Robust dispatching method, system and device for cross-zone frequency support distribution of new energy base

Abstract

The application provides a new energy base cross-region frequency support distribution robust scheduling method, system and equipment, and relates to the field of power system scheduling. The method comprises the following steps: a cross-region frequency response model of energy storage fast frequency response and DC virtual inertia cooperation is constructed, considering the virtual inertia response delay, the frequency change rate, the lowest frequency point and the quasi-steady frequency constraint of the receiving end power grid are established; based on the historical source and load prediction error and the current prediction characteristic screening condition neighbor sample, a Wasserstein-probability box mixed distribution fuzzy set is constructed, and an entropy risk value EVaR is introduced to depict the tail risk of extreme scenarios; on this basis, a new energy base and receiving end power grid risk perception distribution robust cooperative scheduling model is established, a hierarchical adaptive ADMM method fusing multi-breakpoint segmented robust affine rules is used for distributed solution, and a cross-region scheduling scheme is obtained. The application can improve the new energy consumption level and the collaborative scheduling efficiency.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching, and in particular to a method, system and equipment for cross-regional frequency support sub-brush dispatching in new energy bases. Background Technology

[0002] With the continuous advancement of the low-carbon transformation of the power system, large-scale renewable energy (RES) bases in desert and Gobi regions, mainly based on wind and solar power, are developing rapidly and gradually achieving large-scale inter-regional power transmission through high-voltage direct current (HVDC) transmission channels. However, large-scale renewable energy (RES) bases are usually far from load centers and need to rely on HVDC interconnections for inter-regional power transmission. The contradiction between their random fluctuations in output and the declining frequency support capacity of the receiving-end grid is becoming increasingly prominent. On the one hand, the uncertainty of wind and solar power output and receiving-end load exacerbates the system's operational risks, especially under extreme prediction errors or high-loss scenarios. On the other hand, DC connection to the receiving-end grid will reduce the online capacity of conventional units, weaken the system's equivalent inertia and primary frequency regulation capability, and make frequency security risks more prominent. In addition, RES bases and receiving-end grids belong to different dispatching entities, and cross-regional collaborative dispatching faces limited data sharing and difficulties in decentralized decision-making coordination. Therefore, how to construct a collaborative dispatching method for RES bases and receiving-end grids that balances frequency security and distributed solution efficiency under source-load uncertainty is an urgent problem to be solved in the operation of large-scale RES bases.

[0003] To ensure the stable cross-regional power transmission from the Shagohuang renewable energy base via DC interconnection lines, extensive research has been conducted on the cross-regional dispatching problem between the renewable energy base and the receiving-end power grid. However, existing research still has the following shortcomings: First, existing dispatching methods often consider the uncertainty of renewable energy output or the constraints of single-region operation, and are insufficient in characterizing the source-load error distribution characteristics under forecast conditions and the tail risk in extreme scenarios, making it difficult to effectively suppress high-loss risks such as power curtailment and load shedding; Second, after large-scale renewable energy is fed into the receiving-end power grid via DC, the operating space of conventional synchronous machines is compressed, and the system's equivalent inertia and frequency support capability decrease. Existing frequency safety dispatching methods have not fully explored the cross-regional frequency support potential of energy storage and DC interconnection lines; Third, renewable energy bases and receiving-end power grids belong to different dispatching entities, making it difficult to centrally share operational data. Existing cross-regional collaborative dispatching algorithms are prone to problems such as slow convergence, residual oscillation, and insufficient solution stability in multi-entity optimization scenarios with discrete variables. Summary of the Invention

[0004] The purpose of this invention is to provide a method, system and equipment for cross-regional frequency support and distributed scheduling of new energy bases. This method can reduce the operational risks in extreme scenarios and improve the level of new energy consumption and the efficiency of cross-regional collaborative scheduling while ensuring the frequency security of the receiving-end power grid.

[0005] To achieve the above objectives, this invention provides a cross-regional frequency support sub-blob scheduling method for new energy bases, comprising the following steps: Step S1: Construct a cross-regional frequency response model under the synergistic effect of fast frequency response of energy storage and DC virtual inertia, consider the delay of DC virtual inertia response, and establish frequency security constraints of the receiving-end power grid. Step S2: Based on historical source load prediction errors and current prediction features, select conditional nearest neighbor samples, construct a Wasserstein-probability box hybrid distribution fuzzy set, and introduce entropy risk value (EVaR) to characterize tail operation risk in extreme scenarios. Step S3: Based on frequency security constraints and risk measurement, establish a risk perception-based distributed robust collaborative scheduling model between the new energy base and the receiving-end power grid, and use the hierarchical adaptive ADMM method that integrates multi-breakpoint segmented robust affine rules to solve the model in a distributed manner to obtain the cross-regional scheduling scheme.

[0006] This invention also provides a cross-regional frequency support sub-blob scheduling system for new energy bases, comprising: The cross-regional frequency response modeling module is used to construct a cross-regional frequency response model under the synergistic effect of energy storage fast frequency response and DC virtual inertia, taking into account the DC virtual inertia response delay and establishing frequency security constraints of the receiving-end power grid. The distributed fuzzy set construction module is used to filter conditional nearest neighbor samples based on historical source load prediction errors and current prediction features, construct Wasserstein-probability box hybrid distributed fuzzy sets, and introduce entropy risk value (EVaR) to characterize tail running risk in extreme scenarios. The collaborative scheduling solution module is used to establish a risk-aware, distributed, robust collaborative scheduling model between new energy bases and receiving-end power grids based on frequency security constraints and risk metrics. The model is solved in a distributed manner using a hierarchical adaptive ADMM method that integrates multi-disruption segmented robust affine rules to obtain cross-regional scheduling schemes.

[0007] The present invention also provides a computer device, including: a memory and a processor; the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-mentioned cross-regional frequency support sub-brush scheduling method for new energy bases.

[0008] Therefore, the present invention employs the above-mentioned cross-regional frequency support sub-blob scheduling method, system, and equipment for new energy bases, and the beneficial technical effects are as follows: (1) This invention constructs a cross-regional frequency response model under the synergistic effect of fast frequency response of energy storage and DC virtual inertia, and introduces the DC virtual inertia response delay correction of the frequency change rate constraint of the receiving-end power grid, the frequency minimum point constraint and the quasi-steady-state frequency constraint. It fully explores the cross-regional frequency support potential of the energy storage system of the new energy base and the DC interconnection line, effectively alleviates the problem of the reduction of system equivalent inertia and frequency support capability caused by the compression of the start-up space of conventional synchronous machines after large-scale new energy is fed into the receiving-end power grid via DC. While ensuring the frequency security of the receiving-end power grid, it reduces the system's dependence on the frequency support of conventional units and improves the level of new energy consumption.

[0009] (2) This invention constructs representative scenarios by selecting neighboring samples based on historical source-load prediction errors and current prediction features, and introduces probability box constraints to construct Wasserstein-probability box hybrid distribution fuzzy sets, which solves the problem that existing methods cannot adequately characterize the source-load error distribution features under prediction conditions. At the same time, it introduces entropy risk value (EVaR), which characterizes the tail operation risk through the worst expected form under relative entropy constraints, and defines the difference between EVAR and nominal expected cost as risk premium, which effectively suppresses high-loss operation risks such as power curtailment and load shedding under extreme prediction error scenarios, and improves the robustness and reliability of the system under harsh operating scenarios.

[0010] (3) This invention proposes a piecewise robust affine strategy based on multi-breakpoint folding and dimensionality-upgrading features, which overcomes the shortcomings of traditional single-segment affine strategies in characterizing the differences in prediction error direction, interval change features and multi-resource response differences, and improves the adaptability of scheduling decisions to prediction errors. At the same time, it adopts a hierarchical adaptive ADMM method, placing discrete variable updates in the outer layer and continuous variable coordination in the inner layer, and generates initial integer solutions through soft guidance and adaptively adjusts the penalty factor based on the relative size of the original residual and the dual residual. This solves the problems of slow convergence, residual oscillation and insufficient solution stability in multi-agent distributed optimization with discrete variables. While ensuring solution accuracy, it significantly shortens the solution time and improves the computational efficiency of cross-regional collaborative scheduling. Attached Figure Description

[0011] Figure 1 A flowchart illustrating the overall design of the sub-blue bar scheduling method for cross-regional frequency support in new energy bases; Figure 2 This is a graph showing the frequency variation. Figure 2 (a) in the figure is a graph showing the rate of change of frequency (RoCoF); Figure 2 (b) in the figure is a diagram showing the changes in Nadir; Figure 3 This is a power output diagram for new energy bases that only considers the frequency support of conventional generating units; Figure 4The output diagram of the new energy base is taken into account, which is supported by the frequency of energy storage-direct current synergy. Figure 5 A comparison diagram of different affine strategies; Figure 6 A comparison chart of iterative solutions obtained using different algorithms; Figure 7 Figures showing the operational results under different energy storage configurations and capacities; Figure 8 The graph shows the operating results under different DC virtual inertia delays. Detailed Implementation

[0012] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0013] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0014] Example 1 like Figure 1 As shown, the cross-regional frequency support sub-Bluerged scheduling method for new energy bases includes the following steps: Step S1: Construct a cross-regional frequency response model under the synergistic effect of fast frequency response of energy storage and DC virtual inertia, considering the delay of DC virtual inertia response, and establish constraints on the frequency change rate of the receiving-end power grid, the minimum frequency point, and the quasi-steady-state frequency.

[0015] Step S11, Cross-regional frequency support principle based on HVDC: By introducing a frequency-related additional term into the active power reference value of a constant power converter station, the transmission power of the DC tie line can automatically respond to frequency changes in the disturbed area. The specific details are as follows: (1); (2); in, The transmission power after frequency modulation is taken into account for the DC tie line; DC tie-line transmission power without considering frequency support; This is the frequency modulation power support component generated by the frequency controller; Frequency regulation power provided for the conventional units at the RES base; Frequency regulation power provided for the energy storage system at the RES base; This refers to the scheduling period.

[0016] To ensure that the frequency modulation power of the RES base can be stably transmitted through the DC tie line, the capacity constraint of the DC tie line can be reconstructed by introducing upper and lower power margins, as shown in equations (3)-(4): (3); (4); in, and For the vertical adjustment margin of the connecting line; and Adjust the power limit for the connection line.

[0017] Step S12: After a frequency disturbance occurs in the receiving-end power grid, the DC tie line provides virtual inertia support through transient energy exchange of the DC-side equivalent capacitance, which mainly acts in the early stage of the disturbance; the energy storage system provides fast frequency response (FFR) and continuously injects active power in the following few seconds until primary frequency regulation (PFR) is gradually established.

[0018] High-voltage direct current converter stations based on voltage source converters (VSCs) can modulate the DC-side voltage in response to frequency deviations in the receiving-end grid. Dynamic changes in the DC voltage will further cause changes in the energy stored in the DC-side capacitors, and virtual inertia support will be achieved through transient energy exchange. For a VSC-HVDC converter station, the electrostatic energy stored in the DC-side capacitors is shown in equation (5); after differentiating it with respect to time, the expression for instantaneous power exchange can be obtained, as shown in equation (6). (5); (6); in, For capacitor energy; This refers to the capacitance of the DC capacitor. This is a reference value for DC voltage. For capacitor power exchange; To achieve inertia simulation, the HVDC controller modulates the DC voltage reference value based on the frequency deviation of the receiving-end power grid, even with a TB delay. Based on the equivalent relationship between capacitor power exchange and inertial response power, equation (7) can be obtained: (7); in, The time constant of DC virtual inertia; The rated frequency of the power grid; To account for the frequency signal after the DC virtual inertial response delay; Delay for DC virtual inertial response; This refers to the capacity of the DC converter station.

[0019] Equation (6) shows that DC-side capacitor voltage modulation can generate transient power exchange. To characterize the inertial support effect of this process on the frequency dynamics of the receiving-end power grid, this method performs equivalent matching with the inertial power term in the system's equivalent rotor motion equation, thereby obtaining an explicit characterization of the HVDC virtual inertia, as shown in Equation (8): (8); in, The value is the DC virtual inertia (MW·s).

[0020] Synchronous generator units within the RES base can provide primary frequency regulation (PFR) to the receiving-end grid via DC tie lines. Their dynamic response process includes two stages: initial delay and subsequent linear ramp-up. Unlike synchronous generator units, ESS can rapidly inject active power via power electronic interfaces, providing fast frequency response (FFR), as shown in equation (9). (9); in, For energy storage systems at all times Provided FFR; For the FFR delay of the energy storage system; For the FFR capacity of the energy storage system; This refers to the FFR ramp-up time of the energy storage system.

[0021] Step S13: Based on the rotor motion equation, establish the frequency response model of the receiving-end power grid under the coordinated support of energy storage, DC, and thermal power. In the initial period after the disturbance occurs, the virtual inertia of the DC system has not yet been involved, and the system frequency response process is as shown in equation (10). When > At this time, the virtual inertia of the DC tie line begins to participate in the response, increasing the system's equivalent inertia. Its dynamic process can be expressed as equation (11).

[0022] (10); (11); in, The inertia of the receiving-end synchronizing machine; This refers to the frequency deviation of the receiving-end power grid. This refers to the frequency modulation power of the synchronous machine; This represents the load disturbance power.

[0023] Based on the dynamic frequency response process described in equations (10)-(11), the key indicators for frequency security can be further derived: the rate of change of frequency (RoCoF) constraint, the minimum frequency point constraint, and the quasi-steady-state constraint.

[0024] (1) RoCoF constraint.

[0025] Because of the response delay of the virtual inertia of the DC system, the RoCoF in the early stage of the disturbance is determined only by the inertia of the synchronous unit.

[0026] (12); in, The actual frequency change rate of the receiving-end power grid; The initial moment of the disturbance This represents the maximum rate of frequency variation that the receiving-end power grid can accept.

[0027] (2) Minimum frequency constraint.

[0028] When the system is disturbed, in order to prevent the frequency from dropping too much and causing a serious fault, which in turn leads to the system disconnection, a minimum frequency threshold needs to be set to limit the drop, as shown in equation (13): (13); in, This represents the actual frequency change value; This represents the maximum frequency change value.

[0029] To quickly obtain the analytical expression for the minimum frequency constraint, the TB delay of the DC virtual inertia is first ignored, and the minimum frequency constraint ignoring the virtual inertia delay is obtained through the equivalent transformation of equation (12), as shown in equation (14): (14); in, For the delay time of conventional units; Subsequently, the time delay factors represented by equations (15)-(17) are introduced into equation (14) and modified to obtain constraints that better fit the actual operating characteristics.

[0030] (15); (16); (17); in, To account for the actual frequency change caused by the DC virtual inertia delay; and These are the delay times for conventional generating units and energy storage systems, respectively, taking into account the DC virtual inertia delay.

[0031] Taking into account the effects of delayed intervention, through further mathematical derivation and simplification, the minimum frequency constraint considering the virtual inertia response delay can be obtained, as shown in equation (18): (18).

[0032] (3) Quasi-steady-state constraints.

[0033] To ensure that the system frequency can effectively recover to its rated value after a drop, the primary frequency modulation power provided by the system must meet the following constraints: (19).

[0034] Step S2: Based on historical source load prediction errors and current prediction features, select conditional nearest neighbor samples, construct a Wasserstein-probability box hybrid distribution fuzzy set, and introduce entropy risk value (EVaR) to characterize tail operation risk in extreme scenarios.

[0035] Step S21: First, construct the hourly relative error based on the historical actual value and the predicted value, as shown in Equation (20). To reduce the interference of extreme samples on conditional nearest neighbor identification and scene construction, the relative error is truncated by quantiles to obtain robust error samples as shown in Equation (21): (20); (twenty one); in, This represents the historical relative prediction error. , They are the historical first Heavenly Actual and predicted values ​​for the time period; for The lower bound of the denominator for the time period is used to avoid abnormal amplification of errors during periods with small predicted values; These are robust error samples after quantile truncation. The lower quantile (lower limit of error cutoff) is 1%. It is the 99th percentile (the upper limit of the error cutoff).

[0036] Secondly, in order to screen historical samples with similar intraday forecast characteristics, this method selects the curve shape and the total daily forecast to represent the intraday change trend and the overall scale, respectively, constructs a conditional feature vector, and uses Euclidean distance to measure sample similarity, as shown in equation (22): (twenty two); in, The Euclidean distance between the historical samples and the feature vector of the target day; For the first Conditional feature vectors of historical samples; This is the conditional feature vector for the current scheduled day.

[0037] The most similar historical samples within the day are selected using equation (22), and a model is constructed based on this. Each scenario, for the scenarios in the set Based on its distance The nominal conditional probability is obtained by assigning weights using a Gaussian kernel function. As shown in equation (23): (twenty three); in, For the target feature vector and the first The distance between scenes; For kernel bandwidth parameters; For the target feature vector and the first The distance between scenes.

[0038] Finally, to avoid excessive migration of scene probabilities to a few extreme scenarios when only Wasserstein distance constraints are used, causing the worst-case distribution to deviate from the actual probability range, this method further introduces probability box constraints. Let the actual probability vector be... ,make and The upper and lower bound constraints of the probability box are satisfied. Thus, the Wasserstein-probability box mixed distribution fuzzy set is obtained, as shown in equation (25): (twenty four); (25); in, For the scene The actual probability; For transpose; It is a fuzzy set of Wasserstein-probability box mixed distribution; for 3D nonnegative real space; and Let be the upper and lower bounds of the probability box; Let be the radius of the Wasserstein sphere; The distance is the Wasserstein distance.

[0039] Step S22: Source load prediction errors may lead to high operational losses in a few scenarios. When only the expected value is used, high-loss scenarios at the tail are easily averaged out, making it difficult to fully characterize operational risks. Therefore, to enhance the model's ability to identify risks in extreme operational scenarios, this method introduces an entropic value-at-risk (EVaR) measure to characterize the tail risk of operational losses caused by the uncertainty of RES output.

[0040] Based on the above, the set of conditional scenarios and the nominal conditional probabilities are obtained. Based on this, the system operating costs of the RES base station side and the receiving-end grid side under nominal conditional probability can be shown in equations (26) and (27), respectively: (26); (27); in, and They represent the nominal conditional probabilities, respectively. The second phase cost of the downstream RES site and the receiving-end power grid; and Representing the scene respectively The second-stage cost of the lower RES base station side and the receiving-end power grid side; For the second phase cost of the RES base site; This refers to the second-stage cost of the receiving-end power grid.

[0041] The operating costs in each scenario are treated as random loss variables defined on a discrete set of scenarios, with confidence levels of 1- The definition of EVAR is shown in equation (28): (28); in, This is a measure of risk / expected risk. These are the positive parameters to be optimized. For the first The weight or probability of each scenario; is the base of the natural logarithm; For the first Loss / cost values ​​in each scenario; The confidence level.

[0042] To illustrate the role of EVAR in characterizing tail risk, the scenario probability is expressed as a nominal conditional probability. The reweighting is performed in the vicinity, and the reweighting magnitude is limited by the relative entropy constraint. Thus, EVAR can be expressed as the worst-case expectation form based on the relative entropy constraint (29)-(30).

[0043] (29); (30); in, This represents the scene probability vector. It is a fuzzy set of probability distributions based on EVAR constraints; Let be a random loss variable.

[0044] Equations (29)-(30) show that EVAR can be expressed as the worst-case expectation under relative entropy constraints, that is, reweighting high-loss scenarios within a finite deviation range of the probability distribution, thereby achieving a more conservative tail risk assessment. Accordingly, this method defines the difference between EVAR and nominal expected cost as the risk premium, and the risk premiums for the RES base and the receiving-end grid are shown in equations (31)-(32), respectively: (31); (32); in, Risk premium for RES sites; Total cost of the RES site taking into account EVAR risk; Risk premium for the receiving-end power grid; Total cost of the receiving-end power grid considering EVAR risk; Accordingly, the system's overall tail risk premium can be obtained by weighted summation of the EVAR risk premiums on both sides, as shown in equation (33): (33); in, The total risk premium of the system; and These represent the risk preference coefficients of the RES base and the receiving-end power grid, respectively.

[0045] Since EVAR involves exponential operations, it is difficult to directly embed it into a mixed-integer linear programming model. To balance risk characterization and solution efficiency, this method draws on the idea of ​​solvable processing of EVAR risk constraints, constructs a finite number of candidate probability vectors based on the nominal conditional distribution, and uses the worst-case weighted expectation under the candidate probability vectors to approximate the tail risk of EVAR, as shown in Equation (34): (34); in, Let be the fuzzy set of candidate probability distributions for EVAR; The first in the candidate probability distribution set A probability distribution vector; The first in the candidate probability distribution set A probability distribution vector; denoted as the number of conditional probability vectors.

[0046] Step S3: Based on frequency security constraints and risk measurement, a risk-aware, distributed, robust, and collaborative scheduling model between the renewable energy base and the receiving-end power grid is established. A hierarchical adaptive ADMM method incorporating multi-disruption segmented robust affine rules is then used to solve this model in a distributed manner to obtain cross-regional scheduling schemes. This model can enhance cross-regional frequency support, reduce operational risks, and improve the robustness of renewable energy consumption and scheduling.

[0047] Step S31: First, a two-stage sub-Blu-ray robust scheduling model for the coordinated operation of the RES base and the receiving-end power grid was constructed. The first stage determines the start-up and shutdown plan of conventional units, and the second stage performs resource scheduling based on representative scenarios constructed from conditional nearest neighbors, taking into account both the robustness and economy of the scheduling scheme.

[0048] The established model aims to minimize the total system cost. The first stage comprises the start-up and shutdown costs of conventional generating units, while the second stage comprises the operating costs and curtailment costs of the real-time adjustment phase, as well as load shedding penalties and risk penalties. The objective function can be expressed as: (35); in, The overall objective function of the system; and The costs for starting and stopping the first phase of generating units at the RES base and the receiving-end power grid are respectively; A set of random scenes; and Scenes The second-stage rescheduling cost of the two systems below.

[0049] For the RES base, the cost includes the unit operating cost and the wind and solar curtailment penalty, as shown in Equation (36); for the receiving-end grid, the cost includes the unit operating cost, the wind and solar curtailment penalty, and the load shedding penalty, as shown in Equation (37). (36); (37); in, For runtime segment; A collection of conventional generating units; For the conventional units of the RES base Unit operating cost; This is the wind power curtailment penalty coefficient; This refers to the amount of wind power curtailed at the receiving end; This refers to the penalty coefficient for photovoltaic curtailment. This refers to the amount of solar power curtailment at the receiving end; For conventional generating units of the receiving-end power grid Operating costs; This is the penalty coefficient for wind power curtailment at the receiving end; This is the penalty coefficient for photovoltaic curtailment at the receiving end; Receiving-end load penalty factor; and These refer to the amount of wind and solar power curtailed; This refers to the load shedding power.

[0050] Phase 1 operational constraints: The first phase determines the start-up and shutdown status and time constraints of conventional generating units in the RES base and the receiving-end power grid. The minimum start-up and shutdown time constraints are shown in equations (38)-(39): (38); (39); in, and These are conventional units exist Minimum running time and minimum downtime at any given time; and These are conventional units Cumulative running time and downtime; For conventional units exist The start / stop status at any given moment.

[0051] Second-stage operational constraints: (1) Operational constraints of various types of resources in the RES base.

[0052] The constraints for new energy operation are shown in equation (40). The constraints for conventional unit operation include upper and lower limits of output, ramping constraints, etc., which will not be elaborated here. The constraints for energy storage operation include: charging and discharging power constraints (41)-(42) considering frequency regulation reserve capacity, charging and discharging state constraints (43), state of charge constraints (44), and capacity constraints (45)-(46).

[0053] (40); (41); (42); (43); (44); (45); (46); in, For the scene Down RES station during the time period Actual output; For different scenarios Lower RES station exist Maximum output at any given moment; For different scenarios New Energy Power Station exist The amount of wind and solar power curtailed at any given moment; For the scene Downlink energy storage system During the period The charging power; For the scene Downlink energy storage system During the period The charging status; For the scene Downlink energy storage system During the period The discharge power; For the scene Downlink energy storage system During the period The discharge state; , energy storage system The upper and lower limits of charging and discharging power; , These are the energy storage frequency regulation power and the frequency regulation power discharge time, respectively. For the scene Downlink energy storage system During the period The state of charge; The time for energy storage charging and discharging power; , These are the upper and lower limits of energy storage capacity, respectively. , These are the charge and discharge efficiencies, respectively. For the scene Downlink energy storage system At the end of the period The energy storage capacity; For the scene Downlink energy storage system In the initial period The amount of energy stored.

[0054] (2) DC tie line operation constraints.

[0055] DC tie lines can quickly correct DC power commands through active power control links to respond to fluctuations in RES output and changes in the load of the receiving-end grid. Equations (47)-(50) are the step constraints of the DC transmission curve. Equation (51) is the power balance constraint of the RES base transmission line.

[0056] (47); (48); (49); (50); (51); in, DC tie lines in the scenario Down Actual transmission power at any given time; , These are DC tie lines in different scenarios. Down Adjusting the power up and down at any given time; , These are DC tie lines at The indicator variables for both upward and downward power adjustment are binary variables. This represents the maximum number of adjustments allowed for the DC tie line. The minimum stable operating time required for the DC tie line; the left side of equation (51) is the RES base output power, and the right side is the DC tie line transmission power. For DC channels in the scenario Down DC power transmitted at all times; , They are respectively a collection of new energy power stations and a collection of conventional generating units; For energy storage systems A set; For the scene Down RES station during the time period Actual output; Scene Down Standard generating units of the receiving-end power grid during the time period Actual output; Scene Down The power transmitted from the DC channel during a given time period.

[0057] (3) Constraints on the operation of the receiving end power grid.

[0058] The receiving-end power grid needs to meet node power balance constraints (52), line power flow constraints (53)-(54), and frequency security constraints; among which, the frequency security constraints are RoCoF constraints (12), frequency minimum point constraints (18), and quasi-steady-state frequency constraints (19). At the same time, conventional units need to meet output constraints that take into account frequency regulation reserves (55)-(56). Their ramp-up, start-up and shutdown constraints and receiving-end RES output constraints are similar to those of the RES base side, and will not be described in detail in this method.

[0059] (52); (53); (54); (55); (56); in, For receiving-end power grid new energy power stations In the scene Down Efforts made at all times; For conventional generating units of the receiving-end power grid In the scene Down Always put in the effort; This refers to the power deficit when transmitting power via the DC channel. For the load of the receiving end power grid node; For receiving end grid node In the scene Down Constantly cut off load power; , These are respectively the line set and the cross-regional DC interconnection line set; , For nodes With nodes The phase angle; For the line The reactance; The phase angle of the balancing node; For receiving end power grid lines Maximum transmission power capacity; For conventional units Minimum output; For conventional units Maximum output; For conventional units contribution; It is a conventional unit In the scene Down Constant frequency modulation power; It is the frequency regulation coefficient of a conventional unit; For receiving end power grid lines In the scene Down There is always a trend of merit.

[0060] Step S32: To achieve distributed solution of the cross-regional dispatch model, this method uses the power of the DC transmission platform as the boundary coupling variable and constructs a multi-breakpoint robust affine decision rule for the two-stage real-time adjustment variables on both sides under a unified uncertain characteristic space. Based on this, the original problem is decomposed into local optimization problems on the RES base side and the receiving end grid side, and the consistency of boundary variables is coordinated through ADMM. Among them, the boundary coupling variables are shown in equations (57)-(58): (57); (58); in, and Scenes Boundary coupling variables between the lower RES base and the receiving-end power grid; For the scene Time period The RES base contributes; For the scene Time period The power received by the receiving-end power grid.

[0061] In the segmented robust scheduling model, the second stage typically employs affine decision rules for parameterization to avoid dimensionality inflation caused by explicit expansion of scene variables. However, traditional single-segment affine strategies still have shortcomings in characterizing the directional differences, interval variation features, and resource response differences in the impact of errors. Therefore, this method proposes a segmented robust affine strategy based on multi-breakpoint folding and dimensionality-increasing features. First, let the scene... Time period The original prediction error vector As shown in equation (59): (59); in, , , , , These correspond to the prediction errors for base wind power, base photovoltaic, receiving-end wind power, receiving-end photovoltaic, and receiving-end load, respectively.

[0062] For any scalar error component of equation (59), this method pre-sets a multi-breakpoint mesh and constructs a piecewise folding feature accordingly, as shown in equation (60): (60); in, The values ​​are piecewise linear functions / activation functions. For the first Scene, time period ,variable The input variable values; For variables The Each segmented breakpoint; This represents the number of segments; breakpoints are adaptively selected from historical error samples.

[0063] When the sample contains both positive and negative deviations, the zero point is retained and representative positions on the positive and negative sides are selected as internal breakpoints to enhance the characterization of the asymmetry of the error distribution; when the sample does not cross zero, segmented grids are constructed using equidistant breakpoints within the interval.

[0064] Based on equation (60), the original scalar error is expanded into several non-negative piecewise features; by stacking the folded features corresponding to each dimension of error, the dimensionality-upgraded error feature vector can be obtained, as shown in equation (61): (61); in, For different scenarios Down The dimensionality increase error eigenvector at time step 1.

[0065] To address the response characteristics of different regulation resources, this method constructs differentiated error signals: conventional units use hourly power deviation (Equation 59), DC power transmission uses average error within the stepped segment (Equation 62), and energy storage systems use a combined wind and solar load error signal (Equation 63).

[0066] (62); (63); in, For the first In the scenario, the first Class variables in time period set The average error signal within; For use with DC channel The corresponding set of scheduling periods; for In the scenario, the first Class variables in time period The prediction error signal; A collection of DC channels; Provide comprehensive error signals for the RES base; For receiving the comprehensive power error signal from the receiving-end power grid; This is the receiving end load error signal.

[0067] After performing multi-breakpoint folding and dimensionality enhancement on the error distribution characteristics of the different resources mentioned above according to equation (61), the second-stage adjustment decision variables can be uniformly represented as piecewise robust affine forms, as shown in equations (64)-(67): (64); (65); (66); (67); in, For the scene Down Shike New Energy Conventional Units contribution; For the scene Down Conventional generating units at the receiving end of the power grid contribution; For the scene Downline contribution; For the scene Downlink energy storage system contribution; , , and For new energy conventional units Conventional generating units in the receiving-end power grid DC channel Energy storage system The baseline scheduling part; , , They are respectively new energy conventional units Conventional generating units in the receiving-end power grid DC channel The piecewise response coefficient matrix; , , These are the error feature vectors corresponding to the RES base, the receiving-end power grid, and the DC channel, respectively. , , energy storage system The segmented response coefficient matrix of the RES base, receiving-end power grid new energy, and receiving-end power grid load; , , energy storage system The error feature vectors corresponding to the RES base, the error feature vectors corresponding to the new energy sources in the receiving-end power grid, and the error feature vectors corresponding to the loads in the receiving-end power grid.

[0068] After introducing multi-breakpoint folding for dimensionality increase, the worst-case constraint needs to take the maximum value in the dimensional feature space, which makes direct solution difficult. To facilitate subsequent dualization processing, the value range of the folded feature is represented as a polyhedral support set. Specifically, the static tangent plane constraints corresponding to each error dimension are stacked uniformly to obtain the dimensional feature support set shown in Equation (68).

[0069] (68); in, It is an uncertain set; The constraint matrix for an uncertain set; For the feasible set of uncertain quantities; is the constraint boundary vector of the uncertain set.

[0070] Based on this, for any linear term Its supporting set The worst-case value can be rewritten as equation (69) through dualization, which converts the piecewise robust constraint containing the maximization operation into a deterministic constraint that can be directly embedded into the optimization model.

[0071] (69); in, This is the transpose of the coefficient vector of the linear objective function; The constrained transpose matrix of an uncertain set; The dual variable vector; For the transpose of the constraint boundary vector of an uncertain set; is the coefficient vector of the linear objective function.

[0072] Step S33: Since the subproblem contains 0-1 discrete variables such as unit start-up and shutdown, directly using ADMM coordination to solve it can easily lead to boundary power jumps during discrete state switching, causing residual oscillations and convergence difficulties. To address this, this embodiment introduces a hierarchical alternating coordination mechanism, placing discrete variable updates in the outer layer and continuous variable coordination in the inner layer, achieving relative decoupling between mixed integer optimization and boundary consistency constraints.

[0073] No. Discrete decision set of the outermost iteration As shown in equation (70): (70); in, These are discrete decision variables on the RES base side; These are the discrete decision variables of the energy storage system; These are the discrete decision variables of the receiving-end power grid.

[0074] The corresponding boundary consensus quantity is denoted as In the first outer iteration, to avoid infeasibility issues caused by directly fixing the initial boundary values, this method uses a soft-guided approach to generate initial integer solutions, that is, adding deviation penalty terms to both subproblems: (71); in, as the main body The original objective function; as the main body Decision variables; This is the penalty coefficient for soft guidance; These are slack variables; as the main body The response variable; These are the initial coordination variables; Represents a new energy base; This represents the receiving-end power grid. Equation (71) can ensure the feasibility of the subproblem while guiding the solutions on both sides of the boundary to gradually approach the given initial value.

[0075] Starting from the second outer iteration, the boundary consensus value returned by the inner coordination in the previous round is fixed. Solve the mixed integer subproblems on the base side and the receiving side separately. Let the updated set of discrete variables be denoted as . Then the outer convergence criterion is as shown in equation (72): (72); in, For the first Consistency ratio of outer start and stop variables in the next iteration; For the first For each iteration, there is a discrete set of decisions. The threshold for convergence criterion of changes in outer-layer decision variables; Let be the dimension of the discrete decision set. The outer loop terminates when the change ratio of the decision variables in two consecutive rounds of outer iterations meets a threshold.

[0076] After fixing the outer 0-1 variables, the inner problem becomes an optimization problem containing only continuous variables, and is further decomposed into two sub-problems: the RES base side and the receiving-end grid side, as shown in equations (73)-(74): (73); (74); in, and The first Decision variables for the RES base and receiving-end power grid in the next iteration; For the decision variables of the RES base; For the decision variables of the receiving-end power grid; and These are the local boundary variables for the RES base and the receiving-end power grid, respectively; For the first Boundary consensus variables for the next iteration; and Given discrete decision sets The feasible region for the RES base and the receiving-end power grid; and These are the objective functions for the RES base and the receiving-end power grid, respectively. These are unscaled Lagrange multipliers; This is a penalty factor.

[0077] After the local optimization problem on both sides is solved, the consensus variables and multipliers are updated according to equations (75)-(76): (75); (76); in, For the first Consensus variables for the next iteration; For the first Lagrange multipliers in the next iteration; For the first Local boundary variables of the RES base in the next iteration; For the first Local boundary variables of the receiving-end power grid in the next iteration; Subsequently, the original residuals and dual residuals are used to measure the consistency error of the boundary variables and the magnitude of change of the consensus variables, as shown in equations (77)-(78): (77); (78); in, , The first The original residual and dual residual of the next iteration.

[0078] To address the potential issue of two types of residual convergence imbalance during the iteration process, this method adaptively adjusts the penalty factor based on the relative magnitudes of the original residual and the dual residual. The update rule is shown in equation (79): (79); in, This is the scaling factor; , The threshold values ​​are adjusted upwards and downwards respectively; , These are the upper and lower bounds of the penalty factor, respectively; This is the initial convergence tolerance.

[0079] When the inner iteration satisfies the constraint of equation (80), it can be considered that the current inner iteration has reached convergence.

[0080] (80); in, and These are the convergence thresholds for the original residual and the dual residual, respectively.

[0081] Step S4: Simulation is performed using IEEE case studies to analyze the proposed model. This provides a methodological basis for frequency security assurance, operational risk mitigation, and renewable energy consumption enhancement under high-proportion inter-regional renewable energy transmission.

[0082] 1. Example setup.

[0083] This embodiment constructs a simulation system consisting of a RES (Resource Energy Storage) base and a receiving-end power grid, where the receiving-end power grid is simulated using a modified IEEE RTS-79 node system. In the RES base, the installed capacity of wind power and photovoltaic power is 400MW and 800MW respectively, the rated power and capacity of energy storage are 200MW and 400MWh respectively, and the DC transmission capacity is 1000MW. The total installed capacity of the receiving-end power grid is 3700MW, with a peak load of 3500MW. Furthermore, the EVAR risk weighting coefficient is set to 0.1, and the confidence level is set to 0.9. The maximum rate of frequency change is limited to 0.5Hz / s, and the minimum system frequency is not lower than 49.5Hz. The established model is solved using Gurobi, running on Windows 11 64-bit, with an Intel Core i9-14900HX processor.

[0084] To verify the effectiveness of the proposed model, four sets of comparative examples were set up, as shown in Table 1. Example 1 does not consider frequency security constraints; Example 2 considers frequency security constraints but only takes into account the frequency support of conventional units; Example 3 further considers the frequency support of energy storage and DC-DC coordination and uses CVaR to characterize the risk; Example 4, based on Example 3, uses EVAR instead of CVaR to characterize the system operation risk. In addition, separate algorithm comparison experiments were set up to verify the convergence and computational efficiency of the proposed solution method.

[0085] Table 1. Example Settings

[0086] 2. Analysis of Coordinated Scheduling Results.

[0087] This embodiment analyzes the scheduling results from the perspective of operational economy. Table 2 shows that the total cost is lowest without considering frequency security, but this example does not take frequency security requirements into account and is difficult to meet actual operational needs. In Example 2, after introducing frequency security constraints, the system needs to schedule more conventional units online, causing the RES absorption rate to drop from 96% to 89%. In Example 3, after considering the frequency support of energy storage and DC coordination, the system's wind and solar curtailment penalty decreased by 29.52%, indicating that the coordinated response of energy storage and DC can effectively alleviate frequency regulation pressure. In Example 4, after adopting EVAR risk measurement, although the operating cost increased by 4.54% compared to Example 3, its constraints on high-loss tail scenarios were stronger, significantly reducing the load shedding penalty for curtailment, ultimately resulting in a total cost reduction of 8.61%. Overall, the proposed energy storage-DC coordinated frequency support and risk scheduling modeling method can achieve a better balance between safety, economy, and renewable energy absorption.

[0088] Table 2 Comparison of scheduling results for Examples 1-4

[0089] 3. Verification of the frequency support effect of storage-direct current coordination.

[0090] First, we need to verify the necessity of frequency security constraints. Figure 2 Comparative results of RoCoF and frequency minimum point for Example 1 and Example 2 under the same power disturbance are presented. Since Example 1 does not consider the frequency security constraints of the receiving-end power grid, it exhibits frequency limit exceedance risks at multiple time periods, which are more pronounced during the midday and evening peak hours. In contrast, Example 2 shows a lower overall RoCoF than Example 1, and the frequency minimum point exceedance situation is significantly improved. Furthermore, the RoCoF change in Example 2 is more gradual, and the fluctuation range of the frequency minimum point is further narrowed. The results indicate that considering frequency security constraints can effectively improve the frequency security of the system.

[0091] To analyze the impact of storage-DC coordinated frequency support on system frequency support, Figure 3 , Figure 4 The dispatching results of renewable energy bases under Case 2 and Case 3 were compared. The results show that after energy storage and DC collaboratively provide frequency regulation support, the utilization rate of wind and solar power output during the midday peak period is significantly improved, the RES absorption rate increases by 3%, and the DC transmission level also improves accordingly. The above analysis indicates that the participation of energy storage-DC collaboratively in frequency response can enhance the frequency support capability of the receiving-end grid, reduce the system's dependence on the frequency support of conventional units, and thus improve the RES absorption level.

[0092] 4. Risk assessment of the sub-bar based on EVAR.

[0093] (1) Validation of conditional distribution ambiguous fuzzy sets.

[0094] To verify the effectiveness of the proposed conditional nearest neighbor Wasserstein-probability box hybrid fuzzy set, we set up a comparison of the traditional W, W considering the influence of prediction day conditions, and W-PB considering the influence of prediction day conditions. Here, W represents the Wasserstein fuzzy set, and PB represents the probability box constraint. Table 3 shows that the traditional Wasserstein fuzzy set does not consider the error distribution characteristics of the prediction conditions, resulting in a conservative scheduling outcome. After introducing conditional nearest neighbors, the conditional Wasserstein fuzzy set can more accurately characterize the error distribution under the current prediction conditions, reducing the total cost by 9.93%. After adding the probability box constraint, the worst-case distribution shift decreased from 0.4943 to 0.2628, reducing the total cost and increasing the RES absorption rate to 94%; meanwhile, the load shedding penalty is basically the same as the conditional W scheme. The results show that the proposed hybrid fuzzy set can reduce excessive conservatism while suppressing extreme probability redistribution, thereby improving system economy and the level of new energy absorption.

[0095] Table 3. Comparison of scheduling results under different fuzzy set construction methods

[0096] (2) Analysis of the tail risk control effect of EVAR.

[0097] To verify the advantages of EVaR risk measurement, this embodiment conducted robustness tests on 100 random scenarios, and the results are shown in Table 4. The results show that the un-risk-calculated scheme has the highest average curtailment and load shedding, but the lowest robustness qualification rate. Compared to the CVaR scheme, the EVaR scheme improves in all three indicators: average wind and solar curtailment decreases by 10.44%, average load shedding decreases from 79.62 MWh to 28.30 MWh, and the robustness qualification rate increases to 95%. The results indicate that EVaR, by strengthening the characterization of high-loss tail scenarios, can encourage scheduling schemes to reserve more sufficient safety margins, thereby improving the reliability and robustness of system operation.

[0098] Table 4. Out-of-sample robustness test results of different risk measurement methods

[0099] 5. Verification of the effectiveness and computational performance of the solution method.

[0100] (1) Effectiveness analysis of multi-breakpoint segmented robust affine strategy.

[0101] To verify the effectiveness of the proposed multi-breakpoint piecewise robust affine strategy, this method selects single-segment affine decision rules and two-segment piecewise affine decision rules as comparison methods. Figure 5 It is evident that the piecewise affine strategy better characterizes the nonlinear relationship between prediction error and regulation response compared to the single-piece affine strategy, thereby reducing various costs. Specifically, the multi-discontinuity piecewise robust affine strategy proposed in this method further considers the error amplitude range and the response differences of different regulation resources, reducing the total cost by 10.24% compared to the two-piece affine strategy. Although its total solution time increases by 36.77% compared to the two-piece piecewise affine strategy, it is still within an acceptable range, indicating that the proposed strategy can achieve a good balance between scheduling economy and computational efficiency.

[0102] (2) Performance analysis of hierarchical adaptive ADMM solution.

[0103] This embodiment verifies the effectiveness of the proposed hierarchical adaptive ADMM algorithm from both solution efficiency and convergence performance aspects. Table 5 shows that the centralized method has the shortest solution time, but it struggles to meet the data privacy protection and decentralized decision-making requirements of cross-regional scheduling in practical situations. While traditional ADMM can achieve distributed coordination, it suffers from slow residual convergence due to the fixed penalty factor and discrete variable switching, resulting in a relative deviation of 1.14%. In contrast, the proposed hierarchical adaptive ADMM reduces the relative deviation to 0.77% and shortens the solution time by 57.33%. Figure 6It is evident that the traditional ADMM coordination residual decreases slowly and exhibits local fluctuations, while the proposed algorithm can reduce the residual to below the preset convergence threshold within 27 cumulative coordination iterations, indicating that it has better solution efficiency and convergence stability.

[0104] Table 5 Comparison of Results from Different Solution Methods

[0105] 6. Sensitivity analysis.

[0106] (1) Energy storage capacity sensitivity analysis.

[0107] Energy storage can combine RES fluctuation smoothing with rapid frequency regulation. Figure 7 The system operation performance under different energy storage capacities was compared. As the energy storage capacity increased from 0MW to 200MW, the system's flexibility and adjustment capability improved, and the total cost continued to decrease. However, when the capacity further increased to 400MW, the improvement in economic efficiency and renewable energy absorption rate decreased significantly, indicating that the energy storage adjustment benefits exhibit diminishing marginal returns. Therefore, rationally allocating energy storage capacity helps to balance frequency security, operational economy, and renewable energy absorption.

[0108] (2) Sensitivity analysis of DC virtual inertia response delay.

[0109] Virtual inertia response delay directly affects the frequency support capability of DC tie lines. Figure 8 The system operation results under different response delays were compared. As the response delay increases, the intervention of DC virtual inertia lags behind, and the initial frequency response to disturbances relies more on synchronous units. Therefore, it is necessary to increase the online capacity of conventional units to meet frequency safety constraints. When the response delay increases from 100ms to 500ms, the online capacity of the receiving-end conventional units and the online inertia of the synchronous units increase by 17.01% and 13.24%, respectively. This indicates that shortening the response delay can enhance the DC rapid support capability and reduce the system's dependence on the online support of conventional units.

[0110] To address the operational challenges faced by large-scale RES bases connected to the receiving-end power grid via DC tie lines, the proposed method achieves the following technical advantages: (1) By leveraging the virtual inertia support of DC interconnects and the rapid frequency response capability of energy storage, the proposed method fully taps the regulation and support potential of RES bases. With frequency support provided by energy storage-DC synergy, the wind and solar curtailment penalty is reduced by 29.52%, significantly improving the economic efficiency of system operation.

[0111] (2) The conditional nearest neighbor Wasserstein-probability box hybrid distribution fuzzy set can accurately characterize the source load error distribution and suppress extreme probability offsets, making the scheduling results both robust and economical. At the same time, the EVAR risk metric enhances the adaptability of the scheduling scheme to high tail loss scenarios, improves the robustness qualification rate by 14%, and further improves the system's operational robustness.

[0112] (3) The proposed multi-breakpoint segmented robust affine strategy improves the adaptability of scheduling decisions to prediction errors, reducing the total cost by 10.24% compared to the two-segment affine strategy; the hierarchical adaptive ADMM algorithm reduces the solution time by 57.33% while ensuring accuracy, thus improving the convergence stability and computational efficiency of cross-regional collaborative scheduling.

[0113] Example 2 The cross-regional frequency support sub-brush scheduling system for new energy bases includes: The cross-regional dynamic frequency response module is used to construct a cross-regional dynamic frequency response model under the synergistic effect of energy storage fast frequency response and DC virtual inertia, and to characterize the timing response process of DC tie line virtual inertia support, energy storage fast frequency response and conventional unit primary frequency regulation.

[0114] The receiving-end power grid frequency constraint module is used to establish the receiving-end power grid frequency change rate constraint, frequency minimum point constraint, and quasi-steady-state frequency constraint based on the cross-regional dynamic frequency response model and considering the DC virtual inertia response delay.

[0115] The risk perception module is used to construct a Wasserstein-probability box hybrid distribution fuzzy set based on the historical source load prediction error and the current prediction features to select the nearest neighbor samples, and introduces the entropy value risk (EVaR) to characterize the tail operation risk in extreme scenarios.

[0116] The hierarchical adaptive ADMM solution module is used to establish a risk perception-based distributed robust collaborative scheduling model between the new energy base and the receiving-end power grid. It adopts a hierarchical adaptive ADMM method that integrates multi-disruption segmented robust affine rules for distributed solution to obtain unit start-up and shutdown, DC transmission, energy storage regulation and rescheduling schemes.

[0117] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0118] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0119] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0120] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.

[0121] Therefore, the present invention adopts the above-mentioned cross-regional frequency support sub-blob scheduling method, system and equipment for new energy bases. By organically combining energy storage-DC collaborative frequency support with risk-aware sub-blob scheduling, it can effectively cope with the operational risks caused by source-load uncertainty and improve the robustness and computational efficiency of cross-regional collaborative scheduling.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A cross-regional frequency support sub-Blu-ray bar scheduling method for new energy bases, characterized in that, Includes the following steps: Step S1: Construct a cross-regional frequency response model under the synergistic effect of fast frequency response of energy storage and DC virtual inertia, consider the delay of DC virtual inertia response, and establish frequency security constraints of the receiving-end power grid. Step S2: Based on historical source load prediction errors and current prediction features, select conditional nearest neighbor samples, construct a Wasserstein-probability box hybrid distribution fuzzy set, and introduce entropy risk value (EVaR) to characterize tail operation risk in extreme scenarios. Step S3: Based on frequency security constraints and risk measurement, establish a risk perception-based distributed robust collaborative scheduling model between the new energy base and the receiving-end power grid, and use the hierarchical adaptive ADMM method that integrates multi-breakpoint segmented robust affine rules to solve the model in a distributed manner to obtain the cross-regional scheduling scheme.

2. The cross-regional frequency support sub-blob scheduling method for new energy bases according to claim 1, characterized in that, The frequency security constraints of the receiving-end power grid include frequency change rate constraints, frequency minimum point constraints, and quasi-steady-state frequency constraints. The expression constrained by the rate of change of the grid frequency at the receiving end is as follows: ； in, The actual frequency change rate of the receiving-end power grid; This refers to the frequency deviation of the receiving-end power grid. For scheduling periods; The initial moment of the disturbance; The rated frequency of the power grid; This refers to the load disturbance power. The inertia of the receiving-end synchronizing machine; The maximum rate of frequency variation that the receiving-end power grid can accept; The expression for the minimum frequency constraint is as follows: ； in, This is the DC virtual inertia; This represents the maximum frequency change value. This refers to the typical ramp-up time for generator units. Frequency regulation power provided for the energy storage system at the RES base; This refers to the load disturbance power. This refers to the frequency modulation power of the synchronous machine; and These are the delay times for conventional generating units and energy storage systems, respectively, after considering the DC virtual inertia delay. For the FFR ramp-up time of the energy storage system; Delay for DC virtual inertial response; The expression for the quasi-steady-state frequency constraint is as follows: 。 3. The cross-regional frequency support sub-blob scheduling method for new energy bases according to claim 2, characterized in that, The construction process of Wasserstein-probability box mixed distribution fuzzy sets is as follows: First, based on intraday prediction features, the most similar historical samples are selected and representative scenarios are constructed: ； in, The Euclidean distance between the historical samples and the feature vector of the target day; For the first Conditional feature vectors of historical samples; This is the conditional feature vector for the current day to be scheduled; Then, based on the distance between each scene and the current prediction, weights are assigned using a Gaussian kernel function to obtain the nominal conditional probability. : ； in, For the target feature vector and the first The distance between scenes; For kernel bandwidth parameters; For the target feature vector and the first The distance between scenes; Total number of scenes; Then, probability box constraints are introduced to limit the excessive migration of scene probabilities to a few extreme scenarios, constructing a Wasserstein-probability box hybrid distribution fuzzy set: ； ； in, This is the actual probability vector; For the scene The actual probability; For transpose; It is a fuzzy set of Wasserstein-probability box mixed distribution; for 3D nonnegative real space; and Let be the upper and lower bounds of the probability box; The Wasserstein distance; Let be the radius of the Wasserstein sphere.

4. The cross-regional frequency support sub-blob scheduling method for new energy bases according to claim 3, characterized in that, The Entropy Value at Risk (EVaR) is introduced by reweighting the scenario probabilities around the nominal conditional probabilities under relative entropy constraints, thus characterizing the tail-end operational risk in the form of worst-case expectation. ； ； in, This is a measure of risk / expected risk. This represents the scene probability vector. Confidence level; For the first The weight or probability of each scenario; Let it be a random loss variable; It is a fuzzy set of probability distributions based on EVAR constraints; The difference between the entropy value at risk (EVaR) and the nominal expected cost is defined as the risk premium, and the risk premiums for the renewable energy base side and the receiving-end grid side are obtained respectively: ； ； in, For RES base-side risk premium; Total cost of the RES site taking into account EVAR risk; Risk premium on the receiving end of the power grid; Total cost of the receiving-end power grid considering EVAR risk; Costs for the second phase of the RES site; For the second-stage cost of the receiving-end power grid; The system's overall tail risk premium is then obtained by weighted summation of the risk premiums on both sides, where the weighting coefficients reflect the risk preferences of the scheduling entities on both sides: ； in, For system-wide tail risk premium; and These represent the risk preference coefficients of the RES base and the receiving-end power grid, respectively.

5. The cross-regional frequency support sub-blob scheduling method for new energy bases according to claim 4, characterized in that, The risk perception-based collaborative scheduling model between new energy bases and receiving-end power grids aims to minimize system operating costs. ； System operating costs include the first phase of unit start-up and shutdown costs, the second phase of rescheduling costs, and the EVAR tail risk premium for renewable energy bases and receiving-end power grids; The costs associated with new energy bases include unit operating costs and penalties for wind and solar curtailment. ； The receiving-end grid-side cost includes unit operating costs, wind and solar curtailment penalties, and load shedding penalties: ； in, The overall objective function of the system; and The costs for starting and stopping the first phase of generating units at the RES base and the receiving-end power grid are respectively; A set of random scenes; and Scenes The second-stage rescheduling cost of the two lower-side systems; For runtime segment; A collection of conventional generating units; For the conventional units of the RES base Unit operating cost; This is the penalty coefficient for wind power curtailment; This refers to the amount of wind power curtailed at the receiving end; This refers to the penalty coefficient for photovoltaic curtailment. This refers to the amount of solar power curtailment at the receiving end; For conventional generating units of the receiving-end power grid Operating costs; This is the penalty coefficient for wind power curtailment at the receiving end; This is the penalty coefficient for photovoltaic curtailment at the receiving end; Receiving-end load penalty factor; and These refer to the amount of wind and solar power curtailed; This refers to the load shedding power; This represents the total risk premium of the system.

6. The cross-regional frequency support sub-blob scheduling method for new energy bases according to claim 5, characterized in that, The risk perception and collaborative scheduling model between new energy bases and receiving-end power grids includes first-stage operational constraints and second-stage operational constraints. The first stage of operational constraints is used to determine the start-up and shutdown status of conventional units, and includes the minimum operating time constraints and minimum downtime constraints for conventional units. The second phase of operational constraints involves resource scheduling based on representative scenarios. The types of constraints include: output constraints of new energy power plants, upper and lower limits and ramping constraints of conventional unit output, charging and discharging power and state of charge constraints of energy storage, tiered transmission and power balance constraints of DC interconnection lines, power balance constraints of receiving-end grid nodes, and line power flow constraints.

7. The cross-regional frequency support sub-blob scheduling method for new energy bases according to claim 1, characterized in that, The robust affine rule for multi-breakpoint segmentation is as follows: First, construct a five-dimensional original error vector containing the prediction errors of base wind power, base photovoltaic, receiving-end wind power, receiving-end photovoltaic, and receiving-end load. : ； in, , , , , These correspond to the prediction errors for base wind power, base photovoltaic, receiving-end wind power, receiving-end photovoltaic, and receiving-end load, respectively. Furthermore, multiple breakpoints are preset for each scalar error component, and a segmented folding feature is constructed. The breakpoints are adaptively selected from historical error samples. When a sample contains both positive and negative deviations, the zero point is retained, and representative positions on both the positive and negative sides are selected as internal breakpoints. When a sample does not cross zero, equidistant breakpoints within the interval are used. ； in, The values ​​are piecewise linear functions / activation functions. For the first Scene, time period ,variable The input variable values; For variables The Each segmented breakpoint; This represents the number of segments; Then, the folded features of each dimension are stacked to obtain the upgraded error feature vector: ； in, For different scenarios Down The dimensionality-upgrading error eigenvector at time step; Finally, the second-stage adjustment decision variables are uniformly represented as piecewise linear affine functions of the upgraded dimensionality features: ； ； ； ； in, For the scene Down Shike New Energy Conventional Units contribution; For the scene Down Conventional generating units of the receiving end power grid contribution; For the scene Downline contribution; For the scene Downlink energy storage system contribution; , , and For new energy conventional units Conventional generating units in the receiving-end power grid DC channel Energy storage system The baseline scheduling part; , , They are respectively new energy conventional units Conventional generating units in the receiving-end power grid DC channel The piecewise response coefficient matrix; , , These are the error feature vectors corresponding to the RES base, the receiving-end power grid, and the DC channel, respectively. , , energy storage system The segmented response coefficient matrix of the RES base, receiving-end power grid new energy, and receiving-end power grid load; , , energy storage system The error feature vectors corresponding to the RES base, the error feature vectors corresponding to the new energy sources in the receiving-end power grid, and the error feature vectors corresponding to the loads in the receiving-end power grid.

8. The cross-regional frequency support sub-blob scheduling method for new energy bases according to claim 7, characterized in that, The solution process of the hierarchical adaptive ADMM method is divided into two levels: outer iteration and inner iteration. The outer iteration handles 0-1 discrete variables such as unit start-up and shutdown. In the first iteration, a soft-guided method is used to add a deviation penalty term to generate a feasible integer solution. ； in, as the main body The original objective function; as the main body Decision variables; This is the penalty coefficient for soft guidance; These are slack variables; as the main body The response variable; These are the initial coordination variables; Starting from the second iteration, the boundary consensus returned by the inner layer in the previous round is fixed, and the mixed integer subproblems on both sides are solved separately. The convergence criterion for the outer layer is that the proportion of change of the decision variable in two consecutive rounds meets the set threshold. ； in, For the first Consistency ratio of outer start and stop variables in the next iteration; For the first For each iteration, there is a discrete set of decisions. The threshold for convergence criterion of changes in outer-layer decision variables; Let be the dimension of the discrete decision set; For the updated set of discrete variables; After fixing the outer discrete variables, the process moves to the inner iteration, which handles the coordinated optimization of continuous variables, decomposing the original problem into two sub-problems: one on the renewable energy base side and the other on the receiving-end power grid side. ； ； in, and The first Decision variables for the RES base and receiving-end power grid in the next iteration; For the decision variables of the RES base; For the decision variables of the receiving-end power grid; and These are the local boundary variables for the RES base and the receiving-end power grid, respectively; For the first Boundary consensus variables for the next iteration; and Given discrete decision sets The feasible region for the RES base and the receiving-end power grid; and These are the objective functions for the RES base and the receiving-end power grid, respectively. These are unscaled Lagrange multipliers; As a penalty factor; Distributed coordination is achieved through alternating solution and boundary consensus update. The inner convergence criterion is that the original residual and the dual residual satisfy a set threshold. ； in, , The first The original residual and dual residual of the next iteration; and These are the convergence thresholds for the original residual and the dual residual, respectively.

9. A cross-regional frequency support sub-brush scheduling system for new energy bases, characterized in that: include: The cross-regional frequency response modeling module is used to construct a cross-regional frequency response model under the synergistic effect of energy storage fast frequency response and DC virtual inertia, taking into account the DC virtual inertia response delay and establishing frequency security constraints of the receiving-end power grid. The distributed fuzzy set construction module is used to filter conditional nearest neighbor samples based on historical source load prediction errors and current prediction features, construct Wasserstein-probability box hybrid distributed fuzzy sets, and introduce entropy risk value (EVaR) to characterize tail running risk in extreme scenarios. The collaborative scheduling solution module is used to establish a risk-aware, distributed, robust collaborative scheduling model between new energy bases and receiving-end power grids based on frequency security constraints and risk metrics. The model is solved in a distributed manner using a hierarchical adaptive ADMM method that integrates multi-disruption segmented robust affine rules to obtain cross-regional scheduling schemes.

10. A computer device, comprising: Memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, it implements the steps of the new energy base cross-regional frequency support sub-brush scheduling method as described in any one of claims 1-8.

Images