Unified control method for primary frequency regulation and virtual inertia of wind farm considering rotor kinetic energy of multiple wind turbines and minimum energy storage

Abstract

The application discloses a kind of unified control method of wind farm primary frequency regulation and virtual inertia considering multiple wind turbine rotor kinetic energy and minimum energy storage cooperation, the method includes: collecting power grid frequency and wind turbine operating state data, constructs frequency regulation resource digitization configuration model by multi-source heterogeneous data fusion, the frequency regulation power demand is decomposed into fast response component and sustained support component;Based on frequency deviation threshold decision engine, generate basic control strategy, then by multi-objective optimization decision engine, optimization solution is carried out in combination with distributed cooperative control;Using multi-wind turbine-energy storage power-time collaborative allocation algorithm, priority queue is generated by multiple wind turbine kinetic energy aggregation evaluation, multi-machine power dynamic allocation is realized by stage time sequence rotation cooperative control, combined with energy storage dynamic compensation and speed safety boundary monitoring, output multi-machine linkage frequency regulation power instruction set, realize closed-loop control.The application can minimize energy storage configuration capacity, effectively suppress secondary frequency drop, and is suitable for large-scale wind farm grid-connected frequency control.

Description

Technical Field

[0001] This invention belongs to the field of new energy grid connection control technology, specifically involving a unified control method for primary frequency regulation and virtual inertia of wind farms that takes into account the kinetic energy of multiple wind turbine rotors and the coordination of minimum energy storage, and is applicable to the grid connection frequency control of large-scale wind farms. Background Technology

[0002] With the increasing penetration rate of wind power, wind farms' participation in primary frequency regulation has become a mandatory requirement for grid connection. Although wind power generation, as an important form of renewable energy, has developed rapidly in recent years, the volatility of wind turbine output power leads to dynamic imbalances between power source and load in the grid, resulting in increased grid frequency deviation and seriously jeopardizing the safe and stable operation of the power system. To address the problem of the ever-increasing wind power penetration rate in the grid and improve the frequency regulation capability of the power system, it is necessary to optimize the control of wind turbine converters to enable wind turbines to participate in frequency support and inertia response.

[0003] Traditional wind farm frequency regulation technology has the following technical bottlenecks:

[0004] First, due to inconsistent time-series benchmarks and significant differences in communication latency, static threshold judgments are insufficient to adapt to dynamic frequency changes in multi-source heterogeneous operating data, resulting in inadequate accuracy in frequency regulation resource assessment. The timestamps and communication latency of data collected from different data sources (frequency measurement devices, wind turbine SCADA systems, energy storage management systems) differ, and existing technologies do not adequately consider the issues of time-series alignment and dynamic weight allocation in data fusion.

[0005] Secondly, existing frequency regulation systems mostly rely on independent response of a single wind turbine or pure energy storage solutions, lacking in-depth exploration of the kinetic energy aggregation characteristics of multiple wind turbine rotors. This makes it difficult to effectively handle the time dependence and spatial correlation of multi-turbine speed coupling and power timing distribution. Current virtual inertia control strategies focus more on the control capability of a single unit, without fully considering the synergistic effect of frequency regulation of each unit in a wind farm, making it difficult to maximize the virtual inertia support capability provided by the rotor kinetic energy of each wind turbine in a wind farm.

[0006] Third, frequency regulation strategies are mostly based on a single power command or fixed ratio allocation, failing to form a closed loop of frequency sensing, resource assessment, coordinated control, and state recovery. This makes it impossible to achieve coordinated optimization of rotor kinetic energy and energy storage capacity, resulting in excessively high energy storage investment or insufficient frequency regulation performance.

[0007] Fourth, the lack of a clear decision-making mechanism for the connection between speed recovery and energy storage recharge makes it difficult to convert frequency deviation indicators into precise multi-turbine speed recovery parameters and energy storage charging parameters, thus failing to meet the frequency regulation control requirements of wind farms throughout the entire process. The physical mechanism of secondary frequency drops caused by synchronous speed recovery of multiple wind turbines in existing technologies is that each wind turbine simultaneously increases its speed during the frequency recovery phase, and the sudden increase in mechanical power causes a sharp drop in electromagnetic power, forming a power deficit in the same direction as the primary drop.

[0008] In addition, existing wind-storage joint frequency regulation schemes mostly use fuzzy reasoning systems for decision-making, which have problems such as rule conflicts and insufficient interpretability, making it difficult to meet the requirements of determinism and interpretability of control strategies in actual engineering.

[0009] In summary, the shortcomings of the existing technology are: (1) it does not fully consider the spatiotemporal coupling characteristics and timing rotation requirements of the rotor kinetic energy of multiple wind turbines; (2) it does not establish a systematic method for frequency regulation demand decomposition and minimum energy storage optimization configuration; (3) it does not adopt a deterministic decision-making mechanism based on frequency deviation threshold, and the fuzzy reasoning system has problems of rule conflict and insufficient interpretability; (4) it does not effectively solve the problem of secondary frequency drop caused by synchronous speed recovery of multiple wind turbines. Summary of the Invention

[0010] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a unified control method for primary frequency regulation and virtual inertia of wind farms that considers the kinetic energy of multiple wind turbine rotors and the coordination of minimum energy storage. This invention achieves deep coordination between the kinetic energy of multiple wind turbine rotors and the energy storage system by constructing a full-link closed-loop control architecture of perception-evaluation-allocation-compensation-protection-recovery. It minimizes the energy storage capacity while ensuring frequency regulation performance and fundamentally solves the problem of secondary frequency drops caused by the synchronous speed recovery of multiple wind turbines.

[0011] To achieve the above objectives, the present invention adopts the following technical solution: a unified control method for primary frequency regulation and virtual inertia of a wind farm that considers the coordination of multiple wind turbine rotor kinetic energy and minimum energy storage, comprising the following steps:

[0012] Step (1) Collect grid frequency signals and operating status data of multiple wind turbines in the wind farm, fuse the multi-source heterogeneous data, and construct a digital configuration model of wind farm frequency regulation resources with spatiotemporal correlation characteristics; based on frequency regulation demand decomposition, decompose the frequency regulation power demand into fast response component and continuous support component.

[0013] Step (2) Based on the frequency regulation resource digital configuration model, the frequency deviation threshold decision engine generates a basic control strategy according to the frequency deviation. The frequency deviation threshold decision engine divides the frequency state into multiple discrete levels based on the preset frequency deviation threshold range, and each level corresponds to a deterministic control strategy mode. Then, the basic control strategy is received based on the multi-objective optimization decision engine. The multi-objective optimization decision engine takes minimizing the frequency deviation, minimizing the wind turbine speed recovery time, and minimizing the energy storage state of charge return target value as optimization objectives, and takes wind turbine mechanical stress constraints, speed safety constraints, and energy storage power constraints as constraints. It combines distributed collaborative control to perform optimization and solve the problem, and outputs wind turbine speed recovery parameters and energy storage charging and discharging parameters.

[0014] Step (3) uses a multi-wind turbine-energy storage power-time collaborative allocation algorithm to process frequency disturbance events. Based on the wind turbine speed recovery parameters and energy storage charging and discharging parameters output in step (2), the available rotor kinetic energy of each wind turbine is first calculated by the multi-wind turbine kinetic energy aggregation and evaluation module and a priority queue is generated. Then, the multi-machine power dynamic allocation is realized through phased time-series rotation collaborative control. The phased time-series rotation includes: in the initial stage of frequency drop, all wind turbines release virtual inertia synchronously according to the proportion of available rotor kinetic energy; in the continuous frequency regulation stage, one or several wind turbines are time-series dominant and the remaining wind turbines are coordinated to support according to the priority queue, in which the dominant wind turbine undertakes the main part of the slow-changing power; in the steady-state transition stage, the output of the wind turbine cluster is reduced and the energy storage undertakes the main frequency regulation responsibility; then, the energy storage dynamic compensation mechanism is introduced to fill the power deficit, and combined with the speed safety boundary monitoring to prevent chain disconnection; finally, the multi-machine linkage frequency regulation power instruction set is output through power synthesis to realize closed-loop control from frequency disturbance detection to frequency regulation power output.

[0015] Further, in step (1), the fusion processing of the multi-source heterogeneous data includes:

[0016] All data is mapped to a unified time axis using a time alignment algorithm; the frequency measurement device outputs an analog signal, which is converted into a digital frequency signal using a phase-locked loop (PLL); the wind turbine SCADA system transmits speed, power, and torque data via the MODBUS / TCP protocol, and corrects communication delays using timestamps; an adaptive weight fusion method based on variational inference dynamically assigns and fuses the time-aligned operating status dataset to generate a frequency regulation resource feature vector, where the expectation of the variational distribution serves as the dynamic weight of each data source; finally, all data is associated with a unified time-series coordinate system to form a time-aligned operating status dataset.

[0017] The dynamic weights are adaptively adjusted based on the proximity of each fan speed to the minimum safe speed: when the difference between the speed of the i-th fan and the minimum safe speed is less than 10% of the rated speed, the dynamic weight of its corresponding data source is reduced by the attenuation coefficient, so that the contribution of the available kinetic energy of the fan in the frequency regulation resource feature vector decreases as the speed approaches the safe boundary.

[0018] Further, in step (1), the construction of the frequency modulation resource digital configuration model includes:

[0019] The fused multi-physical quantities are mapped to the frequency modulation demand space, and the frequency modulation power reference value satisfies:

[0020]

[0021] in, This is the reference value for frequency modulation power. The droop coefficient is... The inertia coefficient, This is for frequency deviation;

[0022] Decompose the frequency modulation power demand into fast response components. and continuous support components :

[0023]

[0024]

[0025] Where t is the duration calculated from the start of frequency modulation, τf is the fast decay time constant, τs is the slow decay time constant, and α is the transition coefficient, which ranges from 0.2 to 0.4 and is used to prevent power jumps and smooth the power switching between wind turbines and energy storage.

[0026] Establish a minimum energy storage optimization configuration model with the goal of minimizing the rated power and capacity of energy storage:

[0027]

[0028] in, Let be the objective function. Cost per unit power Cost per unit capacity This is the rated power of the energy storage. This refers to the rated capacity of the energy storage.

[0029] The constraints include:

[0030] Power constraints:

[0031] Energy constraints:

[0032] Kinetic energy constraints:

[0033] in, To continuously support the components The maximum value, To aggregate usable rotor kinetic energy for wind farms To provide continuous support for the duration;

[0034] By iteratively solving the optimization model, the minimum energy storage power and capacity are obtained, and a complete digital configuration model for wind farm frequency regulation resources is constructed.

[0035] Furthermore, in step (2), the multi-wind turbine-energy storage power-time collaborative allocation algorithm adopts a five-level progressive structure, wherein:

[0036] The sensing layer collects data on grid frequency, frequency change rate, wind turbine speed, power, and energy storage SOC.

[0037] The evaluation layer uses three layers of feature encoding—numerical encoding, temporal encoding, and spatial encoding—for embedding processing. The multi-wind turbine kinetic energy aggregation evaluation module calculates the available rotor kinetic energy of each wind turbine and generates a priority queue.

[0038] The allocation layer introduces a phased time-series rotation mechanism. In the initial stage of frequency drop, all wind turbines synchronously release rotor kinetic energy to provide virtual inertia support. In the continuous frequency regulation stage, the wind turbine cluster takes the lead and provides coordinated support according to the rotation cycle. In the steady-state transition stage, the output ratio of wind turbines is gradually reduced.

[0039] The compensation layer dynamically fills the power deficit through energy storage;

[0040] The protection layer monitors the rotational speed safety boundary and automatically disconnects faulty fans and updates the queue when the rotational speed is below the safety threshold.

[0041] Furthermore, the multi-wind kinetic energy aggregation assessment and priority queue generation in the assessment layer includes:

[0042] For each wind turbine, assess its available rotor kinetic energy:

[0043]

[0044] in, Let be the available rotor kinetic energy of the i-th wind turbine. Let be the moment of inertia of the i-th wind turbine. The current rotational speed, The minimum safe speed of the i-th fan;

[0045] Calculate priority score And sort them to generate queue Q:

[0046]

[0047] in, The priority score for the i-th wind turbine is given, where w1, w2, and w3 are weighting coefficients, and h... i The health factor ranges from 0.8 to 1.0, p. i As a penalty factor, p is applied when the rotational speed approaches the safety boundary. i Increase;

[0048] The power allocation of the phased time-series rotation mechanism is as follows:

[0049] Phase A: Distribute rapid response power to each wind turbine according to the proportion of available kinetic energy. : ;

[0050] Phase B: Rotation according to priority queue, with the main wind turbine taking over. Auxiliary fan 1 undertakes Auxiliary fan 2 undertakes , where η is the proportion of the main force, and γ1 and γ2 are the proportions of the auxiliary allocation;

[0051] Phase C: Total output of the wind turbine cluster Energy storage undertakes , where β is the depreciation coefficient, ranging from 0.2 to 0.4.

[0052] Furthermore, in step (3), the multi-objective optimization decision engine specifically comprises:

[0053] Establish a multi-objective optimization problem, with the optimization objective being:

[0054] Minimize the integral of the square of the frequency deviation: ;

[0055] Minimize rotational speed recovery time: ;

[0056] Minimize the deviation of energy storage SOC from the target value: ;

[0057] Where J1, J2, and J3 are the optimization objective functions for the square integral of frequency deviation, the rotational speed recovery time, and the deviation of energy storage SOC from the target value, respectively; The frequency deviation at time t; Let i be the moment when the speed of the i-th fan returns to its operating range. Let i be the moment when the speed of the i-th wind turbine begins to recover. The energy storage state of charge at time t; The target state of charge for energy storage;

[0058] The constraints include: the rate of change of wind turbine mechanical power. Speed ​​safety constraints Energy storage power constraints Energy storage SOC constraints ;

[0059] in, For the mechanical power of the fan; This represents the upper limit of the rate of change of mechanical power. Rated power of energy storage; , These are the minimum and maximum values ​​of the energy storage state of charge, respectively;

[0060] The distributed collaborative control adopts a distributed execution and coordinated optimization framework. Each wind turbine controller makes collaborative decisions based on local observations and information from neighboring controllers, while the energy storage controller independently decides on charging and discharging power. The central coordinator evaluates the Pareto front of the joint optimization objective based on the global state and transforms the multi-objective problem into a single objective using a weighted summation method.

[0061]

[0062] in, The weighted overall objective function is dimensionless. These are weighting coefficients, dimensionless. The normalization factor for each objective is dimensionless.

[0063] The output parameters are the speed recovery rate parameters of each fan and the energy storage charging and discharging power parameters.

[0064] Further, in step (3), the frequency deviation threshold decision engine specifically comprises:

[0065] Power demand is calculated using a frequency modulation demand model based on the frequency deviation Δf. , where D() is the frequency modulation function with implicit droop control characteristics, and R is the droop rate;

[0066] Four frequency deviation threshold ranges are set: the first threshold Δf1 is the dividing point between normal frequency and slight frequency drop, the second threshold Δf2 is the dividing point between slight frequency drop and frequency drop, and the third threshold Δf3 is the dividing point between frequency drop and sharp frequency drop.

[0067] When |Δf| ≤ |Δf1|, it is determined to be in the normal frequency range, the control strategy mode is normal power generation mode, the wind turbine operates at maximum power point tracking, and the energy storage remains in standby mode;

[0068] When |Δf1| < |Δf| ≤ |Δf2|, it is determined to be a frequency drop range, the control strategy mode is a slight frequency adjustment mode, the wind turbine releases a small amount of rotor kinetic energy, and the energy storage is compensated with low power.

[0069] When |Δf2| < |Δf| ≤ |Δf3|, it is determined to be a frequency drop range. The control strategy mode is a moderate frequency regulation mode. The wind turbine releases rotor kinetic energy according to the priority queue, and the energy storage is compensated with medium power.

[0070] When |Δf| > |Δf3|, it is determined to be a frequency drop interval. The control strategy mode is emergency frequency regulation mode, all wind turbines release rotor kinetic energy at full power, and the energy storage is output at rated power.

[0071] The control strategy patterns corresponding to the above four intervals constitute a deterministic state machine. The state transitions are triggered by the frequency deviation threshold and do not require a fuzzy reasoning process.

[0072] The present invention also provides a unified control system for primary frequency regulation and virtual inertia of a wind farm that considers the coordination of multi-wind turbine rotor kinetic energy and minimum energy storage to implement the above method, comprising:

[0073] The data acquisition module is used to collect grid frequency, wind turbine speed, power, torque and energy storage SOC data at the wind farm grid connection point and inside the wind turbine through frequency measurement devices, wind turbine SCADA system and energy storage management system;

[0074] The data fusion module is used to perform time-series alignment and dynamic weight allocation fusion processing on multi-source heterogeneous data using an adaptive weight fusion method based on variational inference.

[0075] The frequency regulation resource digital configuration model construction module is used to construct a wind farm frequency regulation resource digital configuration model with spatiotemporal correlation characteristics based on the fused data, and decompose the frequency regulation power demand into a fast response component and a continuous support component.

[0076] The frequency deviation threshold decision engine is used to divide the frequency state into multiple discrete levels based on the frequency modulation power demand output by the frequency modulation resource digital configuration model, according to the preset frequency deviation threshold range. Each level corresponds to a deterministic control strategy mode, and outputs a basic control strategy.

[0077] The multi-objective optimization decision module is used to receive the basic control strategy, take minimizing frequency deviation, minimizing wind turbine speed recovery time and minimizing the energy storage state of charge regression target value as optimization objectives, take wind turbine mechanical stress constraints, speed safety constraints and energy storage power constraints as constraints, and combine distributed cooperative control to perform optimization solution, and output wind turbine speed recovery parameters and energy storage charging and discharging parameters.

[0078] The multi-machine linkage frequency modulation power instruction set construction module is used to process frequency disturbance events based on the wind turbine speed recovery parameters and energy storage charging and discharging parameters using a multi-wind turbine-energy storage power-time collaborative allocation algorithm. First, the multi-wind turbine kinetic energy aggregation and evaluation module evaluates the available rotor kinetic energy of each wind turbine and generates a priority queue. Then, the phased time-sequential rotation collaborative control module realizes the dynamic allocation of multi-machine power. Subsequently, an energy storage dynamic compensation mechanism is introduced to realize deep interaction between rotor kinetic energy and energy storage power, combined with speed safety boundary monitoring to prevent chain disconnection. Finally, the multi-machine linkage frequency modulation power instruction set is output through power synthesis to realize closed-loop control from frequency disturbance detection to frequency modulation power output.

[0079] Furthermore, in the data fusion module, the adaptive weight fusion method based on variational inference dynamically assigns weights to the time-aligned operating status dataset. Each dynamic weight satisfies the normalization condition and is adaptively adjusted according to the proximity of each wind turbine speed to the minimum safe speed: when the difference between the wind turbine speed and the minimum safe speed is less than 10% of the rated speed, the dynamic weight of its corresponding data source is reduced by the attenuation coefficient, so that the contribution of the available kinetic energy of the wind turbine in the frequency regulation resource feature vector attenuates as the speed approaches the safe boundary.

[0080] Furthermore, in the multi-objective optimization decision module, each wind turbine controller makes collaborative decisions based on local observations and information from adjacent controllers, while the energy storage controller independently decides on charging and discharging power. The central coordinator, based on the Pareto front of the joint optimization objective in the global state assessment, transforms the multi-objective into a single objective for solution using a weighted summation method. The optimization objectives include minimizing the integral of the square of the frequency deviation, minimizing the wind turbine speed recovery time, and minimizing the deviation of the energy storage SOC from the target value. The constraints include constraints on the rate of change of wind turbine mechanical power, speed safety constraints, energy storage power constraints, and energy storage SOC constraints.

[0081] Compared with the prior art, the present invention has the following beneficial effects:

[0082] (1) This invention employs precise fusion of multi-source data and adaptive weight allocation through variational inference to improve model reliability. The constructed digital configuration model for frequency regulation resources can accurately reflect the spatiotemporal distribution characteristics of frequency regulation resources such as rotor kinetic energy and energy storage capacity of wind farms.

[0083] (2) This invention introduces a phased time-series rotation mechanism (phase A: synchronous release of multiple turbines, phase B: time-series rotation with time-sharing dominance, and phase C: steady-state transition) through a multi-wind turbine-energy storage power-time collaborative allocation algorithm. This fundamentally solves the problem of secondary frequency drop caused by synchronous speed recovery of multiple turbines. By staggering the turbine speed recovery times, the mechanical power increment is dispersed in the time domain, thus fundamentally eliminating the physical mechanism of secondary frequency drop caused by synchronous speed recovery.

[0084] (3) The present invention adopts a joint architecture of frequency deviation threshold decision engine and multi-objective optimization decision engine. The threshold decision engine provides deterministic state classification and basic control strategy, and the multi-objective optimization engine solves Pareto optimal solution under explicit constraints. Compared with fuzzy reasoning and deep reinforcement learning methods, it has the advantages of strong computational determinism, good parameter interpretation and strong engineering implementation.

[0085] (4) The present invention forms a closed-loop control of frequency sensing, resource assessment, collaborative control and state recovery. By decomposing frequency regulation demand and optimizing the minimum energy storage configuration, it significantly reduces the energy storage configuration capacity compared with the traditional pure energy storage scheme, and has both improved frequency regulation performance and optimized economic benefits. Attached Figure Description

[0086] Figure 1 A block diagram illustrating the implementation principle of the wind farm primary frequency regulation and virtual inertia unified control method considering the kinetic energy of multiple wind turbine rotors and the coordination of minimum energy storage, provided in an embodiment of the present invention.

[0087] Figure 2 This is a block diagram illustrating the implementation principle of the multi-wind turbine-energy storage power-time collaborative allocation algorithm in this embodiment of the invention.

[0088] Figure 3 This is a schematic diagram of wind farm frequency regulation performance monitoring data in an embodiment of the present invention; wherein, (a) is the 24-hour change curve of grid frequency, (b) is the 24-hour change curve of energy storage system SOC, (c) is the time-domain response curve of frequency deviation during the frequency drop event, (d) is the output distribution curve of wind turbine and energy storage, (e) is the phased time-series power distribution curve, and (f) is the SOC change curve of energy storage system during the frequency drop event;

[0089] Figure 4 This is a schematic diagram comparing the utilization rate of frequency regulation resources in an embodiment of the present invention; wherein, (a) is a comparison of the maximum frequency deviation, (b) is a comparison of the frequency recovery time, (c) is a comparison of the secondary frequency drop amplitude, (d) is a comparison of the energy storage capacity configuration, (e) is a comparison of the rotor kinetic energy utilization rate, and (f) is a comparison of the frequency regulation compliance rate. Detailed Implementation

[0090] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0091] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0092] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0093] like Figure 1 As shown in the figure, this embodiment provides a unified control method for primary frequency regulation and virtual inertia of wind farm that takes into account the kinetic energy of multiple wind turbine rotors and the coordination of minimum energy storage. The specific implementation steps are as follows.

[0094] Step (1) Collection and Decomposition

[0095] Collect grid frequency signals and operating status data of multiple wind turbines in the wind farm, fuse multi-source heterogeneous data, and construct a digital configuration model of wind farm frequency regulation resources with spatiotemporal correlation characteristics; based on frequency regulation demand decomposition, decompose frequency regulation power demand into fast response component and continuous support component.

[0096] Step (1.1) Time-series alignment of multi-source heterogeneous data

[0097] Different data sources have different timestamps and communication delays. A time alignment algorithm is used to map all data to a unified time axis. The frequency measurement device outputs an analog signal, which is converted into a digital frequency signal using a phase-locked loop (PLL). The sampling frequency is f. s The wind turbine SCADA system transmits operational data via the MODBUS / TCP protocol with a timestamp accuracy of 1ms, and compensates for communication latency using a network latency correction model. Ultimately, all data is correlated to a unified time-series coordinate system, forming a time-aligned operational status dataset.

[0098] Step (1.2) Adaptive weight fusion based on variational inference

[0099] A probabilistic generation model is constructed, assuming that the true FM resource features are generated by weighted summation from various data sources. An approximate posterior distribution is inferred through variational inference, and the expectation of the variational distribution is used as the dynamic weights. The time-aligned runtime dataset is then fused according to the dynamic weights to form an FM resource feature vector F. fused Each dynamic weight satisfies the normalization condition and is adaptively adjusted according to the proximity of each fan speed to the minimum safe speed: when the difference between the fan speed and the minimum safe speed is less than 10% of the rated speed, the dynamic weight of the corresponding data source is reduced by the attenuation coefficient, so that the contribution of the available kinetic energy of the fan in the frequency regulation resource feature vector decreases as the speed approaches the safe boundary.

[0100] Step (1.3) Decomposition of frequency regulation power demand and optimization of minimum energy storage configuration

[0101] The fused multi-physical quantities are mapped to the frequency modulation demand space to construct a digital configuration model. Frequency modulation power reference value. Satisfying the primary frequency regulation standard of the power grid and energy conservation, the governing equation is the droop-inertia joint response equation:

[0102]

[0103] in, This is the reference value for frequency modulation power. The droop coefficient is... The inertia coefficient, This represents the frequency deviation.

[0104] Decompose the frequency modulation power demand into fast response components. and continuous support components The fast response component is calculated according to the exponential decay law:

[0105]

[0106] Where t is the duration from the start of frequency modulation, and τf is the rapid decay time constant, typically 2 seconds.

[0107] The sustained support component is calculated according to the double exponential decay law:

[0108]

[0109] Where τs is the slow decay time constant; α is the transition coefficient, ranging from 0.2 to 0.4, with a typical value of 0.3, used to prevent power step and smooth the power switching between wind turbine and energy storage.

[0110] The selection of the transition coefficient is based on the following: ensuring that the rate of change of the wind turbine output at the switching moment does not exceed the limit value of the wind turbine mechanical stress, based on the constraint of the power change rate; the stricter the constraint of the power change rate, the smaller the value of α.

[0111] Establish a minimum energy storage optimization configuration model with the goal of minimizing the rated power and capacity of energy storage:

[0112]

[0113] in, Let be the objective function. Cost per unit power Cost per unit capacity This is the rated power of the energy storage. This refers to the rated capacity of the energy storage.

[0114] The constraints include:

[0115] Power constraints:

[0116] Energy constraints:

[0117] Kinetic energy constraints:

[0118] in, To continuously support the components The maximum value, To aggregate usable rotor kinetic energy for wind farms To provide continuous support for the duration.

[0119] By iteratively solving the optimization model (using particle swarm optimization or interior point method), the minimum energy storage power and capacity are obtained, and a complete digital configuration model for wind farm frequency regulation resources is constructed.

[0120] Step (2) Power Allocation and Execution

[0121] Frequency disturbance events are handled using a multi-wind turbine-energy storage power-time collaborative allocation algorithm. For example... Figure 2 As shown, the multi-wind turbine-energy storage power-time collaborative allocation algorithm adopts a five-level progressive structure, as detailed below:

[0122] Step (2.1) Perception layer

[0123] Collect data on the power grid frequency f, the rate of change of frequency df / dt, and the rotational speed ω of N wind turbines. i Power P i And energy storage SOC data, assuming the wind farm contains N wind turbines G = {G1, ..., G...} N The time axis is discretized into T time points, and the input is a multidimensional state vector X. t .

[0124] Step (2.2) Evaluation layer

[0125] Embedding processing is performed using a three-layer feature encoding approach: numerical encoding, temporal encoding, and spatial encoding. For each wind turbine G... i Assess its available rotor kinetic energy ,in Let be the available rotor kinetic energy of the i-th wind turbine. Let be the moment of inertia of the i-th wind turbine. The current rotational speed, Let the minimum safe operating speed be the i-th wind turbine. Calculate the maximum release power and duration. Calculate the priority score. ,in, The priority score for the i-th wind turbine is given, where w1, w2, and w3 are weighting coefficients, and h...i The health factor ranges from 0.8 to 1.0, p. i As a penalty factor, p is applied when the rotational speed approaches the safety boundary. i Increase. Press. Generate a priority queue Q = [G(1), G(2), ..., G(N)] by sorting in descending order.

[0126] Step (2.3) Allocation Layer – Phased Timing Rotation Cooperative Control

[0127] A phased, time-sequential rotation mechanism is introduced, allowing the wind turbine cluster to take turns leading and coordinating support, achieving dynamic power allocation among multiple turbines. A rotation period T is set. ro = 10 seconds, time offset Δt = 5 seconds.

[0128] Phase A (0 to 5 seconds): Initial phase of frequency drop, multiple turbines release power simultaneously. Each turbine distributes its rapid response power according to the proportion of available kinetic energy. .

[0129] Phase B (5 to 30 seconds):

[0130] During the continuous frequency adjustment phase, timing rotation occurs. Main wind turbine index. The main wind turbines bear the responsibility (Main force bearing the burden η = 0.6), auxiliary fan 1 bears the burden (Auxiliary allocation ratio γ1 = 0.25), auxiliary fan 2 undertakes... (Auxiliary allocation ratio γ2 = 0.15).

[0131] Phase C (30 seconds to 15 minutes): Steady-state frequency regulation phase, gradually reducing the proportion of wind turbine output, with energy storage assuming the main frequency regulation responsibility: total output of the wind turbine cluster. Energy storage undertakes , where β is the depreciation coefficient, ranging from 0.2 to 0.4.

[0132] The switching time uses a ramp function for smooth transition:

[0133] In the formula, T ramp = 2 seconds, to avoid power step.

[0134] Step (2.4) Compensation Layer

[0135] A dynamic energy storage compensation algorithm is adopted to compensate based on the coupling relationship between power deficit and State of Charge (SOC). The SOC adaptive compensation coefficient Ksoc is set in segments according to the energy storage's state of charge: Ksoc is 1.2 when SOC is greater than 0.8, 1.0 when SOC is between 0.5 and 0.8, 0.8 when SOC is between 0.2 and 0.5, and 0.3 when SOC is less than or equal to 0.2. Energy storage compensation power... ,in To address the power deficit, the coupled power is synthesized globally from the wind turbine to the energy storage system, and then mapped to frequency regulation power commands through power constraints.

[0136] Step (2.5) Protective Layer

[0137] By monitoring the speed safety boundary, the system automatically disconnects faulty fans and updates the queue when the speed falls below the safety threshold. A safety factor is defined. ,when If the safety factor is ≤ 0, disconnect the wind turbine. Update the priority queue Q and add the safety factor. Fans with a value of 0 are removed from the active queue.

[0138] Step (3) State recovery and storage

[0139] The multi-machine linkage frequency modulation power instruction set output in step (2) is used as the input of the frequency deviation threshold decision engine. The basic control strategy is generated by judging the frequency deviation threshold. Then, the multi-objective optimization decision engine is used for optimization, and the wind turbine speed recovery parameters and energy storage charging and discharging parameters are output to realize closed-loop control.

[0140] Step (3.1) Frequency modulation demand calculation

[0141] Power demand is calculated using a frequency modulation demand model based on the frequency deviation Δf. Where D(·) is the frequency modulation function with implicit droop control characteristics. R is the adjustment coefficient.

[0142] Step (3.2) Frequency Deviation Threshold Decision

[0143] Three frequency deviation thresholds are set: Δf1 = 0.05 Hz (normal / slight drop boundary), Δf2 = 0.15 Hz (slight drop / precipitation boundary), and Δf3 = 0.30 Hz (precipitation / sharp drop boundary). Based on these thresholds, the frequency state is divided into four discrete levels:

[0144] Level I – Normal frequency range: |Δf| ≤ 0.05 Hz, control strategy mode is normal power generation mode, wind turbine operates at maximum power point tracking, energy storage remains in standby state, speed recovery rate is the normal rate v r,nom ;

[0145] Level II – Slight Frequency Drop Range: 0.05 Hz < |Δf| ≤ 0.15 Hz. The control strategy mode is a slight frequency adjustment mode, where the wind turbine releases a small amount of rotor kinetic energy (release ratio 10% to 20%), and energy storage compensates with low power. The speed recovery rate slowed down to ;

[0146] Level III – Frequency drop range: 0.15 Hz < |Δf| ≤ 0.30 Hz. The control strategy mode is moderate frequency regulation mode. The wind turbine releases rotor kinetic energy in priority queues (release ratio 40% to 60%), and the energy storage compensates with medium power. Speed ​​recovery paused (v) r = 0);

[0147] Level IV – Frequency Drop Range: |Δf| > 0.30 Hz. The control strategy mode is emergency frequency regulation mode, with all wind turbines releasing rotor kinetic energy at full power (release ratio 80% to 100%), and energy storage outputting at rated power. Speed ​​recovery paused and, if necessary, reverse power absorption. .

[0148] The above four levels constitute a deterministic state machine. State transitions are triggered only by the frequency deviation threshold. The control parameters corresponding to each level are pre-set deterministic values, requiring no fuzzy reasoning process.

[0149] Step (3.3) Multi-objective optimization solution

[0150] Establish a multi-objective optimization problem, with the optimization objective being:

[0151] Minimize the integral of the square of the frequency deviation: ;

[0152] Minimize rotational speed recovery time: ;

[0153] Minimize the deviation of energy storage SOC from the target value: .

[0154] Where J1, J2, and J3 are the optimization objective functions for the square integral of frequency deviation, the rotational speed recovery time, and the deviation of energy storage SOC from the target value, respectively; The frequency deviation at time t; Let i be the moment when the speed of the i-th fan returns to its operating range. Let i be the moment when the speed of the i-th wind turbine begins to recover. The energy storage state of charge at time t; The target state of charge for energy storage.

[0155] The constraints include: the rate of change of wind turbine mechanical power. Speed ​​safety constraints Energy storage power constraints Energy storage SOC constraints .in, For the mechanical power of the fan; This represents the upper limit of the rate of change of mechanical power. Rated power of energy storage; , These are the minimum and maximum values ​​of the energy storage state of charge, respectively.

[0156] The multi-objective problem is transformed into a single objective using a weighted summation method:

[0157]

[0158] in, The weighted overall objective function is dimensionless. The weighting coefficients are dimensionless and satisfy λ1 + λ2 + λ3 = 1; typical weighting values ​​are: λ1 = 0.5 (frequency deviation first), λ2 = 0.3 (speed recovery second), λ3 = 0.2 (SOC regression). The normalization factor for each objective is dimensionless.

[0159] The distributed collaborative control adopts a distributed execution and coordinated optimization framework: each wind turbine controller makes collaborative decisions based on local observations (machine speed, power) and information from adjacent controllers (adjacent wind turbine speed, output status); the energy storage controller independently decides the charging and discharging power; the central coordinator evaluates the Pareto front of the joint optimization objective based on the global state assessment, and performs calculations per T. coord = Coordinate the global optimization results once per second. Solve the above optimization problem using a sequential quadratic programming algorithm, and output the recovery rate parameter v of each wind turbine speed. ri With energy storage charging and discharging power parameter P b .

[0160] Step (3.4) Instruction Output

[0161] After frequency deviation threshold decision-making and multi-objective optimization, the final wind turbine speed recovery command and energy storage charging and discharging command are generated, realizing closed-loop control from frequency disturbance detection to frequency modulation power output.

[0162] This embodiment also provides a unified control system for primary frequency regulation and virtual inertia of a wind farm to implement the above method, including:

[0163] The data acquisition module is used to collect grid frequency, wind turbine speed, power, torque and energy storage SOC data at the wind farm grid connection point and inside the wind turbine through frequency measurement devices, wind turbine SCADA system and energy storage management system;

[0164] The data fusion module is used to perform time-series alignment and dynamic weight allocation fusion processing on multi-source heterogeneous data using an adaptive weight fusion method based on variational inference.

[0165] The frequency regulation resource digital configuration model construction module is used to construct a wind farm frequency regulation resource digital configuration model with spatiotemporal correlation characteristics based on the fused data, and decompose the frequency regulation power demand into a fast response component and a continuous support component.

[0166] The frequency deviation threshold decision engine is used to divide the frequency state into multiple discrete levels based on the frequency modulation power demand output by the frequency modulation resource digital configuration model, according to the preset frequency deviation threshold range. Each level corresponds to a deterministic control strategy mode, and outputs a basic control strategy.

[0167] The multi-objective optimization decision module is used to receive the basic control strategy, take minimizing frequency deviation, minimizing wind turbine speed recovery time and minimizing the energy storage state of charge regression target value as optimization objectives, take wind turbine mechanical stress constraints, speed safety constraints and energy storage power constraints as constraints, and combine distributed cooperative control to perform optimization solution, and output wind turbine speed recovery parameters and energy storage charging and discharging parameters.

[0168] The multi-machine linkage frequency modulation power instruction set construction module is used to process frequency disturbance events based on the wind turbine speed recovery parameters and energy storage charging and discharging parameters using a multi-wind turbine-energy storage power-time collaborative allocation algorithm. First, the multi-wind turbine kinetic energy aggregation and evaluation module evaluates the available rotor kinetic energy of each wind turbine and generates a priority queue. Then, the phased time-sequential rotation collaborative control module realizes the dynamic allocation of multi-machine power. Subsequently, an energy storage dynamic compensation mechanism is introduced to realize deep interaction between rotor kinetic energy and energy storage power, combined with speed safety boundary monitoring to prevent chain disconnection. Finally, the multi-machine linkage frequency modulation power instruction set is output through power synthesis to realize closed-loop control from frequency disturbance detection to frequency modulation power output.

[0169] The implementation process of this method will be further explained below using a wind farm containing 50 2MW doubly-fed wind turbines as an example. The total rated capacity of this wind farm is 100MW, and the energy storage system is configured with a rated power of 12MW and a rated capacity of 24MWh.

[0170] Step 1: Multi-source heterogeneous data acquisition and fusion processing

[0171] A frequency measurement device is installed at the grid connection point of the wind farm to sample the frequency f. s= 100Hz; Wind turbine speed, power, and torque data are collected via the wind turbine SCADA system using the MODBUS / TCP protocol, with a timestamp accuracy of 1ms; State of Charge (SOC) data is collected via the energy storage management system. Frequency regulation resource feature vector F is generated through time alignment algorithm and variational inference adaptive weight fusion. fused The dynamic weights are adaptively adjusted based on how close each fan speed is to the minimum safe speed. When the speed ω of a certain fan... i With minimum safe speed ω min When the difference is less than 10% of the rated speed, the dynamic weight w of the corresponding data source is... i The attenuation coefficient is reduced to 0.8.

[0172] Step 2: Coordinated Allocation of Multiple Wind Turbines, Energy Storage Power, and Time

[0173] When a frequency drop event occurs (assuming a frequency deviation Δf = -0.3Hz), the execution process of the multi-wind turbine-energy storage power-time collaborative allocation algorithm is as follows:

[0174] Phase A (0 to 5 seconds): 50 wind turbines simultaneously release their rotor kinetic energy. Power is distributed among the turbines according to their available kinetic energy, and the total turbine output... The maximum virtual inertia support is provided when the rate of change of frequency df / dt = -0.15Hz / s. Priority queue Q is based on rating. The values ​​are generated by sorting, where w1 = 0.5, w2 = 0.3, and w3 = 0.2.

[0175] Phase B (5 to 30 seconds): Wind turbines are rotated sequentially according to priority queue Q. The main turbine (highest score) handles η × Ps = 0.6 × 9MW = 5.4MW, auxiliary turbine 1 handles γ1 × Ps = 0.25 × 9MW = 2.25MW, and auxiliary turbine 2 handles γ2 × P... s = 0.15 × 9MW = 1.35MW. Rotation cycle The time offset Δt = 5 seconds ensures no single fan over-releases. A ramp function is used for smooth transition during switching, T... ramp = 2 seconds.

[0176] Phase C (30 seconds to 15 minutes): Steady-state frequency regulation. The total output of the wind turbine cluster decreases to... Energy storage undertakes The fan gradually entered the speed recovery mode.

[0177] Step 3: Construction of Digital Configuration Model for Frequency Modulation Resources

[0178] Set the droop coefficient K d = 30MW / Hz, inertia coefficient KH = 15 pu.s, fast decay time constant τf = 2s, slow decay time constant τs = 300s, transition coefficient α = 0.3.

[0179] Calculate the reference value of frequency modulation power And decomposed into fast response components. With continuous support components .

[0180] Establish a minimum energy storage optimization configuration model and set the continuous support time T. s = 15 minutes, minimum safe speed ω min = 0.7 times the rated speed. The particle swarm optimization algorithm was used iteratively, with a population size of 30, 100 iterations, and an inertia weight of 0.8. The optimization results are shown in Table 1:

[0181] Table 1 Comparison Results of Optimal Energy Storage Configuration

[0182] <![CDATA[Energy storage rated power P b > 30 MW 12 MW 60% <![CDATA[Energy storage rated capacity E b > 60 MWh 24 MWh 60% Rotor kinetic energy utilization rate — 85% — Duration of support 15 minutes 15 minutes flat

[0183] Step 4: Frequency Deviation Threshold Decision and Multi-Objective Optimization

[0184] A frequency deviation of Δf = -0.3Hz falls into Level IV (frequency drop range), triggering emergency frequency regulation mode: all wind turbines release rotor kinetic energy at full power (release ratio 90%), and the energy storage is output at rated power. Speed ​​recovery paused and, if necessary, reverse power absorption. .

[0185] The multi-objective optimization decision engine receives the basic control strategy described above and performs optimization. The optimization objective weights are set as follows: λ1 = 0.5 (frequency deviation priority), λ2 = 0.3 (speed recovery secondary), and λ3 = 0.2 (SOC regression). Constraints include: wind turbine mechanical power change rate constraint. Speed ​​safety constraints Energy storage power constraint |Pb| ≤ 12MW; energy storage SOC constraint 0.1 ≤ SOC ≤ 0.9. A sequential quadratic programming algorithm is used for solution, with a convergence tolerance of 10⁻⁻⁶. 6 The maximum number of iterations is 200. The optimized wind turbine speed recovery rate parameter v... ri With energy storage charging and discharging power parameter P b Output to the execution layer.

[0186] Step 5: Simulation Verification and Performance Comparison

[0187] To verify the effectiveness of the method of the present invention, a simulation model of a wind farm with 50 2MW doubly fed wind turbines was built on the MATLAB / Simulink platform, and the following three schemes were compared and tested: (1) traditional pure energy storage frequency regulation scheme; (2) single wind turbine virtual inertia + energy storage scheme; (3) multi-wind turbine coordinated frequency regulation scheme proposed in this invention. Simulation conditions: the initial output of the wind farm was 80MW, and the load suddenly increased by 20MW, causing the frequency to drop.

[0188] The key performance indicators are compared in Table 2:

[0189] Table 2 Performance Comparison of Frequency Modulation Schemes

[0190] Maximum frequency deviation (Hz) -0.42 -0.28 -0.18 Frequency recovery time (s) 35 22 15 Secondary frequency drop amplitude (Hz) none -0.12 -0.03 Energy storage configuration capacity (MWh) 60 45 24 Rotor kinetic energy utilization rate (%) 0 45 85 Frequency modulation duration (min) 15 10 15 FM compliance rate (%) 78 88 96

[0191] The comparison results show that: the maximum frequency deviation of the present invention is reduced from -0.42Hz (Scheme 1) and -0.28Hz (Scheme 2) to -0.18Hz; Scheme 2 experiences a secondary drop of -0.12Hz during the speed recovery phase, while the present invention controls the secondary drop amplitude to within -0.03Hz through a phased timing rotation mechanism; the energy storage capacity is reduced by 60% compared to Scheme 1 and by 47% compared to Scheme 2; the rotor kinetic energy utilization rate reaches 85%, which is 40 percentage points higher than that of Scheme 2.

[0192] Step 6: Multi-condition adaptability verification

[0193] To verify the adaptability of the strategy of this invention under different disturbance types, simulations were performed on the following two supplementary operating conditions:

[0194] Operating Condition 1: Frequency Surge. A simulated heavy load shedding caused a frequency surge, Δf = +0.25Hz. Results showed that the wind turbine increased its speed by absorbing power, the energy storage system absorbed excess power, and the frequency recovery time was 18s with no overshoot. This verified the bidirectional response capability of the frequency deviation threshold decision engine to positive and negative frequency deviations.

[0195] Operating Condition 2: Dual Disturbance of Source and Load. The wind farm output fluctuated from 80MW to 60MW within 30 seconds, while the load suddenly increased by 10MW. The results show that the data fusion module adjusts the dynamic weight of each wind turbine in real time, the weight of turbines with lower speeds automatically decreases, the frequency regulation task is shifted to higher speed turbines, and the maximum frequency deviation is controlled within -0.15Hz, verifying the effectiveness of multi-source data fusion and dynamic weight allocation.

[0196] Figure 3 The wind farm frequency regulation performance monitoring data in this embodiment is shown.

[0197] Figure 3In the figure, (a) is the 24-hour frequency variation curve of the power grid, and (b) is the 24-hour SOC variation curve of the energy storage system. The frequency fluctuates slightly around the rated value of 50Hz throughout the day. At about 12:00, a sudden increase in load occurs, causing a frequency drop event. The SOC of the energy storage system changes dynamically according to the frequency regulation demand throughout the day, and a significant change in the depth of discharge occurs during the frequency drop event.

[0198] Figure 3 In the diagram, (c) shows the time-domain response curve of the frequency deviation during the frequency drop event, and (d) shows the output distribution curve of the wind turbine and energy storage. After a sudden load increase, the frequency drops rapidly. In the initial stage, the scheme of this invention releases the rotor kinetic energy synchronously through 50 wind turbines, controlling the maximum frequency deviation to -0.18Hz. During the continuous frequency adjustment phase from 5s to 30s, stable power output is maintained through time-sequential switching. After 30s, a steady-state transition phase begins, with the wind turbine output gradually decreasing and the energy storage system smoothly taking over. The frequency recovers to near the rated value within 15s without a secondary drop. The single-wind turbine virtual inertia + energy storage scheme (Scheme 2) experiences a secondary drop of -0.12Hz around 22s. Phase A (0-5s): All wind turbine speeds decrease synchronously. Phase B (5-30s): Main and auxiliary wind turbines take turns performing frequency regulation tasks according to the rotation cycle, with speeds changing in a stepwise manner. Phase C (after 30s): The recovery times of each wind turbine speed are staggered, and the increase in mechanical power is dispersed in the time domain. The minimum speed of a single wind turbine remains above 0.72 pu, which is higher than the safety threshold of 0.7 pu. In the rapid response phase (0-5s), the wind turbine cluster output is approximately 11MW, and energy storage compensates for approximately 3MW. In the continuous frequency regulation phase (5-30s), wind turbine output switches according to the rotation cycle, and energy storage compensates for the remaining deficit. In the steady-state transition phase (after 30s), wind turbine output drops to 2.7MW, and energy storage output rises to 6.3MW, achieving a smooth power handover from wind turbines to energy storage.

[0199] Figure 3 In the diagram, (e) shows the phased time-series power allocation curve, and (f) shows the SOC change curve of the energy storage system during the frequency drop event. In Phase A (0-5s), each wind turbine allocates rapid response power according to its available kinetic energy ratio; in Phase B (5-30s), the main wind turbine, auxiliary turbine 1, and auxiliary turbine 2 take turns undertaking frequency regulation tasks according to priority queues, with energy storage dynamically compensating for power deficits; in Phase C (after 30s), the output of the wind turbine cluster gradually decreases, and energy storage assumes the main responsibility for frequency regulation. The initial SOC is 65%. During Phases A and B, energy storage compensates for power deficits with low power, and the SOC decreases slowly; after Phase C, energy storage assumes the main responsibility for frequency regulation, and the SOC continues to decrease; after the frequency regulation ends, the SOC returns to the target value at the optimal rate.

[0200] Figure 4 A comparison chart of frequency modulation resource utilization in this embodiment is shown.

[0201] Figure 4In the figures, (a) shows the comparison of maximum frequency deviation, (b) shows the comparison of frequency recovery time, and (c) shows the comparison of secondary frequency drop amplitude. The traditional pure energy storage scheme (Scheme 1) has a maximum frequency deviation of 0.42Hz, a frequency recovery time of 35s, and no secondary drop; the single wind turbine virtual inertia + energy storage scheme (Scheme 2) has a maximum frequency deviation of 0.28Hz, a frequency recovery time of 22s, and a secondary drop of -0.12Hz; the scheme of this invention has a maximum frequency deviation of 0.18Hz, a frequency recovery time of 15s, and a secondary drop of -0.03Hz. This invention significantly suppresses the maximum frequency deviation through multi-machine coordinated frequency regulation, shortens the frequency recovery time through a phased timing rotation mechanism, and the secondary drop amplitude is negligible, verifying that the timing rotation mechanism fundamentally eliminates the secondary drop mechanism caused by synchronous speed recovery.

[0202] Figure 4 In the table, (d) compares the energy storage capacity configuration, (e) compares the rotor kinetic energy utilization rate, and (f) compares the frequency regulation compliance rate. Scheme 1 has an energy storage configuration of 60MWh, a rotor kinetic energy utilization rate of 0%, and a frequency regulation compliance rate of 78%; Scheme 2 has an energy storage configuration of 45MWh, a rotor kinetic energy utilization rate of 45%, and a frequency regulation compliance rate of 88%; the scheme of this invention requires only 24MWh of energy storage, has a rotor kinetic energy utilization rate of 85%, and a frequency regulation compliance rate of 96%. Through frequency regulation demand decomposition and minimum energy storage optimization configuration, the energy storage can bear the continuous support component while the rapid response component is covered by the kinetic energy of multiple wind turbine rotors, improving by 40 percentage points compared to Scheme 2. This demonstrates that this invention fully taps the aggregation frequency regulation potential of multi-machine rotor kinetic energy, achieving a balance between frequency deviation suppression, speed recovery, and SOC return.

[0203] The wind farm primary frequency regulation and virtual inertia unified control method proposed in this invention, which takes into account the kinetic energy of multiple wind turbine rotors and the coordination of minimum energy storage, can effectively identify and prioritize the handling of frequency disturbance events, achieve optimal frequency regulation performance under minimum energy storage configuration, and has both rapid response and continuous support capabilities.

[0204] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0205] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0206] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0207] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0208] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A unified control method for primary frequency regulation and virtual inertia of a wind farm that considers the coordination of multiple wind turbine rotor kinetic energy and minimum energy storage, characterized in that, Includes the following steps: Step (1) Collect grid frequency signals and operating status data of multiple wind turbines in the wind farm, fuse the multi-source heterogeneous data, and construct a digital configuration model of wind farm frequency regulation resources with spatiotemporal correlation characteristics; based on frequency regulation demand decomposition, decompose the frequency regulation power demand into fast response component and continuous support component. Step (2) Based on the frequency regulation resource digital configuration model, the frequency deviation threshold decision engine generates a basic control strategy according to the frequency deviation. The frequency deviation threshold decision engine divides the frequency state into multiple discrete levels based on the preset frequency deviation threshold range, and each level corresponds to a deterministic control strategy mode. Then, the basic control strategy is received based on the multi-objective optimization decision engine. The multi-objective optimization decision engine takes minimizing the frequency deviation, minimizing the wind turbine speed recovery time, and minimizing the energy storage state of charge return target value as optimization objectives, and takes wind turbine mechanical stress constraints, speed safety constraints, and energy storage power constraints as constraints. It combines distributed collaborative control to perform optimization and solve the problem, and outputs wind turbine speed recovery parameters and energy storage charging and discharging parameters. Step (3) uses a multi-wind turbine-energy storage power-time collaborative allocation algorithm to process frequency disturbance events. Based on the wind turbine speed recovery parameters and energy storage charging and discharging parameters output in step (2), the available rotor kinetic energy of each wind turbine is first calculated by the multi-wind turbine kinetic energy aggregation and evaluation module and a priority queue is generated. Then, the multi-machine power dynamic allocation is realized through phased time-series rotation collaborative control. The phased time-series rotation includes: in the initial stage of frequency drop, all wind turbines release virtual inertia synchronously according to the proportion of available rotor kinetic energy. During the continuous frequency regulation phase, one or more wind turbines are prioritized and time-sharingly dominated by a priority queue, with the remaining turbines providing coordinated support. The dominant turbine undertakes the main part of the slowly changing power. During the steady-state transition phase, the output of the wind turbine cluster decreases and energy storage assumes the main responsibility for frequency regulation. Subsequently, a dynamic compensation mechanism for energy storage is introduced to fill the power deficit, combined with speed safety boundary monitoring to prevent cascading grid disconnection. Finally, the power synthesis outputs a multi-machine linkage frequency regulation power command set, realizing closed-loop control from frequency disturbance detection to frequency regulation power output.

2. The wind farm primary frequency regulation and virtual inertia unified control method considering the kinetic energy of multiple wind turbine rotors and the minimum energy storage coordination as described in claim 1, characterized in that, In step (1), the fusion processing of the multi-source heterogeneous data includes: All data is mapped to a unified time axis using a time alignment algorithm; the frequency measurement device outputs an analog signal, which is converted into a digital frequency signal using a phase-locked loop (PLL); the wind turbine SCADA system transmits speed, power, and torque data via the MODBUS / TCP protocol, and corrects communication delays using timestamps; an adaptive weight fusion method based on variational inference dynamically assigns and fuses the time-aligned operating status dataset to generate a frequency regulation resource feature vector, where the expectation of the variational distribution serves as the dynamic weight of each data source; finally, all data is associated with a unified time-series coordinate system to form a time-aligned operating status dataset. The dynamic weights are adaptively adjusted based on the proximity of each fan speed to the minimum safe speed: when the difference between the speed of the i-th fan and the minimum safe speed is less than 10% of the rated speed, the dynamic weight of its corresponding data source is reduced by the attenuation coefficient, so that the contribution of the available kinetic energy of the fan in the frequency regulation resource feature vector decreases as the speed approaches the safe boundary.

3. The wind farm primary frequency regulation and virtual inertia unified control method considering the kinetic energy of multiple wind turbine rotors and the minimum energy storage coordination as described in claim 1, characterized in that, In step (1), the construction of the frequency modulation resource digital configuration model includes: The fused multi-physical quantities are mapped to the frequency modulation demand space, and the frequency modulation power reference value satisfies: in, This is the reference value for frequency modulation power. The droop coefficient is... The inertia coefficient, This is for frequency deviation; Decompose the frequency modulation power demand into fast response components. and continuous support components : Where t is the duration calculated from the start of frequency modulation, τf is the fast decay time constant, τs is the slow decay time constant, and α is the transition coefficient, which is used to prevent power jumps and smooth the power switching between wind turbines and energy storage. Establish a minimum energy storage optimization configuration model with the goal of minimizing the rated power and capacity of energy storage: in, Let be the objective function. Cost per unit power Cost per unit capacity This is the rated power of the energy storage. This refers to the rated capacity of the energy storage. The constraints include: Power constraints: Energy constraints: Kinetic energy constraints: in, To continuously support the components The maximum value, To aggregate usable rotor kinetic energy for wind farms To provide continuous support for the duration; By iteratively solving the optimization model, the minimum energy storage power and capacity are obtained, and a complete digital configuration model for wind farm frequency regulation resources is constructed.

4. The wind farm primary frequency regulation and virtual inertia unified control method considering the kinetic energy of multiple wind turbine rotors and the minimum energy storage coordination as described in claim 1, characterized in that, In step (2), the multi-wind turbine-energy storage power-time collaborative allocation algorithm adopts a five-level progressive structure, wherein: The sensing layer collects data on grid frequency, frequency change rate, wind turbine speed, power, and energy storage SOC. The evaluation layer uses three layers of feature encoding—numerical encoding, temporal encoding, and spatial encoding—for embedding processing. The multi-wind turbine kinetic energy aggregation evaluation module calculates the available rotor kinetic energy of each wind turbine and generates a priority queue. The allocation layer introduces a phased time-series rotation mechanism. In the initial stage of frequency drop, all wind turbines synchronously release rotor kinetic energy to provide virtual inertia support. In the continuous frequency regulation stage, the wind turbine cluster takes the lead and provides coordinated support according to the rotation cycle. In the steady-state transition stage, the output ratio of wind turbines is gradually reduced. The compensation layer dynamically fills the power deficit through energy storage; The protection layer monitors the rotational speed safety boundary and automatically disconnects faulty fans and updates the queue when the rotational speed is below the safety threshold.

5. The wind farm primary frequency regulation and virtual inertia unified control method considering the kinetic energy of multiple wind turbine rotors and the minimum energy storage coordination as described in claim 4, characterized in that, The multi-wind kinetic energy aggregation assessment and priority queue generation in the assessment layer include: For each wind turbine, assess its available rotor kinetic energy: in, Let be the available rotor kinetic energy of the i-th wind turbine. Let be the moment of inertia of the i-th wind turbine. The current rotational speed, The minimum safe speed of the i-th fan; Calculate priority score And sort them to generate queue Q: in, The priority score for the i-th wind turbine is given, where w1, w2, and w3 are weighting coefficients, and h... i For health factors, p i As a penalty factor, p is applied when the rotational speed approaches the safety boundary. i Increase; The power allocation of the phased time-series rotation mechanism is as follows: Phase A: Distribute rapid response power to each wind turbine according to the proportion of available kinetic energy. : ; Phase B: Rotation according to priority queue, with the main wind turbine taking over. Auxiliary fan 1 undertakes Auxiliary fan 2 undertakes , where η is the proportion of the main force, and γ1 and γ2 are the proportions of the auxiliary allocation; Phase C: Total output of the wind turbine cluster Energy storage undertakes , where β is the depreciation coefficient.

6. The wind farm primary frequency regulation and virtual inertia unified control method considering the kinetic energy of multiple wind turbine rotors and the coordination of minimum energy storage, as described in claim 1, is characterized in that... In step (3), the multi-objective optimization decision engine specifically refers to: Establish a multi-objective optimization problem, with the optimization objective being: Minimize the integral of the square of the frequency deviation: ; Minimize rotational speed recovery time: ; Minimize the deviation of energy storage SOC from the target value: ; Where J1, J2, and J3 are the optimization objective functions for the square integral of frequency deviation, the rotational speed recovery time, and the deviation of energy storage SOC from the target value, respectively; The frequency deviation at time t; Let i be the moment when the speed of the i-th fan returns to its operating range. Let i be the moment when the speed of the i-th wind turbine begins to recover. The energy storage state of charge at time t; The target state of charge for energy storage; The constraints include: the rate of change of wind turbine mechanical power. Speed ​​safety constraints Energy storage power constraints Energy storage SOC constraints ; in, For the mechanical power of the fan; This represents the upper limit of the rate of change of mechanical power. Rated power of energy storage; , These are the minimum and maximum values ​​of the energy storage state of charge, respectively; The distributed collaborative control adopts a distributed execution and coordinated optimization framework. Each wind turbine controller makes collaborative decisions based on local observations and information from neighboring controllers, while the energy storage controller independently decides on charging and discharging power. The central coordinator evaluates the Pareto front of the joint optimization objective based on the global state and transforms the multi-objective problem into a single objective using a weighted summation method. in, The weighted overall objective function is dimensionless. These are weighting coefficients, dimensionless. The normalization factor for each objective is dimensionless. The output parameters are the speed recovery rate parameters of each fan and the energy storage charging and discharging power parameters.

7. The wind farm primary frequency regulation and virtual inertia unified control method considering the kinetic energy of multiple wind turbine rotors and the minimum energy storage coordination as described in claim 1, characterized in that, In step (3), the frequency deviation threshold decision engine specifically refers to: Power demand is calculated using a frequency modulation demand model based on the frequency deviation Δf. , where D() is the frequency modulation function with implicit droop control characteristics, and R is the droop rate; Four frequency deviation threshold ranges are set: the first threshold Δf1 is the dividing point between normal frequency and slight frequency drop, the second threshold Δf2 is the dividing point between slight frequency drop and frequency drop, and the third threshold Δf3 is the dividing point between frequency drop and sharp frequency drop. When |Δf| ≤ |Δf1|, it is determined to be in the normal frequency range, the control strategy mode is normal power generation mode, the wind turbine operates at maximum power point tracking, and the energy storage remains in standby mode; When |Δf1| < |Δf| ≤ |Δf2|, it is determined to be a frequency drop range, the control strategy mode is a slight frequency adjustment mode, the wind turbine releases a small amount of rotor kinetic energy, and the energy storage is compensated with low power. When |Δf2| < |Δf| ≤ |Δf3|, it is determined to be a frequency drop range. The control strategy mode is a moderate frequency regulation mode. The wind turbine releases rotor kinetic energy according to the priority queue, and the energy storage is compensated with medium power. When |Δf| > |Δf3|, it is determined to be a frequency drop interval. The control strategy mode is emergency frequency regulation mode, all wind turbines release rotor kinetic energy at full power, and the energy storage is output at rated power. The control strategy patterns corresponding to the above four intervals constitute a deterministic state machine. The state transitions are triggered by the frequency deviation threshold and do not require a fuzzy reasoning process.

8. A unified control system for primary frequency regulation and virtual inertia of a wind farm, considering the coordination of multi-wind turbine rotor kinetic energy and minimum energy storage, for implementing the method as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to collect grid frequency, wind turbine speed, power, torque and energy storage SOC data at the wind farm grid connection point and inside the wind turbine through frequency measurement devices, wind turbine SCADA system and energy storage management system; The data fusion module is used to perform time-series alignment and dynamic weight allocation fusion processing on multi-source heterogeneous data using an adaptive weight fusion method based on variational inference. The frequency regulation resource digital configuration model construction module is used to construct a wind farm frequency regulation resource digital configuration model with spatiotemporal correlation characteristics based on the fused data, and decompose the frequency regulation power demand into a fast response component and a continuous support component. The frequency deviation threshold decision engine is used to divide the frequency state into multiple discrete levels based on the frequency modulation power demand output by the frequency modulation resource digital configuration model, according to the preset frequency deviation threshold range. Each level corresponds to a deterministic control strategy mode, and outputs a basic control strategy. The multi-objective optimization decision module is used to receive the basic control strategy, take minimizing frequency deviation, minimizing wind turbine speed recovery time and minimizing the energy storage state of charge regression target value as optimization objectives, take wind turbine mechanical stress constraints, speed safety constraints and energy storage power constraints as constraints, and combine distributed cooperative control to perform optimization solution, and output wind turbine speed recovery parameters and energy storage charging and discharging parameters. The multi-machine linkage frequency modulation power instruction set construction module is used to process frequency disturbance events based on the wind turbine speed recovery parameters and energy storage charging and discharging parameters using a multi-wind turbine-energy storage power-time collaborative allocation algorithm. First, the multi-wind turbine kinetic energy aggregation and evaluation module evaluates the available rotor kinetic energy of each wind turbine and generates a priority queue. Then, the phased time-sequential rotation collaborative control module realizes the dynamic allocation of multi-machine power. Subsequently, an energy storage dynamic compensation mechanism is introduced to realize deep interaction between rotor kinetic energy and energy storage power, combined with speed safety boundary monitoring to prevent chain disconnection. Finally, the multi-machine linkage frequency modulation power instruction set is output through power synthesis to realize closed-loop control from frequency disturbance detection to frequency modulation power output.

9. The unified control system for primary frequency regulation and virtual inertia of a wind farm, considering the kinetic energy of multiple wind turbine rotors and the coordination of minimum energy storage, as described in claim 8, is characterized in that... In the data fusion module, the adaptive weight fusion method based on variational inference dynamically assigns weights to the time-aligned operating status dataset. Each dynamic weight satisfies the normalization condition and is adaptively adjusted according to the proximity of each wind turbine speed to the minimum safe speed: when the difference between the wind turbine speed and the minimum safe speed is less than 10% of the rated speed, the dynamic weight of its corresponding data source is reduced by the attenuation coefficient, so that the contribution of the available kinetic energy of the wind turbine in the frequency regulation resource feature vector attenuates as the speed approaches the safe boundary.

10. The unified control system for primary frequency regulation and virtual inertia of a wind farm, considering the kinetic energy of multiple wind turbine rotors and the coordination of minimum energy storage, as described in claim 8, is characterized in that... In the multi-objective optimization decision-making module, each wind turbine controller makes collaborative decisions based on local observations and information from adjacent controllers, while the energy storage controller independently decides on charging and discharging power. The central coordinator evaluates the Pareto front of the joint optimization objective based on the global state assessment and solves the multi-objective problem by using a weighted summation method. The optimization objectives include minimizing the integral of the square of the frequency deviation, minimizing the wind turbine speed recovery time, and minimizing the deviation of the energy storage SOC from the target value. The constraints include constraints on the rate of change of wind turbine mechanical power, speed safety constraints, energy storage power constraints, and energy storage SOC constraints.

Images