optimization model to obtain the vertices of the new feasible region, which form a new feasible region; The ONN model is defined as follows:

[0151] y(t; w) = y 0 +(1 - e -t )N(t; w) (22)

[0152] where is the initial solution, is the time range, t ∈ [0, T] is the time variable; 1 - e -t is used to ensure that the ONN model always satisfies the initial condition y(0; w) = y 0 ;

[0153] The endpoints of the state solution of the ONN model are approximations of the optimal solution of the local non-linear adjustment and optimization model, that is:

[0154]

[0155] Combining Equations (22)-(23) gives:

[0156] P(y(T; w)) ≈ y * (24)

[0157] where P(·) is the projection function that projects the endpoints onto the feasible set, and Equation (24) indicates that the endpoints of the ONN model together with the projection function are approximate solutions of the optimal solution of the local non-linear adjustment and optimization model;

[0158] The loss function of the ONN model is as follows:

[0159]

[0160] where φ(·) is the ordinary differential equation corresponding to the local non-linear adjustment and optimization model, e -γ*t is used to reallocate the weights in the loss and assign higher weights to times closer to the origin, and γ is a weighting hyperparameter;

[0161] The objective function of the ONN model is as follows:

[0162]

[0163] where the objective function E(w) is the integral of the loss function over the time range [0, T], and the loss function L(T, w) represents the error of the ONN model at time t;

[0164] The model is trained by minimizing the batch loss:

[0165]

[0166] where is a set of times randomly sampled from [0, T], represents the set size;

[0167] After obtaining the new feasible region, calculate its geometric center again, and perform cyclic iteration based on the local non-linear adjustment and optimization model and the ONN model until the preset accuracy requirement is met, and the final feasible region is obtained.

[0168] As an optional embodiment, after obtaining the final feasible region, the method further includes:

[0169] Evaluate the final feasible region respectively using the Hausdorff distance, Minkowski sum and volume ratio, Jaccard index, and mean square error of the support function to obtain an evaluation result.

[0170] where the support function is used to describe the maximum extension distance of a polyhedron in a given direction. Given a polyhedron A and a direction vector λ, the support function h(A, λ) is defined as:

[0171]

[0172] The support function h(A, λ) calculates the projection distance of the farthest point of the polyhedron A in the given direction λ. To compare two polyhedrons A and B, calculate their support function values in multiple different directions and measure the similarity using the mean square error. The smaller the difference of the function, the more similar the two polyhedrons are.

[0173] In a specific example, the IEEE case5 system is used for experiments, and the system source-load information is as follows: The wind turbine is connected to bus B2 with an output of 300 MW. The power grid topology is as Figure 2 shown, where W represents the wind turbine unit, and B1 - B5 represent nodes 1 - 5 respectively. Calculations are performed according to the method of the embodiment of the present invention, and a total of 47 vertices are obtained. The calculation result of the feasible region is as Figure 3 shown, with its Hausdorff distance being 138.97, the Minkowski sum and volume ratio being 4.03, and the Jaccard index being 0.91, meeting the requirements for rapid solution of the power grid feasible region, being able to provide reliable information for power grid dispatching, and effectively improving the efficiency and accuracy of power grid dispatching considering wind power uncertainty.

[0174] The embodiment of the present invention can, through the analytical method of chance constraints based on sensitivity factors and first-order Taylor expansion, maximize the analytical chance constraints while considering wind power uncertainty and ensuring no violation, and transform the chance constraints into a directly processable form; by using the improved double-description method to solve the vertices of the feasible region, it can handle the vertex calculation problem in large-scale high-dimensional cases, reduce the overall complexity of the algorithm, and thus improve the calculation efficiency; by using the ONN model, it can quickly and accurately solve large-scale non-linear optimization models, further improving the accuracy and calculation efficiency. The embodiment of the present invention can quickly calculate the vertices of the feasible region on the basis of ensuring 99.99% safety of the operating points within the feasible region, and can meet the requirements of real-time power grid dispatching.

[0175] Correspondingly, the present invention also provides a power grid dispatching device considering wind power uncertainty, which can implement all processes of the power grid dispatching method considering wind power uncertainty in the above embodiment.

[0176] Please refer to Figure 4 , Figure 4 which is a schematic structural diagram of a power grid dispatching device considering wind power uncertainty provided by the embodiment of the present invention. The power grid dispatching device considering wind power uncertainty includes:

[0177] A chance constraint construction module 401, configured to construct an AC power flow-based chance constraint model according to wind power uncertainty;

[0178] A linear constraint transformation module 402, configured to analyze the chance constraint model by using sensitivity factors and first-order Taylor expansion, and transform the chance constraints in the chance constraint model into linear constraints;

[0179] A feasible region calculation module 403, configured to calculate the original feasible region of the power grid by using the improved double-description method according to the linear constraints;

[0180] The feasible region optimization model 404 is used to construct a local non-linear adjustment optimization model based on the original feasible region, and solve the local non-linear adjustment optimization model by using a deep learning method to obtain the final feasible region, so as to perform power grid scheduling according to the feasible region.

[0181] Preferably, constructing an opportunity-constrained model based on AC power flow according to wind power uncertainty includes:

[0182] Modeling the new energy power as the sum of the ultra-short-term point prediction value of wind power and the wind power uncertainty deviation;

[0183] Based on the wind power uncertainty deviation, construct the generator power balance constraint, the node voltage constraint and the branch power flow constraint;

[0184] Express the generator power balance constraint, the node voltage constraint and the branch power flow constraint as opportunity constraints to obtain an opportunity-constrained model.

[0185] Preferably, analyzing the opportunity-constrained model by using the sensitivity factor and the first-order Taylor expansion, and converting the opportunity constraints in the opportunity-constrained model into linear constraints, includes:

[0186] Determine the sensitivity factor of the wind power uncertainty deviation according to the generator power, voltage and branch power flow;

[0187] According to the sensitivity factor, re-express the opportunity-constrained model to obtain an opportunity constraint containing the sensitivity factor;

[0188] Use the first-order Taylor expansion to analyze the opportunity constraint containing the sensitivity factor to obtain a linear constraint.

[0189] Preferably, the sensitivity factor specifically includes:

[0190]

[0191] Among them, Γ pG and Γ qG respectively represent the sensitivity factors of the active power injection and the reactive power injection of the generator; Γ v is the sensitivity factor of the voltage; Γ pL and Γ qL respectively represent the sensitivity factors of the active power flow and the reactive power flow of the branch, ω is the wind power uncertainty deviation, p G is the active power injection of the generator, q G is the reactive power injection of the generator, v is the node voltage, p L is the active power flow of the branch, q L is the reactive power flow of the branch.

[0192] Preferably, according to the linear constraints, an improved double-description method is adopted to calculate the original feasible region of the power grid, including:

[0193] The linear constraint and the problem of vertex description of the feasible region to be solved are polyhedrons. The problem of vertex description of the polyhedron is transformed into a corresponding polyhedral cone problem.

[0194] Initialize the vertex subset, extreme ray set, and adjacency pair set.

[0195] By gradually adding the inequalities in the linear constraint, update the adjacency pair and vertex. The update of the adjacency pair includes: traversing each extreme ray in the geometry, and performing an adjacency test on each extreme ray and vertex to screen and update the adjacency pair.

[0196] Delete the pairs in the adjacency pair that no longer generate new rays to obtain the final set of polyhedron vertices and determine the original feasible region of the power grid.

[0197] Preferably, based on the original feasible region, a local non-linear adjustment and optimization model is constructed, including:

[0198] Based on the original feasible region, calculate the geometric center of the corresponding polyhedron of the original feasible region. Assume the polyhedron is n-dimensional and has m vertices, where each vertex u i 's coordinate is an n-dimensional vector u i =(u i1 ,u i2 ,…,u in ), then the coordinate of the geometric center C=(C 1 ,C 2 ,…,C n ) is:

[0199]

[0200] Take N points u old on each edge to construct a local non-linear adjustment and optimization model, specifically:

[0201] min‖u new -u old ‖

[0202] s.t. the linear constraint (21)

[0203] where, u new represents the vertex to be solved.

[0204] Preferably, the deep learning method is used to solve the local non-linear adjustment and optimization model to obtain the final feasible region, including:

[0205] A deep learning method based on the optimized neural network ONN is used to batch solve the local non - linear adjustment optimization model to obtain new vertices of the feasible region, which form a new feasible region; the ONN model is defined as follows:

[0206] y(t; w) = y 0 +(1 - e -t )N(t; w) (22)

[0207] Among them, is the initial solution, is the time range, t ∈ [0, T] is the time variable; 1 - e -t is used to ensure that the ONN model always satisfies the initial condition y(0; w) = y 0 ;

[0208] The endpoints of the state solution of the ONN model are approximations of the optimal solution of the local non - linear adjustment optimization model, that is:

[0209]

[0210] Combined with equations (22) - (23), we get:

[0211] P(y(T; w)) ≈ y * (24)

[0212] Among them, P(·) is the projection function that projects the endpoints onto the feasible set, and equation (24) indicates that the endpoints of the ONN model together with the projection function are approximate solutions of the optimal solution of the local non - linear adjustment optimization model;

[0213] The loss function of the ONN model is:

[0214]

[0215] Among them, φ(·) is the ordinary differential equation corresponding to the local non - linear adjustment optimization model, e -γ*t is used to re - distribute the weights in the loss and assign higher weights to the times close to the origin, and γ is the weighted hyperparameter;

[0216] The objective function of the ONN model is:

[0217]

[0218] Among them, the objective function E(w) is the integral of the loss function over the time range [0, T], and the loss function L(T, w) represents the error of the ONN model at time t;

[0219] The model is trained by minimizing the batch loss:

[0220]

[0221] wherein, is a set of times t randomly selected from [0, T], represents the set size;

[0222] After obtaining the new feasible region, calculate its geometric center again, and perform cyclic iteration based on the local non-linear adjustment optimization model and the ONN model until the preset accuracy requirement is met, and the final feasible region is obtained.

[0223] Preferably, the power grid scheduling device considering wind power uncertainty is further configured to:

[0224] Use the Hausdorff distance, Minkowski sum and volume ratio, Jaccard index, and mean square error of the support function to evaluate the final feasible region respectively, and obtain the evaluation result.

[0225] In specific implementation, the working principle, control process, and achieved technical effects of the power grid scheduling device considering wind power uncertainty provided by the embodiments of the present invention are the same as those of the power grid scheduling method considering wind power uncertainty in the above embodiments, and will not be elaborated here.

[0226] The embodiments of the present invention further provide a computer-readable storage medium, which includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the power grid scheduling method considering wind power uncertainty described in any one of the above embodiments.

[0227] The embodiments of the present invention provide a power grid scheduling method, device, and medium considering wind power uncertainty, and the beneficial effects thereof are as follows: Through the chance constraint analytical method based on the sensitivity factor and the first-order Taylor expansion, the chance constraint can be maximally analyzed while considering wind power uncertainty and ensuring no over-limit, and the chance constraint is transformed into a directly processable form; By using the improved double-description method to solve the vertices of the feasible region, the problem of vertex calculation in large-scale high-dimensional cases can be addressed, reducing the overall complexity of the algorithm and thus improving the calculation efficiency; By using the ONN model, the large-scale non-linear optimization model can be solved quickly and accurately, further improving the accuracy and calculation efficiency. The embodiments of the present invention can quickly calculate the vertices of the feasible region on the basis of ensuring the safety of the operating points within the feasible region, and can meet the requirements of real-time power grid scheduling.

[0228] The above is the preferred implementation manner of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present invention.

Claims

1. A power grid dispatching method considering wind power uncertainty, characterized in that: include: According to the uncertainty of wind power, a chance constraint model based on AC power flow is constructed; The opportunity constraint model is analyzed by using sensitivity factor and first-order Taylor expansion, and the opportunity constraint in the opportunity constraint model is converted into a linear constraint; According to the linear constraints, the original feasible region of the power grid is calculated by adopting the improved double description method; Based on the original feasible domain, a local nonlinear adjustment optimization model is constructed, and a deep learning method is used to solve the local nonlinear adjustment optimization model to obtain a final feasible domain, so as to perform power grid dispatch according to the feasible domain.

2. The grid dispatching method considering wind power uncertainty according to claim 1, characterized in that: According to the uncertainty of wind power, an opportunity constraint model based on AC power flow is constructed, including: The renewable energy power is modeled as the sum of the ultra-short-term point forecast value of wind power and the uncertainty deviation of wind power; Based on the wind power uncertainty deviation, generator power balance constraints, node voltage constraints and branch power flow constraints are constructed; The generator power balance constraint, the node voltage constraint and the branch power flow constraint are expressed as chance constraints to obtain a chance constraint model.

3. The grid dispatching method considering wind power uncertainty according to claim 1, characterized in that: The method of analyzing the opportunity constraint model by using the sensitivity factor and the first-order Taylor expansion to convert the opportunity constraint in the opportunity constraint model into a linear constraint includes: Determine the sensitivity factor of wind power uncertainty deviation based on generator power, voltage and branch power flow; According to the sensitivity factor, the opportunity constraint model is re-expressed to obtain the opportunity constraint containing the sensitivity factor; The first-order Taylor expansion is used to analyze the chance constraint containing the sensitivity factor to obtain a linear constraint.

4. The grid dispatching method considering wind power uncertainty according to claim 3, characterized in that: The sensitivity factors specifically include: Among them, Γ pG With Γ qG Respectively represent the sensitivity factors of the generator injected active power and injected reactive power; Γ v is the voltage sensitivity factor; Γ pL With Γ qL They represent the sensitivity factors of active power flow and reactive power flow of the branch, ω is the uncertainty deviation of wind power, and p G Inject active power into the generator, q G Inject reactive power into the generator, v is the node voltage, p L is the branch active power flow, q L It is the branch reactive power flow.

5. The grid dispatching method considering wind power uncertainty according to claim 1, characterized in that: The improved dual description method is used according to the linear constraints to calculate the original feasible region of the power grid, including: The linear constraints and the feasible domain to be solved are a polyhedron vertex description problem, and the polyhedron vertex description problem is converted into a corresponding polyhedral cone problem; Initialize the vertex subset, polar ray set, and adjacency pair set; By gradually adding inequalities in the linear constraints, the adjacency pairs and vertices are updated; the updating of the adjacency pairs includes: traversing each polar ray in the geometry, performing an adjacency test on each polar ray and vertex to screen and update the adjacency pairs; Delete the adjacent pairs that no longer generate new rays to obtain the final set of polyhedral vertices and determine the original feasible domain of the power grid.

6. The grid dispatching method considering wind power uncertainty according to claim 5, characterized in that: The constructing of a local nonlinear adjustment optimization model based on the original feasible region includes: Based on the original feasible domain, the geometric center of the polyhedron corresponding to the original feasible domain is calculated; assuming that the polyhedron is n-dimensional and has m vertices, where each vertex u i The coordinates of is an n-dimensional vector u i =(u i1 ,u i2 ,…,u in ), then the coordinates of the geometric center C=(C1,C2,…,C n )for: Take N points u on each edge old , construct a local nonlinear adjustment optimization model, specifically: min‖u new -in old ‖ st linear constraints Among them, u new Represents the vertex to be solved.

7. The grid dispatching method considering wind power uncertainty according to claim 6, characterized in that: The deep learning method is used to solve the local nonlinear adjustment optimization model to obtain the final feasible domain, including: The local nonlinear adjustment optimization model is solved in batches by a deep learning method based on an optimization neural network ONN to obtain new feasible domain vertices and form a new feasible domain. The ONN model is defined as follows: y(t;w)=y0+(1-e -t )N(t;w) (22) in, is the initial solution, is the time range, t∈[0,T] is the time variable; 1-e -t Used to ensure that the ONN model always meets the initial condition y(0;w)=y0; The endpoint of the state solution of the ONN model is an approximation of the optimal solution of the local nonlinear adjustment optimization model, namely: Combining equations (22)-(23) we get: P(y(T;w))≈y * (24) Wherein, P(·) is the projection function that projects the endpoint onto the feasible set, and formula (24) indicates that the endpoint of the ONN model together with the projection function is an approximate solution to the optimal solution of the local nonlinear adjustment optimization model; The loss function of the ONN model is: Where φ(·) is the ordinary differential equation corresponding to the local nonlinear adjustment optimization model, e -γ*t Used to redistribute the weights in the loss and give higher weights to times close to the origin. γ is a weighted hyperparameter. The objective function of the ONN model is: Among them, the objective function E(w) is the integral of the loss function over the time range [0,T], and the loss function L(T,w) represents the error of the ONN model at time t; Train the model by minimizing the batch loss: in, is a set of times t randomly drawn from [0,T], Indicates the size of the collection; After obtaining the new feasible domain, its geometric center is calculated again, and a loop iteration is performed based on the local nonlinear adjustment optimization model and the ONN model until the preset accuracy requirement is met to obtain the final feasible domain.

8. The grid dispatching method considering wind power uncertainty according to claim 1, characterized in that: After obtaining the final feasible region, the method further includes: The final feasible region is evaluated by using Hausdorff distance, Minkowski sum and volume ratio, Jaccard index and mean square error of support function to obtain evaluation results.

9. A power grid dispatching device considering the uncertainty of wind power, characterized in that: include: The opportunity constraint building module is used to build an AC power flow-based opportunity constraint model according to wind power uncertainty; A linear constraint conversion module, used to analyze the opportunity constraint model by using a sensitivity factor and a first-order Taylor expansion, and convert the opportunity constraint in the opportunity constraint model into a linear constraint; A feasible region calculation module is used to calculate the original feasible region of the power grid using an improved dual description method according to the linear constraints; The feasible domain optimization model is used to construct a local nonlinear adjustment optimization model based on the original feasible domain, and adopt a deep learning method to solve the local nonlinear adjustment optimization model to obtain a final feasible domain, so as to perform power grid dispatch according to the feasible domain.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the device where the computer-readable storage medium is located executes the computer program, the power grid dispatching method considering wind power uncertainty as described in any one of claims 1 to 8 is implemented.

Images