A method for grid-connected scheduling of offshore wind power considering space-time uncertainty of wake

Abstract

The application discloses a kind of considering wake space-time uncertainty offshore wind power grid-connected scheduling method, belong to power system dispatching technical field, the scheduling method includes the following steps: step 1. based on B-P Gaussian wake model construction fan stage incoming flow wind speed mapping;Step 2. establish fan stage power mapping and space-time power deviation sample;Step 3. construct fan stage two-stage distribution robust grid-connected scheduling model;Step 4. construct wasserstein fuzzy distribution set to describe fan stage uncertainty;Step 5. adopt hierarchical space-time clustering to reduce fan stage uncertainty dimension;Step 6. two-stage distribution robust scheduling model is solved using custom column constraint generation algorithm.The application is to power deviation to be described to fan stage or typical fan cluster layer, to more fully reflect spatial difference and time correlation under wake effect;By hierarchical space-time clustering dimension reduction and custom column constraint generation algorithm, provide more fine, robust and practical technical support for large-scale offshore wind farm group grid-connected scheduling.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching technology, specifically to a method for grid-connected dispatching of offshore wind power that considers the spatiotemporal uncertainty of wake. Background Technology

[0002] As the installed capacity of offshore wind power continues to expand, neighboring offshore wind farms are gradually being connected to the power system in the form of clusters, sharing transmission channels and collection networks. The grid-connected dispatch of offshore wind farm clusters is not only affected by system load, conventional units, pumped storage, and transmission constraints, but also by the wake effect within the wind farms.

[0003] The wake effect reduces the incoming wind speed to downstream wind turbines and causes differences in the available power of different turbines at different times. Therefore, the uncertainty of offshore wind power is not only manifested in the fluctuation of the total power of the site, but also includes the spatiotemporal heterogeneity of the turbine level formed by the spatial location of the turbine, changes in wind direction, and the superposition of wake.

[0004] Existing offshore wind power grid-connected scheduling methods often treat a wind farm as a single aggregated power injection node and conduct stochastic optimization, robust optimization, or distributed Bruker optimization scheduling based on the farm-level predicted power or farm-level uncertainty. These methods can describe the impact of overall wind power fluctuations on system operation, but they struggle to reflect the deviation differences and spatiotemporal correlations between different turbines within the same wind farm. Existing wake models, such as the Jensen model and the Bastankhah-Porté-Agel (BP) Gaussian wake model, can be used to describe wake losses between turbines, with the latter providing a more detailed characterization of wake effects through the Gaussian velocity loss distribution. However, current wake-related research focuses more on issues such as internal wind farm control, power enhancement, or energy storage configuration, and has not yet been fully transferred to system-level grid-connected scheduling modeling.

[0005] Furthermore, when uncertainty extends from the site level to the turbine level and covers multiple scheduling periods, the model dimensionality and solution burden increase significantly, limiting the application of refined scheduling methods in large-scale offshore wind farm clusters. Summary of the Invention

[0006] To address the aforementioned technical problems, the present invention aims to provide a method for grid-connected dispatch of offshore wind power that considers the spatiotemporal uncertainty of wake effects. This method can more fully reflect the spatial differences and temporal correlations under the wake effect.

[0007] The technical solution adopted in this invention is: A method for grid-connected dispatch of offshore wind power considering the spatiotemporal uncertainty of wake, characterized by the following steps: Step 1. Construct wind speed mapping for wind turbine level based on BP Gaussian wake model: Establish wake influence relationship between wind turbines based on the spatial coordinates, wind direction and free wind speed of each wind turbine in the offshore wind farm; for horizontal axis wind turbines, establish flow direction coordinate system with wind turbine hub as coordinate origin, and use BP Gaussian wake model to convert free wind speed and wind direction into wake correction effective wind speed of each wind turbine; Step 2. Establish wind turbine-level power mapping and spatiotemporal power deviation samples: Based on the effective incoming wind speed, calculate the output power of each wind turbine according to the wind turbine power curve; input the predicted free incoming wind speed and direction and the measured free incoming wind speed and direction into the wake model and the power mapping model respectively to obtain the predicted power and the measured power, and then generate wind turbine-level spatiotemporal power deviation samples covering all wind turbines and all scheduling periods; Step 3. Construct a two-stage sub-Blu-ray bar grid-connected scheduling model for wind turbines: Based on the spatiotemporal power deviation samples of the wind turbines, establish a two-stage sub-Blu-ray bar scheduling model; wherein, the first stage decision is the day-ahead scheduling decision, including the start-up and shutdown status and standby arrangements of conventional units and pumped storage units; the second stage decision is the real-time rescheduling decision after the realization of uncertainties, including the output of conventional units, the charging and discharging of pumped storage, the wind power injection power and the wind curtailment power; Step 4. Construct a Wasserstein fuzzy distribution set to describe the uncertainty of the wind turbine level: Construct an empirical distribution based on historical wind turbine level spatiotemporal deviation samples, and construct a fuzzy distribution set based on Wasserstein distance to describe the uncertainty of the wind turbine level spatiotemporal power deviation samples; Step 5. Reduce the dimensionality of wind turbine-level uncertainty using hierarchical spatiotemporal clustering: The hierarchical spatiotemporal clustering method is used to reduce the dimensionality of the wind turbine-level spatiotemporal power deviation samples to obtain typical wind turbine clusters and dimensionality-reduced typical deviation samples. The hierarchical spatiotemporal clustering includes: firstly, performing geographically constrained K-means spatial partitioning based on the wind turbine spatial coordinates to ensure that wind turbines within the same partition are geographically adjacent; then, within each geographical partition, performing hierarchical clustering on the historical time series of wind turbines based on dynamic time warping distance to group wind turbines with similar time series patterns into the same typical wind turbine cluster. Step 6. Solve the two-stage sub-Brussels bar scheduling model using a customized column constraint generation algorithm: Based on the typical deviation samples after dimensionality reduction, solve the two-stage sub-Brussels bar scheduling model using a customized column constraint generation algorithm; the customization includes: constructing a three-point candidate support set containing upper and lower bounds for each uncertainty component, explicitly embedding the Wasserstein dual structure in the subproblem, and adding a linear dual cutting plane to the main problem based on the worst-case scenario search, iteratively solving until convergence, and obtaining the grid-connected scheduling scheme for the offshore wind farm group.

[0008] Further, in step 1, the velocity deficit expression of the BP Gaussian wake model is: , in, ; This represents the local average wind speed at a location within the wake region. , , These represent the flow direction, lateral direction, and vertical direction coordinates, respectively. For free-flowing wind speed; This is the thrust coefficient; D k is the diameter of the fan impeller. ★ The wake expansion coefficient; This represents the near-wake width parameter; Indicates the wheel hub height.

[0009] Furthermore, in step 1, the fan... i Effective incoming air velocity U i Determined by both the free-flow wind speed and the wake of the upstream fan: , in, Indicates wind turbine Effective incoming air velocity; Indicates the fan The collection of upstream wind turbines that generate wake effects; Indicates upstream wind turbine The incoming wind speed; This indicates that only wind turbines are being considered. Wake fan Wind speed at that location.

[0010] Furthermore, in step 2, the wind turbine power curve is presented in a piecewise function form: , in, Indicates the cut-in wind speed; Indicates the rated wind speed; Indicates the cut-out wind speed; Indicates the rated power of the fan; This represents the normalized power curve function.

[0011] Furthermore, in step 2, the method for generating the wind turbine-level spatiotemporal power deviation sample is as follows: For the d Heavenly t For each time period, the predicted and measured free-flow wind speeds and directions are input into the wake model and power mapping model, respectively, to obtain the predicted power. and measured power The wind turbine-level hourly power deviation vector is: ;in, Indicates the first Heavenly The power deviation vector of all wind turbines in each time period; Indicates wind turbine The predicted power; Indicates wind turbine The measured power; Indicates the number of fans; express 3D real space; Stacking the deviation vectors for all time periods within a day yields the first... The spatiotemporal power deviation trajectory of wind turbines in the sky: , in, Indicates the first The spatiotemporal power deviation trajectory covers all wind turbines and all scheduling periods; This indicates the number of scheduling periods; further, the spatiotemporal power deviation trajectory of the wind turbine level is recorded as an uncertainty sample: , Indicates the first The spatiotemporal power deviation vector of each historical sample corresponding to the wind turbine level; This indicates the number of historical samples.

[0012] Furthermore, in step 3, the two-stage split-brush scheduling model is expressed as: ,

[0013] in, This represents the day-ahead decision vector for the first stage; Indicates day-ahead scheduling costs; Indicates a given and The optimal rescheduling cost in the second phase; Represents the probability distribution in a fuzzy distribution set; Represents the set of Wasserstein fuzzy distributions; Indicates distribution The expectations below.

[0014] Furthermore, in step 4, the Wasserstein fuzzy distribution set is constructed as follows: Based on historical wind turbine-level spatiotemporal deviation samples, an empirical distribution is constructed. : ,in, Indicates by An empirical distribution consisting of historical samples; Indicates that it is located in the sample Dirac measure at the location; Based on the construction of empirical distributions, a set of Wasserstein fuzzy distributions is constructed. : in, Represents the Wasserstein radius; Indicates support in the set The set of probability measures on; This represents the first-order Wasserstein distance.

[0015] Wasserstein distance is defined as: in, and Represents two uncertain realization vectors; Represents the 1-norm of a vector; Indicates and It is the joint probability measure for marginal distributions.

[0016] Furthermore, in step 5, the deviation aggregation method of the typical wind turbine cluster is as follows: ,in, Indicates the first The number of wind turbines in a typical wind turbine cluster; Indicates the final typical number of wind turbine clusters. For the first Each sample in the time period Typical cluster bias.

[0017] Furthermore, in step 6, after completing the hierarchical spatiotemporal dimensionality reduction, let Representing the historical samples after clustering, the two-stage split-bar model can be transformed into the following semi-infinite form: in, This represents the vector of cost coefficients for the first stage. Represent Wasserstein's dual variable; Indicates the first The supremum auxiliary variable corresponding to each sample; Represents the dimensionality reduction of the first... One historical sample vector; This represents the set of uncertainties, i.e., the set of wind power deviations. This represents the day-ahead decision vector for the first stage; and These represent the first-stage constraint matrix and constraint vector, respectively. express and Multiplying two vectors; The value function for the second stage is expressed as: in, and Represents the constraint matrix; Representation of uncertainty vector The relevant constraint right-hand vector; This represents the cost coefficient vector for the second stage. This represents the second-stage scheduling variable vector; This indicates the cost of rescheduling in the second phase.

[0018] Furthermore, in step 6, to construct a computable three-point candidate support set, upper and lower bounds are defined for each uncertainty component: in, The set of indices representing the uncertain components after dimensionality reduction; and They represent the first The minimum and maximum values ​​of each uncertainty component in the historical sample; Indicates the first In the nth sample One uncertain component; Further construct a three-point candidate support set : in, Indicates that for the first A candidate support set constructed from 1 sample; Represents the candidate uncertainty vector; Indicates the first One uncertain component.

[0019] The beneficial effects of this invention are as follows: Compared with traditional site-level aggregation modeling methods, this invention does not simply equate offshore wind farms to a single uncertain power injection, but rather characterizes power deviations at the turbine level or typical turbine cluster level, thereby more fully reflecting the spatial differences and temporal correlations under the wake effect. At the same time, through hierarchical spatiotemporal clustering dimensionality reduction and customized column constraint generation algorithms, it provides more refined, robust and practical technical support for grid-connected scheduling of large-scale offshore wind farm groups while maintaining the computational feasibility of the scheduling model. Attached Figure Description

[0020] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort: Figure 1 This is a flowchart of the offshore wind power grid connection and dispatch method of the present invention; Figure 2 This is a spatial distribution diagram of the average inflow ratio for OWF1. Figure 3 This is a typical wind turbine cluster distribution map after OWF1 hierarchical spatiotemporal clustering. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of this application can be combined with each other.

[0022] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper surface," "lower surface," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "forward," "reverse," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0023] like Figure 1 As shown, a method for grid-connected dispatch of offshore wind power considering the spatiotemporal uncertainty of wake is characterized by the following steps: Step 1. Construct wind speed mapping for wind turbine level based on BP Gaussian wake model: Establish the wake influence relationship between wind turbines based on the spatial coordinates, wind direction and free wind speed of each wind turbine in the offshore wind farm; for horizontal axis wind turbines, establish a flow direction coordinate system with the wind turbine hub as the coordinate origin, and use BP Gaussian wake model to convert the free wind speed and wind direction into the wake correction effective wind speed of each wind turbine.

[0024] The specific method for constructing the inflow velocity mapping at the wind turbine level is as follows: First, based on the spatial coordinates, wind direction, and free-flow wind speed of each wind turbine within the offshore wind farm, the wake influence relationship between the turbines is established. For horizontal axis turbines, a flow direction coordinate system is established with the turbine hub as the origin. Let the free-flow wind speed be... Location within the wake region The local average wind speed at the location is The speed deficit is: in, This represents the local average wind speed at a location within the wake region. , , These represent the flow direction, lateral direction, and vertical direction coordinates, respectively.

[0025] Under non-yaw and approximately axisymmetric wake conditions, the velocity deficit expression for the BP Gaussian wake model is:

[0026] For free-flowing wind speed; This is the thrust coefficient; D The diameter of the fan impeller; The wake expansion coefficient; This represents the near-wake width parameter; Indicates the wheel hub height.

[0027] The thrust coefficient is defined as: Among them, F th Indicates the thrust of the wind turbine; Indicates air density; This indicates the area swept by the impeller.

[0028] Fan Effective incoming air velocity Determined by both the free-flow wind speed and the wake of the upstream fan: in, Indicates wind turbine Effective incoming air velocity; Indicates the fan The collection of upstream wind turbines that generate wake effects; Indicates upstream wind turbine The incoming wind speed; This indicates that only wind turbines are being considered. Wake fan Wind speed at that location.

[0029] Step 2. Establish wind turbine-level power mapping and spatiotemporal power deviation samples: Based on the effective incoming wind speed, calculate the output power of each wind turbine according to the wind turbine power curve; input the predicted free incoming wind speed and direction and the measured free incoming wind speed and direction into the wake model and the power mapping model respectively to obtain the predicted power and the measured power, and then generate wind turbine-level spatiotemporal power deviation samples covering all wind turbines and all scheduling periods.

[0030] Obtain the effective incoming air velocity for each fan. Then, the fan output power is calculated based on the fan power curve. The fan power is expressed as: in, Indicates wind turbine ; output power; Indicates the incoming wind speed The relevant power factor.

[0031] Based on the power curves from the wind turbine manufacturer, the wind turbine power can be expressed in a segmented form: in, Indicates the cut-in wind speed; Indicates the rated wind speed; Indicates the cut-out wind speed; Indicates the rated power of the fan; This represents the normalized power curve function.

[0032] For the Heavenly The predicted free-flow wind speed and direction will be determined at each time period. And measured free-flow wind speed and direction Input the above wake model and power mapping model to obtain the predicted power. and measured power The wind turbine hourly power deviation vector is then: in, Indicates the first Heavenly The power deviation vector of all wind turbines in each time period; Indicates wind turbine The predicted power; Indicates wind turbine The measured power; Indicates the number of fans; express 3D real space; Indicates the number of scheduling periods.

[0033] Stacking the deviation vectors for all time periods within a day yields the first... The spatiotemporal power deviation trajectory of wind turbines in the sky: , in, Indicates the first The spatiotemporal power deviation trajectory covers all wind turbines and all scheduling periods; Indicates the number of scheduling periods; Furthermore, the spatiotemporal power deviation trajectory of the wind turbine stage is recorded as an uncertainty sample: in, Indicates the first The spatiotemporal power deviation vector of each historical sample corresponding to the wind turbine level; This indicates the number of historical samples.

[0034] Step 3. Construct a two-stage sub-Blu-ray bar grid-connected scheduling model for wind turbines: Based on the spatiotemporal power deviation samples of the wind turbines, establish a two-stage sub-Blu-ray bar scheduling model; wherein, the first stage decision is the day-ahead scheduling decision, including the start-up and shutdown status and standby arrangements of conventional units and pumped storage units; the second stage decision is the real-time rescheduling decision after the realization of uncertainties, including the output of conventional units, the charging and discharging of pumped storage, the wind power injection power and the wind curtailment power.

[0035] The two-phase split-Brow scheduling model is represented as follows: in, This represents the day-ahead decision vector for the first stage; Indicating deviation scenarios The second phase of rescheduling variable vectors; Indicates day-ahead scheduling costs; Indicates a given and The optimal rescheduling cost in the second phase; Represents the probability distribution in a fuzzy distribution set; Represents the set of Wasserstein fuzzy distributions; Indicates distribution The following expectations; This represents the feasible region in the first phase. This represents the feasible region for the second phase. This represents the support set for uncertainty.

[0036] The cost of the second-phase rescheduling is defined as: in, This represents the cost coefficient vector for the second stage. This represents the second-stage scheduling variable vector; This indicates the cost of rescheduling in the second phase.

[0037] Current Costs Written as: in, This represents a collection of conventional generating units; Indicates the unit During the period The start / stop status; and They represent the generating units. During the period Startup and shutdown status; , , They represent the generating units. The costs of no-load operation, startup, and shutdown.

[0038] In deviation scenarios Below, the fan During the period Available wind power for: in, Indicates wind turbine During the period The predicted available power; Indicates wind turbine During the period Power deviation.

[0039] Wind power injection and curtailment constraints are: in, Indicates wind turbine During the period The active power injected; Indicates wind turbine During the period The amount of wind power that is curtailed.

[0040] Step 4. Construct a Wasserstein fuzzy distribution set to describe the uncertainty of the wind turbine level: Construct an empirical distribution based on historical wind turbine level spatiotemporal deviation samples, and construct a fuzzy distribution set based on Wasserstein distance to describe the uncertainty of the wind turbine level spatiotemporal power deviation samples.

[0041] Based on historical wind turbine-level spatiotemporal deviation samples, an empirical distribution is constructed. :

[0042] in, Indicates by An empirical distribution consisting of historical samples; Indicates that it is located in the sample Dirac measure at the location.

[0043] Based on this, construct the Wasserstein fuzzy distribution set. : in, Represents the Wasserstein radius; Indicates support in the set The set of probability measures on; This represents the first-order Wasserstein distance.

[0044] Wasserstein distance is defined as: , in, and Represents two uncertain realization vectors; Represents the 1-norm of a vector; Indicates and It is the joint probability measure for marginal distributions.

[0045] Step 5. Reduce the dimensionality of wind turbine-level uncertainty using hierarchical spatiotemporal clustering: The hierarchical spatiotemporal clustering method is used to reduce the dimensionality of the wind turbine-level spatiotemporal power deviation samples to obtain typical wind turbine clusters and dimensionality-reduced typical deviation samples. The hierarchical spatiotemporal clustering includes: firstly, performing geographically constrained K-means spatial partitioning based on the wind turbine spatial coordinates to ensure that wind turbines within the same partition are geographically adjacent; then, within each geographical partition, performing hierarchical clustering on the historical time series of wind turbines based on dynamic time warping distance to group wind turbines with similar time series patterns into the same typical wind turbine cluster.

[0046] Because the dimension of the spatiotemporal deviation vector of the wind turbine level is When there are a large number of wind turbines, it can lead to a high computational burden. Therefore, this invention uses a hierarchical spatiotemporal clustering method to reduce the dimensionality of wind turbine-level uncertainties.

[0047] First, perform geographically constrained K-means partitioning based on the spatial coordinates of the wind turbines: in, Indicates the number of geographical zones; Indicates the first A collection of wind turbines within a geographical region; Indicates the first The spatial center vector of each partition; Represents the two-dimensional spatial coordinate vector of the wind turbine w; This represents the 2-norm of a vector.

[0048] Geographic partitioning satisfies: The above conditions indicate that different geographical zones do not overlap, and all zones cover all wind turbines.

[0049] Subsequently, within each geographical partition, a dynamic time-warped distance metric is applied to the historical time series of wind turbines, followed by hierarchical clustering. For wind turbine w, its historical sequence vector... Represented as: in, Indicates wind turbine At a historical moment The wake correction wind speed; Indicates the number of historical moments.

[0050] For the final typical wind turbine cluster , No. Each sample in the time period Typical cluster bias for: in, Indicates the first Number of fans in a typical fan cluster; This indicates the final typical number of wind turbine clusters.

[0051] Step 6. Solve the two-stage sub-Brussels bar scheduling model using a customized column constraint generation algorithm: Based on the typical deviation samples after dimensionality reduction, solve the two-stage sub-Brussels bar scheduling model using a customized column constraint generation algorithm; the customization includes: constructing a three-point candidate support set containing upper and lower bounds for each uncertainty component, explicitly embedding the Wasserstein dual structure in the subproblem, and adding a linear dual cutting plane to the main problem based on the worst-case scenario search, iteratively solving until convergence, and obtaining the grid-connected scheduling scheme for the offshore wind farm group.

[0052] After completing the hierarchical spacetime dimensionality reduction, let This represents the historical samples after clustering. The two-stage sub-Bruker model can be transformed into the following semi-infinite form: , , , in, This represents the vector of cost coefficients for the first stage. Represent Wasserstein's dual variable; Indicates the first The supremum auxiliary variable corresponding to each sample; Represents the dimensionality reduction of the first... One historical sample vector; This represents the set of uncertainties, i.e., the set of wind power deviations. This represents the day-ahead decision vector for the first stage; and These represent the first-stage constraint matrix and constraint vector, respectively. express and Multiplying two vectors.

[0053] The value function for the second stage is expressed as: in, and Represents the constraint matrix; Representation of uncertainty vector The relevant constraint right-hand vector; This represents the cost coefficient vector for the second stage. This represents the second-stage scheduling variable vector; This indicates the cost of rescheduling in the second phase.

[0054] To construct a computable three-point candidate support set, upper and lower bounds are defined for each uncertainty component: in, The set of indices representing the uncertain components after dimensionality reduction; and They represent the first The minimum and maximum values ​​of each uncertainty component in the historical sample; Indicates the first In the nth sample One uncertain component.

[0055] Further construct a three-point candidate support set : in, Indicates that for the first A candidate support set constructed from 1 sample; Represents the candidate uncertainty vector; Indicates the first One uncertain component.

[0056] Given the solution to the main problem The subproblem is: , in, Indicates the first In the nth iteration The worst-case assessment value corresponding to each sample; and They represent the first The first-stage decision vector and Wasserstein dual variables obtained from the next iteration.

[0057] By dualizing the second-stage linear programming problem, we obtain: , in, This represents the dual variable vector corresponding to the second-stage constraint.

[0058] Based on the worst-case scenario and dual variables Add a linear dual cut to the main problem: in, Indicates the first In the nth iteration The worst candidate scenario vector obtained from each sample search; This represents the corresponding dual variable vector. By continuously solving the main problem, searching for the worst-case scenario that violates the constraints of the current main problem, and adding dual cuts, until there are no samples that significantly violate the constraints, a grid-connected scheduling scheme for offshore wind farm groups is obtained.

[0059] This embodiment was tested on both an improved IEEE 39-node system and an improved IEEE 118-node system. All models were implemented in Python and solved using Gurobi Optimizer v12.0.1. The platform configuration was an AMD Ryzen 7 7735H CPU with a clock speed of 3.2 GHz and 16 GB of memory. The offshore wind farm used the MingYang MySE5.5-155 Offshore wind turbine, with a rated power of 5.5 MW, a rotor diameter of 155 m, a hub height of 103 m, and cut-in, rated, and cut-out wind speeds of 3.0 m / s, 10.1 m / s, and 25.0 m / s, respectively. The minimum system reserve was set to 10%.

[0060] The improved IEEE 39-bus system comprises 39 nodes, 10 conventional turbines, and 46 transmission lines. The offshore wind farm OWF1 is connected to Bus 21, and the pumped-storage power station is connected to Bus 25. The pumped-storage efficiency is set at 0.80, with a reservoir capacity range of 0–1000 MWh, an initial capacity of 500 MWh, and a maximum pumping and generating capacity of 200 MW each. Figure 2 This demonstrates the differences in the degree to which different wind turbines within a wind farm are affected by wake. Figure 2 The low to medium inflow ratio region mainly appears inside the wind farm, indicating that some wind turbines are continuously blocked by upstream wind turbines, resulting in stronger wake losses.

[0061] use Figure 3 This invention demonstrates that the typical wind turbine clusters formed by geographical constraint partitioning and time series clustering can be compressed into several typical clusters with spatial interpretability.

[0062] To illustrate the effect of wake correction on wind turbine power estimation, Table 1 compares the results of direct prediction and wake correction. The results show that the Bias, MAE, and RMSE of wind speed and power are all reduced after wake correction, indicating that wake correction can improve the consistency between the sample and measured data.

[0063] Table 1. OWF1 wake correction effect

[0064] In the improved IEEE 39-node system, the method of this invention is compared with deterministic scheduling, robust scheduling, site-level DRO, and wake-corrected site-level DRO. Table 2 shows that the method of this invention, while maintaining turbine-level uncertainty modeling, has an average solution time of 476.352 s, demonstrating computability. Furthermore, the total cost and two-stage cost of this invention differ from those of the site-level method, indicating that turbine-level spatiotemporal uncertainties affect the system rescheduling results.

[0065] Table 2 Comparison of Scheduling Results for the IEEE 39-Node System

[0066] In the improved IEEE 118-node system, three offshore wind farms are connected to Bus 20, Bus 59, and Bus 101, respectively, and three pumped-storage power stations are connected to Bus 8, Bus 43, and Bus 92, respectively. OWF1, OWF2, and OWF3 contain 73, 91, and 91 wind turbines, respectively, corresponding to rated capacities of 400 MW, 500 MW, and 500 MW. This embodiment further verifies the applicability of the present invention under the condition of large-scale offshore wind farm cluster integration.

[0067] For the improved IEEE 118-node system, as shown in Table 3, if 365 historical samples are used directly, the average solution time is 3968.14 s. After using 100 weighted representative samples, the average total cost only increases from 3,094,292 to 3,095,346, a change of 0.034%, while the average solution time decreases from 3968.14 s to 1269.88 s, a decrease of 68.0%. This result shows that sample compression can significantly reduce the computational burden with a small cost deviation.

[0068] Table 3. Compression effect of IEEE 118-node system samples

[0069] In the improved IEEE 118-node system, the method of this invention is further compared with deterministic scheduling, robust scheduling, and wake-corrected site-level DRO. The results, shown in Table 4, indicate that the average total cost of the method of this invention is 3,111,871, the two-stage cost is 2,727,941, and the solution time is 1255.78 s. Compared with site-level WFDRO, the total costs are similar. Table 5 further shows that there is a significant difference in daily costs, indicating that turbine-level modeling can reveal the spatiotemporal deviation structure that is difficult for site-level aggregation models to express.

[0070] Table 4 Comparison of Scheduling Results for the IEEE 118-Node System

[0071] Table 5. Cost differences between the wake correction site-level DRO in the IEEE 118-node system and the method of this invention.

[0072] Compared with traditional site-level aggregation modeling methods, this invention does not simply equate offshore wind farms to a single uncertain power injection, but rather characterizes power deviations at the turbine level or typical turbine cluster level, thereby more fully reflecting the spatial differences and temporal correlations under the wake effect. At the same time, through hierarchical spatiotemporal clustering dimensionality reduction and customized column constraint generation algorithms, it provides more refined, robust and practical technical support for grid-connected scheduling of large-scale offshore wind farm groups while maintaining the computational feasibility of the scheduling model.

[0073] Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described herein, as well as the features of those embodiments or examples, without contradiction. Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for grid-connected dispatch of offshore wind power considering the spatiotemporal uncertainty of wake, characterized in that, Includes the following steps: Step 1. Construct wind speed mapping for wind turbine level based on BP Gaussian wake model: Establish wake influence relationship between wind turbines based on the spatial coordinates, wind direction and free wind speed of each wind turbine in the offshore wind farm; for horizontal axis wind turbines, establish flow direction coordinate system with wind turbine hub as coordinate origin, and use BP Gaussian wake model to convert free wind speed and wind direction into wake correction effective wind speed of each wind turbine; Step 2. Establish wind turbine-level power mapping and spatiotemporal power deviation samples: Based on the effective incoming wind speed, calculate the output power of each wind turbine according to the wind turbine power curve; input the predicted free incoming wind speed and direction and the measured free incoming wind speed and direction into the wake model and the power mapping model respectively to obtain the predicted power and the measured power, and then generate wind turbine-level spatiotemporal power deviation samples covering all wind turbines and all scheduling periods; Step 3. Construct a two-stage sub-Blu-ray bar grid-connected scheduling model for wind turbines: Based on the spatiotemporal power deviation samples of the wind turbines, establish a two-stage sub-Blu-ray bar scheduling model; wherein, the first stage decision is the day-ahead scheduling decision, including the start-up and shutdown status and standby arrangements of conventional units and pumped storage units; the second stage decision is the real-time rescheduling decision after the realization of uncertainties, including the output of conventional units, the charging and discharging of pumped storage, the wind power injection power and the wind curtailment power; Step 4. Construct a Wasserstein fuzzy distribution set to describe the uncertainty of the wind turbine level: Construct an empirical distribution based on historical wind turbine level spatiotemporal deviation samples, and construct a fuzzy distribution set based on Wasserstein distance to describe the uncertainty of the wind turbine level spatiotemporal power deviation samples; Step 5. Reduce the dimensionality of wind turbine-level uncertainty using hierarchical spatiotemporal clustering: The hierarchical spatiotemporal clustering method is used to reduce the dimensionality of the wind turbine-level spatiotemporal power deviation samples to obtain typical wind turbine clusters and dimensionality-reduced typical deviation samples. The hierarchical spatiotemporal clustering includes: firstly, performing geographically constrained K-means spatial partitioning based on the wind turbine spatial coordinates to ensure that wind turbines within the same partition are geographically adjacent; then, within each geographical partition, performing hierarchical clustering on the historical time series of wind turbines based on dynamic time warping distance to group wind turbines with similar time series patterns into the same typical wind turbine cluster. Step 6. Solve the two-stage sub-Brussels bar scheduling model using a customized column constraint generation algorithm: Based on the typical deviation samples after dimensionality reduction, solve the two-stage sub-Brussels bar scheduling model using a customized column constraint generation algorithm; the customization includes: constructing a three-point candidate support set containing upper and lower bounds for each uncertainty component, explicitly embedding the Wasserstein dual structure in the subproblem, and adding a linear dual cutting plane to the main problem based on the worst-case scenario search, iteratively solving until convergence, and obtaining the grid-connected scheduling scheme for the offshore wind farm group.

2. The offshore wind power grid-connected dispatch method considering wake spatiotemporal uncertainties according to claim 1, characterized in that: In step 1, the velocity deficit expression of the BP Gaussian wake model is: ， in, ; This represents the local average wind speed at a location within the wake region. , , These represent the flow direction, lateral direction, and vertical direction coordinates, respectively. For free-flowing wind speed; This is the thrust coefficient; D k is the diameter of the fan impeller. ★ The wake expansion coefficient; This represents the near-wake width parameter; Indicates the wheel hub height.

3. The offshore wind power grid-connected dispatch method considering wake spatiotemporal uncertainties according to claim 2, characterized in that, In step 1, the fan i Effective incoming air velocity U i Determined by both the free-flow wind speed and the wake of the upstream fan: ， in, Indicates wind turbine Effective incoming air velocity; Indicates the fan The collection of upstream wind turbines that generate wake effects; Indicates upstream wind turbine The incoming wind speed; This indicates that only wind turbines are being considered. Wake fan Wind speed at that location.

4. The offshore wind power grid-connected dispatch method considering wake spatiotemporal uncertainties according to claim 3, characterized in that, In step 2, the wind turbine power curve is presented in a piecewise function form: ， in, Indicates the cut-in wind speed; Indicates the rated wind speed; Indicates the cut-out wind speed; Indicates the rated power of the fan; This represents the normalized power curve function.

5. The offshore wind power grid-connected dispatch method considering wake spatiotemporal uncertainties according to claim 4, characterized in that, In step 2, the method for generating the wind turbine-level spatiotemporal power deviation sample is as follows: For the d Heavenly t For each time period, the predicted and measured free-flow wind speeds and directions are input into the wake model and power mapping model, respectively, to obtain the predicted power. and measured power The wind turbine-level hourly power deviation vector is: ;in, Indicates the first Heavenly The power deviation vector of all wind turbines in each time period; Indicates wind turbine The predicted power; Indicates wind turbine The measured power; Indicates the number of fans; express 3D real space; Stacking the deviation vectors for all time periods within a day yields the first... The spatiotemporal power deviation trajectory of wind turbines in the sky: ， in, Indicates the first The spatiotemporal power deviation trajectory covers all wind turbines and all scheduling periods; This indicates the number of scheduling periods; further, the spatiotemporal power deviation trajectory of the wind turbine level is recorded as an uncertainty sample: , Indicates the first The spatiotemporal power deviation vector of each historical sample corresponding to the wind turbine level; Indicates the number of historical samples.

6. The offshore wind power grid-connected dispatch method considering wake spatiotemporal uncertainties according to claim 5, characterized in that, In step 3, the two-stage split-brush scheduling model is represented as follows: ， in, This represents the day-ahead decision vector for the first stage; Indicates day-ahead scheduling costs; Indicates a given and The optimal rescheduling cost in the second phase; Represents the probability distribution in a fuzzy distribution set; Represents the set of Wasserstein fuzzy distributions; Indicates distribution The expectations below.

7. The offshore wind power grid-connected dispatch method considering wake spatiotemporal uncertainties according to claim 6, characterized in that... In step 4, the Wasserstein fuzzy distribution set is constructed as follows: Based on historical wind turbine-level spatiotemporal deviation samples, an empirical distribution is constructed. : ,in, Indicates by An empirical distribution consisting of historical samples; Indicates that it is located in the sample Dirac measure at the location; Based on the construction of empirical distributions, a set of Wasserstein fuzzy distributions is constructed. : in, Represents the Wasserstein radius; Indicates support in the set The set of probability measures on; Represents the first-order Wasserstein distance; Wasserstein distance is defined as: in, and Represents two uncertain realization vectors; Represents the 1-norm of a vector; Indicated by and It is a joint probability measure for marginal distributions.

8. The offshore wind power grid-connected dispatch method considering wake spatiotemporal uncertainties according to claim 7, characterized in that, In step 5, the deviation aggregation method of the typical wind turbine cluster is as follows: ， in, Indicates the first The number of wind turbines in a typical wind turbine cluster; Indicates the final typical number of wind turbine clusters. For the first Each sample in the time period Typical cluster bias.

9. The offshore wind power grid-connected dispatch method considering wake spatiotemporal uncertainties according to claim 8, characterized in that: In step 6, after completing the hierarchical spatiotemporal dimensionality reduction, let Representing the historical samples after clustering, the two-stage split-bar model can be transformed into the following semi-infinite form: in, This represents the vector of cost coefficients for the first stage. Represent Wasserstein's dual variable; Indicates the first The supremum auxiliary variable corresponding to each sample; Represents the dimensionality reduction of the first... One historical sample vector; This represents the set of uncertainties, i.e., the set of wind power deviations. This represents the day-ahead decision vector for the first stage; and These represent the first-stage constraint matrix and constraint vector, respectively. express and Multiplying two vectors; The value function for the second stage is expressed as follows: in, and Represents the constraint matrix; Representation of uncertainty vector The relevant constraint right-hand vector; This represents the cost coefficient vector for the second stage. This represents the second-stage scheduling variable vector; This indicates the cost of rescheduling in the second phase.

10. The offshore wind power grid-connected dispatch method considering wake spatiotemporal uncertainties according to claim 9, characterized in that: In step 6, to construct a computable three-point candidate support set, upper and lower bounds are defined for each uncertainty component: in, The set of indices representing the uncertain components after dimensionality reduction; and They represent the first The minimum and maximum values ​​of each uncertainty component in the historical sample; Indicates the first In the nth sample One uncertain component; Further construct a three-point candidate support set : in, Indicates that for the first A candidate support set constructed from 1 sample; Represents the candidate uncertainty vector; Indicates the first One uncertain component.

Images