Offshore wind power cluster-oriented grid-connected power system uncertainty optimization scheduling method
By constructing an uncertainty prediction model and a linearized power flow model, combined with the GlueVaR risk avoidance mechanism and distributed planning, the scheduling deviation and frequency security issues of offshore wind power clusters were solved, achieving efficient and safe power system optimized scheduling.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUADIAN (DANDONG) OFFSHORE WIND POWER CO LTD
- Filing Date
- 2025-12-29
- Publication Date
- 2026-05-01
AI Technical Summary
The power output uncertainty of offshore wind power clusters leads to problems such as increased scheduling deviation costs, narrowed frequency safety margins, complex power flow coupling, and difficulty in data sharing. Traditional methods have failed to effectively balance uncertainty modeling, risk avoidance, and distributed solution.
An uncertainty prediction model and a cluster power flow linearization model for offshore wind farms are constructed. A GlueVaR risk avoidance mechanism and frequency security constraints are designed. A distributed dynamic programming solution strategy is adopted to achieve safe and economical dispatch of offshore wind power clusters.
It improves the dispatch accuracy and system security of offshore wind power clusters under uncertain conditions, enhances resource utilization and grid connection flexibility, ensures power coordination and line load balance among multiple wind farms, and strengthens the operational stability and dispatch flexibility of the power system.
Smart Images
Figure CN121965787A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of power system optimization operation, specifically relating to an uncertainty optimization scheduling method for grid-connected power systems oriented towards offshore wind power clusters. Background Technology
[0002] In recent years, the capacity of offshore wind turbines has been continuously increasing, with multiple wind farms clustering and centrally connected to the power system. Due to the randomness, spatiotemporal correlation, and rapid fluctuations of offshore wind speeds, the power output of offshore wind farms exhibits high uncertainty, leading to the following prominent problems: dispatching deviation costs rise sharply, making it difficult for wind farms to accurately track intraday / real-time planning curves issued by dispatching agencies; frequency safety margins narrow, as the low inertia of offshore wind power makes the system more prone to frequency exceedances; complex power flow coupling within multi-wind farm clusters, with traditional methods equating wind farms to single turbines resulting in large errors; and immense computational pressure for centralized optimization, as different offshore wind farms belong to different owners, making data sharing difficult and hindering centralized dispatching. Traditional methods mostly employ scenario-based or deterministic dispatching, failing to consider data-driven uncertainty descriptions, frequency safety, active power flow coordination within wind farms, and distributed solution requirements. Therefore, a method for optimizing the dispatching of offshore wind-connected power systems is needed, combining uncertainty modeling, risk mitigation, frequency safety constraints, and distributed solution capabilities. Summary of the Invention
[0003] To address the shortcomings of existing offshore wind power grid-connected systems in terms of uncertainty, frequency security, power flow coordination, and distributed solution, this invention provides an uncertainty optimization scheduling method for offshore wind power grid-connected power systems. This invention can improve the system's uncertainty resistance and frequency security level while ensuring the system's economic operation.
[0004] The technical solution adopted in this invention is as follows: A method for uncertain optimization scheduling of grid-connected power systems for offshore wind power clusters, comprising the following steps:
[0005] S1: Construct an uncertainty prediction model and a cluster power flow linearization model for offshore wind farms to provide a computable basic structure for subsequent distributed bar scheduling;
[0006] S2: Design a robust scheduling strategy that incorporates GlueVaR risk avoidance mechanism and frequency security constraints to achieve safe and stable operation of offshore wind power clusters under extreme scenarios;
[0007] S3: Integrates uncertainty prediction models, risk-avoidance scheduling models, and distributed dynamic programming solution strategies into a unified offshore wind power grid connection control framework to achieve safe and economical scheduling and scalable solutions for large-scale wind power clusters.
[0008] Furthermore, the specific process of constructing the uncertainty prediction model and cluster power flow linearization model for offshore wind farms in step S1 is as follows:
[0009] First, based on the randomness, spatial correlation, and rapid fluctuation of wind speed in offshore wind farms, a joint modeling of wind speed sequences from multiple wind farms is performed. By statistically analyzing the sample distribution of random wind speed disturbances, a model for modeling and discrete prediction of wind speed-power uncertainty in offshore wind farm clusters is constructed using Wasserstein distance. The second-order Wasserstein radius between the empirical probability distribution and the true distribution is calculated, and a fuzzy sphere that can be used for decompositional bar optimization is established.
[0010] Then, based on the topology of the power collection system of offshore wind farms, the voltage and current relationship between the primary and secondary sides of the multi-feed wind farm output is linearized in the synchronous rotating dq coordinate system, and the forward Euler method is used to discretize the power flow equation, so that the active power flow of the offshore power collection network can be written as a linear mapping between the node injected power and the line voltage.
[0011] Finally, a cluster power deviation value function is constructed, and the planned deviation is characterized by linear absolute value programming. At the same time, the baseline trajectory is corrected by combining single-step prediction and two-step long-time domain prediction models and second-order Lagrange extrapolation prediction reference auxiliary calculation tools, so that the deviation function can measure the output risk of the wind farm under uncertainty conditions. The prediction model, linear power flow model and deviation model are integrated to uniformly describe the operation behavior of offshore wind power clusters in a state space framework that can be used for subsequent sub-Bruker optimization.
[0012] Furthermore, the models involved in step S1 are characterized by the following formulas: the modeling and discrete prediction model for wind speed-power uncertainty of offshore wind power clusters is shown in formulas (1)-(8); the single-step prediction and two-step long-time domain prediction models are shown in formulas (9)-(12); the cluster power deviation value function is shown in formula (13); and the second-order Lagrange extrapolation prediction reference auxiliary calculation tool is shown in formulas (14)-(15).
[0013] (1)
[0014] (2)
[0015] (3)
[0016] (4)
[0017] (5)
[0018] (6)
[0019] (7)
[0020] (8)
[0021] (9)
[0022] (10)
[0023] (11)
[0024] (12)
[0025] (13)
[0026] (14)
[0027] (15)
[0028] In the formula: For historical samples The empirical distribution constituted; For the landing point The Dirac measure; A data-driven fuzzy set constructed based on Wasserstein distance; The second-order Wasserstein distance; The radius of the fuzzy set; For the first Each wind field at any time The wind speed vector; This is the mean vector of wind speed; Let be the covariance matrix of wind speed; For the first Typhoon machines at all times Maximum available active power output; The unit power curve; , The power linearization coefficient; Inject power vectors into the collection network nodes; The node voltage vector; , To linearize the power flow coefficient matrix and constant terms; Contribute to energy storage; The injected power at the grid connection point; For constraint functions; , , , , These are the coefficients used for regression prediction; , To predict residual terms; For the first Deviation penalty coefficient for each wind field As a weight for network loss costs; Let be the active power loss function of the power grid; The time sampling interval; , It is a first-order difference; , These are reference predictions derived through second-order Lagrange extrapolation.
[0029] Furthermore, the specific process of designing a robust scheduling strategy in step S2 is as follows:
[0030] First, a GlueVaR risk avoidance mechanism is introduced into the scheduling model. By considering the combined risk indicators between VaR and CVaR, higher weight is given to the high deviation costs that may occur in extreme scenarios. By using risk measurement and risk avoidance objective function, GlueVaR is embedded into the sub-bar target, so that the scheduling model can still flexibly match the system risk preference by adjusting the combination coefficient when facing extreme events such as isolated strong winds and sudden drops, and maintain a balance between economy and reliability.
[0031] Subsequently, frequency security constraints for the offshore wind power system are established. The relationship between frequency dynamics and disturbance power injection is established using the system frequency response model. The frequency extrema are obtained based on the system frequency dynamic equation and the expression for the lowest frequency point. Then, the nonlinear constraints such as the lowest frequency point and the primary frequency regulation requirement are uniformly converted into linear constraints through frequency security linearization approximation and primary frequency regulation capability constraints, and embedded into the scheduling optimization problem.
[0032] Finally, a sub-Bluerg risk avoidance scheduling model for offshore wind power clusters is constructed. The system cost, planning deviation, network loss cost, frequency regulation capacity cost and GlueVaR risk term are uniformly incorporated into a sub-Bluerg optimization framework. The optimal scheduling scheme is solved under the probability distribution of maximizing loss, while taking into account the spatial correlation between wind farms and the grid topology.
[0033] Furthermore, the model involved in step S2 is characterized by the following formulas: the risk measurement and risk aversion objective function are shown in formulas (16)-(21), the system frequency dynamic equation and the expression for the minimum frequency point are shown in formulas (22)-(25), and the frequency safety linearization approximation and primary frequency modulation capability constraint are shown in formulas (26)-(29):
[0034] (16)
[0035] (17)
[0036] (18)
[0037] (19)
[0038] (20)
[0039] (twenty one)
[0040] (twenty two)
[0041] (twenty three)
[0042] (twenty four)
[0043] (25)
[0044] (26)
[0045] (27)
[0046] (28)
[0047] (29)
[0048] In the formula: It is a random cost variable; Value at risk; Conditional risk value; , The GlueVaR combination coefficients; This is the scheduling decision vector; Weights for risk items; For uncertainties such as wind speed and load; For the first The operating cost function of resources; A set of discrete scenes; For scenarios that satisfy the Wasserstein constraint; This is the system's equivalent moment of inertia. The damping coefficient; This refers to the power disturbance. The lower limit of the allowable frequency deviation; , The coefficients are obtained by approximating the nonlinear nadir function as a piecewise linear function; A set of linear segment indices; For the first The unit's available primary frequency modulation capacity; This is the proportionality coefficient; This is the margin factor for total reserve requirements.
[0049] Furthermore, the specific process of integration and solution in step S3 is as follows:
[0050] First, the uncertainty prediction model, linear power flow model, frequency security constraints and GlueVaR risk avoidance mechanism are unified and constructed in a large-scale distributed bar optimization scheduling framework, and the solution is achieved by distributed approximate dynamic programming; the solution is decomposed among the wind farms based on the value function decomposition model of offshore wind power cluster scheduling, and each farm retains only local data privacy, and the main control center ensures system-level optimality through coordination;
[0051] Then, each wind farm independently constructs local sub-problems based on local wind speed prediction, unit constraints, and power flow parameters, and uses linear basis functions to approximate the local value function to form an estimate of future scheduling costs;
[0052] Finally, the main control center integrates the local results from each field, iteratively solves the subproblems of coordinating power allocation, frequency security and cluster power flow based on the distributed alternating direction multiplier method, and uses the value function rolling approximation update rule to correct the basis function weights online, so that the solution of the entire scheduling problem can be completed efficiently and in parallel.
[0053] Furthermore, the model involved in step S3 is characterized by the following formulas: the value function decomposition model for offshore wind power cluster scheduling is shown in formulas (30)-(33), the distributed alternating direction multiplier method iteratively solves the subproblems as shown in formulas (34)-(37), and the value function rolling approximation update rule is shown in formulas (38)-(39):
[0054] (30)
[0055] (31)
[0056] (32)
[0057] (33)
[0058] (34)
[0059] (35)
[0060] (36)
[0061] (37)
[0062] (38)
[0063] (39)
[0064] In the formula: For the system at time The state; For the first The local state of a wind farm; For a moment The set of scheduling decisions; For the first The dispatch cost of a wind farm; For the first The parameter vector of the value function of a wind farm; , , To form the power flow and power balance coefficients constrained by the alternating direction multiplier method; The coordinating variable of the main control center; The Lagrange factor describes the coupling between each wind farm and the main control center; The penalty coefficient for the alternating direction multiplier method; The learning rate; Update the weights based on value.
[0065] This invention also provides an uncertainty-based optimization scheduling system for grid-connected power systems oriented towards offshore wind power clusters, used to implement the uncertainty-based optimization scheduling method for grid-connected power systems oriented towards offshore wind power clusters as described above. The system is characterized by including a model building module, a scheduling strategy design module, and an integrated solution module, with each module working collaboratively.
[0066] The model building module is used to build an uncertainty prediction model, a cluster power flow linearization model, and a wind power deviation function for offshore wind farms, providing a computable basic structure for subsequent sub-Bruker optimal scheduling.
[0067] The scheduling strategy design module is used to design robust scheduling strategies that incorporate GlueVaR risk avoidance mechanisms and frequency security constraints, and to construct a robust risk avoidance scheduling model to ensure the safe and stable operation of offshore wind power clusters under extreme scenarios.
[0068] The integrated solution module is used to integrate the uncertainty prediction model and the distributed bar risk avoidance scheduling model into a unified offshore wind power grid connection control framework. It adopts distributed approximate dynamic programming and alternating direction multiplier method to achieve efficient parallel solution, thereby achieving safe and economical scheduling and scalable solution for large-scale wind power clusters.
[0069] The present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement an uncertainty-based optimization scheduling method for grid-connected power systems oriented towards offshore wind power clusters as described above.
[0070] The present invention also provides a computer program product, which, when executed by a processor, implements the aforementioned method for uncertain optimization scheduling of grid-connected power systems for offshore wind power clusters.
[0071] Advantages and benefits of this invention: This invention achieves coordinated and optimized scheduling of offshore wind power clusters under uncertain conditions, improving scheduling accuracy, resource utilization, and grid connection flexibility while maintaining safe and stable system operation. By constructing a distributed, decomposable scheduling model to replace the traditional centralized solution method, this invention enables each wind farm to achieve autonomous optimization locally, thereby significantly improving the real-time performance and scalability of large-scale offshore wind power cluster scheduling. By introducing risk constraints and a multi-scenario rolling optimization mechanism, this invention effectively mitigates the impact of wind speed fluctuations, power uncertainty, and offshore grid disturbances on system operation, making offshore wind power grid connection characteristics more stable, power distribution more reasonable, and maintaining the feasibility and reliability of the scheduling scheme under different operating conditions. Furthermore, this invention achieves power coordination and line load balancing among multiple wind farms through a system-level collaborative strategy, significantly improving the operational stability and scheduling flexibility of the power system under conditions of high-proportion offshore wind power integration. Attached Figure Description
[0072] Figure 1 This is a flowchart of the present invention;
[0073] Figure 2 The following are wind field and wind speed correlation analysis diagrams; (a) is a wind speed spatial correlation coefficient matrix diagram, and (b) is a wind speed joint distribution diagram for a specific time period.
[0074] Figure 3 is a verification diagram of the effectiveness of uncertainty modeling and the rationality of risk robustness design of the present invention. Among them, (a) is the wind speed probability distribution diagram of each wind field, (b) is the available power distribution diagram, and (c) is the available power guarantee rate curve.
[0075] Figure 4 This is a verification diagram of the effectiveness of the risk measurement mechanism and the rationality of the weight allocation of the sub-Bru bar scheduling in this invention. Among them, (a) is a diagram of each risk measurement index, and (b) is a diagram of the sub-Bru bar optimized weight distribution in a typical time period.
[0076] Figure 5 The diagram shows the economic efficiency and multi-wind field coordination optimization effect of the scheduling scheme of the present invention. In this diagram, (a) is a bar chart of the superimposed deviation cost and network loss cost, and (b) is the power allocation ratio among the wind fields. Detailed Implementation
[0077] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0078] Example 1:
[0079] like Figure 1 As shown, an uncertainty-based optimization scheduling method for grid-connected power systems oriented towards offshore wind power clusters includes the following steps:
[0080] S1: Construct an uncertainty prediction model and a cluster power flow linearization model for offshore wind farms to provide a computable infrastructure for subsequent distributed bar scheduling.
[0081] Given the significant randomness, spatial correlation, and rapid fluctuations in wind speed at offshore wind farms, this embodiment first performs joint modeling of wind speed sequences from multiple wind farms. By statistically analyzing the sample distribution of random wind speed disturbances, a fuzzy set of prediction errors is constructed using Wasserstein distance, ensuring that all potential wind speed distributions are included in the uncertainty set. Utilizing the offshore wind power cluster's wind speed-power uncertainty modeling and discrete prediction model, the second-order Wasserstein radius between the empirical probability distribution and the actual distribution is embedded in a fuzzy sphere, guaranteeing the feasibility and safety margin of the scheduling model even under adverse scenarios.
[0082] Furthermore, to balance real-time performance and forward-looking capabilities, single-step and two-step long-time-domain prediction models were established to capture short-term abrupt changes and medium-term trends in the wind speed-power curve, providing a rolling time-domain baseline trajectory for optimized scheduling. Subsequently, a linearized power flow model was established within the offshore wind farm cluster. Based on the topology of the offshore wind farm's collection system, the voltage and current relationships between the primary and secondary sides of the multi-infeed wind farm outputs were linearized in a synchronously rotating dq coordinate system. The forward Euler method was used to discretize the power flow equations, enabling the active power flow of the offshore collection network to be written as a linear mapping between node-injected power and line voltage. This linear model significantly reduces computational complexity while maintaining accuracy, allowing power flow solutions for large-scale wind farm clusters to be executed in real-time within the optimization framework.
[0083] To quantify the economic risks arising from deviations from the plan, a wind power deviation function is constructed. Using the cluster power deviation value function, the expected mismatch cost under uncertainty is characterized by linear absolute value programming. A second-order Lagrange extrapolation prediction reference is introduced to perform high-order corrections on the baseline trajectory in the rolling optimization, reducing cumulative prediction errors.
[0084] Among them, the modeling and discrete prediction model of wind speed-power uncertainty of offshore wind power cluster is shown in formula (1)-(8), the single-step prediction and two-step long time domain prediction model is shown in formula (9)-(12), the cluster power deviation value function is shown in formula (13), and the second-order Lagrange extrapolation prediction reference calculation tool is shown in formula (14)-(15).
[0085] (1)
[0086] (2)
[0087] (3)
[0088] (4)
[0089] (5)
[0090] (6)
[0091] (7)
[0092] (8)
[0093] (9)
[0094] (10)
[0095] (11)
[0096] (12)
[0097] (13)
[0098] (14)
[0099] (15)
[0100] In the formula: For historical samples The empirical distribution constituted; For the landing point The Dirac measure; A data-driven fuzzy set constructed based on Wasserstein distance; The second-order Wasserstein distance; The radius of the fuzzy set; For the first Each wind field at any time The wind speed vector; This is the mean vector of wind speed; Let be the covariance matrix of wind speed; For the first Typhoon machines at all times Maximum available active power output; The unit power curve; , The power linearization coefficient; Inject power vectors into the collection network nodes; The node voltage vector; , To linearize the power flow coefficient matrix and constant terms; Contribute to energy storage; The injected power at the grid connection point; For constraint functions; , , and , These are the coefficients used for regression prediction; , To predict residual terms; For the first Deviation penalty coefficient for each wind field As a weight for network loss costs; Let be the active power loss function of the power grid; The time sampling interval; , It is a first-order difference; , These are reference predictions derived through second-order Lagrange extrapolation.
[0101] S2: A robust scheduling strategy incorporating GlueVaR risk avoidance mechanism and frequency security constraints is designed to achieve safe and stable operation of offshore wind power clusters under extreme scenarios.
[0102] To address the issues of frequency overruns, insufficient local support, and inadequate frequency regulation capabilities in offshore wind power grid-connected systems under high wind speed fluctuations, this embodiment introduces a GlueVaR risk avoidance mechanism into the scheduling model. This mechanism considers the combined risk indicators of VaR and CVaR simultaneously, giving higher weight to potentially high deviation costs in extreme scenarios. By embedding GlueVaR into a distributed bar objective function using risk measurement and risk avoidance, the scheduling model can flexibly match the system's risk preference by adjusting the combined coefficients when facing extreme events such as isolated sudden gusts of wind, maintaining a balance between economy and reliability.
[0103] Subsequently, frequency safety constraints for offshore wind power systems were established. Considering the characteristics of wind power—low inertia, low short-circuit ratio at grid connection, and rapid frequency fluctuations—the nonlinear relationship between frequency sag and disturbance power injection was explicitly analyzed using the system's frequency dynamic equation and the expression for the frequency minimum point. Furthermore, through frequency safety linearization approximation and primary frequency regulation capability constraints, nonlinear indicators such as the frequency minimum point and frequency regulation demand were uniformly transformed into a linear constraint family that can be embedded in an optimization problem. This frequency safety model ensures that the system does not experience excessive frequency sag under any predicted scenario and is suitable for high-proportion wind power integration under weak grid conditions.
[0104] Based on the aforementioned risk and safety constraints, this embodiment constructs a sub-Bluerg risk-avoidance scheduling model for offshore wind power clusters. This model integrates system cost, planning deviation, network loss cost, frequency regulation capacity cost, and the GlueVaR risk term into a single sub-Bluerg optimization framework. The model solves for the optimal scheduling scheme by maximizing the probability distribution of losses, ensuring that it minimizes overall operating costs even under extreme wind speed disturbances. Furthermore, this scheduling framework considers the spatial correlation between wind farms and the grid topology, resulting in scheduling outcomes with high physical feasibility and security.
[0105] Among them, the risk measurement and risk avoidance objective function are shown in formulas (16)-(21), the system frequency dynamic equation and the expression of the lowest frequency point are shown in formulas (22)-(25), and the frequency security linearization approximation and primary frequency modulation capability constraint are shown in formulas (26)-(29).
[0106] (16)
[0107] (17)
[0108] (18)
[0109] (19)
[0110] (20)
[0111] (twenty one)
[0112] (twenty two)
[0113] (twenty three)
[0114] (twenty four)
[0115] (25)
[0116] (26)
[0117] (27)
[0118] (28)
[0119] (29)
[0120] In the formula: It is a random cost variable; Value at risk; Conditional risk value; , The GlueVaR combination coefficients; This is the scheduling decision vector; Weights for risk items; For uncertainties such as wind speed and load; For the first The operating cost function of resources; A set of discrete scenes; For scenarios that satisfy the Wasserstein constraint; This is the system's equivalent moment of inertia. The damping coefficient; This refers to the power disturbance. The lower limit of the allowable frequency deviation; , The coefficients are obtained by approximating the nonlinear nadir function as a piecewise linear function; A set of linear segment indices; For the first The unit's available primary frequency modulation capacity; This is the proportionality coefficient; This is the margin factor for total reserve requirements.
[0121] S3: Integrates uncertainty prediction models, risk-avoidance scheduling models, and distributed dynamic programming solution strategies into a unified offshore wind power grid connection control framework to achieve safe and economical scheduling and scalable solutions for large-scale wind power clusters.
[0122] In the hybrid offshore wind power cluster scheduling system, this embodiment integrates the aforementioned uncertainty prediction model, linear power flow model, frequency security constraints, and GlueVaR risk avoidance mechanism into a large-scale distributed bar optimization scheduling framework, and employs distributed approximate dynamic programming to achieve the solution. Utilizing the offshore wind power cluster scheduling value function decomposition model, the global cost is decomposed into independently computable local sub-value functions based on spatial topology and uncertainty scenarios, laying the structural foundation for distributed approximation.
[0123] In the actual implementation process, each wind farm independently constructs local sub-problems based on local wind speed prediction, unit constraints, and power flow parameters, and approximates the local value function through linear basis functions to estimate future scheduling costs. The main control center is responsible for integrating the local solution results of the wind farms, iterating the sub-problems through the distributed alternating direction multiplier method, coordinating power allocation, frequency security constraints, and cluster power flow balance among wind farms, so that the solution of the entire scheduling problem can be completed efficiently and in parallel.
[0124] By deeply integrating risk-avoidance scheduling with a distributed approximate dynamic programming solution strategy, the proposed unified scheduling and control framework ensures stable operation of the system under various wind power conditions, grid strengths, and operating conditions. Particularly in weak grid environments, the added frequency constraints prevent frequency drops; under extreme wind speed fluctuations, the GlueVaR risk-avoidance mechanism avoids high scheduling costs; and under large-scale wind power cluster conditions, distributed approximate dynamic programming makes the solution time engineering-ready. This framework significantly improves the operational flexibility, safety, and economy of offshore wind power grid-connected systems under conditions of high renewable energy integration. Specifically, the rolling approximation update rule of the value function can adjust the basis function weights online at each real-time rolling step, ensuring that the approximation accuracy continuously evolves with operating conditions.
[0125] Among them, the value function decomposition model of offshore wind power cluster scheduling is shown in formulas (30)-(33), the distributed alternating direction multiplier method iteratively solves the subproblems as shown in formulas (34)-(37), and the value function rolling approximation update rule is shown in formulas (38)-(39).
[0126] (30)
[0127] (31)
[0128] (32)
[0129] (33)
[0130] (34)
[0131] (35)
[0132] (36)
[0133] (37)
[0134] (38)
[0135] (39)
[0136] In the formula: For the system at time The state; For the first The local state of a wind farm; For a moment The set of scheduling decisions; For the first The dispatch cost of a wind farm; For the first The parameter vector of the value function of a wind farm; , , To form the power flow and power balance coefficients constrained by the alternating direction multiplier method; The coordinating variable of the main control center; The Lagrange factor describes the coupling between each wind farm and the main control center; The penalty coefficient for the alternating direction multiplier method; The learning rate; Update the weights based on value.
[0137] Example 2
[0138] Figure 2 (a) is the correlation coefficient matrix of wind speed in the three wind fields; (b) is the joint probability distribution of the time period 11:00, verifying the feasibility of generating scene samples by the "multivariate normal model with spatial correlation" in S1. Figure 3 (a) Display the probability density curves of wind speeds in the three wind fields; Figure 3 (b) Displays the statistical distribution of total available power during key periods, intuitively reflecting the mapping relationship between wind speed and power. Figure 3 (c) indicates that under the weight adjustment of the GlueVaR optimization, the system’s robustness to low availability scenarios is enhanced, supporting the implementation of “GlueVaR risk measurement” and “worst-case scenario weight amplification”. Figure 4 (a) Comparison of VaR0 among the four curves. 86 ,CVaR0. 95 The changes in GlueVaR and the weighted expectation of the split-bar optimization over time demonstrate that GlueVaR has a stronger ability to capture tail risks, and the weighted expectation of the split-bar optimization is significantly increased in extreme periods. Figure 4 (b) The scene distribution curves of the three time periods are arranged in descending order of weight. They show an exponential decay characteristic, indicating that a few high-cost scenes have been significantly amplified in weight. Figure 5(a) shows a superimposed bar chart of deviation cost and network loss cost within 24 hours, proving that the composition of scenario cost conforms to the description of local deviation cost and network loss approximation, and that the proportion of network loss is relatively small, and the scheduling strategy mainly focuses on reducing the risk of plan deviation. Figure 5 (b) indicates that the power distribution exhibits a dynamic adjustment mode, effectively demonstrating the process of power distribution optimization between wind farms over time.
Claims
1. A method for uncertain optimization scheduling of grid-connected power systems for offshore wind power clusters, characterized in that, Includes the following steps: S1: Construct an uncertainty prediction model and a cluster power flow linearization model for offshore wind farms to provide a computable basic structure for subsequent distributed bar scheduling; S2: Design a robust scheduling strategy that incorporates GlueVaR risk avoidance mechanism and frequency security constraints to achieve safe and stable operation of offshore wind power clusters under extreme scenarios; S3: Integrates uncertainty prediction models, risk-avoidance scheduling models, and distributed dynamic programming solution strategies into a unified offshore wind power grid connection control framework to achieve safe and economical scheduling and scalable solutions for large-scale wind power clusters.
2. The method for uncertain optimization scheduling of grid-connected power systems for offshore wind power clusters according to claim 1, characterized in that, The specific process of constructing the uncertainty prediction model and cluster power flow linearization model for offshore wind farms in step S1 is as follows: First, based on the randomness, spatial correlation, and rapid fluctuation of wind speed in offshore wind farms, a joint modeling of wind speed sequences from multiple wind farms is performed. By statistically analyzing the sample distribution of random wind speed disturbances, a model for modeling and discrete prediction of wind speed-power uncertainty in offshore wind farm clusters is constructed using Wasserstein distance. The second-order Wasserstein radius between the empirical probability distribution and the true distribution is calculated, and a fuzzy sphere that can be used for decompositional bar optimization is established. Then, based on the topology of the power collection system of offshore wind farms, the voltage and current relationship between the primary and secondary sides of the multi-feed wind farm output is linearized in the synchronous rotating dq coordinate system, and the forward Euler method is used to discretize the power flow equation, so that the active power flow of the offshore power collection network can be written as a linear mapping between the node injected power and the line voltage. Finally, a cluster power deviation value function is constructed, and the planned deviation is characterized by linear absolute value programming. At the same time, the baseline trajectory is corrected by combining single-step prediction and two-step long-time domain prediction models and second-order Lagrange extrapolation prediction reference auxiliary calculation tools, so that the deviation function can measure the output risk of the wind farm under uncertainty conditions. The prediction model, linear power flow model and deviation model are integrated to uniformly describe the operation behavior of offshore wind power clusters in a state space framework that can be used for subsequent sub-Bruker optimization.
3. The method for uncertain optimization scheduling of grid-connected power systems for offshore wind power clusters according to claim 2, characterized in that, The model involved in step S1 is characterized by the following formulas: the model for modeling and discrete prediction of wind speed-power uncertainty in offshore wind power clusters is shown in formulas (1)-(8); the single-step prediction and two-step long time domain prediction models are shown in formulas (9)-(12); the cluster power deviation value function is shown in formula (13); and the second-order Lagrange extrapolation prediction reference auxiliary calculation tool is shown in formulas (14)-(15). （1） （2） （3） （4） （5） （6） （7） （8） （9） （10） （11） （12） （13） （14） （15） In the formula: For historical samples The empirical distribution constituted; For the landing point The Dirac measure; A data-driven fuzzy set constructed based on Wasserstein distance; The second-order Wasserstein distance; The radius of the fuzzy set; For the first Each wind field at any time The wind speed vector; This is the mean vector of wind speed; Let be the covariance matrix of wind speed; For the first Typhoon machines at all times Maximum available active power output; The unit power curve; , The power linearization coefficient; Inject power vectors into the collection network nodes; The node voltage vector; , To linearize the power flow coefficient matrix and constant terms; Contribute to energy storage; The injected power at the grid connection point; For constraint functions; , , , , These are the coefficients used for regression prediction; , To predict residual terms; For the first Deviation penalty coefficient for each wind field As a weight for network loss costs; Let be the active power loss function of the power grid; The time sampling interval; , It is a first-order difference; , These are reference predictions derived through second-order Lagrange extrapolation.
4. The uncertainty-based optimal scheduling method for grid-connected power systems oriented towards offshore wind power clusters as described in claim 1, characterized in that, The specific process of designing a robust scheduling strategy in step S2 is as follows: First, a GlueVaR risk avoidance mechanism is introduced into the scheduling model. By considering the combined risk indicators between VaR and CVaR, higher weight is given to the high deviation costs that may occur in extreme scenarios. By using risk measurement and risk avoidance objective function, GlueVaR is embedded into the sub-bar target, so that the scheduling model can still flexibly match the system risk preference by adjusting the combination coefficient when facing extreme events such as isolated strong winds and sudden drops, and maintain a balance between economy and reliability. Subsequently, frequency security constraints for the offshore wind power system are established. The relationship between frequency dynamics and disturbance power injection is established using the system frequency response model. The frequency extrema are obtained based on the system frequency dynamic equation and the expression for the lowest frequency point. Then, the nonlinear constraints such as the lowest frequency point and the primary frequency regulation requirement are uniformly converted into linear constraints through frequency security linearization approximation and primary frequency regulation capability constraints, and embedded into the scheduling optimization problem. Finally, a sub-Bluerg risk avoidance scheduling model for offshore wind power clusters is constructed. The system cost, planning deviation, network loss cost, frequency regulation capacity cost and GlueVaR risk term are uniformly incorporated into a sub-Bluerg optimization framework. The optimal scheduling scheme is solved under the probability distribution of maximizing loss, while taking into account the spatial correlation between wind farms and the grid topology.
5. The uncertainty-based optimization scheduling method for grid-connected power systems oriented towards offshore wind power clusters according to claim 4, characterized in that, The model involved in step S2 is characterized by the following formulas: the risk measurement and risk avoidance objective function are shown in formulas (16)-(21), the system frequency dynamic equation and the expression of the frequency minimum point are shown in formulas (22)-(25), and the frequency safety linearization approximation and primary frequency modulation capability constraint are shown in formulas (26)-(29): （16） （17） （18） （19） （20） （21） （22） （23） （24） （25） （26） （27） （28） （29） In the formula: It is a random cost variable; Value at risk; Conditional risk value; , The GlueVaR combination coefficients; This is the scheduling decision vector; Weights for risk items; For uncertainties such as wind speed and load; For the first The operating cost function of resources; A set of discrete scenes; For scenarios that satisfy the Wasserstein constraint; This is the system's equivalent moment of inertia. The damping coefficient; This refers to the power disturbance. The lower limit of the allowable frequency deviation; , The coefficients are obtained by approximating the nonlinear nadir function as a piecewise linear function; A set of linear segment indices; For the first The unit's available primary frequency modulation capacity; This is the proportionality coefficient; This is the margin factor for total reserve requirements.
6. The uncertainty-based optimization scheduling method for grid-connected power systems oriented towards offshore wind power clusters according to claim 1, characterized in that, The specific process of integration and solution in step S3 is as follows: First, the uncertainty prediction model, linear power flow model, frequency security constraints and GlueVaR risk avoidance mechanism are unified and constructed in a large-scale distributed bar optimization scheduling framework, and the solution is achieved by distributed approximate dynamic programming; the solution is decomposed among the wind farms based on the value function decomposition model of offshore wind power cluster scheduling, and each farm retains only local data privacy, and the main control center ensures system-level optimality through coordination; Then, each wind farm independently constructs local sub-problems based on local wind speed prediction, unit constraints, and power flow parameters, and uses linear basis functions to approximate the local value function to form an estimate of future scheduling costs; Finally, the main control center integrates the local results from each field, iteratively solves the subproblems of coordinating power allocation, frequency security and cluster power flow based on the distributed alternating direction multiplier method, and uses the value function rolling approximation update rule to correct the basis function weights online, so that the solution of the entire scheduling problem can be completed efficiently and in parallel.
7. The method for uncertain optimization scheduling of grid-connected power systems for offshore wind power clusters according to claim 6, characterized in that, The model involved in step S3 is characterized by the following formulas: the value function decomposition model of offshore wind power cluster scheduling is shown in formulas (30)-(33), the distributed alternating direction multiplier method iterative solution of subproblems is shown in formulas (34)-(37), and the value function rolling approximation update rule is shown in formulas (38)-(39): （30） （31） （32） （33） （34） （35） （36） （37） （38） （39） In the formula: For the system at time The state; For the first The local state of a wind farm; For a moment The set of scheduling decisions; For the first The dispatch cost of a wind farm; For the first The parameter vector of the value function of a wind farm; , , To form the power flow and power balance coefficients constrained by the alternating direction multiplier method; The coordinating variable of the main control center; The Lagrange factor describes the coupling between each wind farm and the main control center; The penalty coefficient for the alternating direction multiplier method; The learning rate; Update the weights based on value.
8. A grid-connected power system for offshore wind power clusters with uncertainties, used to implement the grid-connected power system for offshore wind power clusters with uncertainties as described in any one of claims 1-7, characterized in that, It includes a model building module, a scheduling strategy design module, and an integrated solution module, and these modules work together. The model building module is used to build an uncertainty prediction model, a cluster power flow linearization model, and a wind power deviation function for offshore wind farms, providing a computable basic structure for subsequent sub-Bruker optimal scheduling. The scheduling strategy design module is used to design robust scheduling strategies that incorporate GlueVaR risk avoidance mechanisms and frequency security constraints, and to construct a robust risk avoidance scheduling model to ensure the safe and stable operation of offshore wind power clusters under extreme scenarios. The integrated solution module is used to integrate the uncertainty prediction model and the distributed bar risk avoidance scheduling model into a unified offshore wind power grid connection control framework. It adopts distributed approximate dynamic programming and alternating direction multiplier method to achieve efficient parallel solution, thereby achieving safe and economical scheduling and scalable solution for large-scale wind power clusters.
9. An electronic device, characterized in that, include: The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement an uncertainty-based optimization scheduling method for grid-connected power systems for offshore wind power clusters as described in any one of claims 1-7.
10. A computer program product, characterized in that, When the computer program / instruction is executed by the processor, it implements the uncertainty optimization scheduling method for grid-connected power systems for offshore wind power clusters as described in any one of claims 1-7.