Data-driven distribution robust joint opportunity constraint scheduling method for alternating-current and direct-current hybrid power system

Abstract

The invention discloses a data-driven distribution robust joint opportunity constraint scheduling method for an AC / DC hybrid power system. The method comprises the following steps: constructing a flexible consumption model of a DC tie line; building an AC / DC hybrid system data driving distribution robust joint opportunity constraint scheduling model based on the DC tie line flexible absorption model; and performing reconstruction based on optimal condition value-at-risk approximation OCA on the data-driven distributed robust joint opportunity constraint scheduling model, and solving the reconstructed data-driven distributed robust joint opportunity constraint scheduling model through an iterative sequence convex programming SCP algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a data-driven distributed robust joint opportunity-constrained scheduling method for an AC / DC hybrid power system. Background Art

[0002] In the past few decades, global wind power generation has developed rapidly, but since some wind farms are far away from the load center, cross-regional power transmission is crucial for absorbing large-scale wind power. At present, high-voltage direct current transmission (HVDC) and high-voltage alternating current transmission (HVAC) technologies provide sufficient technical support for large-scale renewable energy transmission between two remote areas. HVDC has less loss in long-distance transmission and enhances the controllability of the power grid. Due to its flexible power adjustment capability, HVDC transmission technology can make the operation of the entire AC / DC hybrid system more economical and efficient by interconnecting the AC power grid through high-voltage DC interconnection lines.

[0003] At present, modeling methods such as stochastic optimization, two-stage robust optimization, and data-driven distributed robust opportunity constraints are generally used to deal with the uncertainty of high-proportion renewable energy output. It is difficult for stochastic optimization methods to obtain accurate probability distributions of uncertain variables, and the solution of two-stage robust optimization is too conservative. Due to the difficulty in obtaining probability distributions suitable for describing uncertainty and the availability of historical data, the data-driven distributed robust opportunity constraint modeling method has gradually attracted attention in recent years. It eliminates the dependence of stochastic optimization on accurate probability distributions and solves the conservative problem of robust optimization solutions. However, the current distributed robust scheduling models all ignore the spatiotemporal correlation between the uncertain variables of renewable energy output, and the solutions are too conservative. Moreover, the strong non-convexity of the distributed robust joint opportunity constraint model makes the model difficult to solve, and the computational feasibility is more challenging. It is urgent to study efficient and advanced approximate reconstruction methods of data-driven distributed robust joint opportunity constraint scheduling models that consider the spatiotemporal correlation of uncertain variables, so as to realize data-driven distributed robust joint opportunity constraint scheduling of AC / DC hybrid power systems. Summary of the invention

[0004] In view of this, the present invention provides a data-driven distributed robust joint opportunity constrained scheduling method for an AC / DC hybrid power system, which is used to at least solve the problem that the model in the prior art is difficult to solve.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] A data-driven distributed robust joint opportunity-constrained scheduling method for an AC / DC hybrid power system comprises the following steps:

[0007] Construct a flexible consumption model for DC interconnection lines;

[0008] A data-driven distributed and robust joint opportunity-constrained dispatching model is constructed based on the flexible consumption model of DC tie lines;

[0009] The data-driven distributed robust joint opportunity-constrained scheduling model is reconstructed and solved, including:

[0010] S1. Rewrite the data-driven distributed robust joint opportunity-constrained scheduling model into a compact form:

[0011] Objective function:

[0012] Joint opportunity constraints:

[0013] In the formula, c 0,1 、c 0,2 and C0 represent the coefficient matrix of the objective function, and the elements in C0 are either 0 or 1. and represents the power regulation factor of the thermal power unit at time t, and t∈T, T represents the total number of scheduling periods; the matrix and Indicates that α t Other decision variables besides represents the probability distribution of wind power output, represents the set of all possible probability distributions of wind power output, represents the expected value, ξ represents the wind power prediction error, Indicates The risk level of the joint opportunity constraint is joint opportunity constraints, and No. The coefficient matrix of the joint opportunity constraints;

[0014] S2. According to the optimal conditional value at risk approximation OCA processing method, the compact form of the data-driven distributed robust joint opportunity constraint scheduling model is reconstructed to obtain the reconstructed model Φ(Δ), where the objective function represents the sum of the power generation cost under the wind power forecast scenario and the power generation adjustment cost under the wind power error scenario:

[0015]

[0016] In the formula, For the The newly introduced scaling parameter is constrained by a joint opportunity, ∈ is the risk level of the joint opportunity constraint, τ is an auxiliary variable, and L(·) represents the loss function;

[0017] S3. Find the optimal solution to equation (3) through the iterative sequential convex programming SCP algorithm:

[0018] S31. Initialize iterative residual Constant φ>0, let For the The number of constraints in a joint opportunity constraint;

[0019] S32. Introduce slack variables to reconstruct equation (3) into equation (4), and iteratively solve equation (4):

[0020]

[0021] In the formula, For the Auxiliary slack variables introduced by joint chance constraints, and ω is the penalty coefficient; A represents the expected term in the objective function (1) After the equivalent convex cone programming problem, B represents the convex cone constraint after the joint opportunity constraint (2) is processed by the optimal conditional risk value approximation OCA;

[0022] S33. If satisfied Or if the preset maximum number of iterations is reached, the current solution (x, Λ, ν) is the optimal solution, otherwise proceed to S34;

[0023] S34. Return to S32 by solving (5) to update

[0024]

[0025] Among them, δ is the scaling parameter, Π ++ Represents the relative interior of the probability simplex.

[0026] Preferably, the DC tie line flexible consumption model includes:

[0027] DC tie line power limit:

[0028]

[0029] DC tie line power adjustment rate constraints:

[0030]

[0031] DC tie line power adjustment direction constraints:

[0032]

[0033] Limitation on the number of times the DC tie line power can be adjusted:

[0034]

[0035] DC tie line power adjustment direction restrictions:

[0036]

[0037] Step-by-step operation characteristics of DC tie line transmission power:

[0038]

[0039] In the formula, represents the power of the ith DC tie line at time t, P i DC and They represent the lower and upper limits of the transmission power of the ith DC tie line, represents the set of DC tie lines, Represents a set of time periods, and They represent the upward and downward power regulation status flags of the ith DC tie line at time t, respectively. and denote the upward and downward power regulation rates of the ith DC tie line, κ DC Indicates the number of power adjustments allowed by the DC tie line, M + and M - represents a positive number, τ represents the shortest duration after the DC tie line adjusts the power once, and Indicates whether the i-th DC tie line starts and ends power regulation at time t.

[0040] Preferably, the data-driven distributed robust joint opportunity-constrained scheduling model is:

[0041]

[0042] The objective function is to minimize the expected power generation cost of the AC regional power grid under the worst wind power conditions:

[0043]

[0044] (18)-(19)(31)

[0045] Among them, (21) represents the AC power balance constraint, (22) represents the power injection limit of the DC interconnection line, (23) represents the power injection limit of the AC interconnection line, (24)-(25) represent the minimum start-stop time constraint of the thermal power unit, (26)-(27) represent the start-stop action constraint of the thermal power unit, and (28)-(30) are joint opportunity constraints;

[0046] In the formula, and They represent the output of thermal power unit i at time t after the uncertainty occurs and in the prediction scenario, respectively, it represents the power regulation factor of thermal power unit i at time t, ξ jt represents the prediction error of wind farm j at time t, and They represent the load, thermal power unit, line, wind farm, DC tie line node and AC tie line node set in AC power grid a respectively. represents the AC power grid, a i , b i and c i They represent the consumption characteristic coefficient of thermal power unit i, u it 、v it and w it They represent the start / stop status, start-up action and shutdown action of thermal power unit i at time t respectively. represents the startup cost of thermal power unit i, d it represents the output adjustment cost coefficient of thermal power unit i, represents the predicted value of wind farm i at time t, represents the load of node i at time t, represents the power of the ith AC tie line at time t, P i AC and They represent the lower and upper limits of the transmission power of the i-th AC tie line, MU i and MD i are the minimum start-up time and minimum shutdown time of thermal power unit i, respectively, P i G , Represents the lower and upper limits of the output of thermal power unit i, R i represents the ramp rate limit of thermal power unit i, ∈1, ∈2 and ∈3 represent the risk levels of joint opportunity constraints, represents the transmission power upper limit of line l, w represents the wind farm, d represents the load in AC grid a, g represents the thermal power unit, j represents the jth AC tie line node, M gl 、M wl 、M dl 、M il and M jl They represent the power transfer distribution factors of thermal power units, wind farms, loads, DC interconnection lines and AC interconnection lines respectively.

[0047] Preferably, the wind power prediction error ξ is constrained by a polyhedron constraint Ω, and Ω=ξ:Lξ≤l, L and l represent the parameters constituting the polyhedron Ω, and the expected term in the objective function (1) is It is equivalent to solving the convex cone programming problem, namely A:

[0048]

[0049] In the formula, θ represents the Wasserstein radius, N represents the number of samples of wind power prediction error, represents the sample value of the i-th wind power prediction error, κ o , They are all auxiliary variables.

[0050] Preferably, the specific content of S2 includes:

[0051] The joint opportunity constraint (2) is reconstructed by the optimal conditional value at risk approximation OCA processing method:

[0052] The joint opportunity constraint (2) can be further expressed as:

[0053]

[0054] Among them, C k and c k Respectively and The kth row of

[0055] Refactor (33) as follows:

[0056]

[0057] Using the conditional value at risk approximation (34), we get:

[0058]

[0059] according to For a given scaling parameter δ, Equation (35) is further transformed into the following convex cone programming form, namely B:

[0060]

[0061] In the formula, κ c , σ c , β c , γ c is an auxiliary variable;

[0062] Then (1)-(2) are rewritten into the form based on OCA reconstruction as shown in formula (3).

[0063] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a data-driven distributed and robust joint opportunity-constrained scheduling method for an AC / DC hybrid power system, which has the following beneficial effects:

[0064] (1) The present invention makes full use of the flexible adjustment capability of DC interconnection lines and proposes a flexible consumption model for high-voltage DC interconnection lines to achieve joint optimization of the transmission power of DC interconnection lines and the output of power grid generators, thereby promoting the cross-regional consumption of a high proportion of new energy.

[0065] (2) A data-driven distributed robust joint chance-constrained scheduling model is established to cope with the uncertainty of renewable energy output and the spatiotemporal correlation of prediction errors. An optimized conditional value at risk algorithm (OCA) method is proposed to reconstruct the distributed robust joint chance-constrained scheduling model that takes into account the spatiotemporal correlation of renewable energy prediction errors.

[0066] (3) A sequential convex programming (SCP) algorithm with an iterative optimization strategy is proposed to solve the distributed robust joint chance-constrained scheduling model reconstructed by OCA, which can reduce the burden of model solution and the conservatism of the distributed robust model. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0068] Figure 1 A topological diagram of an AC / DC hybrid power system containing a high proportion of new energy provided by an embodiment of the present invention;

[0069] Figure 2 A wind power and load output curve diagram provided by an embodiment of the present invention;

[0070] Figure 3 A flexible consumption curve of a DC tie line provided in an embodiment of the present invention;

[0071] Figure 4 The operating costs of the distributed robust joint chance-constrained scheduling model based on BA reconstruction provided in the embodiment of the present invention and the distributed robust joint chance-constrained scheduling model based on OCA reconstruction and solved by the SCA algorithm are compared. DETAILED DESCRIPTION

[0072] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] The present invention provides a data-driven distributed and robust joint opportunity-constrained scheduling method for an AC / DC hybrid power system, which is oriented to objects such as Figure 1 As shown, AC grid A containing a high proportion of renewable energy transmits large-scale wind power in AC grid A to AC grids B and C over long distances through two high-voltage DC interconnection lines.

[0074] 1. Construct a flexible consumption model for DC interconnection lines

[0075] In order to make full use of the flexible adjustment capability of DC interconnection lines and promote the long-distance cross-regional consumption of high-proportion new energy, a DC interconnection line flexible consumption model is established. (6) represents the power limit of the DC interconnection line, (7)-(8) represent the power adjustment rate of the DC interconnection line, (9)-(11) represent the power adjustment direction of the DC interconnection line, and ensure that the transmission power of the DC interconnection line cannot be reversed in adjacent time periods, (12) represents the number of power adjustments of the DC interconnection line, (13)-(14) represent the power adjustment direction limit of the DC interconnection line, and (15)-(17) represent the step-by-step operation characteristics of the DC interconnection line transmission power.

[0076] DC tie line power limit:

[0077]

[0078] DC tie line power adjustment rate constraints:

[0079]

[0080] DC tie line power adjustment direction constraints:

[0081]

[0082] Limitation on the number of times the DC tie line power can be adjusted:

[0083]

[0084] DC tie line power adjustment direction restrictions:

[0085]

[0086] Step-by-step operation characteristics of DC tie line transmission power:

[0087]

[0088]

[0089] In the formula, represents the set of DC tie lines, represents the time period set, κ DC Indicates the number of power adjustments allowed by the DC tie line, P i DC , Indicates the lower and upper limits of the transmission power of the DC tie line, M + 、M - represents a large positive number, τ represents the shortest duration after the DC tie line adjusts its power once, Indicates the upward and downward power regulation rate of the DC tie line, represents the power of the DC tie line at time t, Indicates the upward and downward power regulation status of the DC tie line at time t. Indicates whether the DC tie line starts or ends power regulation at time t.

[0090] 2. Constructing a data-driven distributed robust joint opportunity-constrained scheduling model

[0091] In order to alleviate the impact of wind power forecast error on power generation plan, a data-driven distributed robust joint opportunity constrained scheduling model is constructed, and the affine adjustable strategy shown in formula (6) is used to adjust the power of thermal power units to ensure that the forecast error can be fully absorbed.

[0092]

[0093] Objective function (20) represents minimizing the expected power generation cost of the AC regional power grid under the worst case of wind power, (21) represents the power balance constraint of the AC power grid, (22) represents the power injection limit of the DC interconnection line, (23) represents the power injection limit of the AC interconnection line, (24)-(25) represent the minimum start and stop time constraint of the thermal power unit, (26)-(27) represent the start and stop action constraint of the thermal power unit, and the combined opportunity constraints (28)-(30) ensure that even under the worst probability distribution of wind power, the maximum / minimum output limit, climbing / slipping rate and line transmission capacity constraints of the thermal power unit can be met.

[0094]

[0095]

[0096] (18)-(19)(31)

[0097] In the formula, represents the set of loads, thermal power units, lines, wind farms, DC tie line nodes, and AC tie line nodes in the AC power grid a. represents the AC power grid, R i represents the ramp rate limit of thermal power unit i, represents the load of node i at time t, P i G , Indicates the lower and upper limits of the output of thermal power unit i, u it 、v it 、w it They represent the start / stop status, start-up action, and shutdown action of thermal power unit i at time t, respectively. represents the startup cost of thermal power unit i, MU i ,MD i are the minimum start-up time and minimum shutdown time of thermal power unit i, represents the predicted value of wind farm i at time t, ∈ 1-3 represents the risk level of the joint opportunity constraint, M gl 、M wl 、M dl 、M il 、M jl It represents the power transfer distribution factor of thermal power units, wind farms, loads, DC interconnection lines, and AC interconnection lines. represents the upper limit of the transmission power of line l, a i , b i 、c i They represent the consumption characteristic coefficient of thermal power unit i, d it represents the output adjustment cost coefficient of thermal power unit i, They represent the output of thermal power unit i at time t after the uncertainty occurs and in the prediction scenario, respectively, it represents the power regulation factor of thermal power unit i at time t, ξ jt represents the prediction error of wind farm j at time t.

[0098] 3. Data-driven distributed robust joint opportunity-constrained scheduling model reconstruction

[0099] First, the data-driven distributed robust joint opportunity-constrained scheduling problem is written into a compact form as shown in Equation (10). This model considers different nodes and different times Wind power prediction error ξ j,t The spatiotemporal correlation of wind power forecasting is not required, and there is no need to assume that wind power forecasting errors are independent and identically distributed.

[0100] Objective function:

[0101] Joint opportunity constraints:

[0102] In the formula, c 0,1 、c 0,2 and C0 represent the coefficient matrix of the objective function, and the elements in C0 are either 0 or 1. and represents the power regulation factor of the thermal power unit at time t, and t∈T, T represents the total number of scheduling periods; the matrix and Indicates that α t Other decision variables besides represents the probability distribution of wind power output, represents the set of all possible probability distributions of wind power output, represents the expected value, ξ represents the wind power prediction error, Indicates The risk level of the joint opportunity constraint is joint opportunity constraints, and No. The coefficient matrix of the joint opportunity constraints.

[0103] Without loss of generality, assume that the wind power prediction error ξ is subject to a polyhedral constraint Ω, and Ω = ξ:Lξ≤l, L and l represent the parameters constituting the polyhedron Ω, then the expected term in the objective function (1) is It is equivalent to solving the following convex cone programming problem, namely A:

[0104]

[0105] In the formula, θ represents the Wasserstein radius, N represents the number of samples of wind power prediction error, represents the sample value of the i-th wind power prediction error, κ o , They are all auxiliary variables.

[0106] The distributed robust joint opportunity constraint (2) can be further expressed as:

[0107]

[0108] Among them, C k and c k are the k-th row of the coefficient matrices C and c respectively.

[0109] A classic method to deal with the distributed robust joint chance constraints shown in equation (33) is to use Bonferroni approximation (BA) to transform the joint chance constraints into multiple independent chance constraints as follows:

[0110]

[0111] Then, the worst-case value at risk is used to express the individual independent opportunity constraints:

[0112]

[0113] in, is defined as:

[0114] However, in the above classic BA processing method, ∈ k The value of will significantly affect the accuracy of the distributed robust joint chance constraint model, so it is necessary to choose a suitable k But ∈ k It is very difficult to choose the value of However, this value will make the model results too conservative.

[0115] To this end, the present invention proposes an optimized conditional value at risk (OCA) processing method to transform the distributed robust joint chance constraints (28)-(30) into a single tractable form.

[0116] First, reconstruct (33) as:

[0117]

[0118] Among them, δ is the scaling parameter.

[0119] Using the conditional value at risk approximation (34), we get:

[0120]

[0121] When the scaling parameter δ is chosen optimally, the approximation of Equation (35) becomes very accurate, and thus the worst-case CVaR constraint can be effectively evaluated.

[0122] according to For a given scaling parameter δ, Equation (35) can be further transformed into the following convex cone programming form, namely B:

[0123]

[0124] In the formula, κ c , σ c , β c , γ c is an auxiliary variable.

[0125] Therefore, the data-driven distributed robust joint chance-constrained scheduling problem (10) that considers the uncertainty of wind power output and the spatiotemporal correlation of prediction errors can be written as the OCA-based reconstruction form shown in formula (3).

[0126]

[0127] Problem (3) provides an upper bound for the distributed robust joint chance-constrained scheduling problem (10) by optimizing the scaling parameter δ∈Π ++ To find the optimal solution to problem (3), it is:

[0128]

[0129] However, when the scaling parameter δ is a variable, problem (4) becomes a non-convex optimization problem. To this end, the present invention further proposes an iterative sequential convex programming (SCP) algorithm to handle non-convex problems by sequentially optimizing (x, Λ) and Δ in an alternating direction. The iterative process of the SCP algorithm is as follows.

[0130] Step 1: Initialize the iterative residual A large constant φ>0, let the scaling parameter For the The number of constraints in a joint opportunity constraint;

[0131] Step 2: Solve the following problem:

[0132]

[0133] (13), (19)

[0134] Step 3: If If it is satisfied or reaches the maximum number of iterations set by the SCP algorithm, the optimal solution (x, Λ, ν) is returned, otherwise go to step 4;

[0135] Step 4: Update by solving (5) And return to step 2.

[0136]

[0137] In the SCP algorithm, an auxiliary slack variable ν ≥ 0 is introduced in problem (4) and penalized by the penalty coefficient ω in the objective function to ensure the feasibility of the algorithm when the initialization value of the scaling parameter is poorly selected.

[0138] The effect of the present invention will be verified by means of test examples.

[0139] In this embodiment, the test environment is carried out in a personal computer equipped with Intel Core i7-12650H, 2.3GHz, 16GB RAM, and 10 cores. The parallel computing toolbox is called in the MATLAB R2021a platform, and Gurobi10.0 is used for solving.

[0140] The test case consists of three 6-node AC grids A, B, and C. AC grid A is connected to AC grids B and C through two DC interconnection lines, and AC grid B is connected to AC grid C through an AC interconnection line. In AC grid A, two wind farms are located at nodes 2 and 4 respectively. The output of each wind farm and load in AC grid A is shown in the figure. Figure 2 The transmission capacity upper and lower limits of the DC interconnection line are [50, 150] MW, the transmission capacity upper and lower limits of the AC interconnection line are [0, 100] MW, the power adjustment rate of the DC interconnection line is [10, 30] MW, the number of power adjustments allowed per day is 6 times, and the minimum power adjustment duration time interval is 2 hours.

[0141] (1) Effectiveness of the flexible consumption model of DC interconnection lines

[0142] The role of the proposed DC interconnection line flexible consumption model in promoting the cross-regional consumption of high-proportion renewable energy is as follows: Figure 3 As shown in the figure, the transmission power of DC tie lines AB and AC is consistent with the change of wind power output in AC grid A. In periods 2 to 9, the wind power output increases, and the transmission power of DC tie lines also increases, transmitting more power to AC grids B and C. Similarly, in periods 21 to 24, when the load of the AC grid increases, the transmission power of DC tie lines increases to support the load demand of AC grids B and C. It can be seen that the proposed DC tie line flexible consumption model can realize the joint optimization of DC tie line transmission power and AC grid generator output, and better adapt to the fluctuation of large-scale wind power output.

[0143] (2) Effectiveness of the data-driven distributed robust joint opportunity-constrained scheduling model considering the uncertainty of renewable energy output and the spatiotemporal correlation of prediction errors

[0144] The proposed distributed robust model based on OCA reconstruction and iterative solution using the SCP algorithm is compared with the classic distributed robust model based on BA. The system operation cost comparison under different sample numbers N is shown in Figure 2. Figure 4 As shown in Figure 2, it can be found that the size of the system operation cost is negatively correlated with the size of the sample size N, that is, when the sample size N is larger, the system operation cost is smaller. Figure 4 It can be seen from the above that under a certain sample size N, the system operation cost of the classic BA-based distributed robust model is higher than that of the proposed distributed robust model based on OCA reconstruction and SCP solution, that is, the proposed OCA reconstruction model can reduce the conservatism of the distributed robust scheduling model. Moreover, when the Wasserstein radius θ is close to 10 -3 When θ is less than 10, the operating cost of the BA-based system increases dramatically. It can be concluded that when the Wasserstein radius θ is large, the solution obtained by BA is more conservative than the proposed model based on OCA reconstruction and SCP solution. In addition, when the Wasserstein radius θ is less than 10 -2 When , the operating cost of the proposed model reconstructed based on OCA and solved by SCP does not change significantly, which means that the proposed distributed robust scheduling model based on OCA reconstruction, which considers uncertainty and spatiotemporal correlation of prediction errors, has a critical Wasserstein radius θ, which corresponds to better solution performance.

[0145] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A data-driven distributed robust joint opportunity-constrained scheduling method for AC / DC hybrid power systems, characterized in that: The following steps are involved: Construct a flexible consumption model for DC interconnection lines; A data-driven distributed and robust joint opportunity-constrained dispatching model is constructed based on the flexible consumption model of DC tie lines; The data-driven distributed robust joint opportunity-constrained scheduling model is reconstructed and solved, including: S1. Rewrite the data-driven distributed robust joint opportunity-constrained scheduling model into a compact form: Objective function: Joint opportunity constraints: In the formula, c 0,1 、c 0,2 and C0 represent the coefficient matrix of the objective function, and the elements in C0 are either 0 or 1. and represents the power regulation factor of the thermal power unit at time t, and t∈T, T represents the total number of scheduling periods; the matrix and Indicates that α t Other decision variables besides represents the probability distribution of wind power output, represents the set of all possible probability distributions of wind power output, represents the expected value, ξ represents the wind power prediction error, ∈ l represents the risk level of the l-th joint opportunity constraint. For the l-th joint opportunity constraint, C l (·) and c l (·) The coefficient matrix of the lth joint opportunity constraint; S2. According to the optimal conditional value at risk approximation OCA processing method, the compact form of the data-driven distributed robust joint opportunity constraint scheduling model is reconstructed to obtain the reconstructed model Φ(Δ), where the objective function represents the sum of the power generation cost under the wind power forecast scenario and the power generation adjustment cost under the wind power error scenario: In the formula, δ l The newly introduced scaling parameter for the lth joint opportunity constraint, ∈ is the risk level of the joint opportunity constraint, τ is an auxiliary variable, and L(·) represents the loss function; S3. Find the optimal solution to equation (3) through the iterative sequential convex programming SCP algorithm: S31. Initialize iterative residual Constant φ>0, let For the The number of constraints in a joint opportunity constraint; S32. Introduce slack variables to reconstruct equation (3) into equation (4), and iteratively solve equation (4): Where, ν l is the auxiliary slack variable introduced for the lth joint opportunity constraint, and ν l ≥0, ω is the penalty coefficient; A represents the expected term in the objective function (1) After the equivalent convex cone programming problem, B represents the convex cone constraint after the joint opportunity constraint (2) is processed by the optimal conditional risk value approximation OCA; S33. If satisfied Or if the preset maximum number of iterations is reached, the current solution (x, Λ, ν) is the optimal solution, otherwise proceed to S34; S34. Update δ by solving (5) and returning to S32 l ; Among them, δ is the scaling parameter, Π ++ Represents the relative interior of the probability simplex.

2. According to claim 1, a data-driven distributed robust joint opportunity-constrained scheduling method for an AC / DC hybrid power system is characterized in that: The flexible consumption model of DC tie lines includes: DC tie line power limit: DC tie line power adjustment rate constraints: DC tie line power adjustment direction constraints: Limitation on the number of times the DC tie line power can be adjusted: DC tie line power adjustment direction restrictions: Step-by-step operation characteristics of DC tie line transmission power: In the formula, represents the power of the ith DC tie line at time t, P i DC and They represent the lower and upper limits of the transmission power of the ith DC tie line, represents the set of DC tie lines, Represents a set of time periods, and They represent the upward and downward power regulation status flags of the ith DC tie line at time t, respectively. and denote the upward and downward power regulation rates of the ith DC tie line, κ DC Indicates the number of power adjustments allowed by the DC tie line, M + and M - represents a positive number, τ represents the shortest duration after the DC tie line adjusts the power once, and Indicates whether the i-th DC tie line starts and ends power regulation at time t.

3. The data-driven distributed robust joint opportunity-constrained scheduling method for an AC / DC hybrid power system according to claim 2 is characterized in that: The data-driven distributed robust joint opportunity-constrained scheduling model is: The objective function is to minimize the expected power generation cost of the AC regional power grid under the worst wind power conditions: Among them, (21) represents the AC power balance constraint, (22) represents the power injection limit of the DC interconnection line, (23) represents the power injection limit of the AC interconnection line, (24)-(25) represent the minimum start-stop time constraint of the thermal power unit, (26)-(27) represent the start-stop action constraint of the thermal power unit, and (28)-(30) are joint opportunity constraints; In the formula, and They represent the output of thermal power unit i at time t after the uncertainty occurs and in the prediction scenario, respectively, it represents the power regulation factor of thermal power unit i at time t, ξ jt represents the prediction error of wind farm j at time t, and They represent the load, thermal power unit, line, wind farm, DC tie line node and AC tie line node set in AC power grid a respectively. represents the AC power grid, a i 、b i and c i They represent the consumption characteristic coefficient of thermal power unit i, u it 、v it and w it They represent the start / stop status, start-up action and shutdown action of thermal power unit i at time t respectively. represents the startup cost of thermal power unit i, d it represents the output adjustment cost coefficient of thermal power unit i, represents the predicted value of wind farm i at time t, represents the load of node i at time t, represents the power of the ith AC tie line at time t, P i AC and They represent the lower and upper limits of the transmission power of the i-th AC tie line, MU i and MD i are the minimum start-up time and minimum shutdown time of thermal power unit i, respectively, P i G , Represents the lower and upper limits of the output of thermal power unit i, R i represents the ramp rate limit of thermal power unit i, ∈1, ∈2 and ∈3 represent the risk levels of joint opportunity constraints, represents the transmission power upper limit of line l, w represents the wind farm, d represents the load in AC grid a, g represents the thermal power unit, j represents the jth AC tie line node, M gl 、M wl 、M dl 、M il and M jl They represent the power transfer distribution factors of thermal power units, wind farms, loads, DC interconnection lines and AC interconnection lines respectively.

4. The data-driven distributed robust joint opportunity-constrained scheduling method for an AC / DC hybrid power system according to claim 1, characterized in that: The wind power prediction error ξ is subject to a polyhedral constraint Ω, and Ω=ξ:Lξ≤l, L and l represent the parameters constituting the polyhedron Ω. The expected term in the objective function (1) is It is equivalent to solving the convex cone programming problem, namely A: In the formula, θ represents the Wasserstein radius, N represents the number of samples of wind power prediction error, represents the sample value of the i-th wind power prediction error, κ o , They are all auxiliary variables.

5. The data-driven distributed robust joint opportunity-constrained scheduling method for an AC / DC hybrid power system according to claim 1, characterized in that: The specific contents of S2 include: The joint opportunity constraint (2) is reconstructed by the optimal conditional value at risk approximation OCA processing method: The joint opportunity constraint (2) can be further expressed as: Among them, C k and c k Respectively represent C l (·) and c l The kth row of (·); Refactor (33) as follows: Using the conditional value at risk approximation (34), we get: according to For a given scaling parameter δ, Equation (35) is further transformed into the following convex cone programming form, namely B: In the formula, κ c , σ c , β c , γ c is an auxiliary variable; Then (1)-(2) are rewritten into the form based on OCA reconstruction as shown in formula (3).

Images