A distributed wind power cluster sub-cluster adaptive frequency support control method

Abstract

The application discloses a kind of dispersed wind power cluster sub-group adaptive frequency support control methods, through gathering unit operation data set and common connection point frequency characteristic sequence, extract operating condition, electrical coupling and inertia reserve characteristics are fused into comprehensive feature vector, and initial subgroup division result and subgroup feature center vector are generated by adaptive density clustering;Frequency deviation and frequency change rate are extracted based on the common connection point frequency characteristic sequence and the frequency response stage is identified, and then the virtual inertia and damping adjustment coefficient are calculated according to the frequency response stage identification signal and the subgroup feature center vector, the system total frequency support power demand is distributed to each subgroup according to the support ability evaluation index, and then distributed to each unit according to the speed safety margin and wind speed interval difference, and incrementally update subgroup division, finally superimposed to the basic active power reference value execution, realize the full excavation of cluster frequency support potential and the coordinated protection of unit operation safety.

Description

Technical Field

[0001] This invention relates to the field of wind power generation control technology, and in particular to a distributed wind power cluster group adaptive frequency support control method. Background Technology

[0002] With the continued advancement of the "dual carbon" target, the installed capacity of new energy power generation, represented by wind power, continues to expand. Distributed wind power, with its characteristics of dispersed deployment, low access voltage levels, and wide geographical distribution, occupies an increasingly important position in regional power grids with a high proportion of new energy connected to the grid. However, traditional variable-speed constant-frequency wind turbines operate under maximum power point tracking (MPPT) control strategies, and there is no direct coupling relationship between the rotor speed and the grid frequency. Unlike traditional synchronous generators, they cannot automatically respond to frequency changes and provide inertia support. As the penetration rate of distributed wind power clusters continues to increase, the system's equivalent rotational inertia decreases significantly, and the risk of frequency instability becomes increasingly prominent.

[0003] To address this, virtual synchronous generator (VSR) control technology has been introduced into renewable energy grid-connected systems. By simulating the rotor motion equations of a synchronous generator in the converter control stage, wind turbines can acquire inertia and damping response characteristics similar to synchronous machines. However, most existing VSR control strategies focus on a single unit or a single field, employing uniform fixed control parameters. This fails to fully consider the significant heterogeneity of units within a distributed wind power cluster in terms of operating conditions, electrical location, and inertia reserves. When each unit responds independently with the same control parameters, some units approach the lower speed limit protection due to excessive release of rotor kinetic energy, while others with support potential fail to fully utilize their capabilities, making it difficult to effectively tap the overall frequency support potential of the cluster. Furthermore, existing control strategies often employ simple average or capacity ratio allocation methods for power distribution, which are ill-suited to the differentiated requirements for inertia support and damping suppression at different stages of frequency response. Summary of the Invention

[0004] In view of this, the purpose of this invention is to propose a distributed wind power cluster grouping adaptive frequency support control method, which dynamically groups heterogeneous units within the cluster and allocates support tasks differently according to the frequency response stage, thereby fully exploring the frequency support potential of the distributed wind power cluster and coordinating and ensuring the safe operation of the units.

[0005] To achieve the above-mentioned technical objectives, the technical solution adopted by this invention is: a distributed wind power cluster grouping adaptive frequency support control method, comprising: Collect operating status data of each wind turbine in the distributed wind power cluster and frequency measurement data of the grid common connection point to generate a unit operation dataset and a frequency feature sequence of the common connection point; Based on the unit operation dataset, the operating condition characteristics, electrical coupling characteristics and inertia reserve characteristics of each wind turbine are extracted and fused into a comprehensive feature vector of the unit. An adaptive density clustering algorithm is used to cluster and divide the integrated feature vector of the unit, generating the initial subgroup division results and the feature center vector of the subgroup; Based on the frequency feature sequence of common connection point, the frequency deviation and frequency change rate are extracted, input into the stage identification state machine to determine the frequency response stage, and generate the frequency response stage identification signal. Based on the frequency response stage identifier signal and the subgroup characteristic center vector, calculate the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup, and generate adaptive control parameters. The total frequency support power requirement of the system is calculated based on the frequency characteristic sequence of the common connection point, a comprehensive evaluation index of support capability is generated, the total frequency support power requirement of the system is allocated to each subgroup according to the comprehensive evaluation index of support capability, and subgroup power support instructions are generated. Based on the rotor speed safety margin and wind speed range division results of each unit, the subgroup power support command is differentiated and allocated to each wind turbine in the subgroup to generate unit power support command; Monitor the evolution of the operating status of each subgroup and the frequency response stage switching event. When the subgroup boundary update condition is triggered, incrementally update the initial subgroup division result based on the latest unit operating status data to generate the updated subgroup division result. The unit power support command is superimposed onto the basic active power reference value of each wind turbine to generate the total active power reference value of the unit, which is then executed by the active power control link of the converter.

[0006] In some embodiments, based on the unit operation dataset, the operating condition characteristics, electrical coupling characteristics, and inertia reserve characteristics of each wind turbine are extracted and fused into a comprehensive unit feature vector, including: The ratio of active power output to rated power, the ratio of rotor speed to rated speed, the ratio of current wind speed to rated wind speed, and the pitch angle position of each wind turbine are extracted from the unit operation dataset to form the operating condition feature sub-vector. The equivalent electrical distance between each wind turbine's access point and the point of common coupling, the equivalent impedance modulus of the access line, and the voltage level of the access point are extracted from the unit operation dataset to form an electrical coupling feature sub-vector. Extract the rotational kinetic energy reserve corresponding to the current rotor speed of each wind turbine, the virtual rotational inertia reference value of the virtual synchronous control configuration, and the historical output depth of participating frequency support within the preset time window from the unit operation dataset to form an inertia reserve feature sub-vector. The operating condition feature vector, electrical coupling feature vector, and inertia reserve feature vector are weighted and fused according to preset weight coefficients to generate the comprehensive feature vector of each wind turbine unit.

[0007] In some embodiments, an adaptive density clustering algorithm is used to cluster the integrated feature vector of the units, generating initial subgroup partitioning results and subgroup feature center vectors, including: Calculate the Euclidean distance matrix between the comprehensive feature vectors of each wind turbine unit, and automatically determine the initial value of the neighborhood radius based on the statistical distribution characteristics of the Euclidean distance matrix; The number of neighboring wind turbines within the neighborhood radius of each wind turbine is used as a density metric to identify high-density core turbine points in the feature space. For non-core unit points, they are assigned to the corresponding density reachable subgroups based on their reachability to each core unit point. For boundary unit points located in the boundary regions of multiple subgroups, obtain the electrical topology connectivity of the boundary unit points. If a boundary unit point has a direct line connection with a unit in a certain subgroup in terms of electrical topology, then the boundary unit point is assigned to that subgroup. The feature center vector of each subgroup is calculated based on the integrated feature vector of the units within each subgroup, and the subgroup affiliation identifier of each unit and the feature center vector of each subgroup are output as the initial subgroup division result.

[0008] In some embodiments, based on the frequency feature sequence of the common connection point, the frequency deviation and the rate of change of frequency are extracted, input into the stage identification state machine for frequency response stage discrimination, and a frequency response stage identification signal is generated, including: The frequency characteristic sequence of the common junction point is subjected to central difference operation to obtain the frequency change rate estimate, and the frequency change rate estimate is smoothed by low-pass filtering to generate the frequency change rate. The frequency deviation is generated by subtracting the real-time frequency value from the rated frequency value in the frequency characteristic sequence of the common junction point; The frequency deviation and frequency change rate are input into the stage identification state machine, which has preset inertial response start conditions, damping coordination start conditions, power recovery start conditions and recovery end conditions. When the absolute value of the rate of change of frequency is greater than the inertial response start threshold and the frequency deviation is in a continuous increasing trend, the stage identification state machine outputs the inertial response stage identifier. When the absolute value of the frequency change rate decreases from greater than the damping coordination start threshold to less than the damping coordination start threshold and the frequency deviation reaches a local extreme value, the stage identification state machine outputs a damping coordination stage identifier. When the absolute value of the frequency deviation begins to decrease continuously and the sign of the frequency change rate is opposite to the sign of the frequency deviation, the stage identification state machine outputs a power recovery stage indicator. When the absolute value of the frequency deviation is less than the steady-state deviation tolerance and the absolute value of the frequency change rate is less than the steady-state change rate tolerance, the stage identification state machine outputs a recovery end flag. Among them, when the stage identification state machine outputs a stage identifier switch, it confirms the switch only after the corresponding switching conditions are met for multiple consecutive control cycles, and generates a frequency response stage identifier signal.

[0009] In some embodiments, based on the frequency response stage identifier signal and the subgroup characteristic center vector, the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are calculated to generate adaptive control parameters, including: Based on the inertia reserve characteristic component in the characteristic center vector of each subgroup, the reference virtual inertia value of each subgroup is determined, and based on the electrical coupling characteristic component in the characteristic center vector of each subgroup, the reference damping coefficient value of each subgroup is determined. When the frequency response stage identifier signal is the inertia response stage identifier, the virtual inertia adjustment coefficient of each subgroup is determined according to the absolute value of the frequency change rate. The virtual inertia adjustment coefficient increases with the increase of the absolute value of the frequency change rate. The virtual inertia adjustment coefficient is corrected according to the operating condition characteristic components in the characteristic center vector of each subgroup. At the same time, the damping adjustment coefficient of each subgroup is set to a level lower than the reference damping coefficient value. When the frequency response stage identifier signal is the damping coordination stage identifier, the damping adjustment coefficient of each subgroup is determined according to the degree to which the frequency deviation is close to the local extremum. The damping adjustment coefficient increases as the degree to which the frequency deviation is close to the local extremum increases. The damping adjustment coefficient is corrected according to the electrical coupling characteristic component in the characteristic center vector of each subgroup. At the same time, the virtual inertia adjustment coefficient of each subgroup is reduced to a level below the reference virtual inertia value. When the frequency response stage indicator signal is the power recovery stage indicator, the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are smoothly adjusted according to the frequency recovery rate. The faster the frequency recovery rate, the faster the virtual inertia adjustment coefficient and damping adjustment coefficient decrease. The rotor speed of the units in each subgroup is monitored. If the rotor speed of the units in the subgroup that exceeds the preset proportion approaches the safety lower limit threshold, the virtual inertia adjustment coefficient of that subgroup is forcibly reduced. The virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are output as adaptive control parameters.

[0010] In some embodiments, the total frequency support power requirement of the system is calculated based on the frequency characteristic sequence of the point of common coupling, and a comprehensive evaluation index of support capability is generated, including: Multiply the frequency deviation by a preset droop coefficient to generate a frequency deviation term, and multiply the frequency change rate by a preset virtual inertia coefficient to generate a frequency change rate term. Determine the weights of the frequency deviation term and the frequency change rate term based on the frequency response stage identifier signal. Increase the weight of the frequency change rate term when in the inertia response stage, and increase the weight of the frequency deviation term when in the damping coordination stage or power recovery stage. Sum the weighted frequency deviation term and the frequency change rate term to generate the total frequency support power requirement of the system. Based on the inertia reserve characteristic component in the characteristic center vector of each subgroup, the available inertia reserve assessment value of each subgroup is generated; based on the electrical coupling characteristic component in the characteristic center vector of each subgroup, the response agility assessment value of each subgroup is generated; based on the safety margin between the rotor speed and the lower speed limit protection value of the units within each subgroup, the state health assessment value of each subgroup is generated. The available inertia reserve assessment value, response agility assessment value, and state health assessment value are weighted and summed according to preset weight coefficients to generate a comprehensive assessment index of the support capability of each subgroup.

[0011] In some embodiments, the total system frequency support power requirement is allocated to each subgroup based on a comprehensive support capability evaluation index, and subgroup power support instructions are generated, including: The comprehensive evaluation index of the support capacity of each subgroup is normalized to generate the initial power allocation weight of each subgroup. Based on the inertia reserve characteristic component and the operating condition characteristic component in the characteristic center vector of each subgroup, the maximum supportable power limit of each subgroup is determined. The initial power allocation weight of each subgroup is multiplied by the total frequency support power requirement of the system to generate the initial power support command for each subgroup. Determine whether the initial power support command of each subgroup exceeds the corresponding maximum support power limit. If it does, redistribute the power difference exceeding the limit according to the proportion of the remaining support capacity of the other subgroups to generate a corrected subgroup power support command. The modified subgroup power support commands for each subgroup are output as subgroup power support commands.

[0012] In some embodiments, based on the rotor speed safety margin of each unit and the wind speed range division results, the subgroup power support command is differentially allocated to each wind turbine within the subgroup to generate unit power support commands, including: Obtain the current rotor speed and the lower limit protection value of each wind turbine in the subgroup, and calculate the rotor speed safety margin of each wind turbine. The rotor speed safety margin is the ratio of the difference between the current rotor speed and the lower limit protection value to the rated speed range. Based on the current wind speed of each wind turbine, the wind turbines are divided into low wind speed zone units, medium wind speed zone units, and high wind speed zone units. When the subgroup power support command is an increased power support command, a power support upper limit constraint is set for the low wind speed zone units, and the subgroup power support command is allocated to each unit according to the rotor speed safety margin ratio of the medium wind speed zone units and the high wind speed zone units, generating an increased power unit support command. When the subgroup power support command is a power reduction recovery command, the support power of each unit is reduced in the following order: low wind speed zone units first, medium wind speed zone units second, and high wind speed zone units last, to generate a power reduction unit recovery command. The rotor speed of each wind turbine is monitored in real time. If the rotor speed of a certain turbine drops to the warning threshold that is close to the lower speed limit protection value, the support power limiting protection of that turbine is triggered, and the support power share originally undertaken by that turbine is transferred to other turbines in the same subgroup with sufficient rotor speed safety margin, and the corresponding turbine power support command is updated.

[0013] In some embodiments, the evolution of the operating status of each subgroup and the frequency response phase switching event are monitored. When the subgroup boundary update condition is triggered, the initial subgroup division result is incrementally updated based on the latest unit operating status data to generate an updated subgroup division result, including: Monitor the switching events of the frequency response phase identifier signal. When the frequency response phase identifier signal switches phases, trigger the cluster boundary assessment. Monitor the rotor speed status of units in each subgroup and the comprehensive evaluation index of the support capacity of each subgroup. When the rotor speed of units in any subgroup exceeding a preset proportion approaches the safety lower limit or the comprehensive evaluation index of the support capacity of the subgroup decreases by more than a preset threshold, the subgroup boundary assessment is triggered. When the cluster boundary assessment is triggered, the latest operating status data of each wind turbine is obtained, the latest comprehensive feature vector of each wind turbine is calculated, and wind turbines whose changes exceed a preset threshold compared with the corresponding historical comprehensive feature vector are marked as turbines with significant status changes. Only for units whose status has changed significantly, their latest integrated feature vector is recalculated and adaptive density clustering is re-executed to determine their subgroup affiliation. Units whose status has not changed significantly retain their original subgroup affiliation, generating updated subgroup division results. For units whose subgroup affiliation changes, the power support command weighted by the original subgroup and the new subgroup is executed within the current control cycle. The weighting coefficient gradually transitions to the power support command value of the new subgroup as the control cycle progresses until the power support command of the new subgroup is fully switched.

[0014] In some embodiments, the unit power support command is superimposed onto the baseline active power reference value of each wind turbine to generate a total active power reference value for the unit, which is then executed through the converter active power control loop, including: The unit power support command of each wind turbine is summed with the corresponding basic active power reference value to generate the total active power reference value of each wind turbine. Apply a power change rate limiting constraint to the total active power reference value of the unit to generate a limited total active power reference value; The total reference value of the limited active power is input to the active power control loop of the converter of each wind turbine. The active power control loop of the converter adjusts the switching state of the converter through the inner loop current control and pulse width modulation strategy to control the active power injected into the grid by each wind turbine. The reactive power control of each wind turbine unit operates independently and does not change its reactive power control mode with the adjustment of active power.

[0015] Compared with the prior art, the present invention, using the above technical solution, has the following beneficial effects: The present invention provides a distributed wind power cluster grouping adaptive frequency support control method. By collecting unit operation datasets and frequency feature sequences of the common coupling point, it extracts operating conditions, electrical coupling, and inertia reserve features and fuses them into a comprehensive feature vector. Adaptive density clustering is used to generate initial subgroup division results and subgroup feature center vectors. Based on the frequency feature sequence of the common coupling point, it extracts frequency deviation and frequency change rate and identifies frequency response stages. Then, based on the frequency response stage identifier signal and the subgroup feature center vector, it calculates virtual inertia and damping adjustment coefficients. The total system frequency support power demand calculated based on the frequency feature sequence of the common coupling point is allocated to each subgroup according to the support capacity assessment index, and then allocated to each unit according to the speed safety margin and wind speed range differences. The subgroup division is updated incrementally and finally superimposed on the basic active power reference value for execution. This realizes the full exploitation of the cluster frequency support potential and the coordinated guarantee of unit operation safety. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of steps S101 to S109 of the method described in the specific implementation embodiment; Figure 2 This is a schematic diagram of steps S201 to S204 of the method described in the specific implementation. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be particularly noted that the following embodiments are for illustrative purposes only and do not limit the scope of the invention. Similarly, the following embodiments are only some, not all, embodiments of the present invention, and all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 This embodiment provides a distributed wind power cluster adaptive frequency support control method, including: S101. Collect the operating status data of each wind turbine in the distributed wind power cluster and the frequency measurement data of the grid common connection point, and generate the unit operation dataset and the frequency feature sequence of the common connection point; S102. Based on the unit operation dataset, extract the operating condition characteristics, electrical coupling characteristics and inertia reserve characteristics of each wind turbine unit, and fuse them into a comprehensive feature vector of the unit. S103. Adaptive density clustering algorithm is used to cluster and divide the unit's comprehensive feature vector to generate initial subgroup division results and subgroup feature center vectors. S104. Based on the frequency feature sequence of the common connection point, extract the frequency deviation and frequency change rate, input the stage identification state machine to perform frequency response stage discrimination, and generate frequency response stage identification signal. S105. Based on the frequency response stage identifier signal and the subgroup characteristic center vector, calculate the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup, and generate adaptive control parameters. S106. Calculate the total frequency support power requirement of the system based on the frequency characteristic sequence of the common connection point, generate a comprehensive evaluation index of support capability, allocate the total frequency support power requirement of the system to each subgroup according to the comprehensive evaluation index of support capability, and generate subgroup power support instructions. S107. Based on the rotor speed safety margin and wind speed range division results of each unit, the subgroup power support command is differentially allocated to each wind turbine unit in the subgroup to generate unit power support command. S108. Monitor the evolution of the operating status of each subgroup and the frequency response stage switching event. When the subgroup boundary update condition is triggered, incrementally update the initial subgroup division result based on the latest unit operating status data to generate the updated subgroup division result. S109. The unit power support command is superimposed onto the basic active power reference value of each wind turbine to generate the total active power reference value of the unit, which is then executed through the converter active power control link.

[0020] In step S101, the unit operation dataset is compiled from local measurement information of each wind turbine in the distributed wind power cluster, covering multiple operating parameters reflecting the current steady-state operating status of the units. The point of common coupling frequency characteristic sequence is composed of frequency trajectory data continuously collected at a fixed sampling period by a measurement device installed at the grid's point of common coupling. These two types of data respectively characterize the individual operating status of each unit within the cluster and the frequency dynamic characteristics of the external power grid, serving together as the basic input for subsequent cluster calculations and frequency support control.

[0021] In step S102, operating condition characteristics, electrical coupling characteristics, and inertia reserve characteristics are extracted from the unit operation dataset. The operating condition characteristics represent the unit's current steady-state operating point and its power regulation margin; the electrical coupling characteristics reflect the tightness of the electrical connection between the unit's connection point and the point of common coupling; and the inertia reserve characteristics characterize the level of rotational kinetic energy that the unit can release in its current state. These three types of characteristics are weighted and fused to form a comprehensive unit feature vector, quantifying the comprehensive response characteristics of each unit under frequency support scenarios in a unified data format.

[0022] In step S103, the adaptive density clustering algorithm automatically identifies high-density areas and forms clusters based on the distribution density of the comprehensive feature vectors of each unit in the feature space. There is no need to pre-specify the number of subgroups. By grouping units with similar characteristics into the same subgroup, the units within each subgroup have a high degree of consistency in frequency response characteristics. The feature center vector of the subgroup serves as a quantitative representation of the overall characteristics of each subgroup, providing a reference benchmark for subsequent differentiated control.

[0023] In step S104, the frequency deviation and the frequency change rate are obtained by comparison and differential operation of the frequency characteristic sequence of the common connection point and the rated frequency, respectively. The stage identification state machine divides the frequency response process into several characteristic stages and outputs frequency response stage identification signals according to the dynamic trajectory of the two and their interrelationship, providing a timing basis for the adaptive adjustment of control parameters.

[0024] In step S105, the virtual inertia adjustment coefficient and the damping adjustment coefficient are used to adjust the inertia response strength and oscillation suppression strength of each subgroup during the frequency support process, respectively. The two are calculated in conjunction with the current control requirements indicated by the frequency response stage identifier signal and the differential characteristics of each subgroup reflected by the subgroup feature center vector. The resulting adaptive control parameters enable different subgroups to obtain matching control characteristics at different frequency response stages.

[0025] In step S106, the total frequency support power requirement of the system is determined by comprehensively considering the frequency deviation and frequency change rate reflected by the frequency characteristic sequence of the common connection point. The comprehensive evaluation index of support capability quantifies the support potential of each subgroup from the dimensions of inertia reserve level, response speed and operational safety margin, and allocates the total requirement according to the actual support capability of each subgroup to generate subgroup power support instructions.

[0026] In step S107, the rotor speed safety margin represents the remaining safety space between the current speed of each unit and the lower speed protection value. The wind speed interval division result reflects the wind speed conditions of the unit and its corresponding kinetic energy reserve level. The combination of the two is used to determine the amount of support power that each unit can safely bear, so that the subgroup power support command is allocated among the units in the group according to the principle of differentiation, and the unit power support command is generated.

[0027] In step S108, the cluster boundary update conditions include two types of trigger sources: frequency response stage switching events and abnormal evolution of subgroup operating states. Meeting either trigger source will trigger the cluster boundary assessment. Specifically, a frequency response stage switching event refers to a jump in the frequency response stage identifier signal output by the stage identification state machine, indicating that the system's frequency dynamic characteristics have entered a new stage, and the role of each subgroup in subsequent control may change. Abnormal evolution of subgroup operating states refers to the rotor speed safety margin or support capacity-related indicators of units within the subgroup deviating from the normal range, indicating that the actual support capacity of the subgroup no longer matches the assessment results at the initial clustering time. During incremental updates, only units whose states have changed significantly are re-assigned, generating updated subgroup division results, allowing the clustering results to adaptively adjust to the dynamic evolution of unit operating states.

[0028] In step S109, the basic active power reference value is the active power output target value determined by each wind turbine according to the maximum power tracking control or preset power dispatch strategy when no frequency disturbance occurs. The unit power support command is superimposed on this basic value to generate the total active power reference value of the unit. The switching state and power output of the converter are adjusted by the converter active power control link, so that each unit can inject or recover active power into the grid according to the command.

[0029] This embodiment collects data on the operating status of the cluster units and the grid frequency measurement data, extracts multi-dimensional features and clusters them, identifies the frequency response stage and adaptively adjusts the control parameters, combines support capability assessment to perform power allocation between and within the cluster, and supplements it with dynamic updates of cluster boundaries and superimposed execution of power commands. This achieves orderly and coordinated support for distributed wind power clusters under frequency disturbance scenarios, avoiding the waste of support capabilities and operational safety hazards caused by differences in the characteristics between units.

[0030] Please see Figure 2In some embodiments, based on the unit operation dataset, the operating condition characteristics, electrical coupling characteristics, and inertia reserve characteristics of each wind turbine are extracted and fused into a comprehensive unit feature vector, including: S201. Extract the ratio of active power output to rated power, the ratio of rotor speed to rated speed, the ratio of current wind speed to rated wind speed, and the pitch angle position of each wind turbine from the unit operation dataset to form the operating condition feature sub-vector. S202. Extract the equivalent electrical distance between the access point and the common connection point of each wind turbine, the equivalent impedance modulus of the access line, and the voltage level of the access point from the unit operation dataset to form an electrical coupling feature subvector. S203. Extract the rotational kinetic energy reserve corresponding to the current rotor speed of each wind turbine, the virtual rotational inertia reference value of the virtual synchronous control configuration, and the historical output depth of participating frequency support within the preset time window from the unit operation dataset to form an inertia reserve feature sub-vector. S204. The operating condition feature vector, electrical coupling feature vector, and inertia reserve feature vector are weighted and fused according to preset weight coefficients to generate the comprehensive feature vector of each wind turbine unit.

[0031] In step S201, the ratio of active power output to rated power represents the proportion of the unit's current output level relative to its rated capacity; the ratio of rotor speed to rated speed represents the relative position of the unit's speed; the ratio of current wind speed to rated wind speed represents the relative abundance of wind resources at the unit's location; and the blade pitch angle position reflects the current blade pitch angle value. These parameters are presented as relative quantities to form a characteristic sub-vector of operating conditions, eliminating dimensional differences between units of different capacity levels and allowing for comparison of the operating states of each unit on the same scale. Each component data is directly extracted from the corresponding measurement item in the unit's operating dataset, requiring no additional calculation.

[0032] In step S202, the equivalent electrical distance refers to the equivalent impedance value between the unit's connection point and the point of common coupling (PCC), considering line impedance, transformer impedance, and intermediate electrical equipment impedance. This is distinct from geographical distance and is used to quantify the tightness of the electrical connection between the unit and PCC. The smaller the equivalent electrical distance, the more directly the unit perceives frequency disturbances at PCC, and the shorter the response delay. The magnitude of the equivalent impedance of the connection line is taken as the magnitude of the product of the unit's impedance per unit length and the line length, reflecting the electrical loss level and voltage drop of the power transmission channel. The voltage level of the connection point characterizes the voltage level to which the unit is connected; different voltage levels correspond to different electrical coupling strengths and dispatch levels. These three components characterize the electrical coupling relationship between the unit and PCC from three aspects: path impedance, line parameters, and connection level. The data for each component are calculated from the unit's connection topology and line parameters.

[0033] In step S203, the rotational kinetic energy reserve is calculated from the unit's equivalent moment of inertia and the current rotor speed, reflecting the amount of rotational kinetic energy that the unit can release at the current speed. The virtual moment of inertia reference value is the virtual synchronous generator control parameter preset in the unit's converter control loop. The historical output depth is determined by the ratio of the maximum power change when the unit participates in frequency support within a preset time window to the rated capacity. The three components characterize the unit's inertia reserve from three dimensions: the current releaseable kinetic energy, the control configuration capability, and the historical participation level.

[0034] In step S204, the preset weighting coefficients are pre-set based on the degree of influence of operating condition characteristics, electrical coupling characteristics and inertia reserve characteristics on the frequency support characteristics of the unit. During weighted fusion, the three feature sub-vectors are multiplied by the corresponding preset weighting coefficients and then summed to generate the unit comprehensive feature vector, which serves as the unified input for subsequent clustering.

[0035] This embodiment maps the raw measurement parameters in the unit operation dataset to three dimensions: operating conditions, electrical coupling, and inertia reserve. It eliminates the differences in capacity and voltage levels between units in the form of relative quantities and realizes a unified quantitative representation of multi-dimensional information through weighted fusion. This provides an input basis that can comprehensively reflect the frequency support characteristics of each unit for subsequent clustering based on the unit's comprehensive feature vector.

[0036] In some embodiments, an adaptive density clustering algorithm is used to cluster the integrated feature vector of the units, generating initial subgroup partitioning results and subgroup feature center vectors, including: Calculate the Euclidean distance matrix between the comprehensive feature vectors of each wind turbine unit, and automatically determine the initial value of the neighborhood radius based on the statistical distribution characteristics of the Euclidean distance matrix; The number of neighboring wind turbines within the neighborhood radius of each wind turbine is used as a density metric to identify high-density core turbine points in the feature space. For non-core unit points, they are assigned to the corresponding density reachable subgroups based on their reachability to each core unit point. For boundary unit points located in the boundary regions of multiple subgroups, obtain the electrical topology connectivity of the boundary unit points. If a boundary unit point has a direct line connection with a unit in a certain subgroup in terms of electrical topology, then the boundary unit point is assigned to that subgroup. The feature center vector of each subgroup is calculated based on the integrated feature vector of the units within each subgroup, and the subgroup affiliation identifier of each unit and the feature center vector of each subgroup are output as the initial subgroup division result.

[0037] In this embodiment, the dimension of the Euclidean distance matrix is ​​determined by the total number of wind turbine units. The element in the i-th row and j-th column of the matrix represents the Euclidean distance between the comprehensive feature vectors of the i-th and j-th turbine units. The smaller this distance value, the closer the comprehensive characteristics of the two units are in terms of operating conditions, electrical coupling, and inertia reserve. The initial value of the neighborhood radius is automatically determined based on the statistical distribution characteristics of all off-diagonal elements in the Euclidean distance matrix. It can be automatically generated by multiplying statistical quantities such as the mean, median, or quantile of the distance by a preset scaling factor, without the need for manual specification. This allows the neighborhood radius to adapt to the feature distribution scale of the current cluster, avoiding the impact on the accuracy of subsequent density determination due to differences in turbine characteristics causing the neighborhood radius to be too large or too small.

[0038] Core unit locations are identified using the number of neighboring units within their neighborhood radius as a density metric. When the number of neighboring units reaches a preset density threshold, the unit is marked as a high-density core unit. The reachability distance between non-core units and core units is determined based on the Euclidean distance between the unit's comprehensive feature vectors. Non-core units are assigned to the subgroup containing the core units with the smallest reachability distance that also meet the density reachability criteria.

[0039] Boundary unit points refer to non-core unit points that simultaneously fall within the density coverage of multiple subgroups. The subgroup feature center vector is obtained by averaging the comprehensive feature vectors of all units within each subgroup. The subgroup affiliation identifier is recorded in the form of a correspondence between unit number and subgroup number. Together, they constitute the initial subgroup partitioning result.

[0040] This embodiment adaptively determines the neighborhood radius by utilizing the statistical properties of the Euclidean distance matrix, avoiding the blindness of manually setting clustering parameters; it identifies core unit points by using density thresholds and sequentially assigns non-core points, achieving automatic clustering without needing to know the number of subgroups in advance; and it performs secondary determination on boundary unit points by using electrical topology connectivity, ensuring that the clustering results simultaneously consider feature similarity and topological coupling, providing a reasonable grouping framework for subsequent differentiated control.

[0041] In some embodiments, based on the frequency feature sequence of the common connection point, the frequency deviation and the rate of change of frequency are extracted, input into the stage identification state machine for frequency response stage discrimination, and a frequency response stage identification signal is generated, including: The frequency characteristic sequence of the common junction point is subjected to central difference operation to obtain the frequency change rate estimate, and the frequency change rate estimate is smoothed by low-pass filtering to generate the frequency change rate. The frequency deviation is generated by subtracting the real-time frequency value from the rated frequency value in the frequency characteristic sequence of the common junction point; The frequency deviation and frequency change rate are input into the stage identification state machine, which has preset inertial response start conditions, damping coordination start conditions, power recovery start conditions and recovery end conditions. When the absolute value of the rate of change of frequency is greater than the inertial response start threshold and the frequency deviation is in a continuous increasing trend, the stage identification state machine outputs the inertial response stage identifier. When the absolute value of the frequency change rate decreases from greater than the damping coordination start threshold to less than the damping coordination start threshold and the frequency deviation reaches a local extreme value, the stage identification state machine outputs a damping coordination stage identifier. When the absolute value of the frequency deviation begins to decrease continuously and the sign of the frequency change rate is opposite to the sign of the frequency deviation, the stage identification state machine outputs a power recovery stage indicator. When the absolute value of the frequency deviation is less than the steady-state deviation tolerance and the absolute value of the frequency change rate is less than the steady-state change rate tolerance, the stage identification state machine outputs a recovery end flag. Among them, when the stage identification state machine outputs a stage identifier switch, it confirms the switch only after the corresponding switching conditions are met for multiple consecutive control cycles, and generates a frequency response stage identifier signal.

[0042] In this embodiment, the central difference operation uses the difference between the frequency measurements of one sampling point before and after the current sampling time in the frequency characteristic sequence of the point of common coupling (PCC) divided by twice the sampling period to obtain the estimated frequency change rate at the current time. Compared with one-sided difference, the central difference operation responds more closely to the actual frequency change trend. Low-pass filtering and smoothing processes use first-order inertial filtering or moving average filtering to attenuate the high-frequency noise components in the estimated frequency change rate, generating the frequency change rate. The real-time frequency value is taken from the frequency measurement at the current sampling time in the PCC frequency characteristic sequence, and the rated frequency is the nominal frequency of the power grid. The sign of the frequency deviation obtained by subtracting the two values ​​indicates whether the current frequency is higher or lower than the rated value.

[0043] The phase identification state machine judges state transitions based on the dynamic trajectory of frequency deviation and frequency change rate. The inertial response initiation condition, damping coordination initiation condition, power recovery initiation condition, and recovery termination condition correspond to the typical characteristics of the initial stage of frequency drop, frequency approaching an extreme point, frequency starting to rise, and frequency returning to stability, respectively. Specifically, in the inertial response initiation condition, the absolute value of the frequency change rate is greater than the inertial response initiation threshold, and the frequency deviation is continuously increasing, indicating that the system is in the initial stage of rapid frequency deviation from the rated value. In the damping coordination initiation condition, the absolute value of the frequency change rate decreases from greater than the damping coordination initiation threshold to less than the damping coordination initiation threshold, and the frequency deviation reaches a local extreme, indicating that the frequency deviation has reached an inflection point. In the power recovery initiation condition, the absolute value of the frequency deviation begins to continuously decrease, and the sign of the frequency change rate is opposite to the sign of the frequency deviation, indicating that the frequency has passed the extreme point and is returning to the rated value. In the recovery termination condition, the absolute value of the frequency deviation is less than the steady-state deviation tolerance, and the absolute value of the frequency change rate is less than the steady-state change rate tolerance, indicating that the frequency has entered the steady-state range. The inertial response start threshold, damping coordination start threshold, steady-state deviation tolerance, and steady-state rate of change tolerance are pre-set according to the system frequency safety operation requirements and the accuracy level of the frequency measurement device.

[0044] To prevent the frequency response stage identifier from switching frequently near the threshold boundary due to measurement noise interference, the stage identification state machine only confirms the switch after the corresponding switching conditions are met for several consecutive control cycles when the output stage identifier is switched. The control cycle is synchronized with the sampling cycle of the frequency measurement data at the common connection point. After the switch is confirmed, the frequency response stage identifier signal is updated to the identifier value corresponding to the current stage.

[0045] This embodiment extracts the rate of change of frequency from the frequency feature sequence of the common connection point by combining center differential and low-pass filtering, which takes into account both response speed and noise resistance. The frequency dynamic process is segmented and judged by four preset conditions, so that the frequency response stage indicator signal can accurately reflect the dynamic stage of the system frequency. The hysteresis mechanism that confirms the switching for multiple consecutive control cycles effectively avoids frequent chattering of the stage indicator near the boundary, and provides a stable and reliable timing reference for subsequent adaptive control based on the frequency response stage indicator signal.

[0046] In some embodiments, based on the frequency response stage identifier signal and the subgroup characteristic center vector, the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are calculated to generate adaptive control parameters, including: Based on the inertia reserve characteristic component in the characteristic center vector of each subgroup, the reference virtual inertia value of each subgroup is determined, and based on the electrical coupling characteristic component in the characteristic center vector of each subgroup, the reference damping coefficient value of each subgroup is determined. When the frequency response stage identifier signal is the inertia response stage identifier, the virtual inertia adjustment coefficient of each subgroup is determined according to the absolute value of the frequency change rate. The virtual inertia adjustment coefficient increases with the increase of the absolute value of the frequency change rate. The virtual inertia adjustment coefficient is corrected according to the operating condition characteristic components in the characteristic center vector of each subgroup. At the same time, the damping adjustment coefficient of each subgroup is set to a level lower than the reference damping coefficient value. When the frequency response stage identifier signal is the damping coordination stage identifier, the damping adjustment coefficient of each subgroup is determined according to the degree to which the frequency deviation is close to the local extremum. The damping adjustment coefficient increases as the degree to which the frequency deviation is close to the local extremum increases. The damping adjustment coefficient is corrected according to the electrical coupling characteristic component in the characteristic center vector of each subgroup. At the same time, the virtual inertia adjustment coefficient of each subgroup is reduced to a level below the reference virtual inertia value. When the frequency response stage indicator signal is the power recovery stage indicator, the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are smoothly adjusted according to the frequency recovery rate. The faster the frequency recovery rate, the faster the virtual inertia adjustment coefficient and damping adjustment coefficient decrease. The rotor speed of the units in each subgroup is monitored. If the rotor speed of the units in the subgroup that exceeds the preset proportion approaches the safety lower limit threshold, the virtual inertia adjustment coefficient of that subgroup is forcibly reduced. The virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are output as adaptive control parameters.

[0047] In this embodiment, the reference virtual inertia value is obtained by mapping the inertia reserve characteristic component in the characteristic center vector of each subgroup. Specifically, it can be proportionally calculated based on the rotational kinetic energy reserve in the inertia reserve characteristic component, and the weighted contribution of the virtual rotational inertia reference value is superimposed. The reference damping coefficient value is obtained by mapping the electrical coupling characteristic component, and can be proportionally calculated based on the reciprocal of the equivalent electrical distance, and the correction of the equivalent impedance modulus of the access line is superimposed. The reference value is determined once after grouping and remains unchanged during this frequency response process.

[0048] When the frequency response stage identifier signal is the inertia response stage identifier, the virtual inertia adjustment coefficient is determined as follows: The absolute value of the frequency change rate is used as input. A base value is obtained by multiplying by a preset proportional coefficient or by looking up a table using a piecewise linear function. Then, the correction direction and magnitude are determined based on the characteristic components of the operating condition. The correction direction depends on the range of output level and speed position. Subgroups with higher output levels and speeds within the safe range receive positive correction, while subgroups with limited output levels or speeds close to the safe boundary receive negative correction. The damping adjustment coefficient is calculated by multiplying the base damping coefficient value by a reduction factor less than 1. This reduction factor decreases as the absolute value of the frequency change rate increases.

[0049] When the frequency response stage identifier signal is the damping coordination stage identifier, the damping adjustment coefficient is determined as follows: Calculate the absolute value of the difference between the frequency deviation and its historical observed extreme values. The smaller the absolute value of this difference, the higher the damping adjustment coefficient. This mapping relationship can be described using an inverse proportional function or a piecewise linear function. Then, determine the correction direction based on the electrical coupling characteristic components. Subgroups with closer equivalent electrical distances receive positive corrections. The virtual inertia adjustment coefficient is calculated by multiplying the reference virtual inertia value by a reduction factor less than 1. The reduction factor decreases as the frequency deviation approaches a local extreme value.

[0050] When the frequency response phase identifier signal is the power recovery phase identifier, the frequency recovery rate is calculated by dividing the absolute value of the frequency deviation between two adjacent control cycles by the control cycle duration. The virtual inertia adjustment coefficient and damping adjustment coefficient continuously decrease as the frequency recovery rate increases, with the adjustment step size constrained by a preset maximum rate of change limit. If the rotor speed of a unit exceeding a preset proportion within a subgroup approaches the lower safety threshold, the virtual inertia adjustment coefficient of that subgroup is directly and forcibly reduced to a preset safety level. This operation takes precedence over the frequency recovery rate adjustment logic.

[0051] This embodiment establishes differentiated starting points for control parameters of each subgroup by using a benchmark value, drives the phased switching of the adjustment coefficient by using a frequency response phase identifier signal, and makes the adjustment coefficient fit the actual operating state and electrical position of each subgroup by correcting the operating condition characteristic components and electrical coupling characteristic components. It also ensures the safe operation of the unit under extreme conditions through a safety protection mechanism, thus realizing adaptive and coordinated control of the distributed wind power cluster in the multi-stage frequency response process.

[0052] In some embodiments, the total frequency support power requirement of the system is calculated based on the frequency characteristic sequence of the point of common coupling, and a comprehensive evaluation index of support capability is generated, including: Multiply the frequency deviation by a preset droop coefficient to generate a frequency deviation term, and multiply the frequency change rate by a preset virtual inertia coefficient to generate a frequency change rate term. Determine the weights of the frequency deviation term and the frequency change rate term based on the frequency response stage identifier signal. Increase the weight of the frequency change rate term when in the inertia response stage, and increase the weight of the frequency deviation term when in the damping coordination stage or power recovery stage. Sum the weighted frequency deviation term and the frequency change rate term to generate the total frequency support power requirement of the system. Based on the inertia reserve characteristic component in the characteristic center vector of each subgroup, the available inertia reserve assessment value of each subgroup is generated; based on the electrical coupling characteristic component in the characteristic center vector of each subgroup, the response agility assessment value of each subgroup is generated; based on the safety margin between the rotor speed and the lower speed limit protection value of the units within each subgroup, the state health assessment value of each subgroup is generated. The available inertia reserve assessment value, response agility assessment value, and state health assessment value are weighted and summed according to preset weight coefficients to generate a comprehensive assessment index of the support capability of each subgroup.

[0053] In this embodiment, the preset droop coefficient is adjusted based on the static correspondence between system frequency deviation and power deficit, and the preset virtual inertia coefficient is adjusted based on the dynamic correspondence between system equivalent inertia level and frequency change rate. Preferably, the weight allocation of the frequency deviation term and the frequency change rate term is implemented through a preset weight mapping table. This mapping table uses the frequency response stage identifier signal as an index and pre-stores the weight values ​​corresponding to each stage. The relative magnitude of the weight values ​​is set as follows: in the inertia response stage, the weight of the frequency change rate term is greater than the weight of the frequency deviation term; in the damping coordination stage and the power recovery stage, the weight of the frequency deviation term is greater than the weight of the frequency change rate term; and the sum of the weight values ​​is always 1. The sum of the weighted frequency deviation term and the frequency change rate term is the total frequency support power requirement of the system.

[0054] The available inertia reserve assessment value is obtained by summing the rotational kinetic energy reserve in the inertia reserve characteristic component according to the units within the subgroup. The response agility assessment value is obtained by taking the reciprocal of the equivalent electrical distance in the electrical coupling characteristic component and averaging it among the units within the subgroup. The condition health assessment value is obtained by averaging the difference between the rotor speed and the lower speed limit protection value of each unit in the subgroup divided by the rated speed range. After normalization, the three assessment values ​​are weighted and summed according to preset weighting coefficients to generate a comprehensive assessment index of the support capacity of each subgroup. The preset weighting coefficients are pre-tuned based on the contribution of the three assessment values ​​to the frequency support effect.

[0055] This embodiment uses a weighted mapping table to achieve phased adaptive switching of the weights of the frequency deviation term and the frequency change rate term, so that the total frequency support power demand of the system has different dominant components in different frequency response stages. The support potential of each subgroup is quantified from three dimensions: inertia reserve, response speed and operational safety through three evaluation values. After normalization and weighted summation, a unified comprehensive evaluation index of support capability is formed, which provides a quantitative basis for subsequent inter-group power allocation.

[0056] In some embodiments, the total system frequency support power requirement is allocated to each subgroup based on a comprehensive support capability evaluation index, and subgroup power support instructions are generated, including: The comprehensive evaluation index of the support capacity of each subgroup is normalized to generate the initial power allocation weight of each subgroup. Based on the inertia reserve characteristic component and the operating condition characteristic component in the characteristic center vector of each subgroup, the maximum supportable power limit of each subgroup is determined. The initial power allocation weight of each subgroup is multiplied by the total frequency support power requirement of the system to generate the initial power support command for each subgroup. Determine whether the initial power support command of each subgroup exceeds the corresponding maximum support power limit. If it does, redistribute the power difference exceeding the limit according to the proportion of the remaining support capacity of the other subgroups to generate a corrected subgroup power support command. The modified subgroup power support commands for each subgroup are output as subgroup power support commands.

[0057] In this embodiment, the normalization process is performed by dividing the comprehensive evaluation index of the support capability of each subgroup by the sum of the comprehensive evaluation indices of the support capability of all subgroups, so that the sum of the initial power allocation weights of each subgroup is 1. The maximum supportable power limit is obtained by converting the rotational kinetic energy reserve in the inertia reserve characteristic component into a power value according to a preset energy conversion efficiency, and then adding it to the available adjustment space corresponding to the ratio of active power in the operating condition characteristic component.

[0058] The remaining support capacity of the remaining subgroups is the difference between the maximum support power limit of each subgroup and its initial power support command. The redistribution process can be executed iteratively until the power support commands of all subgroups do not exceed the corresponding maximum support power limit. Finally, the power support commands actually executed by each subgroup are output as the subgroup power support commands.

[0059] This embodiment transforms the support capability assessment results into weighted coefficients that can be directly used for power allocation through normalization. It constrains the power bearing boundaries of each subgroup by limiting the maximum support power upper limit and ensures that the total frequency support power demand of the system is fully allocated under the premise of meeting the safety constraints of each subgroup through the proportional redistribution mechanism of the excess part, thus realizing the rationality and feasibility of inter-group power allocation.

[0060] In some embodiments, based on the rotor speed safety margin of each unit and the wind speed range division results, the subgroup power support command is differentially allocated to each wind turbine within the subgroup to generate unit power support commands, including: Obtain the current rotor speed and the lower limit protection value of each wind turbine in the subgroup, and calculate the rotor speed safety margin of each wind turbine. The rotor speed safety margin is the ratio of the difference between the current rotor speed and the lower limit protection value to the rated speed range. Based on the current wind speed of each wind turbine, the wind turbines are divided into low wind speed zone units, medium wind speed zone units, and high wind speed zone units. When the subgroup power support command is an increased power support command, a power support upper limit constraint is set for the low wind speed zone units, and the subgroup power support command is allocated to each unit according to the rotor speed safety margin ratio of the medium wind speed zone units and the high wind speed zone units, generating an increased power unit support command. When the subgroup power support command is a power reduction recovery command, the support power of each unit is reduced in the following order: low wind speed zone units first, medium wind speed zone units second, and high wind speed zone units last, to generate a power reduction unit recovery command. The rotor speed of each wind turbine is monitored in real time. If the rotor speed of a certain turbine drops to the warning threshold that is close to the lower speed limit protection value, the support power limiting protection of that turbine is triggered, and the support power share originally undertaken by that turbine is transferred to other turbines in the same subgroup with sufficient rotor speed safety margin, and the corresponding turbine power support command is updated.

[0061] In this embodiment, the rated speed range is the difference between the upper limit protection value and the lower limit protection value of the rotor speed. The current rotor speed and the lower limit protection value of the rotor speed are both obtained from the unit operation data. The lower limit protection value of the rotor speed is preset by the wind turbine manufacturer according to the mechanical structure and electrical constraints of the unit.

[0062] Wind speed zones are defined by the cut-in wind speed and the rated wind speed. The rated wind speed is determined based on the aerodynamic characteristics of the blades, the rated capacity of the generator, and the load-bearing capacity of the transmission system. It serves as the minimum wind speed required for the unit to achieve its rated power output. This value is stored in the factory configuration parameters of the unit's control system and can be directly retrieved from the unit's operating data set. Units with a current wind speed lower than the cut-in wind speed are classified as low-wind-speed units; units with a current wind speed between the cut-in wind speed and the rated wind speed are classified as medium-wind-speed units; and units with a current wind speed higher than the rated wind speed are classified as high-wind-speed units. The current wind speed data is extracted from the unit's operating data set.

[0063] When the subgroup power support command is an increase power support command, the upper limit of power support for units in the low wind speed zone is set according to a preset proportion of their rated power. Units in the medium wind speed zone and the high wind speed zone participate in the allocation of remaining power. The allocation proportion for each unit is its rotor speed safety margin divided by the sum of the rotor speed safety margins of the units participating in the allocation. When the subgroup power support command is a decrease power recovery command, the power reduction order of each unit in the same wind speed range is arranged in ascending order of rotor speed safety margin. The magnitude of each power reduction is a preset proportion of the current support power, until the support power of the unit is reduced to zero.

[0064] The warning threshold is set as the rotor speed lower limit protection value plus a preset safety margin. After the support power limiting protection is triggered, the power support command of the limited unit is frozen at the current value, and its original power support share is allocated to other units in the same subgroup with sufficient rotor speed safety margin according to the rotor speed safety margin ratio of the receiving unit.

[0065] This embodiment establishes a graded evaluation system for unit-level support capabilities by quantitatively calculating the rotor speed safety margin and dividing wind speed ranges. Through a combination strategy of limiting power output in low wind speed zones and prioritizing high-margin units in power-increasing scenarios, and a recovery order that prioritizes the withdrawal of low-margin units in power-reducing scenarios, it achieves a safe and differentiated allocation of subgroup power support commands among units within the group. Through support power limiting protection and power share transfer mechanisms, it ensures that individual units can withdraw support in a timely manner when approaching the speed safety boundary without affecting the overall power support capability of the subgroup.

[0066] In some embodiments, the evolution of the operating status of each subgroup and the frequency response phase switching event are monitored. When the subgroup boundary update condition is triggered, the initial subgroup division result is incrementally updated based on the latest unit operating status data to generate an updated subgroup division result, including: Monitor the switching events of the frequency response phase identifier signal. When the frequency response phase identifier signal switches phases, trigger the cluster boundary assessment. Monitor the rotor speed status of units in each subgroup and the comprehensive evaluation index of the support capacity of each subgroup. When the rotor speed of units in any subgroup exceeding a preset proportion approaches the safety lower limit or the comprehensive evaluation index of the support capacity of the subgroup decreases by more than a preset threshold, the subgroup boundary assessment is triggered. When the cluster boundary assessment is triggered, the latest operating status data of each wind turbine is obtained, the latest comprehensive feature vector of each wind turbine is calculated, and wind turbines whose changes exceed a preset threshold compared with the corresponding historical comprehensive feature vector are marked as turbines with significant status changes. Only for units whose status has changed significantly, their latest integrated feature vector is recalculated and adaptive density clustering is re-executed to determine their subgroup affiliation. Units whose status has not changed significantly retain their original subgroup affiliation, generating updated subgroup division results. For units whose subgroup affiliation changes, the power support command weighted by the original subgroup and the new subgroup is executed within the current control cycle. The weighting coefficient gradually transitions to the power support command value of the new subgroup as the control cycle progresses until the power support command of the new subgroup is fully switched.

[0067] In this embodiment, the switching event of the frequency response stage identifier signal is monitored by comparing the values ​​of the frequency response stage identifier signal in adjacent control cycles. A difference in values ​​between adjacent control cycles indicates a stage switching has occurred. The method for determining whether the rotor speed is approaching the safety lower limit is as follows: if the difference between the rotor speed and the safety lower limit is less than a preset margin, it is considered approaching. The calculation method for the decrease in the comprehensive support capability evaluation index is: the difference between the comprehensive support capability evaluation index of the current control cycle and the comprehensive support capability evaluation index at the initial clustering time, divided by the comprehensive support capability evaluation index at the initial clustering time. The above monitoring is continuously executed according to a preset cycle.

[0068] When the cluster boundary assessment is triggered, the latest integrated characteristic vector of each wind turbine is recalculated based on the latest operating status data. The change magnitude is calculated as the Euclidean distance between the latest integrated characteristic vector and the historical integrated characteristic vector, which is the characteristic vector calculated and stored during the initial subgroup division. The change magnitude threshold is preset based on the dimensions and numerical distribution range of the integrated characteristic vector. The reassignment determination for turbines with significant status changes is based on the distance between their latest integrated characteristic vector and the characteristic center vector of each subgroup. The determination rule is consistent with the density attainment rule used during the initial subgroup division. The updated subgroup division results require recalculation of the characteristic center vector of each subgroup.

[0069] For units whose subgroup affiliation changes, the weighted mixed value is obtained by multiplying the power support command of the original subgroup by the original affiliation weight and adding the power support command of the new subgroup by the new affiliation weight. The original affiliation weight decreases from an initial value of 1 each control cycle, and the new affiliation weight increases from 0 each control cycle. The sum of the two is always 1, and the number of transition steps is preset to a fixed value.

[0070] This embodiment ensures that the grouping boundary update conditions can reflect the dynamic characteristics of the system frequency and the evolution of the unit operating status in a timely manner by continuously monitoring the phase switching events and subgroup state anomalies; by re-executing the affiliation determination only for units with significant state changes, it avoids the computational overhead and control oscillations caused by full regrouping; and by using a weighted mixed transition of power support commands, it eliminates the risk of sudden power command changes at the moment of subgroup affiliation switching, ensuring the continuity and smoothness of the frequency support process.

[0071] In some embodiments, the unit power support command is superimposed onto the baseline active power reference value of each wind turbine to generate a total active power reference value for the unit, which is then executed through the converter active power control loop, including: The unit power support command of each wind turbine is summed with the corresponding basic active power reference value to generate the total active power reference value of each wind turbine. Apply a power change rate limiting constraint to the total active power reference value of the unit to generate a limited total active power reference value; The total reference value of the limited active power is input to the active power control loop of the converter of each wind turbine. The active power control loop of the converter adjusts the switching state of the converter through the inner loop current control and pulse width modulation strategy to control the active power injected into the grid by each wind turbine. The reactive power control of each wind turbine unit operates independently and does not change its reactive power control mode with the adjustment of active power.

[0072] In this embodiment, the power change rate limiting constraint is achieved through a preset maximum power change rate limiting value, which is pre-set based on the thermal capacity of the converter power devices and the mechanical stress bearing capacity of the unit shaft system. The limited total active power reference value serves as the input command for the converter active power control loop. The difference between this value and the current actual output power is adjusted by proportional-integral regulation of the inner loop current control to generate an inner loop current reference value. This value is then converted into a switching drive signal for the converter power devices via pulse width modulation, ensuring that the actual output power of the unit tracks the total active power reference value of the unit at a rate that does not exceed the limiting value.

[0073] This embodiment avoids the impact of sudden power command changes on the converter and generator shaft system by limiting the power change rate. It achieves accurate tracking and execution of active power commands by adopting inner loop current control and pulse width modulation strategy, and ensures stable voltage maintenance at each access point during frequency support by operating reactive power control independently.

[0074] By adopting the above technical solutions, this invention differs from existing technologies and possesses the following beneficial effects: By collecting operating status data of each wind turbine in a distributed wind power cluster and frequency measurement data from the grid's common connection point, operating condition characteristics, electrical coupling characteristics, and inertia reserve characteristics are extracted and fused into a comprehensive feature vector for the turbine. An adaptive density clustering algorithm is then used for grouping, causing turbines with similar frequency response characteristics to be clustered into the same subgroup. Based on the frequency feature sequence of the common connection point, frequency deviation and frequency change rate are extracted. A stage identification state machine is used to determine the frequency response stage, and based on this, the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are calculated, achieving inertia support and oscillation suppression. Dynamic coordination at different stages; the total frequency support power demand of the system is allocated to each subgroup through a comprehensive evaluation index of support capacity, and then differentiatedly allocated to each wind turbine based on the rotor speed safety margin and wind speed range division results, taking into account both the tapping of support potential and the safety of unit operation; when the group boundary update condition is triggered, the initial subgroup division results are incrementally updated so that the group affiliation adapts to the evolution of the operating state, and finally the unit power support command is superimposed on the basic active power reference value and executed through the converter active power control link, realizing the full release of the frequency support potential of the distributed wind power cluster and the orderly coordination among multiple subgroups, and improving the frequency stability level of the regional power grid under the high proportion of new energy access.

[0075] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0076] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0077] The above description is only a part of the embodiments of the present invention and does not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made based on the content of the present invention specification and drawings, or direct or indirect application in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A distributed wind power cluster adaptive frequency support control method, characterized in that, include: Collect operating status data of each wind turbine in the distributed wind power cluster and frequency measurement data of the grid common connection point to generate a unit operation dataset and a frequency feature sequence of the common connection point; Based on the unit operation dataset, the operating condition characteristics, electrical coupling characteristics and inertia reserve characteristics of each wind turbine are extracted and fused into a comprehensive feature vector of the unit. An adaptive density clustering algorithm is used to cluster and divide the integrated feature vector of the unit, generating the initial subgroup division results and the feature center vector of the subgroup; Based on the frequency feature sequence of common connection point, the frequency deviation and frequency change rate are extracted, input into the stage identification state machine to determine the frequency response stage, and generate the frequency response stage identification signal. Based on the frequency response stage identifier signal and the subgroup feature center vector, calculate the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup, and generate adaptive control parameters. Calculate the total frequency support power requirement of the system based on the frequency characteristic sequence of the common connection point, generate a comprehensive evaluation index of support capability, allocate the total frequency support power requirement of the system to each subgroup according to the comprehensive evaluation index of support capability, and generate subgroup power support instructions. Based on the rotor speed safety margin and wind speed range division results of each unit, the subgroup power support command is differentiatedly allocated to each wind turbine unit within the subgroup to generate unit power support command; Monitor the evolution of the operating status of each subgroup and the frequency response stage switching event. When the subgroup boundary update condition is triggered, incrementally update the initial subgroup division result based on the latest unit operating status data to generate the updated subgroup division result. The power support command of the unit is superimposed on the basic active power reference value of each wind turbine to generate the total active power reference value of the unit, which is then executed by the active power control link of the converter.

2. The distributed wind power cluster adaptive frequency support control method according to claim 1, characterized in that, Based on the unit operation dataset, the operating condition characteristics, electrical coupling characteristics, and inertia reserve characteristics of each wind turbine are extracted and fused into a comprehensive feature vector of the unit, including: The ratio of active power output to rated power, the ratio of rotor speed to rated speed, the ratio of current wind speed to rated wind speed, and the pitch angle position of each wind turbine are extracted from the unit operation dataset to form an operating condition feature sub-vector. The equivalent electrical distance between each wind turbine's access point and the common connection point, the equivalent impedance modulus of the access line, and the voltage level of the access point are extracted from the unit operation dataset to form an electrical coupling feature sub-vector. The rotational kinetic energy reserve corresponding to the current rotor speed of each wind turbine, the virtual rotational inertia reference value of the virtual synchronous control configuration, and the historical output depth of participating in frequency support within the preset time window are extracted from the unit operation dataset to form an inertia reserve feature sub-vector. The operating condition feature vector, electrical coupling feature vector, and inertia reserve feature vector are weighted and fused according to preset weighting coefficients to generate a comprehensive feature vector for each wind turbine unit.

3. The distributed wind power cluster adaptive frequency support control method according to claim 1, characterized in that, An adaptive density clustering algorithm is used to cluster the integrated feature vectors of the units, generating initial subgroup partitioning results and subgroup feature center vectors, including: Calculate the Euclidean distance matrix between the comprehensive feature vectors of each wind turbine unit, and automatically determine the initial value of the neighborhood radius based on the statistical distribution characteristics of the Euclidean distance matrix; The number of neighboring wind turbine units within the neighborhood radius of each wind turbine unit is used as a density metric to identify high-density core turbine unit points in the feature space. For non-core unit points, based on their reachability distance to each core unit point, the non-core unit points are assigned to the corresponding density reachability subgroup. For a boundary unit point located in the boundary region of multiple subgroups, obtain the electrical topology connectivity of the boundary unit point. If the boundary unit point has a direct line connection with a unit in a certain subgroup in terms of electrical topology, then the boundary unit point is assigned to that subgroup. The feature center vector of each subgroup is calculated based on the integrated feature vector of the units within each subgroup, and the subgroup affiliation identifier of each unit and the feature center vector of each subgroup are output as the initial subgroup division result.

4. The distributed wind power cluster adaptive frequency support control method according to claim 1, characterized in that, Based on the frequency feature sequence of the common connection point, the frequency deviation and frequency change rate are extracted, input to the stage identification state machine for frequency response stage discrimination, and frequency response stage identification signals are generated, including: The frequency feature sequence of the common connection point is subjected to central difference operation to obtain the frequency change rate estimate, and the frequency change rate estimate is subjected to low-pass filtering smoothing to generate the frequency change rate. The frequency deviation is generated by subtracting the real-time frequency value from the rated frequency value in the frequency feature sequence of the common connection point. The frequency deviation and frequency change rate are input into the stage identification state machine, which is preset with inertial response start-up conditions, damping coordination start-up conditions, power recovery start-up conditions and recovery end conditions. When the absolute value of the frequency change rate is greater than the inertial response activation threshold and the frequency deviation is in a continuous increasing trend, the stage identification state machine outputs an inertial response stage identifier. When the absolute value of the frequency change rate decreases from greater than the damping coordination start threshold to less than the damping coordination start threshold and the frequency deviation reaches a local extreme value, the stage identification state machine outputs a damping coordination stage identifier. When the absolute value of the frequency deviation begins to decrease continuously and the sign of the frequency change rate is opposite to the sign of the frequency deviation, the stage identification state machine outputs a power recovery stage identifier. When the absolute value of the frequency deviation is less than the steady-state deviation tolerance and the absolute value of the frequency change rate is less than the steady-state change rate tolerance, the stage identification state machine outputs a recovery end flag. When the stage identification state machine outputs a stage identifier switch, it confirms the switch only after the corresponding switching conditions are met for multiple consecutive control cycles, and generates a frequency response stage identifier signal.

5. The distributed wind power cluster adaptive frequency support control method according to claim 1, characterized in that, Based on the frequency response stage identifier signal and the subgroup feature center vector, the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are calculated to generate adaptive control parameters, including: Based on the inertia reserve characteristic component in the characteristic center vector of each subgroup, the reference virtual inertia value of each subgroup is determined, and based on the electrical coupling characteristic component in the characteristic center vector of each subgroup, the reference damping coefficient value of each subgroup is determined. When the frequency response stage identifier signal is the inertia response stage identifier, the virtual inertia adjustment coefficient of each subgroup is determined according to the absolute value of the frequency change rate. The virtual inertia adjustment coefficient increases as the absolute value of the frequency change rate increases, and the virtual inertia adjustment coefficient is corrected according to the operating condition characteristic components in the characteristic center vector of each subgroup. At the same time, the damping adjustment coefficient of each subgroup is set to a level lower than the reference damping coefficient value. When the frequency response stage identifier signal is the damping coordination stage identifier, the damping adjustment coefficient of each subgroup is determined according to the degree to which the frequency deviation is close to the local extremum. The damping adjustment coefficient increases as the degree to which the frequency deviation is close to the local extremum increases. The damping adjustment coefficient is corrected according to the electrical coupling characteristic component in the characteristic center vector of each subgroup. At the same time, the virtual inertia adjustment coefficient of each subgroup is reduced to a level below the reference virtual inertia value. When the frequency response stage identifier signal is the power recovery stage identifier, the virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are smoothly adjusted according to the frequency recovery rate. The faster the frequency recovery rate, the faster the virtual inertia adjustment coefficient and damping adjustment coefficient decrease. The rotor speed of the units in each subgroup is monitored. If the rotor speed of the units in the subgroup that exceeds a preset proportion approaches the safety lower limit threshold, the virtual inertia adjustment coefficient of that subgroup is forcibly reduced. The virtual inertia adjustment coefficient and damping adjustment coefficient of each subgroup are output as adaptive control parameters.

6. The distributed wind power cluster adaptive frequency support control method according to claim 1, characterized in that, Based on the frequency characteristic sequence of the common connection point, the total frequency support power requirement of the system is calculated, and a comprehensive evaluation index of support capability is generated, including: The frequency deviation is multiplied by a preset droop coefficient to generate a frequency deviation term, and the frequency change rate is multiplied by a preset virtual inertia coefficient to generate a frequency change rate term. The weights of the frequency deviation term and the frequency change rate term are determined according to the frequency response stage identifier signal. When in the inertia response stage, the weight of the frequency change rate term is increased. When in the damping coordination stage or power recovery stage, the weight of the frequency deviation term is increased. The weighted frequency deviation term and the frequency change rate term are summed to generate the total frequency support power requirement of the system. Based on the inertia reserve characteristic component in the characteristic center vector of each subgroup, the available inertia reserve assessment value of each subgroup is generated; based on the electrical coupling characteristic component in the characteristic center vector of each subgroup, the response agility assessment value of each subgroup is generated; based on the safety margin between the rotor speed and the lower speed limit protection value of the units within each subgroup, the state health assessment value of each subgroup is generated. The available inertia reserve assessment value, response agility assessment value, and state health assessment value are weighted and summed according to preset weight coefficients to generate a comprehensive assessment index of the support capability of each subgroup.

7. The distributed wind power cluster adaptive frequency support control method according to claim 1, characterized in that, Based on the comprehensive evaluation index of the support capability, the total frequency support power requirement of the system is allocated to each subgroup, and subgroup power support instructions are generated, including: The comprehensive evaluation index of the support capacity of each subgroup is normalized to generate the initial power allocation weight of each subgroup. Based on the inertia reserve characteristic component and the operating condition characteristic component in the characteristic center vector of each subgroup, determine the maximum supportable power limit of each subgroup. The initial power allocation weight of each subgroup is multiplied by the total frequency support power requirement of the system to generate the initial power support command for each subgroup. Determine whether the initial power support command of each subgroup exceeds the corresponding maximum support power limit. If it does, redistribute the power difference exceeding the limit according to the proportion of the remaining support capacity of the other subgroups to generate a corrected subgroup power support command. The modified subgroup power support commands for each subgroup are output as subgroup power support commands.

8. The distributed wind power cluster adaptive frequency support control method according to claim 1, characterized in that, Based on the rotor speed safety margin and wind speed range division results of each unit, the subgroup power support command is differentiatedly allocated to each wind turbine within the subgroup, generating unit power support commands, including: Obtain the current rotor speed and the lower limit protection value of each wind turbine unit in the subgroup, and calculate the rotor speed safety margin of each wind turbine unit. The rotor speed safety margin is the ratio of the difference between the current rotor speed and the lower limit protection value to the rated speed range. Based on the current wind speed of each wind turbine, the wind turbines are divided into low wind speed zone units, medium wind speed zone units, and high wind speed zone units. When the subgroup power support command is a power increase support command, a power support upper limit constraint is set for the low wind speed zone unit, and the subgroup power support command is allocated to each unit according to the rotor speed safety margin ratio of the medium wind speed zone unit and the high wind speed zone unit, generating a power increase unit support command. When the subgroup power support command is a power reduction recovery command, the support power of each unit is reduced in the following order: low wind speed zone units first, medium wind speed zone units second, and high wind speed zone units last, to generate a power reduction unit recovery command. The rotor speed of each wind turbine is monitored in real time. If the rotor speed of a certain turbine drops to the warning threshold that is close to the lower speed limit protection value, the support power limiting protection of that turbine is triggered, and the support power share originally undertaken by that turbine is transferred to other turbines in the same subgroup with sufficient rotor speed safety margin, and the corresponding turbine power support command is updated.

9. The distributed wind power cluster adaptive frequency support control method according to claim 1, characterized in that, Monitor the evolution of operating status and frequency response phase switching events of each subgroup. When the subgroup boundary update condition is triggered, incrementally update the initial subgroup division result based on the latest unit operating status data to generate an updated subgroup division result, including: Monitor the switching events of the frequency response phase identifier signal, and trigger cluster boundary assessment when the frequency response phase identifier signal undergoes phase switching; Monitor the rotor speed status of units in each subgroup and the comprehensive evaluation index of the support capacity of each subgroup. When the rotor speed of units in any subgroup exceeding a preset proportion approaches the safety lower limit or the comprehensive evaluation index of the support capacity of the subgroup decreases by more than a preset threshold, the subgroup boundary assessment is triggered. When the cluster boundary assessment is triggered, the latest operating status data of each wind turbine is obtained, the latest comprehensive feature vector of each wind turbine is calculated, and wind turbines whose changes exceed a preset threshold compared with the corresponding historical comprehensive feature vector are marked as turbines with significant status changes. Only for units whose status has changed significantly, their latest integrated feature vector is recalculated and adaptive density clustering is re-executed to determine their subgroup affiliation. Units whose status has not changed significantly retain their original subgroup affiliation, and an updated subgroup division result is generated. For units whose subgroup affiliation changes, the power support command weighted by the original subgroup and the new subgroup is executed within the current control cycle. The weighting coefficient gradually transitions to the power support command value of the new subgroup as the control cycle progresses until the power support command of the new subgroup is fully switched.

10. The distributed wind power cluster adaptive frequency support control method according to claim 1, characterized in that, The power support command of the wind turbine is superimposed on the basic active power reference value of each wind turbine to generate a total active power reference value for the wind turbines, which is then executed by the active power control loop of the converter, including: The unit power support command of each wind turbine is summed with the corresponding basic active power reference value to generate the total active power reference value of each wind turbine. A power change rate limiting constraint is applied to the total active power reference value of the unit to generate a limited total active power reference value; The total reference value of the limited active power is input to the active power control loop of the converter of each wind turbine. The active power control loop of the converter adjusts the switching state of the converter through the inner loop current control and pulse width modulation strategy to control the active power injected into the grid by each wind turbine. The reactive power control of each wind turbine unit operates independently and does not change its reactive power control mode with the adjustment of active power.

Images