A power grid dispatching method, device and storage medium based on elastic computing resources

Through the grid scheduling method based on elastic computing resources, the high-dimensional convex hull economic operation domain is approximated by effective dimensions and extended relaxation technology, and combined with the large M preprocessing method, the problem of low solution efficiency of convex hull economic operation domain in power grid scheduling is solved. A more efficient grid scheduling calculation is achieved.

CN118589496BActive Publication Date: 2024-11-19ZHEJIANG ZHENENG TECHN RES INST CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411060503.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-11-19
Estimated Expiration
2044-08-05

AI Technical Summary

Technical Problem

When the existing power grid scheduling method deals with large-scale power grids, the solution efficiency of the convex hull economic operation domain is low, resulting in an increase in the computational time complexity. Especially when the number of dispatchable units and the number of scheduling periods is large, it is difficult for conventional methods to effectively deal with new energy. Determinism.

Method used

Using the grid scheduling method based on elastic computing resources, by establishing effective dimension determination method and extending slack method, the low-dimensional subconvex hull economic operation domain is used to approximate the high-dimensional convex hull economic operation domain, and improve the expansion point through the large M preprocessing method. Solve the speed and reduce the time complexity and total number of update convex hull calculations.

Benefits of technology

It significantly reduces the complexity and total number of update convex hull calculation times, speeds up the solution speed of the subconvex hull economic operation domain, is suitable for scenarios with a large number of parallel computing resources, and improves the robustness and computing efficiency of the scheduling plan.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118589496B_ABST
    Figure CN118589496B_ABST
Patent Text Reader

Abstract

The present application discloses a power grid dispatching method, device and storage medium based on elastic computing resources; it belongs to the field of power grid dispatching technology. The method includes establishing a method for determining the effective dimension of the economic operation domain of the power grid and a method for expanding and relaxing the convex hull economic operation domain considering the effective dimension, using multiple effective low-dimensional sub-convex hull economic operation domains to approximate the high-dimensional convex hull economic operation domain, and the expanded relaxation result is the expanded convex hull economic operation domain; establishing a method for solving the sub-convex hull economic operation domain considering elastic computing resources, realizing that each time multiple extension points are generated, an updated convex hull calculation is performed to speed up the solution of the sub-convex hull economic operation domain; establishing a large M preprocessing method, preprocessing the large number M contained in the sub-convex hull extension point solution model, and improving the solution speed of the extension point; obtaining the dispatching result. The method for solving the extended convex hull economic operation domain of the power grid in the present application can be decomposed into multiple sub-problems with low computational complexity and independent of each other, and is suitable for scenarios with a large number of parallel computing resources.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of power grid dispatching, and in particular to a power grid dispatching method, device and storage medium based on elastic computing resources. Background Art

[0002] The power grid dispatching center is the core organization for managing and operating the power system. Its main responsibility is to coordinate and optimize various power resources to meet the electricity needs of power users under the premise of ensuring the safe and stable operation of the power grid.

[0003] The power grid needs to meet a large number of safety constraints during actual operation, including branch power flow thermal stability limit constraints, unit ramp rate constraints, bus voltage range constraints, and unit start and stop constraints. Since the precise power flow model of the power grid has nonlinear and non-convex characteristics, power grid dispatching is actually an extremely complex problem. In order to improve the ability of the dispatching plan to cope with the uncertainty of new energy output, that is, to improve the robustness of the dispatching plan, scholars from all walks of life have conducted a lot of research in the field of dispatching plans based on uncertain optimization. At present, the commonly used uncertain optimization modeling methods mainly include stochastic optimization method, chance constrained optimization method, robust optimization method, and distributed robust optimization method.

[0004] When the scale of the power grid system is large and the number of dispatchable units is large, the overall solution speed of the conventional convex hull economic operation domain will slow down. There are three main reasons: When the number of dispatchable units increases, the spatial dimension of the corresponding convex hull economic operation domain will also increase, resulting in a significant increase in the time complexity of a single update of the convex hull calculation. In the worst case, the time complexity of updating the convex hull calculation is , where is the number of point sets involved in the calculation and is the spatial dimension. When the number of dispatchable units and the number of dispatch periods are large, the number of expansion points of the demand solution will also increase. Summary of the invention

[0005] The main technical problem solved by the present application is to provide a power grid dispatching method based on elastic computing resources to solve the problem of low efficiency in solving the conventional convex hull economic operation domain.

[0006] In order to solve the above technical problems, a technical solution adopted in this application is to provide a power grid dispatching method based on elastic computing resources, comprising the steps of:

[0007] Firstly, a method for determining the effective dimension of the economic operation domain of the power grid and an extended relaxation method for the convex hull economic operation domain considering the effective dimension are established. Multiple effective low-dimensional sub-convex hull economic operation domains are used to approximate the high-dimensional convex hull economic operation domain. The extended relaxation result is the extended convex hull economic operation domain.

[0008] Secondly, a method for solving the sub-convex hull economic operation domain considering elastic computing resources is established, so that the convex hull calculation is updated once each time multiple extension points are generated, thus speeding up the solution of the sub-convex hull economic operation domain.

[0009] Finally, a large M preprocessing method is established to preprocess the large number M contained in the sub-convex hull extension point solution model to improve the solution speed of the extension point; and the scheduling result is obtained.

[0010] Furthermore, the objective function of the sub-convex hull economic operation domain solution model is:

[0011]

[0012] Where: is the active output of generator g at time t; N T N is the total number of time periods for scheduling. T The time scale of each period can be set according to the needs of the dispatcher; G is the number of dispatchable generating units; N S is the number of energy storage systems; a g , b g and c g is the operating cost coefficient of generator g; and are the charging and discharging power of the energy storage system s at time t respectively; and are the penalty coefficients for the charging and discharging power of the energy storage system s respectively.

[0013] Furthermore, the constraints on the objective function in the sub-convex hull economic operation domain solution model are:

[0014] 1) Technical output constraints of dispatchable units

[0015]

[0016] Where: and are the minimum and maximum active power that generator g can output respectively;

[0017] 2) Climbing / slipping constraints

[0018]

[0019] Where: and are the maximum landslide and maximum climbing rates of generator g, respectively; is the initial value of the active output of generator g at time t=0;

[0020] 3) Constraints on abandoned power from new energy sources

[0021]

[0022] Where: Predict power for new energy sources; Abandoned power for new energy sources;

[0023] 4) Line flow constraints

[0024]

[0025] Where: f l max is the maximum transmission power of line l; is the active power of load d at time t; N W is the number of new energy units; N D is the load quantity; and are the power transfer distribution coefficients adapted to the dispatchable generating units, new energy generating units, energy storage systems and load nodes respectively;

[0026] 5) Energy storage system charging and discharging power constraints

[0027]

[0028] Where: and are the maximum charging and discharging power of the energy storage system s, respectively;

[0029] 6) Energy storage system power constraints

[0030]

[0031] Where: and are the charging and discharging efficiencies of the energy storage system s, respectively; e s,t is the storage capacity of energy storage system s at time t; is the capacity of the energy storage system s; Represents the initial power of the energy storage system s in the next scheduling cycle;

[0032] 7) Power balance constraints

[0033]

[0034] The optimal solution of the sub-convex hull economic operation domain solution model is the predicted value of the active output of the new energy unit w The corresponding optimal scheduling solution; the objective function is and When the sub-convex hull economic operation domain solves the model to obtain the optimal value, it will automatically satisfy the constraint of preventing the energy storage system from charging and discharging at the same time.

[0035] Furthermore, in order to simplify the model input, the sub-convex hull economic operation domain solution model is written in the following compact form;

[0036] min p G T Ap G +B T p G +C T x+D T ×1

[0037] E1p G +F1p w +G1x+H1≤0

[0038] E2p G +F2p W +G2 x +H2=0,

[0039] Where: p G Output of dispatchable generators The column vector composed of W Providing power for new energy units A column vector composed of e s,t and The matrix A, E1, E2, F1, F2, G1, G2 and the column vector B, C, H1, H2, D respectively represent the coefficient information corresponding to the sub-convex hull economic operation domain solution model; 1 is a column vector whose all elements are 1; the superscript T represents transposition;

[0040] The equivalent KKT condition of the quadratic convex optimization model (i.e., the compact form of the sub-convex hull economic operation domain solution model) is as follows:

[0041] λ ≥ 0

[0042] diag(λ)·(E1p G +F1p W +G1x+H1)=0

[0043]

[0044] Where: λ and v are column vectors consisting of Lagrange multipliers with dimensions q1 and q2 respectively; diag(λ) represents a diagonal matrix with λ as diagonal element.

[0045] Furthermore, in order to determine the effective dimension of the economic operation domain of the power grid, the following two two-level optimization models are constructed based on the KKT condition;

[0046] The first-level optimization model is expressed as follows:

[0047] Objective function:

[0048] max

[0049] Constraints:

[0050]

[0051] [p G , x]∈arg{min st},

[0052] The second-level optimization model is expressed as follows:

[0053] Objective function:

[0054] min

[0055] Constraints:

[0056]

[0057] [p G , x]∈arg{min s.t.},

[0058] Where: and They are respectively the lower and upper bounds of the predicted output of the new energy unit w at time t.

[0059] Furthermore, the specific extended relaxation method is as follows:

[0060] 1) From Take any K dispatchable generating units corresponding to the effective dimensions of the economic operation domain and obtain Different generator combinations;

[0061] 2) Solve a K-dimensional subconvex hull economic operation domain for the K generators in each combination and write it in a half-space representation;

[0062] 3) Solve The K-dimensional subconvex hull economic operation domain is uniformly mapped back In the dimensional space, the intersection is the extended convex hull economic operation domain of the power grid.

[0063] Furthermore, the solution model for the i-th sub-convex hull extension point of the extended convex hull economic operation domain of the power grid at time t is as follows;

[0064] Objective function:

[0065]

[0066] Constraints:

[0067]

[0068] [p G , x]∈arg{min s.t.},

[0069] Where: the operator min&max represents the minimum and maximum values ​​of the upper objective function of the two-layer optimization model, respectively. γ Representation combination The index of the γth generator in ; Indicates that the index at time t is i γ Active output of dispatchable generating units; and df t k They represent the coefficients and biases in the half-space inequality constraints corresponding to the k-th face of the K-dimensional subconvex hull to be updated at time t, respectively. represents the γth coefficient;

[0070] The two-level optimization model is transformed into the following single-level linear mixed integer programming form that can be directly solved by commercial solvers;

[0071]

[0072] Furthermore, the specific process of the large M preprocessing method is as follows:

[0073] 1) Randomly generate N in the uncertain set of renewable energy output sce New energy output scenarios k = 1, 2, ..., N sce , and solve the optimal dispatch plan corresponding to each renewable energy output scenario based on the sub-convex hull extension point solution model Based on the following linear optimization model, solve the Lagrange multiplier λ corresponding to each renewable energy output scenario k and v k , k = 1, 2, ..., N sce ;

[0074] Objective function: min1 T λ k ,

[0075] Constraints:

[0076]

[0077] Formula min 1 T λ k Can make λ k The value of satisfies the bilinear constraint. To solve, is a known quantity, and the KKT condition of the subconvex hull extension point solution model is used to solve the corresponding optimal Lagrange multiplier λ opt,k and v opt,k ;

[0078] 2) Solve the optimal Lagrange multiplier λ corresponding to all renewable energy output scenarios opt,k and v opt,k Then, calculate according to the following formula and

[0079]

[0080] Where: m1, m2 and m3 are normal coefficients set in advance; τ1 and τ2 are the relevant inequality constraint and equality constraint indexes respectively; and They are the τ1th row of E, F, G, and H respectively; is opt,k The τ1th element of ; v opt,t The τ2th element of ;

[0081] 3) When solving the extension point of the sub-convex hull economic operation domain, replace the M value of the corresponding index position in the constraint condition of the single-layer optimization model with And the constraints Added to the single-layer optimization model.

[0082] The present application also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method when executing the computer program.

[0083] The present application also provides a computer-readable storage medium having a computer program stored thereon, and the steps of the method described above are implemented when the computer program is executed by a processor.

[0084] The beneficial effects of this application are:

[0085] 1) In terms of reducing the time complexity of updating the convex hull calculation, this application proposes a relaxation method for expanding the convex hull economic operation domain considering the effective dimension to generate an expanded convex hull economic operation domain; by eliminating invalid dimension units and using multiple dimensional sub-convex hull economic operation domain constraints that only consider the effective dimension to approximate the original high-dimensional convex hull economic operation domain, the time complexity of updating the high-dimensional convex hull calculation is reduced;

[0086] 2) In terms of updating convex hull calculation, this application theoretically proves that when the dimension of the original convex hull economic operation domain is high, solving multiple dimensional sub-convex hull economic operation domains generally has a lower time complexity; in addition, the solutions of each dimensional sub-convex hull economic operation domain are independent of each other, and can be solved sequentially or in parallel, which is suitable for scenarios with a large number of parallel computing resources;

[0087] 3) In terms of reducing the total number of times the convex hull calculation is updated, the present application proposes a method for solving the sub-convex hull economic operation domain that takes into account elastic computing resources; by decoupling the solution order of the extension points, the convex hull is updated once each time multiple extension points are generated, thereby reducing the total number of times the convex hull calculation is updated during the solution process; in the new solution algorithm, the extension points can be solved either sequentially or in parallel, which is suitable for scenarios with a large number of elastic parallel computing resources;

[0088] 4) In terms of accelerating the speed of solving the extension point, the present application proposes a large M preprocessing method, which significantly improves the speed of solving the extension point by evaluating and eliminating the number of large numbers M in the extension point solution model in advance; using the optimization scheduling model containing the energy storage system, through example analysis, the effectiveness of the method proposed in the present application is verified, and the extended convex hull economic operation domain generated by the method proposed in the present application completely covers the optimal scheduling plan set;

[0089] 5) The method for solving the extended convex hull economic operation domain of the power grid proposed in this application can be decomposed into multiple independent sub-problems with low computational complexity, and is suitable for scenarios with a large number of parallel computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 It is a logic block diagram of a method for generating an example convex hull economic operation domain according to an embodiment of the present application;

[0091] Figure 2 1 is a schematic diagram of the spatial relationship between the extended convex hull economic operation domain and the convex hull economic operation domain according to an embodiment of the present application; wherein (a) is a schematic diagram of the scheduling period 1, and (b) is a schematic diagram of the scheduling period 2;

[0092] Figure 3 It is a schematic diagram of the spatial relationship between the expanded convex hull economic operation domain and the generator operating area according to an embodiment of the present application; wherein (a) is a schematic diagram of scheduling period 1, (b) is a schematic diagram of scheduling period 2, (c) is a schematic diagram of scheduling period 5, and (d) is a schematic diagram of scheduling period 20. DETAILED DESCRIPTION

[0093] In order to facilitate the understanding of the present application, the present application is described in more detail below in conjunction with the accompanying drawings and specific embodiments. The preferred embodiments of the present application are provided in the accompanying drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described in this specification. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive.

[0094] It should be noted that, unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art of the present application. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and / or" used in this specification includes any and all combinations of one or more related listed items.

[0095] Figure 1-Figure 3 The embodiment of the power grid dispatching method based on elastic computing resources of the present application is shown, including: first, establishing a method for determining the effective dimension of the economic operation domain of the power grid and a method for extending and relaxing the convex hull economic operation domain considering the effective dimension, using multiple effective low-dimensional sub-convex hull economic operation domains to approximate the high-dimensional convex hull economic operation domain, and the extended relaxation result is the extended convex hull economic operation domain;

[0096] Secondly, a method for solving the sub-convex hull economic operation domain considering elastic computing resources is established, so that the convex hull calculation is updated once each time multiple extension points are generated, thus speeding up the solution of the sub-convex hull economic operation domain.

[0097] Finally, a large M preprocessing method is established to preprocess the large number M contained in the sub-convex hull extension point solution model to improve the solution speed of the extension point; and the scheduling result is obtained.

[0098] In terms of reducing the time complexity of updating the convex hull calculation, this application proposes an extended relaxation method for the convex hull economic operation domain that takes into account the effective dimensions to generate an extended convex hull economic operation domain. The original high-dimensional convex hull economic operation domain is approximated by eliminating invalid dimension units and using multiple sub-convex hull economic operation domain constraints that only consider the effective dimensions, thereby reducing the time complexity of updating the high-dimensional convex hull calculation. In terms of updating the convex hull calculation, it is theoretically proved that when the dimension of the original convex hull economic operation domain is high, solving multiple sub-convex hull economic operation domains generally has a lower time complexity. In addition, the solutions of each sub-convex hull economic operation domain are independent of each other, and can be solved sequentially or in parallel, which is suitable for scenarios with a large number of parallel computing resources.

[0099] In order to reduce the total number of times the convex hull is updated, a method for solving the sub-convex hull economic operation domain that takes into account elastic computing resources is proposed. By decoupling the solution order of the extension points, the convex hull is updated once each time multiple extension points are generated, thereby reducing the total number of times the convex hull is updated during the solution process. In the new solution algorithm, the extension points can be solved either sequentially or in parallel, which is suitable for scenarios with a large number of elastic parallel computing resources.

[0100] In terms of speeding up the solution of extension points, a large M preprocessing method is proposed, and the physical meaning and setting range of relevant parameters are introduced. The proposed large M preprocessing method significantly improves the solution speed of extension points by evaluating and eliminating the number of large numbers M in the extension point solution model in advance. Using the optimization scheduling model containing the energy storage system, the effectiveness of the method proposed in this application and the complete coverage of the optimal scheduling plan set by the extended convex hull economic operation domain generated by the method proposed in this application are verified through example analysis. The method for solving the extended convex hull economic operation domain of the power grid proposed in this application can be decomposed into multiple independent sub-problems with low computational complexity, and is suitable for scenarios with a large number of parallel computing resources.

[0101] Optimal dispatching model for power grid considering energy storage and power abandonment of new energy sources: The economic operation domain of the power grid is closely related to the optimal dispatching model and the uncertainty set. In order to further illustrate the applicability of the economic operation domain characterization method proposed in this application to more complex scenarios, and to analyze the impact of the energy storage system on the economic operation domain of the generator active power, this application adds energy storage system (ESS) operation constraints and new energy unit power abandonment constraints. The dispatching goal is to absorb as much new energy as possible at the lowest operating cost, and obtain the sub-convex hull economic operation domain solution model.

[0102] The objective function of the sub-convex hull economic operation domain solution model is:

[0103]

[0104] Where: is the active output of generator g at time t; N T N is the total number of time periods for scheduling. T The time scale of each period can be set according to the needs of the dispatcher; G is the number of dispatchable generating units; N S is the number of energy storage systems; a g 、b g and c g is the operating cost coefficient of generator g; and are the charging and discharging power of the energy storage system s at time t respectively; and are the penalty coefficients for the charging and discharging power of the energy storage system s respectively.

[0105] The constraints on the objective function in the sub-convex hull economic operation domain solution model are:

[0106] 1) Technical output constraints of dispatchable units

[0107]

[0108] Where: and are the minimum and maximum active power that generator g can output respectively;

[0109] 2) Climbing / slipping constraints

[0110]

[0111] Where: and are the maximum landslide and maximum climbing rates of generator g, respectively; is the initial value of the active output of generator g at time t=0;

[0112] 3) Constraints on abandoned power from new energy sources

[0113]

[0114] Where: Predicting power for new energy sources; Abandoned power for new energy sources;

[0115] 4) Line flow constraints

[0116]

[0117] Where: f l max is the maximum transmission power of line l; is the active power of load d at time t; N W is the number of new energy units; N D is the load quantity; and are the power transfer distribution coefficients adapted to the dispatchable generating units, new energy generating units, energy storage systems and load nodes respectively;

[0118] 5) Energy storage system charging and discharging power constraints

[0119]

[0120] Where: and are the maximum charging and discharging power of the energy storage system s, respectively;

[0121] 6) Energy storage system power constraints

[0122]

[0123] Where: and are the charging and discharging efficiencies of the energy storage system s, respectively; e s,t is the storage capacity of energy storage system s at time t; is the capacity of the energy storage system s; Represents the initial power of the energy storage system s in the next scheduling cycle;

[0124] 7) Power balance constraints

[0125]

[0126] The optimal solution of the sub-convex hull economic operation domain solution model is the predicted value of the active output of the new energy unit w The corresponding optimal scheduling solution; since the objective function is and Therefore, when the sub-convex hull economic operation domain solves the model to obtain the optimal value, the constraint of preventing the energy storage system from charging and discharging at the same time will be automatically satisfied.

[0127] For the convenience of expression, the sub-convex hull economic operation domain solution model is written in the following compact form;

[0128] min p G T Ap G +B T p G +C T x+D T ×1

[0129] E1p G +F1p W +G1N+H1≤0

[0130] E2p G +F2p W +G2x+H2=0,

[0131] Where: p G Output of dispatchable generators The column vector composed of W Providing power for new energy units A column vector composed of e s,t and The matrix A, E1, E2, F1, F2, G1, G2 and the column vector B, C, H1, H2, D respectively represent the coefficient information corresponding to the sub-convex hull economic operation domain solution model; 1 is a column vector whose all elements are 1; the superscript T represents transposition;

[0132] It is easy to see that the compact form of the sub-convex hull economic operation domain solution model is a quadratic convex optimization model. To facilitate the subsequent elaboration of the relevant algorithm in this application, the equivalent KKT conditions of the quadratic convex optimization model are written as follows:

[0133] λ ≥ 0

[0134] diag(λ)·(E1p G +F1p W +G1x+H1)=0

[0135]

[0136] Where: λ and v are column vectors consisting of Lagrange multipliers with dimensions q1 and q2 respectively; diag(λ) represents a diagonal matrix with diagonal elements λ. To facilitate direct solution, the complementary relaxation condition is further transformed into the following equivalent linear mixed integer form.

[0137]

[0138] Where: M is a sufficiently large constant; θ is a column vector consisting of 0-1 variables.

[0139] The overall solution method of the extended convex hull economic operation domain considering the effective dimension: When the number of dispatchable power generation units in the power grid is large, the corresponding convex hull economic operation domain of the power grid will also have a higher spatial dimension. When the dimension of the convex hull is high, solving the convex hull problem requires a long computing time. In order to reduce the time complexity of updating the convex hull calculation, this application first proposes a method for determining the effective dimension of the economic operation domain of the power grid, and proposes an extended relaxation method for the convex hull economic operation domain that only considers the effective dimension. By updating multiple low-dimensional convex hulls instead of the high-dimensional convex hull, the extended convex hull economic operation domain that considers the effective dimension is finally obtained.

[0140] Method for determining the effective dimension of the economic operation domain of the power grid: When the new energy unit has active output When changes occur, the optimal output of each dispatchable unit does not always change with That is to say, when the active output of the new energy unit fluctuates within its uncertainty set, there are some dispatchable generator sets whose optimal output always remains constant in some time periods. In the economic operation domain of the corresponding time period, the dimensions corresponding to these dispatchable generator sets whose optimal output remains constant will degenerate. Therefore, when solving the economic operation domain, the constant active output information of these generators can be determined and saved in advance, and then the invalid dimensions corresponding to these generators can be eliminated, and finally the economic operation domain is solved only for the generator sets corresponding to the remaining valid dimensions. This application uses represents the number of dispatchable generating units corresponding to the effective dimension of the economic operation domain of the power grid at time t, that is, the number of effective dimensions of the economic operation domain; Represents the index set of the generator sets corresponding to the effective dimensions of the economic operation domain of the power grid at time t. Eliminating the invalid dimensions of the economic operation domain in advance is conducive to reducing the time complexity of updating the convex hull calculation.

[0141] In order to determine the effective dimensions of the economic operation domain of the power grid, the following two two-layer optimization models are first constructed.

[0142] The first-level optimization model is expressed as follows:

[0143] Objective function:

[0144] max

[0145] Constraints:

[0146]

[0147] [p G , x]∈arg{min s.t.},

[0148] The second-level optimization model is expressed as follows:

[0149] Objective function:

[0150] min

[0151] Constraints:

[0152]

[0153] [p G , x]∈arg{min s.t.},

[0154] Where: and are the lower and upper bounds of the predicted output of the new energy unit w at time t. Taking the interval number form of the new energy output uncertainty set shown as an example, the uncertainty set can be further replaced by any other convex set form that considers more factors, and the algorithms proposed in this application are applicable.

[0155] Constraints G , x]∈arg{mni s.t.} is the lower optimization in the two-level optimization, the purpose is to make the decision variable [p G , x] is always the optimal solution for solving the model in the sub-convex hull economic operation domain. Therefore, it is easy to see that the constraint and [p G , x]∈arg{min s.t.} G Limited to the economic operation domain of the power grid considering the uncertainty of new energy output t The essence of the two-level optimization model is to t Internal solution The maximum value of and minimum value like Then explain The optimal output of is constant, and the corresponding dimension is the invalid dimension of the economic operation domain, which can be eliminated in advance when solving the economic operation domain. However, it is difficult for existing solvers to directly solve the two-layer optimization model. Therefore, based on the equivalent form of KKT conditions and bilinear constraints, this application writes the two-layer optimization model into the following two equivalent single-layer linear mixed integer programming forms, which can be directly solved using commercial solvers.

[0156]

[0157] Based on the optimization model, this application further proposes the following algorithm 1 for determining the effective dimension of the economic operation domain.

[0158]

[0159] In Algorithm 1, δ1 is generally a small positive number, which is mainly used to eliminate the influence of numerical calculation errors on the results. In addition, in Algorithm 1, the optimization models corresponding to each moment and each unit are completely independent, and can be solved either sequentially or in parallel. When parallel computing resources are available, parallel solutions can be used to speed up the determination of the economic operation domain dimensions.

[0160] Overall solution method for the extended convex hull economic operation domain of the power grid: When the characterization accuracy of the economic operation domain of the power grid can be appropriately relaxed, in order to improve the solution efficiency, this application further proposes an extended relaxation method for the convex hull economic operation domain for the effective dimension of the economic operation domain, generates an extended convex hull economic operation domain, and uses the relaxation coefficient to determine the degree of relaxation (generally, And K is a positive integer. The smaller the value of K, the greater the degree of relaxation. The specific expansion relaxation method is to first start from Take any K dispatchable generating units corresponding to the effective dimensions of the economic operation domain and obtain different generator combinations; then solve a K-dimensional subconvex hull economic operation domain for the K generators in each combination and write it in half-space representation; finally, the solved The K-dimensional subconvex hull economic operation domain is uniformly mapped back In the dimensional space, the intersection is the extended convex hull economic operation domain of the power grid. The overall solution process is shown in Algorithm 2.

[0161]

[0162] In Algorithm 2, the final set Ψ t E-conv The intersection of all half-space constraints in is the effective dimension at time t. In addition, the solution for the K-dimensional sub-convex hull economic operation domain is given in line 6.

[0163] Direct solution dimensional effective dimension of the convex hull economic operation domain, if The time complexity of updating the convex hull calculation in the worst case is Where N f-dim express The number of points involved in the calculation in the K-dimensional space. Solve the K-dimensional subconvex hull economic operation domain When K ≥ 4, the time complexity of updating the convex hull calculation in the worst case is reduced to If k = 2 or 3, the time complexity of updating the convex hull calculation in the worst case is further reduced to O(N f-K logN f-N ), where N f-K Indicates the number of points involved in the calculation in K-dimensional space. Generally, N f-K Will be much smaller than N f-dim Therefore, when the effective dimension When it is high, although it is necessary to solve the extended convex hull economic operation domain first The K-dimensional sub-convex hull economic operation domain can significantly reduce the time complexity of solving the convex hull problem in updating the convex hull calculation, especially when the value of K is small. In addition, the time complexity of each scheduling period in Algorithm 2 is The solutions of the K-dimensional sub-convex hull economic operation domains are independent of each other and can be solved sequentially or in parallel. When there are sufficient parallel computing resources, the overall solution speed can be accelerated by solving each sub-convex hull economic operation domain in parallel. It should be noted that when the effective dimension of the economic operation domain is When is higher, in most cases, the time complexity of the extended convex hull economic operation domain in updating the convex hull calculation will be lower than that of the convex hull economic operation domain. However, there are some special cases, such as when When the number of vertices of the economic operation domain of the convex hull is small, directly solve The extended convex hull economic operation domain may also have a lower computational time complexity. The extended convex hull economic operation domain is more suitable for scenarios with a large number of parallel computing resources and where the characterization accuracy can be appropriately relaxed.

[0164] A method for solving the sub-convex hull economic operation domain taking into account elastic computing resources; it enables it to perform an updated convex hull calculation every time multiple extension points are generated, thereby reducing the total number of updated convex hull calculations performed, thereby improving the solution efficiency of the sub-convex hull economic operation domain.

[0165] Sequentially decoupled method for solving extension points of sub-convex hull economic operation domain: The proposed optimization model uses auxiliary variables to solve the extension points of the convex hull economic operation domain, and simultaneously considers the convex hull constraints of all time periods when solving. The advantage of this model is that it can expand the initial convex hull as quickly as possible, and the disadvantage is that it can only be solved sequentially, that is, each time an extension point is solved, the convex hull must be updated. The proposed two-layer optimization model is adjusted to decouple the solution order of each extension point to a certain extent, so that the convex hull is updated once each time multiple extension points are solved. The i-th sub-convex hull of the extended convex hull economic operation domain of the power grid at time t (corresponding to the combination )The extended point solution model is as follows.

[0166] Objective function:

[0167] Constraints:

[0168]

[0169] [p G , x]∈arg{min s.t.},

[0170] Where: i γ Representation combination The index of the γth generator in ; Indicates that the index at time t is i γ Active output of dispatchable generating units; and df t kThey represent the coefficients and biases in the half-space inequality constraints corresponding to the k-th face of the K-dimensional subconvex hull to be updated at time t, respectively. represents the γth coefficient. Constraint The essence of the objective function is to limit the active output of the dispatchable generators to the economic operation domain of the power grid taking into account the uncertainty of the output of renewable energy. The essence of solving the objective function is to find the hyperplane from the economic operation domain. As far as possible above the point, this application uses Although the objective function only contains the generator output in a single dispatch period, the constraint [p G , x]∈arg{min s.t.} contains the climbing / slipping constraints of each generator and the operating constraints of the energy storage system, so its solution can still reflect the coupling relationship between the decision variables in each time period. Since it is difficult for existing solvers to directly solve the two-layer optimization model, this application converts the two-layer optimization model into the following single-layer linear mixed integer programming problem that can be directly solved by commercial solvers.

[0171]

[0172] It is not difficult to see that the solution model only depends on one face of the sub-convex hull to be updated, so the extension points can be solved for multiple faces at the same time, and the solutions of each extension point are independent of each other, and can be solved sequentially or in parallel. In addition, the solution results of the model have the following properties 1.

[0173] Property 1: If the vector Not orthogonal to the final generated K-dimensional subconvex hull economic operation domain Any face of , the model is based on the coefficients The expansion point to be solved is the final generated convex hull economic operation domain The vertex.

[0174] Proof: This application uses proof by contradiction to prove the above properties. The specific idea is to first assume that the extension point solved based on the model is not a vertex of the economic operation domain of the convex hull of the dimensional sub-point, and then find a counterexample in the neighborhood of the extension point to prove that the extension point is not the optimal solution of the model. The detailed proof process is as follows.

[0175] Assume that the extension point based on the model solution is and no The vertex of . May be in inside, or possibly in On a face or edge of There must be a sufficiently small neighborhood ρ around , and the neighborhood ρ does not contain It is not difficult to know that the extension point based on the model solution On the hyperplane above (otherwise It cannot become an extension point, see Algorithm 4 for details), that is, Easy to calculate points To the hyperplane The distance is

[0176] like exist On the edge of , let the direction vector of the edge be v, then there must be two points in the neighborhood ρ and Where τ is a sufficiently small positive number. It is easy to calculate that points p1 and p2 are on the hyperplane The distances are, and Because the vector cf t k Not with Any face of is orthogonal, so cf t kT ·v≠0. If cf t kT ·v>0. Then d1=d0+τ·cf t kT ·v / ||cf t k ||>d0; if cf t kT v<0, then d2=d0+τ·cf t kT ·v / ||cf t k ||>d0. Therefore, Not a distance hyperplane The farthest point is not the optimal solution of the optimization model. exist Inside or on a surface, you can also find a t k Orthogonal direction vectors can be proved in a similar way.

[0177] In summary, if cf t k Not orthogonal to Any aspect of The extension point to be solved is the final generated K-dimensional subconvex hull economic operation domain The vertex.

[0178] When solving the extension points based on the double-layer optimization model proposed in this application, since the upper-layer optimization objective function only considers a single face of the sub-convex hull, a small number of sub-convex hull economic operation domain vertices may be repeatedly solved, that is, different faces of the same sub-convex hull correspond to the same vertex. However, since the objective function of the optimization model proposed in this application is relatively simple, the solution speed is usually faster, and the extension points corresponding to different faces of each sub-convex hull can be solved sequentially or in parallel, so it is very suitable for scenarios with a large number of parallel computing resources.

[0179] The sub-convex hull economic operation domain generation method considering extension point decoupling; the overall logic block diagram is as follows Figure 1 As shown in Figure 2. The main idea is to first generate an initial K-dimensional sub-convex hull based on Algorithm 3, then solve the extension points based on the proposed optimization model, and update the current sub-convex hull according to the obtained extension points until the final K-dimensional sub-convex hull economic operation domain is generated. For the detailed solution process, see the overall algorithm 4.

[0180] In order to demonstrate the feasibility of the proposed method for generating the sub-convex hull economic operation domain considering the decoupling of the extension points, we first solve the model based on the sub-convex hull economic operation domain and prove that the sub-convex hull economic operation domain has the following properties2.

[0181] Property 2: If the output of new energy is uncertain, set Y W is a bounded convex polyhedron, then the subconvex hull economic operation domain considering the uncertainty of new energy output is It is also a bounded convex polyhedron.

[0182] Proof: This proof process is mainly divided into three steps. First, based on the uncertain set of new energy output Y W and the KKT condition of the optimal scheduling model, using the exhaustive bilinear constraints λ j By taking the value of zero, it is proved that the economic operation domain Ψ considering all time periods and all dimensions is the union of a finite number of bounded convex polyhedra (the Ψ considering all time periods is the union of all Ψ t Then, based on the definition of convex combinations, we prove that the convex hull Ψ of Ψ conv is a bounded convex polyhedron. Finally, it is further proved that the sub-convex hull economic operation domain of each period It is also a bounded convex polyhedron. The detailed proof process is as follows.

[0183] Due to the uncertainty of new energy output, W is a bounded convex polyhedron, so it can be expressed as a half-space constraint intersection, that is, Y W ={p W |Jp W +K≤0}, where matrix J and vector K represent the corresponding coefficients. According to the constraints, the active output of the generator p G,tis always bounded. To facilitate the proof process, the set is defined as follows, which represents the corresponding generator output set when the Lagrange multiplier is exhausted.

[0184]

[0185] Where: E 1i 、F 1i , G 1i and H 1i are the i-th row of matrices E1, F1, G1 and H1 respectively; λ j represents the j-th element of vector λ; is the index set and of q1 equations. It is easy to know that the q1 equations have Different α - combination, α = 0, 1.2, ..., q1; Indicates the βth α - The set of equality indices contained in the combination, Obviously, when α and β take different values, the set Ψ α,β is either an empty set or a bounded convex polyhedron. And because So Ψ is the union of a finite number of bounded convex polyhedra with a finite number of vertices.

[0186] When α,β When it is not empty, let V α,β For α,β The vertex set and V α,β The i-th vertex in . Let Ψ conv For all α,β The convex hull of the vertices, V conv For conv The vertex set and V conv The jth vertex in . From the definition of convex combination, we know that for a bounded convex polyhedron Ψ α,β Any point v in α,β , you can use α,β It is represented by a convex combination of the vertices of because You can use the set V conv It can be represented by a convex combination of the points in therefore,

[0187]

[0188] because So α,β ∈Ψ conv , we can further get Ψ∈Ψ conv . And becauseconv Yes all α,β The convex hull of the vertices, and That is Ψ α,β All vertices of are inside Ψ, so In summary, conv That is the convex hull of Ψ, which is a bounded convex polyhedron.

[0189] Because the sub-convex hull economic operation domain

[0190] That is Ψ conv In subspace So It is also a bounded convex polyhedron.

[0191] According to properties 1 and 2, the sub-convex hull economic operation domain of the sub-convex hull economic operation domain solution model considering the uncertainty of new energy output can be obtained by solving the extension points finitely and can be accurately described using the intersection of half-space constraints. In order to standardize the process of generating the initial sub-convex hull, this application first introduces Algorithm 3 for generating the initial sub-convex hull of the dimension.

[0192] First, Algorithm 3 randomly selects a scenario from the uncertain set of renewable energy output and solves an initial optimal dispatch plan based on the sub-convex hull economic operation domain solution model. Then, in line 46, for the initial optimal scheduling plan Combination The corresponding dimensions introduce the set deviation amount δ to generate K deviation points. At time t, the combination The column vector composed of the generator output, is a K-dimensional column vector whose θth dimension element is equal to δ and whose other dimension elements are all 0. In line 78, the set is solved based on the fast convex hull algorithm. The convex hull of the K+1 initial points contained in the convex hull and its half-space constraint coefficients Save to collection In, collection The half-space constraints in correspond to the initial subconvex hull.

[0193]

[0194] Based on Algorithm 3 and the model, this application further proposes a method for generating a sub-convex hull economic operation domain that considers extension point decoupling. The overall solution process is shown in Algorithm 4.

[0195] For the i-th sub-convex hull economic operation domain at time t, first in line 1 of Algorithm 4, initialize the set and In line 2, a sufficiently small K-dimensional subconvex hull is initialized and its vertices are stored in the set In the half-space constraint Save to collection In line 3, update the collection and for Prepare for the subsequent calculation of the extension points of the sub-convex hull economic operation domain.

[0196]

[0197] Lines 4 to 14 are the outer loop, which checks the set The state of is used to determine whether the solution of all the extension points in the economic operation domain of the sub-convex hull has been completed. Lines 5 to 10 are inner loops, which are mainly responsible for solving the extension points for the current sub-convex hull. For collection Specifically, in line 6, based on the set The kth hyperplane in Solving Extension Points And calculate To the hyperplane If d>δ2, then Add to Collection In which δ2 is the preset threshold value for generating extension points, which is usually set to a small positive number according to the calculation error of the solver. In line 11, the set is calculated based on the fast convex hull algorithm. The half-space constraints of the convex hull of the points contained in And update the set according to all the half-space constraints currently obtained In lines 12-13, update the collection and in Used to determine whether it is necessary to continue solving the extension point. Used to record all half-space constraints generated during the solution process The purpose is to ensure that the extension point is solved at most once for each face of the subconvex hull. The half-space constraints contained in correspond to the sub-convex hull economic operation domain corresponding to the i-th combination at time t.

[0198] It should be noted that the inner loops of Algorithm 4 are independent of each other and can be solved sequentially or in parallel. It can update the convex hull once for each generated multiple extension points, thus significantly reducing the total number of convex hull calculations, thereby indirectly improving the speed of solving the sub-convex hull economic operation domain. In addition, Algorithm 4 is more suitable for scenarios where the scale of the power grid system is relatively large, the number of dispatch periods considered is relatively large, and there are a large number of parallel computing resources.

[0199] A large M preprocessing method to improve the speed of solving extension points; In order to improve the speed of solving extension points in the sub-convex hull economic operation domain, the present application proposes a large M preprocessing method to accelerate the speed of solving extension points, thereby further improving the overall solution efficiency of the sub-convex hull economic operation domain. The large M preprocessing method is a method of finding an initial basis feasible solution after using the artificial variable method when the constraint condition of the linear programming problem is (=) equality or (≥) greater than type.

[0200] Large M preprocessing method; The two-layer optimization model for solving the extension points of the sub-convex hull economic operation domain is difficult to solve directly using existing commercial solvers. Therefore, for the convenience of calculation, this application converts it into a single-layer linear mixed integer programming problem. However, a large number of constants M and 0-1 variables with large values ​​are introduced in the process of conversion, which causes the solver to increase the amount of calculation during the solution process due to excessive relaxation of the calculation results of the branch nodes, thereby affecting the solution speed of the extended vertices. In order to alleviate the impact of large numbers M on the solution speed of the model, this application proposes a large M preprocessing method. The main idea is to first independently evaluate the range of M values ​​required in each constraint condition, and then set a reasonable size of M for each constraint condition separately. τ values, rather than all using the same large number M, thereby increasing the speed of solving the extension points of the economic operation domain of the sub-convex hull.

[0201] First, randomly generate N sce New energy output scenarios And based on the sub-convex hull economic operation domain solution model, the optimal dispatch plan corresponding to each new energy output scenario is solved Then, based on the following linear optimization model, the Lagrange multiplier λ corresponding to each renewable energy output scenario is further solved: k and v k , k = 1, 2, ..., N sce .

[0202] Objective function:

[0203] min 1 T λ k ,

[0204] Constraints:

[0205]

[0206] Formula min 1 T λ k Can make λ k The value of satisfies the bilinear constraint. To solve, is a known quantity, so the formula min 1 T λ k For only λ k The essence of solving the model is to and the corresponding optimal scheduling plan The optimal Lagrange multiplier λ corresponding to the KKT condition solution based on the optimization scheduling model opt′k and v opt,k .

[0207] Solve the optimal Lagrange multiplier λ corresponding to all renewable energy output scenarios opt,k and v opt,k Then, further calculate according to the following formula and

[0208]

[0209] Where: m1, m2 and m3 are normal coefficients set in advance; τ1 and τ2 are the relevant inequality constraint and equality constraint indexes respectively; and They are the τ1th row of E, F, G, and H respectively; is opt,k The τ1th element of ; v opt,k The τ2th element of .

[0210] Finally, when solving the extension point of the sub-convex hull economic operation domain, the M value of the corresponding index position in the constraint condition of the single-layer optimization model is replaced by And the constraints Just add it to the single-layer optimization model. The value of will be much smaller than the large number M. Therefore, after applying the large M preprocessing method, the solver will ease the relaxation of the branch node calculation results during the solution process, which is beneficial to improve the solution speed.

[0211] Physical meaning and setting of related parameters: In the proposed large M preprocessing method, the constant coefficient m1 is the reliability coefficient, which is mainly used to and Linear amplification is performed, and the value range is generally 1.1-2.0; m2 is the offset, and its main function is when and When it is 0, it provides a minimum bias value, which is generally a multiple of the unit value, such as 10; m3 is mainly used to limit and The value boundary of and The value is too large or too small, and the value range is generally. It should be noted that the parameters m1, m2 and m3 should not be set too large, otherwise it will slow down the solution speed. In the worst case, it will be close to the model solution speed before applying the large M preprocessing method. A more appropriate setting is to increase the number of randomly sampled new energy output scenes N as much as possible. sce On this basis, m1 and m2 are set to relatively small numbers, and m3 is set to a relatively large number (which may be the same as the large number M).

[0212] Case analysis: This application verifies the effectiveness of the extended convex hull economic operation domain solution method proposed in this application based on modified IEEE-9, IEEE-30, IEEE-57 and IEEE-118 test systems. Specifically, it includes parameter settings of each test system, visual verification of the effectiveness of the extended convex hull economic operation domain generation method considering effective dimensions proposed in this application, the impact of the energy storage system on the extended convex hull economic operation domain, verification of the full coverage of the optimal scheduling plan set by the extended convex hull economic operation domain generated based on the method proposed in this application, and analysis of the effect of the large M preprocessing method on the speed of solving the extension points of the sub-convex hull economic operation domain. This application mainly uses a sequential solution method to solve the relevant example results. All simulation experiments are completed on a computer platform equipped with a CPU with a base frequency of 2.90GHz and 16GB of running memory. The commercial solver used is Gurobi 9.5.1.

[0213] System parameter setting: In order to verify the effectiveness of the extended convex hull economic operation domain solution method proposed in this application and analyze the impact of the energy storage system on the extended convex hull economic operation domain, this application connects two new energy sites (W1 and W2) and one energy storage system to the four test systems. The bus information of the new energy site, energy storage system and dispatchable generator connected to each test system is shown in Table 1. The maximum climbing / slope rate of the dispatchable generator and the rated capacity, maximum charge and discharge power and charge and discharge efficiency of the energy storage system are shown in Table 2. It is uniformly assumed that the initial storage energy of each energy storage system is half of its rated capacity.

[0214] Uniformly set the total number of scheduling periods N T=24, each period is 1h. For the convenience of designing examples, each test system generates the predicted output and load demand level curves of each renewable energy station based on the coefficients shown in Table 3. Specifically, for each dispatch period, the predicted power generation of the renewable energy station is the product of its given basic power and the corresponding coefficient in Table 3, and the overall load demand is the product of the basic load power and the corresponding coefficient in Table 3. Assume that the basic power of the renewable energy station of the modified IEEE-9, IEEE-30, IEEE-57 and IEEE-118 test systems is 30MW, 40MW, 100MW and 200MW respectively. For the IEEE-9 test system, the basic load power connected to busbars 5, 7 and 9 is modified to 80MW, 60MW and 60MW respectively, and the rated power of each dispatchable generator is modified to 100MW. The rated transmission capacity of each branch is shown in Table 4. The basic load power of the IEEE-30, IEEE-57 and IEEE-118 test systems can be found in case30, case57 and case 118 of MATPOWER. For the IEEE-9, IEEE-30, and IEEE-57 test systems, the minimum output power of each dispatchable generator is set to 30% of its rated power. In addition, for the IEEE-30 test system, the rated transmission capacity of each branch is modified to twice the original capacity. For the IEEE-57 test system, the rated transmission capacity of each branch is set to 200MW. For the IEEE-118 test system, the rated transmission capacity of each branch is set to 300MW, and the maximum output power of each dispatchable generator set is set to 800MW. Other related parameters are set to, δ1=0.02,δ2=0.1,N sce =1000.

[0215] Table 1 Bus numbers for connecting new energy stations, energy storage systems and dispatchable generators

[0216] Table 2 Operation parameters of dispatchable generator sets and energy storage systems

[0217] Table 3 Prediction coefficients of active output and load demand level of new energy stations

[0218] Table 4 Rated capacity of each branch of IEEE-9 test system

[0219] The spatial morphological relationship between the extended convex hull economic operation domain and the convex hull economic operation domain:

[0220] The extended convex hull economic operation domain is obtained by extending and relaxing the convex hull economic operation domain. This application first sets the relaxation coefficient K=2 based on the IEEE-9 test system to visualize the morphological relationship between the extended convex hull economic operation domain and the convex hull economic operation domain considering the effective dimension. For convenience, it is assumed that the energy storage system is shut down, and the scheduling period 1 and scheduling 2 are taken as examples. Figure 2 As shown. It is easy to see that in dispatch period 1, the effective dimension of the economic operation domain is 3, and the extended convex hull economic operation domain completely includes the convex hull economic operation domain, and there is a certain degree of relaxation. In dispatch period 2, the effective dimension of the economic operation domain is 2, which is the same as the relaxation system of the extended convex hull economic operation domain. At this time, the result of the extended convex hull economic operation domain completely overlaps with the convex hull economic operation domain. In general, when the effective dimension of the economic operation domain is greater than the relaxation coefficient K of the extended convex hull economic operation domain, the obtained extended convex hull economic operation domain will completely include the convex hull economic operation domain. When the effective dimension of the economic operation domain is equal to K, the extended convex hull economic operation domain will only contain one sub-convex hull, and the sub-convex hull is the convex hull economic operation domain. Therefore, the obtained extended convex hull economic operation domain completely overlaps with the convex hull economic operation domain, which is equivalent to not relaxing the convex hull economic operation domain.

[0221] The impact of energy storage system on the extended convex hull economic operation domain:

[0222] This application is based on the modified IEEE-9 test system, and analyzes the impact of energy storage systems on the economic operation domain from three perspectives: effective dimension, shape, and space share of the extended convex hull economic operation domain. First, it is assumed that the active output prediction error of each new energy station is ±65%, that is, the uncertainty set Y W It is the ±65% range of the predicted output curve of new energy. Then, for the two working conditions of the energy storage system in operation and outage, the effective dimension of the economic operation domain of each scheduling period and the extended convex hull economic operation domain considering the effective dimension are solved respectively. In addition, in order to further analyze the spatial proportion of the extended convex hull economic operation domain, this application compares the spatial proportion of the extended convex hull economic operation domain with the convex hull economic operation domain and the generator operable area of ​​the energy storage system outage.

[0223] In terms of the effective dimension of the economic operation domain, as shown in Table 5, when the energy storage system is shut down, the effective dimension of the economic operation domain drops to 2 in dispatch periods 2, 4, 6, and 24. When the energy storage system is put into operation, the effective dimension of the economic operation domain drops to 2 in dispatch periods 1 to 9. In contrast, the effective dimension of the generator operable area is always 3. The dimensionality reduction phenomenon of the economic operation domain shows that in certain dispatch periods, the optimal output power of the dispatchable generator set of the corresponding dimension remains constant (in this example, the generator set connected to bus 1 has an optimal output power of 30MW), and will not be affected by the active output fluctuation of the new energy station. This conclusion that the optimal output power of the generator is constant can be used to guide the real-time dispatch of the power grid.

[0224] Table 5 Comparison of effective dimensions of various operating domains

[0225]

[0226] In terms of expanding the convex hull economic operation domain, such as Figure 3 As shown, representative scheduling periods 1, 2, 5 and 20 are taken as examples. The main difference between the generator operable area and the extended convex hull economic operation domain is that the extended convex hull economic operation domain additionally considers the optimality condition. Therefore, it is easy to understand that the extended convex hull economic operation domain of the energy storage system outage will always be within the generator operable area of ​​the energy storage system outage. However, when the energy storage system is put into operation, since the energy storage system can transfer electric energy between scheduling periods through charging and discharging, it will change the distribution of the economic operation domain of the dispatchable units, thereby causing the corresponding extended convex hull economic operation domain to exceed the generator operable area of ​​the energy storage system outage in certain scheduling periods. As Figure 3 As shown in the figure, it is easy to see that the operating area of ​​the generator with the energy storage system shut down occupies most of the space in the adjustable area of ​​the generator output. However, when the optimality condition is considered, the spatial range occupied by the extended convex hull economic operation domain is significantly reduced, and the generator output power at bus 1 is always at a low level (because the operating cost of the generator is high). In the dispatch period 1, the extended convex hull economic operation domain of the energy storage system is even shrunk into a very narrow area. Therefore, the commissioning of the energy storage system will change the shape and spatial distribution of the economic operation domain of the dispatchable units.

[0227] In terms of the spatial proportion of the extended convex hull economic operation domain, as shown in Tables 5 and 6, this application uses the Monte Carlo method to quantitatively evaluate the spatial proportion of the generator's operable area, the convex hull economic operation domain, and the extended convex hull economic operation domain (i.e., the spatial proportion of the generator's adjustable area). It can be seen from the new energy output prediction coefficient in Table 3 and the prediction error of ±65% that the active output range of the new energy station accounts for 7.45% to 77.54% of the total load and 5.13% to 44.17% of the dispatchable generator installed capacity. When the new energy station is uncertain about the active output within the above range, the spatial proportion of the generator's operable area in each dispatching period that occupies the generator's adjustable area is in the range of 9.83% to 75.03%.

[0228] Table 6 Comparison of the space occupied by various operating domains

[0229]

[0230] However, for the convex hull economic operation domain of the energy storage system shutdown, when the effective dimension is 3, its spatial size is 0.14% to 3.27% of the generator adjustable area, and when the effective dimension is 2, its spatial size is 5.47% to 11.79% of the generator adjustable area. For the convex hull economic operation domain of the energy storage system in operation, when the effective dimension is 3, its spatial size is 0.02% to 1.50% of the generator adjustable area, and when the effective dimension is 2, its spatial size is 0.18% to 11.57% of the generator adjustable area. For the extended convex hull economic operation domain of the energy storage system shutdown, when the effective dimension is 3, its spatial size is 0.18% to 3.84% of the generator adjustable area, which is slightly larger than the convex hull economic operation domain, and when the effective dimension is 2, its spatial size is the same as the convex hull economic operation domain. For the extended convex hull economic operation domain of the energy storage system, when the effective dimension is 3, its spatial size is 0.10% to 1.93% of the adjustable area of ​​the generator, which is slightly larger than the convex hull economic operation domain. When the effective dimension is 2, its spatial size is the same as the convex hull economic operation domain. Therefore, it is easy to see that the convex hull economic operation domain and the extended convex hull economic operation domain of the dispatchable unit usually only occupy a small proportion of the adjustable area of ​​the generator, and are significantly smaller than the operable area of ​​the generator. The commissioning of the energy storage system helps to further reduce the economic operation domain and even reduce its effective dimension.

[0231] Economic analysis of the extended convex hull economic operation domain: This application is based on a modified IEEE-9 test system, and analyzes the economics of the extended convex hull economic operation domain solved by the method proposed in this application, that is, the coverage of the optimal dispatch plan, and compares it with the box-type operating region (BC-OR) solved based on the robust optimization algorithm. For the convenience of designing examples, this application assumes that the prediction errors of renewable energy power generation are ±65%, ±55%, ±45%, ±35% and ±25%, and constructs a new energy output uncertainty set based on this. For the two working conditions of energy storage system shutdown and commissioning, the box-type operating domain and the extended convex hull economic operation domain corresponding to the prediction errors of each renewable energy output are solved. Finally, 500 renewable energy output scenarios are randomly selected in each renewable energy output uncertainty set, and the optimal dispatch plan of the power grid corresponding to each scenario, the optimal dispatch plan considering the box-type operating domain, and the optimal dispatch plan considering the extended convex hull economic operation domain are solved respectively. The average operating costs of the above-mentioned scheduling plans and their ratios to the optimal average operating costs are compared and analyzed. The relevant results are shown in Table 7.

[0232] According to the calculation results shown in Table 7, the average operating cost of the dispatch plan considering the extended convex hull economic operation domain is the same as the optimal average operating cost, indicating that the extended convex hull economic operation domain solved based on the method proposed in this application can cover the optimal dispatch plan corresponding to all new energy output scenarios. However, the average operating cost of the dispatch plan considering the box-type operation domain is higher than the optimal average operating cost, and the excess ratio increases with the increase of the new energy output prediction error, indicating that the box-type operation domain solved based on the two-stage robust optimization cannot completely cover the optimal dispatch plan corresponding to all new energy output scenarios. This is because the box-type operation domain solved based on the two-stage robust optimization is mainly to cover the optimal dispatch plan under the worst scenario, so it has a certain degree of conservatism. In addition, it is easy to see that the optimal average operating cost of the energy storage system when it is put into operation is lower than when the energy storage system is shut down.

[0233] Table 7 Economic analysis of the extended convex hull economic operation domain

[0234]

[0235] In summary, compared with the box-type operation domain solved based on the two-stage robust optimization, the extended convex hull economic operation domain solved based on the method proposed in this application has better coverage of the optimal dispatch plan of the power grid, that is, it shows better economy. In addition, the commissioning of the energy storage system helps to further reduce the operating cost of the power grid.

[0236] Effectiveness analysis of the big M preprocessing method: This application is based on the modified IEEE-9, IEEE-30, IEEE-57 and IEEE-118 test systems, and compares and analyzes the speed of solving the sub-convex hull economic operation domain extension point using the traditional big M method and the big M preprocessing method proposed in this application. For the big M preprocessing method, set m1=1.5, m2=10, m3=105. For the traditional big M method, set the large number M=10 5 Assuming that the prediction error of the new energy output of each test system is ±65%, the relaxation coefficient K is set to 2, and the extended convex hull economic operation domain under the two conditions of the energy storage system being put into operation and shut down is solved respectively. The average solution time of the extended points of each test system is shown in Table 8.

[0237] Table 8 Average solution time for extended vertices

[0238]

[0239] It can be seen from Table 8 that the application of the large M preprocessing method proposed in this application can significantly improve the solution speed of the extension point of the sub-convex hull economic operation domain. When the energy storage system is put into operation, the operation constraints of the energy storage system are introduced, which increases the calculation time. In addition, after applying the traditional large M method, when the energy storage system is shut down, the solution time of the IEEE-9 test system is longer than that of the IEEE-30 test system. After applying the large M preprocessing method, when the energy storage system is put into operation, the solution time of the IEEE-57 test system may also be longer than that of the IEEE-118 test system. In general, the solution time of the extension point will increase with the expansion of the scale of the power grid system, but occasionally there will be exceptions, that is, the smaller power grid system consumes longer solution time than the larger power grid system. This is because the solution efficiency of existing commercial solvers for mixed integer programming problems is affected by many factors, such as problem scale, branching strategy, node relaxation, etc., and has certain uncertainties.

[0240] In general, for the traditional big M method, if the selected M value is too large, it will cause the solver to relax the calculation results of the branch nodes excessively during the solution process, thereby greatly increasing the number of branch nodes to be tested, and then increasing the solution time. However, if the selected M value is too small, it may cause the optimization model to have no feasible solution. Therefore, it is generally difficult to select a suitable M value for the traditional big M method. In contrast, the parameters m1, m2 and m3 of the big M pretreatment method proposed in this application have a universal range of values ​​for different optimization models, and it is very easy to determine the value. Moreover, after applying the big M pretreatment method, the number of large numbers M in the optimization model can be significantly reduced, thereby alleviating the excessive relaxation of the solver to the calculation results of the branch nodes during the solution process, and then reducing the number of branch nodes to be tested and shortening the solution time. In summary, according to Table 8, it is easy to see that the big M pretreatment method proposed in this application can significantly improve the solution efficiency of the extension point of the sub-convex hull economic operation domain (the average calculation time is increased by more than 17 times).

[0241] This application deeply analyzes the three key factors that affect the efficiency of solving the economic operation domain of the power grid convex hull: 1. The time complexity of a single update of the convex hull calculation; 2. The total number of executions of the update convex hull calculation; 3. The speed of solving the extension point. In view of the above key factors that restrict the solution speed, this application further establishes a solution method for the economic operation domain of the power grid extended convex hull considering the effective dimension. By appropriately reducing the characterization accuracy, the efficiency of solving the economic operation domain is effectively improved. The main summary is as follows:

[0242] (1) In terms of reducing the time complexity of updating the convex hull calculation, a relaxation method for expanding the convex hull economic operation domain considering effective dimensions is established to generate an expanded convex hull economic operation domain. The original high-dimensional convex hull economic operation domain is approximated by eliminating invalid dimension units and using multiple dimensional sub-convex hull economic operation domain constraints that only consider effective dimensions, thereby reducing the time complexity of updating the high-dimensional convex hull calculation. In terms of updating the convex hull calculation, it is theoretically proved that when the dimension of the original convex hull economic operation domain is high, solving multiple dimensional sub-convex hull economic operation domains generally has a lower time complexity. In addition, the solutions of each dimensional sub-convex hull economic operation domain are independent of each other and can be solved sequentially or in parallel, which is suitable for scenarios with a large number of parallel computing resources.

[0243] (2) In terms of reducing the total number of times the convex hull is updated, a method for solving the sub-convex hull economic operation domain that takes into account elastic computing resources is established. By decoupling the solution order of the extension points, the convex hull is updated once each time multiple extension points are generated, thereby reducing the total number of times the convex hull is updated during the solution process. In the new solution algorithm, the extension points can be solved either sequentially or in parallel, which is suitable for scenarios with a large number of elastic parallel computing resources.

[0244] (3) In terms of speeding up the solution of extension points, a large M preprocessing method is established, and the physical meaning and setting range of relevant parameters are introduced. The proposed large M preprocessing method significantly improves the solution speed of extension points by pre-evaluating and eliminating the number of large numbers M in the extension point solution model.

[0245] In addition, this application uses an optimized scheduling model that includes an energy storage system. The access of the energy storage system will change the form of the economic operation domain, generally shrinking the economic operation domain or even reducing its dimension.

[0246] The above are merely embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structural transformations made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A power grid dispatching method based on elastic computing resources, characterized in that: Includes steps: S1. Establish a method for determining the effective dimension of the economic operation domain of the power grid and an extended relaxation method for the convex hull economic operation domain considering the effective dimension, use multiple effective low-dimensional sub-convex hull economic operation domains to approximate the high-dimensional convex hull economic operation domain, and the extended relaxation result is the extended convex hull economic operation domain; S2. Establish a sub-convex hull economic operation domain solution model that takes into account elastic computing resources, and implement an update convex hull calculation every time multiple extension points are generated, thereby accelerating the solution speed of the sub-convex hull economic operation domain; S3, establish a large M preprocessing method, preprocess the large number M contained in the sub-convex hull extension point solution model, and improve the solution speed of the extension point; obtain the scheduling result; The specific process of the big M preprocessing method is as follows: 1) Randomly generate N in the uncertain set of renewable energy output sce New energy output scenarios k=1.2,…,N sce , and solve the optimal dispatch plan corresponding to each renewable energy output scenario based on the sub-convex hull extension point solution model k = 1, 2, ..., N sce ; Based on the following linear optimization model, solve the Lagrange multiplier λ corresponding to each renewable energy output scenario k and v k , k=1.2,…,N sce ; Objective function: min 1 T λ k , Constraints: Formula min 1 T λ k Can make λ k The value of satisfies the bilinear constraint, and the KKT condition of the sub-convex hull extension point solution model is used to solve the corresponding optimal Lagrange multiplier λ opt,k and v opt,k ; 2) Solve the optimal Lagrange multiplier λ corresponding to all renewable energy output scenarios opt,k and v opt,k Then, calculate according to the following formula and Where: m1, m2 and m3 are normal coefficients set in advance; τ1 and τ2 are the relevant inequality constraint and equality constraint indexes respectively; and They are the τ1th row of E, F, G, and H respectively; is opt,k The τ1th element of ; v opt,k The τ2th element of ; 3) When solving the extension point of the sub-convex hull economic operation domain, replace the M value of the corresponding index position in the constraint condition of the single-layer optimization model with And the constraints Added to the single-layer optimization model.

2. The power grid dispatching method based on elastic computing resources according to claim 1, characterized in that: The objective function of the sub-convex hull economic operation domain solution model is: Where: is the active output of generator g at time t; N T N is the total number of time periods for scheduling. T The time scale of each period can be set according to the needs of the dispatcher; G is the number of dispatchable generating units; N S is the number of energy storage systems; a g , b g and c g is the operating cost coefficient of generator g; and are the charging and discharging power of the energy storage system s at time t respectively; and are the penalty coefficients for the charging and discharging power of the energy storage system s respectively.

3. The power grid dispatching method based on elastic computing resources according to claim 2, characterized in that: The constraints on the objective function in the sub-convex hull economic operation domain solution model are: 1) Technical output constraints of dispatchable units Where: and are the minimum and maximum active power that generator g can output respectively; 2) Climbing / slipping constraints Where: and are the maximum landslide and maximum climbing rates of generator g, respectively; 3) Constraints on abandoned power from new energy sources Where: Predict power for new energy sources; Abandoned power for new energy sources; 4) Line flow constraints Where: is the maximum transmission power of line 1; is the active power of load d at time t; N W is the number of new energy units; N D is the load quantity; and are the power transfer distribution coefficients adapted to the dispatchable generating units, new energy generating units, energy storage systems and load nodes respectively; 5) Energy storage system charging and discharging power constraints Where: and are the maximum charging and discharging power of the energy storage system s, respectively; 6) Energy storage system power constraints Where: and are the charging and discharging efficiencies of the energy storage system s, respectively; e s,t is the storage capacity of energy storage system s at time t; is the capacity of the energy storage system s; Represents the initial power of the energy storage system s in the next scheduling cycle; 7) Power balance constraints The optimal solution of the sub-convex hull economic operation domain solution model is the predicted value of the active output of the new energy unit w The corresponding optimal scheduling solution; The objective function is and When the sub-convex hull economic operation domain solves the model to obtain the optimal value, it will automatically satisfy the constraint of preventing the energy storage system from charging and discharging at the same time.

4. The power grid dispatching method based on elastic computing resources according to claim 3, characterized in that: In order to simplify the input of the model, the sub-convex hull economic operation domain solution model is written in the following compact form; Where: p G Output of dispatchable generators The column vector composed of W Providing power for new energy units A column vector composed of e s,t and The matrix A, E1, E2, F1, F2, G1, G2 and the column vector B, C, H1, H2, D respectively represent the coefficient information corresponding to the sub-convex hull economic operation domain solution model; 1 is a column vector whose all elements are 1; the superscript T represents transposition; The equivalent KKT conditions for solving the compact form of the sub-convex hull economic operation domain model are as follows: Where: λ and v are column vectors composed of Lagrange multipliers, with dimensions q1 and q2 respectively; diag(λ) represents a diagonal matrix with λ as its diagonal element.

5. The power grid dispatching method based on elastic computing resources according to claim 4, characterized in that: In order to determine the effective dimension of the economic operation domain of the power grid, based on the KKT condition, the following two two-level optimization models are constructed; The first-level optimization model is expressed as follows: Objective function: Constraints: The second-level optimization model is expressed as follows: Objective function: Constraints: Where: and They are respectively the lower and upper bounds of the predicted output of the new energy unit w at time t.

6. The power grid dispatching method based on elastic computing resources according to claim 5, characterized in that: The specific extended relaxation method is as follows: 1) From Take any K dispatchable generating units corresponding to the effective dimensions of the economic operation domain and obtain Different generator combinations; 2) Solve a K-dimensional subconvex hull economic operation domain for the K generators in each combination and write it in a half-space representation; 3) Solve The K-dimensional subconvex hull economic operation domain is uniformly mapped back In the dimensional space, the intersection is the extended convex hull economic operation domain of the power grid.

7. The power grid dispatching method based on elastic computing resources according to claim 6, characterized in that: The solution model for the i-th sub-convex hull extension point of the extended convex hull economic operation domain of the power grid at time t is as follows; Objective function: Constraints: Where: the operator min&max represents the minimum and maximum values ​​of the upper objective function of the two-layer optimization model, respectively. γ Representation combination The index of the γth generator in ; Indicates that the index at time t is i γ Active output of dispatchable generating units; and They represent the coefficients and biases in the half-space inequality constraints corresponding to the k-th face of the K-dimensional subconvex hull to be updated at time t, respectively. represents the γth coefficient; The two-level optimization model is transformed into the following single-level linear mixed integer programming form that can be directly solved by commercial solvers; 8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the power grid scheduling method based on elastic computing resources described in any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for power grid scheduling based on elastic computing resources described in any one of claims 1 to 7 are implemented.