A cooperative response management method, system, device and medium for low voltage ride through of a wind power station

Abstract

The application relates to a cooperative response management method, system, device and medium for low-voltage ride-through of a wind power station. The method comprises: acquiring real-time operation state data of multiple wind turbine generators in the wind power station, extracting a voltage dynamic response feature sequence by performing feature extraction on a change time sequence of terminal voltage of each generator within a preset time window, performing distributed consistency synchronization and alignment on a voltage support action time sequence of each generator based on the sequence to generate a station-level cooperative time sequence matrix, analyzing a contribution degree feature of each generator to voltage recovery of a power grid based on the matrix to obtain a cooperative contribution degree index, inputting the cooperative contribution degree index into an efficiency evaluation model to quantitatively evaluate cooperative efficiency to obtain cooperative efficiency evaluation data. The method can solve the problems of asynchronous response time sequence of the generators, unbalanced distribution of reactive power support and lack of quantitative evaluation of cooperative efficiency, improve voltage recovery speed and stability, reduce oscillation amplitude, and realize accurate synchronous cooperation and balanced distribution of contributions of multiple generators.

Description

Technical Field

[0001] This invention belongs to the field of wind power technology, and in particular relates to a collaborative response management method, system, equipment and medium for low-voltage ride-through of wind farms. Background Technology

[0002] As wind power accounts for an increasingly larger share of the power system, the low-voltage ride-through (LVRT) capability of wind farms has become a key requirement for ensuring the safe and stable operation of the power grid. To achieve grid connection reliability for large-scale wind farms, modern wind farms generally require coordinated low-voltage response capabilities at the farm level. This means that multiple wind turbines can coordinate and provide reactive power support and voltage recovery assistance to the grid according to a unified time base and operating logic during grid voltage dips.

[0003] In traditional technologies, wind farms typically manage low-voltage ride-through processes using independent monitoring and control of each individual turbine. Each turbine independently determines whether a fault has occurred based on the magnitude of its own terminal voltage drop and autonomously executes reactive power compensation or active power restoration actions according to a preset control strategy. The farm monitoring system only collects and records the operating status of each turbine discretely, lacking unified alignment and collaborative analysis methods for the response timing of multiple turbines.

[0004] However, the aforementioned traditional methods have the following technical problems: Because the actual timing and depth of voltage dips experienced by each unit during voltage sag events differ, and the communication and clock synchronization accuracy between units is insufficient, site managers cannot accurately determine whether the timing of each unit's actions meets the grid dispatch requirements for the overall site response time. Simultaneously, existing monitoring systems struggle to quantify the actual contribution of different units during voltage recovery, making it impossible to accurately assess the coordinated response efficiency of site-level low-voltage ride-through, thus hindering the improvement of digital management at wind farms. Summary of the Invention

[0005] Therefore, it is necessary to provide a collaborative response management method, system, equipment, and medium for low-voltage ride-through of wind farms to address the aforementioned technical issues.

[0006] Firstly, this application provides a collaborative response management method for low-voltage ride-through at wind farms, including:

[0007] S1. Obtain real-time operating status data of multiple wind turbine units within the wind farm;

[0008] S2. Based on real-time operating status data, feature extraction is performed on the timing of the terminal voltage change of each wind turbine within a preset time window to obtain the voltage dynamic response feature sequence of the wind turbine.

[0009] S3. Based on the voltage dynamic response characteristic sequence, the voltage support action timing of each wind turbine is synchronized and aligned in a distributed manner to generate a station-level collaborative timing matrix.

[0010] S4. Based on the site-level collaborative time series matrix, analyze the contribution characteristics of different wind turbine units to grid voltage recovery, and obtain the collaborative contribution index of each wind turbine unit.

[0011] S5. Input the collaborative contribution index of each wind turbine into the performance evaluation model to quantitatively evaluate the overall low-voltage ride-through collaborative action of the wind farm and obtain collaborative performance evaluation data.

[0012] Secondly, this application also provides a collaborative response management system for low-voltage ride-through at wind farms, including:

[0013] The data acquisition and monitoring module is used to acquire real-time operating status data of multiple wind turbine units within the wind farm.

[0014] The unit dynamic feature extraction module is used to extract features of the terminal voltage change sequence of each wind turbine within a preset time window based on real-time operating status data, and obtain the voltage dynamic response feature sequence of the wind turbine.

[0015] The station-level timing coordination and alignment module is used to perform distributed consistency synchronization and alignment of the voltage support action timing of each wind turbine based on the voltage dynamic response characteristic sequence, and generate a station-level coordination timing matrix.

[0016] The unit collaboration contribution assessment module is used to analyze the contribution characteristics of different wind turbine units to grid voltage recovery based on the site-level collaboration time series matrix, and obtain the collaboration contribution index of each wind turbine unit.

[0017] The collaborative performance evaluation module is used to input the collaborative contribution index of each wind turbine into the performance evaluation model to quantify the performance of the overall low-voltage ride-through collaborative action of the wind farm and obtain collaborative performance evaluation data.

[0018] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0019] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect.

[0020] The aforementioned collaborative response management method, system, equipment, and medium for low-voltage ride-through at wind farms acquires real-time operating status data of multiple wind turbines within the wind farm. It extracts features from the timing of voltage changes at the turbine terminals within a preset time window to construct a voltage dynamic response feature sequence. Based on this feature sequence, it performs distributed consistency synchronization and alignment of the voltage support action timing of each turbine, generating a farm-level collaborative timing matrix. Subsequently, it analyzes the contribution characteristics of each turbine to grid voltage recovery based on the collaborative timing matrix and quantifies them into a collaborative contribution index. Finally, it inputs the collaborative contribution index into an efficiency evaluation model to perform a closed-loop quantitative evaluation of the overall low-voltage ride-through collaborative action efficiency of the wind farm. This improves the overall voltage recovery speed and support stability during low-voltage ride-through, reduces the oscillation amplitude during voltage recovery, and minimizes redundancy and conflicts in reactive power response among turbines. It enables precise synchronization and coordinated action of multiple turbine timings and balanced contribution distribution. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating a collaborative response management method for low-voltage ride-through at a wind farm, as described in one embodiment.

[0023] Figure 2 This is a schematic diagram of the structure of a collaborative response management system for low-voltage ride-through at a wind farm, as described in one embodiment.

[0024] Figure 3 This is a schematic diagram of a computer device in one embodiment. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0026] refer to Figure 1 The document presents a flowchart illustrating a collaborative response management method for low-voltage ride-through at wind farms, as provided in this application. The method includes the following steps:

[0027] S1. Obtain real-time operating status data of multiple wind turbine units within the wind farm.

[0028] For example, a data acquisition scheme combining a synchronous phasor measurement unit (TPMU) and a data acquisition and monitoring system can be adopted. The TPMU can be used to acquire high-precision voltage and current phasor data to meet the requirements for instantaneous dynamic capture of voltage sags; the data acquisition and monitoring system can simultaneously acquire unit operating parameters, balancing data volume and real-time performance. The types of data acquired must cover both electrical and mechanical quantities to ensure data comprehensiveness.

[0029] To ensure data validity, a three-level preprocessing process is required. The first level is data format standardization, which converts heterogeneous data from the synchronous phasor measurement unit and the data acquisition and monitoring system into a standard format to ensure data exchange compatibility. The second level is missing value imputation, which can be achieved using linear interpolation based on data from adjacent time points. For cases where consecutive missing data points do not exceed a set number, imputation values ​​are calculated using formulas.

[0030] The formula for linear interpolation can be:

[0031]

[0032] in, Indicates time The fill voltage value, Indicates time voltage value, Indicates time voltage value, Indicates time At that time The time interval, Indicates time At that time The time interval.

[0033] If more than a set number of consecutive data points are missing, the data for that period is marked as invalid and will be removed in subsequent analysis. The third level is outlier removal, based on the 3σ criterion. First, the mean and standard deviation of a certain parameter are calculated. Data that exceeds the set range is identified as outliers and replaced with the mean of that parameter.

[0034] Data integrity verification uses the following formula:

[0035]

[0036] in, This indicates that the data is efficient. This indicates the number of valid data points after preprocessing, including both the original valid data and the imputed valid data. This indicates the total number of data points collected. The set efficiency requirements must be met before proceeding to the next step. If these requirements are not met, re-collection or inspection of the relevant sensors is necessary to ensure that subsequent analysis is based on reliable data.

[0037] S2. Based on real-time operating status data, feature extraction is performed on the timing of the terminal voltage change of each wind turbine within a preset time window to obtain the voltage dynamic response feature sequence of the wind turbine.

[0038] Optionally, real-time operating status data of the wind turbine can be used to characterize its operating conditions and electrical characteristics. This data may include voltage amplitude, frequency, current, active power, and reactive power. The voltage dynamic response characteristic sequence can be extracted by analyzing the change of voltage amplitude over time, and its expression is:

[0039]

[0040] In the formula, Indicates at time The voltage dynamic response characteristic sequence; Indicates at time The Voltage amplitude at each sampling point This represents the total number of sampling points; This is a time variable used to identify the sampling time. Furthermore, to more accurately describe the dynamic trend of voltage change, the voltage change rate can be calculated, and its formula is as follows:

[0041]

[0042] in, Indicates the rate of change of voltage; and These represent two adjacent sampling times. The voltage amplitude; The sampling interval time reflects the temporal resolution of the sampling.

[0043] S3. Based on the voltage dynamic response characteristic sequence, the voltage support action timing of each wind turbine is synchronized and aligned in a distributed manner to generate a station-level collaborative timing matrix.

[0044] For example, based on the collected voltage dynamic response feature sequence, a station-level collaborative time-series matrix is ​​constructed to achieve collaborative control of various devices. The voltage dynamic response feature sequence can be used to describe the response characteristics of the wind farm during voltage fluctuations. This sequence includes the voltage amplitude change sequence and phase change sequence of each monitoring node. The voltage amplitude change sequence is as follows: ,in Indicates the monitoring node number, ( (Total number of monitoring nodes) Indicates the sampling time. ( (This represents the number of sampling points corresponding to the total sampling duration); the phase change sequence is... Station-level collaborative time series matrix The expression can be:

[0045]

[0046] In this matrix, each row corresponds to the voltage amplitude and phase data of all monitoring nodes at a given sampling time; each column corresponds to the voltage amplitude or phase data of a monitoring node at all sampling times. This collaborative time-series matrix comprehensively reflects the voltage dynamic response characteristics of each node in the wind farm at different times, providing a data foundation for the subsequent formulation of collaborative control strategies.

[0047] S4. Based on the site-level collaborative time series matrix, analyze the contribution characteristics of different wind turbine units to grid voltage recovery, and obtain the collaborative contribution index of each wind turbine unit.

[0048] For example, to analyze the contribution characteristics of wind turbines to grid voltage recovery and obtain a collaborative contribution index, quantitative calculations can be performed based on the operating parameters of the wind turbines and grid voltage change data. The active and reactive power output data of the wind turbines can be used to reflect their support capability for grid voltage recovery; this data may include real-time active power during fault periods. and real-time reactive power Collaborative contribution index The expression can be:

[0049]

[0050] in, and , which is a weighting coefficient used to adjust the proportion of active power contribution and reactive power contribution in the collaborative contribution index; This represents the average change in active power of a single wind turbine unit during a fault. It is the sum of the changes in active power of all wind turbines in the wind farm; This represents the average value of the reactive power change of a single wind turbine unit during a fault. This represents the sum of reactive power changes across all wind turbines within the wind farm. The collaborative contribution index calculated using this formula can comprehensively assess the contribution of each wind turbine during grid voltage recovery, providing a quantitative basis for formulating subsequent collaborative response management strategies.

[0051] S5. Input the collaborative contribution index of each wind turbine into the performance evaluation model to quantitatively evaluate the overall low-voltage ride-through collaborative action of the wind farm and obtain collaborative performance evaluation data.

[0052] For example, by constructing an effectiveness evaluation model, the effectiveness of low-voltage ride-through coordinated response management at wind farms can be quantitatively evaluated. The expression of the effectiveness evaluation model can be:

[0053]

[0054] in, This indicates the comprehensive performance evaluation results of the coordinated response management for low-voltage ride-through at wind farms; The number of indicators involved in the evaluation; For the first The weights of each evaluation indicator, and satisfying , This reflects the relative importance of each indicator in the evaluation system; For the first Standardized scores for each evaluation indicator It can be calculated by comparing the actual collected data with the set standard value, and is used to reflect the performance of the indicator in the evaluation.

[0055] Furthermore, to more accurately assess the synergistic effectiveness under different operating conditions, a correction factor can be introduced. The effectiveness assessment formula for the correction factor can be:

[0056]

[0057] In the formula, The revised overall performance evaluation results As a correction factor, its value is determined based on the operating conditions of the wind farm (such as the degree of grid voltage drop, fault duration, and wind turbine type), and is used to correct the comprehensive performance evaluation results to improve the accuracy and applicability of the evaluation.

[0058] In the aforementioned collaborative response management method for low-voltage ride-through at wind farms, real-time operating status data of multiple wind turbines within the wind farm are acquired. Features of the terminal voltage change timing of each turbine within a preset time window are extracted to construct a voltage dynamic response feature sequence. Based on this feature sequence, distributed consistency synchronization and alignment are performed among the turbines, unifying the originally asynchronous and discrete voltage support timing under a global reference timing and generating a farm-level collaborative timing matrix. This eliminates support cancellation and voltage oscillations caused by asynchronous responses. Furthermore, based on collaborative... The time-series matrix analysis identifies and quantifies the contribution characteristics of each generating unit to grid voltage recovery into a collaborative contribution index. This identifies and corrects the imbalance in reactive power support distribution among generating units, bringing the contributions closer to equilibrium. Finally, the collaborative contribution index is input into the performance evaluation model to conduct a closed-loop quantitative evaluation of the overall low-voltage ride-through collaborative performance of the wind farm. This improves the overall voltage recovery speed and support stability during low-voltage ride-through at wind farms, reduces the oscillation amplitude during voltage recovery, and minimizes redundancy and conflicts in reactive power response among generating units. It enables precise time-series synchronization and collaborative performance and balanced contribution distribution among multiple generating units.

[0059] In an optional embodiment, S2 includes:

[0060] S21. Based on real-time operating status data, extract the terminal voltage timing data of each wind turbine within a preset time window.

[0061] Optionally, the preset time window should cover the entire lifecycle of the voltage sag event, including the pre-sag steady-state phase, the sag drop phase, and the sag recovery phase, ensuring coverage of the entire process of voltage recovery to steady state. The window duration is determined based on the typical duration of the grid voltage sag and can be dynamically optimized through an adaptive window adjustment mechanism to avoid prematurely truncating recovery process data.

[0062] Based on the collected instantaneous three-phase voltage values ​​at the turbine terminals, the instantaneous positive-sequence voltage values ​​at the turbine terminals of each wind turbine can be calculated using the symmetrical component method, serving as the foundational data for time-series analysis. The symmetrical component method can decompose three-phase unbalanced voltage into positive-sequence, negative-sequence, and zero-sequence components, with the positive-sequence component best reflecting the normal operating state and voltage sag characteristics of the power grid. The calculation formula for the symmetrical component method is as follows:

[0063]

[0064] in, Indicates the first Taiwanese crew at all times The instantaneous value of the positive sequence voltage at the machine terminal. Indicates the first Taiwanese crew at all times of Instantaneous value of phase voltage Indicates the first Taiwanese crew at all times of Instantaneous value of phase voltage Indicates the first Taiwanese crew at all times of Instantaneous value of phase voltage Indicates the rotation factor. This represents the square of the rotation factor.

[0065] The timestamps of the timing data can be based on the GPS synchronization clock of the synchronized phasor measurement unit to ensure the consistency of the time base for all units. The positive sequence voltage timing data at the generator terminals includes elements such as the start time, data acquisition time interval, and number of data points. The data acquisition time interval is consistent with the sampling frequency of the synchronized phasor measurement unit, and the number of data points is calculated from the window duration and the data acquisition time interval.

[0066] S22. Perform time-domain noise reduction on the terminal voltage timing data to obtain smoothed voltage timing data for each wind turbine.

[0067] Optionally, since measurement noise exists in the acquired data, this noise can cause spikes in the voltage time-series curve, affecting the accuracy of subsequent feature extraction. Therefore, adaptive Kalman filtering is required for time-domain noise reduction. Compared to traditional Kalman filtering, adaptive Kalman filtering can adapt to the abrupt changes in dynamic characteristics during voltage sags by adjusting the process noise variance and observation noise variance in real time, thus avoiding signal distortion caused by over-filtering.

[0068] The state equation and observation equation for the filtering process are as follows:

[0069]

[0070]

[0071] in, Indicates the first The state vector at any given time can be taken as the instantaneous value of the terminal voltage. Indicates the first The state vector at time t, The state transition matrix is ​​defined based on the first-order dynamic characteristics of voltage changes. Indicates the first The process noise at time t, follows the mean of variance is The Gaussian distribution characterizes the random fluctuations of the voltage itself. Indicates the first The observed value at a given time, i.e., the collected terminal voltage data. Represents the observation matrix. Indicates the first The observation noise at any given time follows a mean of variance is The Gaussian distribution represents the error in the measurement process.

[0072] The adaptive adjustment mechanism is implemented as follows: the observation residual is calculated, the variance of the residual is calculated using the sliding window method, and the observation noise variance is adjusted according to the relationship between the variance and the preset threshold. At the same time, the process noise variance is dynamically updated to ensure that the process noise variance adapts to the dynamic changes of the voltage.

[0073] The formula for calculating the observation residual can be:

[0074]

[0075] in, Indicates the first The observation residual at time, Indicates the first The voltage estimate at time t.

[0076] The formula for calculating residual variance is as follows:

[0077]

[0078] in, The variance represents the residual. Indicates the length of the sliding window. Indicates the first Observe the square of the residual at all times.

[0079] The process noise variance update formula can be:

[0080]

[0081] in, Indicates the forgetting factor, Indicates the first Voltage estimate at time [time]. Indicates the first The voltage estimate at time t.

[0082] The final output of smoothed voltage timing data must have noise power reduced to below a set percentage of the original data; the noise power calculation formula can be:

[0083]

[0084] in, Indicates noise power. This indicates the number of data points within the time window. Indicates the first Taiwanese crew at all times The original voltage value, Indicates the first Taiwanese crew at all times The smoothed voltage value.

[0085] The formula for calculating the power of the raw data can be:

[0086]

[0087] in, This indicates the power of the raw data.

[0088] If the noise power requirement is not met, the initial parameters of the adaptive Kalman filter need to be adjusted and the filter re-filtered.

[0089] S23. Based on smoothed voltage time series data, calculate the instantaneous voltage drop depth and recovery rate of each wind turbine during a voltage sag event to obtain a preliminary dynamic feature set.

[0090] Optionally, the instantaneous voltage drop depth can be used to characterize the severity of the voltage sag, directly affecting the timing and intensity of the unit's reactive power support strategy. It is defined as the ratio of the instantaneous voltage drop to the rated voltage during the sag period.

[0091] The formula for calculating the instantaneous voltage drop depth can be:

[0092]

[0093] in, Indicates the first Taiwanese crew at all times The instantaneous voltage drop depth, This indicates the rated terminal voltage of the generator unit. Indicates the first Taiwanese crew at all times The smoothed voltage value.

[0094] When the instantaneous voltage drop depth reaches the set threshold, it is determined that the low voltage ride-through state has been entered, triggering subsequent reactive power support actions; when the instantaneous voltage drop depth is lower than the set threshold and continues for a set duration, it is determined that the voltage recovery is complete.

[0095] The voltage recovery rate characterizes how quickly the voltage recovers from its lowest point, reflecting the effectiveness of the unit's reactive power support. It is defined as the magnitude of voltage recovery per unit time and can be calculated using the sliding window method to avoid misjudgments caused by single-point abrupt changes. The expression for the voltage recovery rate can be:

[0096]

[0097] in, Indicates the first The voltage recovery rate of the generator unit at time t. Indicates the first Taiwanese crew at all times Smooth voltage value, Indicates the first Taiwanese crew at all times Smooth voltage value, Indicates the duration of the sliding window.

[0098] To avoid the influence of outliers, the calculated voltage recovery rate needs to be limited: if the voltage recovery rate exceeds the unit's maximum possible recovery rate, it is set to the unit's maximum possible recovery rate; if the voltage recovery rate is negative, i.e. the voltage drops instead of rising, it is set to 0. The unit's maximum possible recovery rate is determined based on the converter capacity, and the value of the sliding window duration needs to balance response sensitivity and stability.

[0099] The initial dynamic feature set is a two-dimensional feature vector set composed of the drop depth and recovery rate at each moment. This set needs to include the features of the entire voltage sag process, from the depth increase stage from the drop start point to the lowest point, the stabilization stage at the lowest point (if it exists), and the rate change stage of the recovery stage, to ensure that the subsequent feature fusion can fully reflect the dynamic response of the unit.

[0100] S24. Input the preliminary dynamic feature set into the feature fusion model, perform feature splicing and feature mapping on the preliminary dynamic feature set, and obtain the comprehensive dynamic response vector of each wind turbine.

[0101] Optionally, the feature fusion model can be implemented using a fully connected neural network to compress the temporal features of fall depth and recovery rate into a fixed-dimensional vector, facilitating subsequent temporal synchronization and contribution analysis. The structural design of the fully connected neural network needs to be tailored to practical engineering requirements. The number of neurons in the input layer should be related to the number of data points within the time window. Two hidden layers should be used to avoid overfitting. The number of neurons in the first and second hidden layers should be set according to a preset ratio. The number of neurons in the output layer should be determined based on engineering needs, ensuring sufficient feature information is retained while reducing subsequent computational load.

[0102] First, feature concatenation can be performed on the initial dynamic feature set, combining multiple two-dimensional feature vectors along the time dimension into a one-dimensional vector in chronological order. This operation preserves the temporal correlation of the features. Subsequently, feature mapping is performed through hidden layers. The activation function of the first hidden layer can be a rectified linear unit (RCU). The first hidden layer contains a weight matrix and bias vector from the input layer to the first hidden layer, effectively mitigating the vanishing gradient problem. The expression for the rectified linear unit activation function can be:

[0103]

[0104] in, This represents the input value of the activation function.

[0105] The second hidden layer can use the Leaky ReLU activation function. This second hidden layer contains the weight matrix and bias vector from the first hidden layer to the second hidden layer, avoiding the problem of the gradient being zero in the negative region for the corrected linear unit. The Leaky ReLU activation function expression can be:

[0106]

[0107] in, This represents the input value of the activation function.

[0108] The output layer outputs a composite dynamic response vector with a set dimension. A linear activation function can be used to avoid compressing the feature magnitude. The expression for the composite dynamic response vector can be:

[0109]

[0110] in, Indicates the first The overall dynamic response vector of the unit, This represents the weight matrix from the second hidden layer to the output layer. This represents the output value of the second hidden layer. This represents the output layer bias vector.

[0111] Weights and biases are determined through offline training. The training dataset uses historical low-voltage ride-through event data from wind farms, including different scenarios, and the sample size meets the specified requirements. The training objective is to minimize the mean squared error between the integrated dynamic response vector and the manually labeled feature vector. The expression for the mean squared error can be:

[0112]

[0113] in, Indicates mean square error. Indicates the number of samples. Indicates the first The comprehensive dynamic response vector of each sample. Indicates the first The manually labeled feature vector of each sample. This represents the Euclidean norm.

[0114] The number of training iterations is executed within a set range until the mean square error converges. The convergence criterion is that the change is lower than a set threshold.

[0115] S25. Arrange the comprehensive dynamic response vectors in chronological order to obtain the voltage dynamic response characteristic sequence of each wind turbine.

[0116] Optionally, the comprehensive dynamic response vectors are arranged according to their corresponding timestamps to generate a voltage dynamic response feature sequence. Each element in this sequence is the comprehensive dynamic response vector at the corresponding time point, and the entire sequence includes both time and feature dimensions. The voltage dynamic response feature sequence retains the temporal characteristics of voltage dynamic changes while compressing the data dimension through feature fusion and eliminating redundant information, providing efficient input for subsequent time-series synchronization. If the original two-dimensional feature sequence is used directly for synchronization, the low feature dimension will lead to inaccurate identification of key turning points. The comprehensive feature vector with defined dimensions contains richer dynamic information, improving synchronization accuracy.

[0117] To ensure the validity of the sequence, its temporal consistency needs to be verified by calculating the cosine similarity of feature vectors at adjacent time points. If the cosine similarity is lower than a set threshold for a set number of consecutive time points, it indicates that feature mutations are too frequent, which may be due to problems in the filtering or feature fusion process. In this case, it is necessary to return to the previous denoising steps and redo the denoising, or adjust the number of hidden layer neurons in the fully connected neural network and retrain. The expression for cosine similarity can be:

[0118]

[0119] in, Indicates the first Taiwanese crew at all times With time Cosine similarity of eigenvectors Indicates the first Taiwanese crew at all times The comprehensive dynamic response vector, Indicates the first Taiwanese crew at all times The comprehensive dynamic response vector, This represents the Euclidean norm.

[0120] In an optional embodiment, S3 includes:

[0121] S31. Based on the voltage dynamic response feature sequence, extract the starting point and key turning point of the dynamic response time sequence of each wind turbine, and generate a fault time sequence event tag set for each wind turbine.

[0122] Optionally, the sliding window variance method can be used to extract the starting point and key inflection points of the voltage dynamic response feature sequence. The sliding window variance can effectively identify abrupt changes in the feature sequence, thus corresponding to key events of voltage sags. The expression for the sliding window variance can be:

[0123]

[0124] in, Indicates the first Taiwanese crew at all times The variance of the sliding window reflects the degree of dispersion of the feature vectors within the window. Indicates the length of the sliding window. Indicates the first Taiwanese crew at all times The comprehensive dynamic response vector, Indicates the first Taiwanese crew at all times The mean of the overall dynamic response vector within the window. The calculation formula can be , This represents the Euclidean norm, used to calculate the magnitude of a vector, reflecting the magnitude of the vector.

[0125] The event identification rules must be clear and operable. The condition for determining the starting point of a voltage sag is that when the variance of the sliding window first exceeds a preset threshold, and the variance of the sliding window remains above the preset threshold for a set number of subsequent moments, then that moment is determined as the starting point, and the event type is marked as a voltage sag.

[0126] The criteria for determining the lowest voltage drop point are as follows: after the starting point is identified, when the sliding window variance reaches its peak value, that is, the current sliding window variance is greater than the sliding window variance of the previous and next time moments, and the instantaneous voltage drop depth at the corresponding moment is the maximum value after the starting point, then the moment is determined to be the lowest voltage drop point, and the event type is marked as the lowest voltage drop point identifier.

[0127] The condition for determining the recovery rate mutation point is that after the lowest drop point is identified, when the sliding window variance reaches a peak again, and the voltage recovery rate at the corresponding moment is the maximum value after the lowest point, then that moment is determined to be the recovery rate mutation point, and the event type is marked as recovery rate mutation identifier.

[0128] The criteria for determining the recovery completion point are as follows: after identifying the abrupt change in the recovery rate, if the sliding window variance drops below a preset threshold, and the sliding window variance remains below the preset threshold for a set number of subsequent time points, while the instantaneous voltage drop depth is lower than a set value, then that time point is determined to be the recovery completion point, and the event type is marked as a recovery completion identifier. The preset threshold is determined based on historical data statistics and is usually related to the mean of the sliding window variance during normal operation.

[0129] These characteristic moments and their corresponding event types are combined into a fault timing event tag set, where each element contains the event's moment and type identifier. If a unit does not identify a certain type of event, the corresponding identifier is marked at the location of that event in the tag set, and the timing alignment of that event is ignored during subsequent synchronization.

[0130] S32. Based on the electrical connection topology of each wind turbine unit in the wind farm, a distributed information interaction network is constructed.

[0131] Optionally, the main electrical wiring diagram of the wind farm can be obtained first, including the connection relationships of the turbines, busbars, and transmission lines, and each wind turbine can be regarded as a network node. If two turbines are directly connected to the same busbar, or connected to adjacent busbars and the busbars are connected by a short-distance cable, then the two nodes are considered to have an information exchange link.

[0132] The adjacency matrix of a distributed information exchange network is ,in, Represents the adjacency matrix of the th Line number Column elements, and All represent the unit serial number, with values ​​ranging from the interval corresponding to the total number of units. The adjacency matrix is ​​a symmetric matrix, i.e. diagonal elements This indicates that the unit itself does not interact with itself. When, it indicates that there is an information exchange link between any two units. When this occurs, it indicates that there is no information exchange link between any two generating units.

[0133] To improve network robustness, network connectivity needs to be verified. A depth-first search can be used to traverse network nodes. If all nodes can be traversed, indicating a connected graph, the requirement is met. If isolated nodes exist, meaning a unit has no electrical connection to any other unit, an interaction link with the unit corresponding to the nearest busbar needs to be added to that node. The corresponding element in the adjacency matrix is ​​then set to 1 to ensure that information can be transmitted to all units.

[0134] S33. Using the fault time-series event tag set of each wind turbine as the initial state, perform multiple rounds of iterative consensus in the distributed information interaction network to update the time-series event estimates of each wind turbine.

[0135] Optionally, the time values ​​in the fault timing event label set of each unit are used as the initial timing event estimates. For the same event type, each unit has its own initial detection time, and the average of the initial detection times of the event for all units is used as the initial estimate of the event for that unit.

[0136] The formula for calculating the initial estimate is: ,in, Indicates the first Taiwan unit Initial estimates of class events, Indicates the total number of generating units. Indicates the first Taiwan unit The detection time for class events.

[0137] In a distributed information exchange network, when executing a multi-round iterative consensus algorithm, the iterative update formula can adopt the distributed consensus formula:

[0138]

[0139] in, Indicates the first During the first iteration For the timing event estimation of a set of machines, for a certain type of event, each event needs to be iterated separately. Indicates the first Taiwanese unit in Estimated time series events after rounds of iteration. Indicates the relationship with the first The set of neighboring units directly connected to a given unit is determined by the adjacency matrix. If the corresponding element in the adjacency matrix is ​​1, then that unit belongs to the set of neighboring units. Indicates the first Taiwanese unit to the first Weighting coefficients for the Taiwanese unit.

[0140] The calculation of the weighting coefficients needs to take into account the electrical distance, firstly the computer group With the unit The electrical distance is calculated based on the transmission line impedance. The greater the electrical distance, the greater the impedance, and the greater the signal transmission loss. The formula for calculating the electrical distance is: ,in, Indicates the unit With the unit electrical distance, Indicates the unit With the unit The resistance of the transmission lines between them Indicates the unit With the unit The reactance of the transmission lines between them The imaginary unit is represented; then the weights are calculated using the normalization formula: ,in, Indicates the relationship with the first The formula assigns the serial number of the neighboring units directly connected to the unit to ensure that the closer the electrical distance between the units, the greater the weight. This satisfies the convex combination characteristic that the sum of the weight coefficients is 1, thus avoiding the excessive influence of the estimated value of one unit on other units.

[0141] During the iteration process, pre-defined standard messages can be used for inter-unit information exchange. The duration of information exchange in each iteration must meet real-time requirements. Data transmission consists of each unit sending only the estimated timing events of the current iteration to its neighboring units, without sending raw data, thus reducing communication bandwidth consumption. The exception handling mechanism is as follows: if a unit does not receive information from its neighboring units within a set number of iterations, the neighboring unit is considered to have a communication failure. Its weight coefficient is set to 0, and the weights of other neighboring units are renormalized to ensure that the sum of their weight coefficients remains 1.

[0142] S34. When the change in the estimated time series events of all wind turbine units is less than the preset convergence threshold, stop the iteration and use the finally agreed time series event estimates as the global baseline time series.

[0143] Optionally, the criterion for iterative convergence is that the changes in the estimated time-series events of all units are less than a preset convergence threshold, i.e., the following conditions are met:

[0144]

[0145] in, Indicates the first Taiwan unit The absolute difference between the estimated value of round k and the estimated value of round k. This indicates taking the maximum value of the differences among all units. This indicates a preset convergence threshold to ensure that the estimated values ​​for all generating units have stabilized. The preset convergence threshold is set based on the grid dispatching requirements for timing synchronization accuracy.

[0146] If the number of iterations exceeds the preset maximum number of iterations and still fails to converge, it indicates that there is an abnormal unit in the network. An alarm mechanism needs to be triggered, the type of non-converged event needs to be marked, and the mean of the estimated value of the preset round of iterations needs to be used as a temporary global baseline time series.

[0147] When the convergence condition is met, the iteration stops. At this point, the set of time sequence event estimates that all units agree on is the global reference time sequence. Each element is the time of the corresponding global reference event. This time sequence is the unified time reference for all units, eliminating the time sequence differences caused by clock deviation and detection delay.

[0148] S35. Based on the global reference time series, perform time axis alignment and interpolation resampling on the original voltage dynamic response feature sequences of each wind turbine to obtain aligned voltage dynamic response feature sequences. Arrange all aligned voltage dynamic response feature sequences according to the wind turbine number to obtain the site-level collaborative time series matrix.

[0149] Optionally, based on the global reference timing sequence, cubic spline interpolation can be used to align the time axis and resample the original voltage dynamic response characteristic sequences of each unit. Cubic spline interpolation can be used to maintain the continuity of the first and second derivatives at the interpolation points, avoid inflection points in the interpolation curve, and ensure that the aligned characteristic sequences can truly reflect the dynamic response characteristics of the unit.

[0150] Specifically, for the first The characteristic sequences of the kilowatt group are used to construct cubic polynomials on each sub-interval, with the global reference time series as the target time axis:

[0151]

[0152] in, Indicates the first Taiwanese unit in The cubic interpolation polynomial for each subinterval, , , , All represent interpolation coefficients. Represents a time variable.

[0153] The interpolation polynomial must satisfy interpolation conditions, continuity conditions, and boundary conditions. The interpolation condition is that the interpolation curve passes through the original feature points; that is, the function value of the cubic polynomial at the endpoints of the sub-interval is equal to the original eigenvector value. The continuity condition is that the first and second derivatives of the interpolation curves in adjacent intervals are continuous, ensuring a smooth transition. The boundary condition uses natural spline boundaries, meaning the second derivative of the interpolation curves at the beginning and end of the sub-intervals is 0, avoiding curve distortion at the boundaries.

[0154] By solving the above system of equations, interpolation coefficients are obtained, and then the feature vectors at each global reference time are calculated to obtain the aligned voltage dynamic response feature sequence. The time axis of this sequence is completely consistent with the global reference time series, and includes the time dimension corresponding to the number of global reference events and the feature dimension corresponding to the dimension of the integrated dynamic response vector.

[0155] By arranging the alignment feature sequences of all units according to their unit numbers, a site-level collaborative time series matrix is ​​constructed. ,in This indicates the total number of wind turbines in the wind farm. Indicates the number of global baseline events. The dimension of the composite dynamic response vector is represented by each element of the matrix. Indicates the first Taiwanese unit in The comprehensive dynamic response vector at each global baseline event moment.

[0156] To verify the validity of the matrix, its temporal consistency index needs to be calculated. For each global reference time, the variance of the eigenvectors of all units is calculated. The expression for the variance can be:

[0157]

[0158] in, Indicates the first The variance of the eigenvectors of all units at a global reference time. Indicates the first The mean of the characteristic vectors of all units at a global reference time. This represents the Euclidean norm.

[0159] If the variance at any given moment exceeds a set multiple of the mean variance during normal operation, it indicates that the alignment of some units is poor at that moment, and it is necessary to return to the previous iterative synchronization steps to re-perform iterative synchronization.

[0160] In an optional embodiment, S4 includes:

[0161] S41. Based on the station-level collaborative time sequence matrix, calculate the product of the real-time reactive power output increment and voltage deviation change of each wind turbine during the voltage recovery process to obtain the instantaneous power-voltage product sequence of each wind turbine.

[0162] Optionally, based on the smoothed voltage time-series data corresponding to the station-level collaborative time-series matrix and the collected reactive power output data, the real-time reactive power output increment of each unit is first calculated. The reactive power output increment can be used to reflect the reactive power support strength of the unit during low-voltage ride-through. The expression for the real-time reactive power output increment can be:

[0163]

[0164] in, Indicates the first Taiwanese crew at all times Real-time reactive power output increment, Indicates the first Taiwanese crew at all times Real-time reactive power output, Indicates the first Steady-state reactive power output of the generator unit before voltage dip.

[0165] The steady-state reactive power output before a voltage sag is taken as the average value over a set period before the start of the sag, and the number of average samples is determined based on the data acquisition frequency. If the real-time reactive power output increment is negative, it indicates that the unit is not providing reactive power support, or even absorbing reactive power, and in this case, it is set to 0.

[0166] The voltage deviation change is defined as the magnitude of voltage recovery relative to the lowest point of the voltage drop. It can be used to reflect the effectiveness of voltage recovery. The expression for the voltage deviation change can be:

[0167]

[0168] in, Indicates the first Taiwanese crew at all times The change in voltage deviation. Indicates the first Taiwanese crew at all times Smooth voltage value, Indicates the first The lowest voltage value of the unit during this temporary voltage drop, the first The lowest voltage value of the unit during this temporary sag was extracted from the smoothed voltage timing data. The voltage deviation change is non-negative; if it is negative, it is set to 0.

[0169] The instantaneous power-voltage product sequence can be calculated using the following formula:

[0170]

[0171] in, Indicates the first Taiwanese crew at all times The instantaneous contribution product value is the product of reactive power support strength and voltage recovery effect. If the reactive power increment provided by the unit is large and the corresponding voltage recovery is large, the instantaneous contribution product value is large, indicating that the unit's collaborative contribution at that moment is significant; conversely, if the reactive power increment provided by the unit is small, or the voltage does not recover, the instantaneous contribution product value is small, and the contribution is weak.

[0172] If the frequency of voltage data acquisition is higher than the frequency of reactive power data acquisition, the reactive power data needs to be interpolated and frequency-upgraded. A linear interpolation method can be used to ensure that the timestamps of the real-time reactive power increment and voltage deviation change are completely aligned, thus avoiding product calculation errors caused by frequency inconsistency.

[0173] S42. Integrate the instantaneous power-voltage product sequence within a preset time window to obtain the total energy contribution of each wind turbine during the entire low-pressure ride-through event.

[0174] Optionally, the instantaneous power-voltage product sequence can be integrated within a preset time window to obtain the total energy contribution value. The total energy contribution value of different units can be used to reflect the differences in their cumulative contributions throughout the low-voltage ride-through process. Integration can accumulate instantaneous contributions into a total contribution, comprehensively reflecting the role of the unit in the entire low-voltage ride-through event. The trapezoidal integration method can be used to calculate the total energy contribution value.

[0175]

[0176] in, Indicates the first The total energy contribution of the unit. Indicates the start time of the preset time window. Indicates the duration of the preset time window. Indicates the first Taiwanese crew at all times The instantaneous contribution product value, Indicates the number of sampling points within the time window. Indicates the first Each sampling time, Indicates the first Taiwanese crew at all times The instantaneous contribution product value, Indicates the first Each sampling time, Indicates the first Taiwanese crew at all times The instantaneous contribution product value, This indicates the sampling interval. The number of sampling points within a time window is calculated from the preset time window duration and the sampling interval. ,in This indicates the aligned sampling interval.

[0177] Accuracy verification of the trapezoidal integral method requires calculating the relative error between the integral result and the analytical solution. If there is an instantaneous contribution product value from a simple function fitting, the relative error must be within a set range. If the error is too large, the sampling interval needs to be reduced, i.e., the data acquisition frequency needs to be increased, or the Simpson integral method can be used for recalculation. The Simpson integral method is suitable for cases where the number of sampling points is even. The expression for calculating the total contribution energy value using the Simpson integral method can be:

[0178]

[0179] in, Indicates the first The total energy contribution of the unit. Indicates the aligned sampling interval. Indicates the start time of the preset time window. Indicates the first Taiwanese crew at all times The instantaneous contribution product value, Indicates the number of sampling points within the time window. Indicates the first Each sampling time, Indicates the first Taiwanese crew at all times The instantaneous contribution product value, Indicates the first Each sampling time, Indicates the sampling interval. This indicates the end time of the preset time window. Indicates the first Taiwanese crew at all times The instantaneous contribution product value.

[0180] S43. Input the total contribution energy value into the contribution normalization model, calculate the normalized weight of each unit relative to the total contribution of all units in the station, and obtain the initial weight.

[0181] Optionally, the contribution normalization model is implemented through a summation layer and a division layer to convert the total contribution energy value of each unit into a percentage weight. The summation layer can be used to calculate the sum of the total contribution energy values ​​of all units, and the expression for the sum of the total contribution energy values ​​can be... ,in, This represents the total energy contribution of the station. This represents the total number of wind turbines, reflecting the overall collaborative contribution made by the wind farm during this low-voltage ride-through event.

[0182] The division layer can be used to divide the total energy contribution of each unit by the sum of its total energy contribution values ​​to obtain the initial weights. The expression for the initial weights can be: ,in, Indicates the first The initial weights of each generating unit range from [0,1], satisfying the normalization property that the sum of the initial weights of all generating units is 1. The physical meaning of the initial weights is the proportion of the contribution of a single generating unit to the total contribution of the station.

[0183] If the total energy contribution of all generating units is 0, meaning no generating unit provides effective reactive power support and voltage has not been restored, then the initial weight of all generating units is set to the average allocation value. This triggers an alarm mechanism, indicating that the low-voltage ride-through coordinated response of the wind farm has failed.

[0184] S44. The initial weights and the response speed factors of each wind turbine in the critical stage of voltage recovery are weighted and fused to obtain the comprehensive contribution factor.

[0185] Optionally, considering only the total energy contribution value is insufficient to comprehensively evaluate the coordinated performance of the units. Some units may have a high total energy contribution but a slow response speed, resulting in insufficient support in the initial stage of voltage recovery; other units may have a moderate total energy contribution but a fast response speed, effectively curbing further voltage drops. Therefore, a response speed factor needs to be introduced and weighted with the initial weights to obtain a comprehensive contribution factor.

[0186] The response rate factor is the reciprocal of the time required for the unit to recover from the voltage minimum point to the steady-state voltage set ratio. The expression for the response rate factor can be: ,in, Indicates the first The response speed factor of the Taiwanese unit Indicates the first Voltage recovery time of the kilowatt unit.

[0187] No. The voltage recovery time of the generator unit is calculated based on aligned voltage timing data. The minimum moment at which the smoothed voltage value reaches the set steady-state voltage ratio is found is the voltage recovery time. The difference between this moment and the moment of lowest voltage drop is the voltage recovery time. A larger response rate factor indicates faster voltage recovery and a stronger driving effect on the overall recovery process.

[0188] The comprehensive contribution factor can be obtained through weighted fusion, and its expression is as follows: ,in, This represents the overall contribution factor of the i-th generating unit. Indicates the weighting coefficient. Indicates the first The initial weights of the tandem units Represents the normalized i-th The response speed factor of the unit.

[0189] The range of weighting coefficients can be set according to the importance that grid dispatch places on energy contribution and response speed. If dispatch focuses more on the final recovery effect, the value tends to be larger; if it focuses more on the timeliness of recovery, the value tends to be smaller. The normalized response speed factor maps the original response speed factor to the [0,1] interval to avoid the influence of dimensional differences on the fusion effect. The expression for the normalization of the response speed factor can be: ,in, This represents the minimum value of the response speed factor for all units. This represents the maximum value of the response speed factor for all units.

[0190] The determination of weighting coefficients needs to be verified using the Analytic Hierarchy Process (AHP). A hierarchical structure is constructed, consisting of a target layer, a criterion layer, and a solution layer. Experts in relevant fields are invited to conduct pairwise comparisons and scoring of the criterion layer, constructing a judgment matrix. Weighting coefficients are then calculated and a consistency check is performed to ensure their rationality. The criterion for consistency checking is that the consistency ratio is lower than a set threshold.

[0191] S45. Normalize the comprehensive contribution factor to obtain the collaborative contribution index of each wind turbine.

[0192] Optionally, the comprehensive contribution factor can be normalized using a min-max method to eliminate the influence of differences in factor amplitudes under different scenarios, thus obtaining a collaborative contribution index. The expression for the collaborative contribution index can be:

[0193]

[0194] in, Indicates the first The collaborative contribution index of the unit has a value range of [0,1]. Indicates the first The overall contribution factor of the Taiwanese unit This represents the minimum value of the overall contribution factor of all generating units. This represents the maximum value of the overall contribution factor of all generating units.

[0195] The core purpose of normalization is to make the indicators comparable. Regardless of the differences in total energy contribution and response speed among different low-pressure ride-through events, the collaborative contribution index will remain within the range of [0,1], facilitating the establishment of a unified evaluation standard. The closer the collaborative contribution index is to 1, the better the unit's collaborative contribution performance, indicating a larger energy contribution and faster response speed; the closer it is to 0, the worse the performance. For units with poor performance, their control parameters need to be optimized.

[0196] To ensure the reliability of the indicators, their discrimination ability needs to be verified. The standard deviation of all unit indicators needs to be calculated. The formula for calculating the standard deviation is: ,in, This represents the standard deviation of the collaborative contribution index of all generating units. This represents the average of all unit collaborative contribution indicators.

[0197] If the standard deviation is lower than the set threshold, it indicates that the indicator has insufficient discrimination and the performance of all units is small. It is necessary to adjust the value of the weight coefficient or redefine the response speed factor to improve the indicator's discrimination ability.

[0198] In an optional embodiment, the performance evaluation model includes an input layer, a hidden layer, and an output layer, and S5 includes:

[0199] S51. Based on the collaborative contribution index of each wind turbine, the overall collaborative index of the wind farm is calculated.

[0200] Optionally, the coordination index can be used to quantify the balance of the contribution distribution among the units. An ideal coordinated response should result in a relatively balanced contribution from all units, avoiding situations where some units are overloaded and others are idle. The expression for the coordination index can be:

[0201]

[0202] in, This represents the degree of synergy, with a value range of [0,1]. Indicates the first The collaborative contribution index of the Taiwanese unit. This represents the total number of generating units. When the contributions of all generating units are perfectly balanced, that is, when the collaborative contribution index of all generating units is equal to... When the summation term results in 0, the coordination index equals 1, indicating optimal coordination and the most balanced resource utilization, meaning all units contribute equally. When some units contribute extremely much while others contribute very little (or 0), the summation term increases, the coordination index approaches 0, and coordination is at its worst.

[0203] Optionally, different levels of coordination can be divided according to the values ​​of the coordination index. If the coordination index reaches the set high threshold, it indicates excellent coordination; if it is in the set medium-high threshold range, the coordination is good; if it is in the set medium-low threshold range, the coordination is average; if it is below the set low threshold, the coordination is poor and the reactive power support strategy of the unit needs to be adjusted.

[0204] S52. Based on the site-level collaborative time series matrix, extract the overall voltage recovery curve of the wind farm during the voltage recovery phase, and calculate the time required for the average voltage of the site to recover to the preset voltage, thus obtaining the overall recovery time index.

[0205] Optionally, the overall recovery time metric measures the total time required for a wind farm to recover to a stable operating state after experiencing a voltage dip fault. The formula for this metric can be expressed as follows: ,in, Indicates the overall recovery time of the wind farm; Representing the Recovery time for each stage, This is the stage number, with a value range from... arrive , This refers to the total number of stages that a wind farm goes through from the occurrence of a fault to its restoration to stability.

[0206] In addition, the recovery time for a single phase The start time of this phase can be used as a reference. and end time To calculate it, its expression can be:

[0207]

[0208] In the formula, Indicates the first The starting point of each phase Indicates the first The formula above allows for the precise calculation of the overall recovery time of wind farms during low-voltage ride-through, providing crucial data support for optimizing subsequent coordinated response management strategies.

[0209] S53. Input the multi-dimensional feature vector composed of the synergy index, the overall recovery time index, and the maximum voltage drop depth of the wind farm during the voltage sag into the input layer.

[0210] Optionally, the synergy index, the overall recovery time index, and the maximum voltage drop depth of the wind farm during the voltage dip can be combined into a multi-dimensional feature vector. The input is fed into the input layer. Among them, the coordination index... The formula used to measure the degree of coordination among wind turbines in a wind farm during low-voltage ride-through is as follows: , Indicates the first Taiwanese unit in Operating parameters at each sampling time point; overall recovery time index The expression used to characterize the total time from the occurrence of a voltage sag to the recovery of the system voltage to a normal level is: , This represents the initial moment of the voltage sag. The moment when the voltage returns to normal levels; maximum voltage drop depth indicator. Record the maximum decrease in voltage amplitude during the voltage sag, calculated using the following formula: , It is the minimum voltage value during the voltage sag. This is the system's rated voltage.

[0211] S54. Input the multidimensional feature vector into the hidden layer, and perform nonlinear transformation and feature fusion on the multidimensional feature vector through the hidden layer to obtain the hidden state representation.

[0212] Optionally, the multidimensional feature vector Input is fed into the hidden layer. The hidden layer is activated by a function. After performing nonlinear transformation and feature fusion on the multidimensional feature vectors, the expression for the hidden state representation is as follows:

[0213]

[0214] in, Here is the weight matrix of the hidden layer, with dimension 1. , Indicates the number of neurons in the hidden layer. The dimension of the multidimensional feature vector; Let be the bias vector of the hidden layer, with dimension . ; For activation functions, such as the commonly used ReLU function. This is used to introduce nonlinear relationships, enabling the model to learn complex patterns; The input multidimensional feature vector contains key data during the low-voltage ride-through process of wind farms, such as voltage drop depth, duration, and reactive power demand. This is the hidden state representation obtained after the hidden layer transformation, which is used for subsequent model calculation and decision-making.

[0215] S55. Input the hidden state representation into the output layer, perform linear weighting and activation function mapping on the hidden state representation through the output layer, and output the co-performance score.

[0216] Optionally, the hidden state representation is input to the output layer, and the output layer processes the hidden state representation. Let the hidden state representation be... ,in For the dimension of the hidden state, Indicates the hidden state of the first There are 10 elements. The output layer performs a linear weighting, and the expression for the linear weighting is: ,in It is the weight matrix of the output layer, with dimension 1. , The number of neurons in the output layer. It is a dimension of The bias vector, It is the result vector after linear weighting.

[0217] Then, to Perform activation function mapping, the activation function is The result after processing by the activation function is: The final output is the collaborative effectiveness score. for After further processing, the calculation formula can be obtained as follows: ,in For the corresponding The weighting coefficients are used to adjust the contribution of different output layer neuron results to the final collaborative efficacy score, and the final output collaborative efficacy score is obtained.

[0218] S56. Use the collaborative effectiveness score as collaborative effectiveness evaluation data.

[0219] Optionally, the collaboration effectiveness score can be used as data for collaboration effectiveness evaluation. Specifically, the collaboration effectiveness score can be calculated using the following formula:

[0220]

[0221] in, The representative collaborative performance score is the final collaborative performance evaluation data, which can be used to measure the collaborative response effect of various equipment in a wind farm during low-voltage ride-through. This indicates the number of indicators involved in the collaborative response assessment; For the first The weights of each evaluation indicator range from [0,1] and satisfy the following conditions: The weight can reflect the relative importance of the indicator in the overall evaluation; For the first The score for each evaluation indicator is based on a comparison of the actual operating parameters of wind farm equipment during low-voltage ride-through with preset standards. A higher score indicates better coordinated response performance under that indicator. The score is calculated using the formula described above. This data will serve as the final collaborative performance evaluation data, providing a quantitative basis for subsequent optimization of low-voltage ride-through collaborative response management strategies for wind farms.

[0222] In an optional embodiment, S52 includes:

[0223] S521. Based on the site-level collaborative time series matrix, extract the voltage time series data segments of each wind turbine during the voltage recovery phase, and perform weighted aggregation according to the rated capacity ratio of each wind turbine to obtain the weighted average voltage recovery sequence of the wind farm.

[0224] Optionally, based on the voltage time-series data of each unit corresponding to the station-level collaborative time-series matrix, weighted aggregation is performed according to the rated capacity ratio of the units. Units with larger rated capacities have a greater impact on the overall station voltage due to their voltage changes; therefore, they need to be assigned higher weights to ensure that the weighted average voltage accurately reflects the overall station voltage situation. The expression for the weighting coefficients can be:

[0225]

[0226] in, Indicates the first Weighting coefficients for the Taiwanese unit Indicates the first The rated capacity of the unit This represents the total rated capacity of the wind farm, with the weighting factor satisfying that the sum of the weighting factors of all units is 1.

[0227] The expression for the weighted average voltage recovery sequence can be:

[0228]

[0229] in, Indicates the time of the wind farm station The weighted average voltage, Indicates the first Taiwanese crew at all times Smoothed voltage timing data.

[0230] S522. Input the weighted average voltage recovery sequence into the parametric recovery trajectory fitting model, and perform parametric curve fitting on the weighted average voltage recovery sequence to obtain the overall voltage recovery curve and its fitting parameters; wherein, the expression of the overall voltage recovery curve is:

[0231]

[0232] in, This indicates the overall voltage recovery curve at time [time]. voltage value, Indicates the target value of steady-state voltage. Indicates the first The linear combination coefficients of exponentially decaying basis functions, Indicates the first The time constant of an exponentially decaying basis function Indicates the start time of the voltage sag. This represents the number of exponentially decaying basis functions.

[0233] Optionally, a multi-exponential decay model can be used to fit the weighted average voltage recovery sequence. This model can accurately characterize the exponential decay characteristics of the voltage recovery process. The process of voltage recovery from the lowest point to steady state usually conforms to a multi-exponential law, including a rapid recovery phase and a slow approach phase. The expression of the multi-exponential decay model is as follows:

[0234]

[0235] in, This indicates the overall voltage recovery curve at time [time]. voltage value, This represents the target steady-state voltage value, taken as the rated voltage of the wind farm. Indicates the first The linear combination coefficients of the exponentially decaying basis functions characterize the magnitude contribution of the basis functions. Indicates the first The time constant of the exponentially decaying basis function characterizes the decay rate. Represents a time variable. Indicates the start time of the voltage sag. This represents the number of exponentially decaying basis functions.

[0236] The number of exponentially decaying basis functions is typically set within a certain range, and the optimal number is determined using the Akaike Information Criterion. The core idea of ​​the Akaike Information Criterion is to balance goodness of fit with model complexity, selecting the number of basis functions with the smallest Akaike Information Criterion value as the optimal number. The expression for the Akaike Information Criterion can be: ,in, This indicates that the number of exponentially decaying basis functions is The Akaike Information Criterion Value at that time This indicates that the number of exponentially decaying basis functions is The likelihood function value of the model represents the degree to which the model fits the data; the larger the likelihood function value, the better the fit.

[0237] The fitting process can employ the nonlinear least squares method, solving for the linear combination coefficients and time constant by minimizing the objective function. The expression for the objective function can be:

[0238]

[0239] in, This represents the sum of squared residuals, characterizing the deviation between the fitted curve and the actual voltage recovery sequence. This represents the weighted average voltage. This represents the voltage predicted by the model.

[0240] The solution process can employ the Levenberg-Marquardt algorithm, an improved version of the nonlinear least squares method, which converges faster than the gradient descent method. The initial parameter settings are as follows: the initial value of the linear combination coefficients is the difference between the steady-state voltage target value and the lowest point value of the weighted average voltage, divided by the number of exponentially decaying basis functions; the initial value of the time constant is allocated according to a predetermined rule.

[0241] To verify the fit, the coefficient of determination needs to be calculated. The closer the coefficient of determination is to 1, the better the fit. The formula for calculating the coefficient of determination is:

[0242]

[0243] in, The coefficient of determination is represented by the coefficient of determination. This represents the mean of the weighted average voltage.

[0244] S523. Based on the fitting parameters of the overall voltage recovery curve, calculate the equivalent recovery time constant, and solve for the time corresponding to the overall voltage recovery curve reaching the preset voltage recovery threshold according to the equivalent recovery time constant, to obtain the theoretical recovery time.

[0245] Optionally, the core function of the equivalent recovery time constant is to transform the complex time characteristics of the multi-exponential decay model into a single parameter characterizing the overall recovery rate, facilitating rapid evaluation of recovery efficiency in engineering applications. Its calculation is based on the fitting parameters of the multi-exponential decay model, fusing the time constants of each exponential basis function through a weighted average, with the weights being the proportions of the corresponding linear combination coefficients. The expression for the equivalent recovery time constant can be: ,in, It represents the equivalent recovery time constant, reflecting the average speed of overall voltage recovery; the smaller the value, the faster the recovery speed. Indicates the first The linear combination coefficients of an exponentially decaying basis function characterize the contribution of the basis function to the magnitude of the recovery curve. Indicates the first The time constant of an exponentially decaying basis function represents the decay rate corresponding to that basis function; This indicates the number of exponentially decaying basis functions. The larger the linear combination coefficient of a basis function, the greater the weight of its time constant on the equivalent recovery time constant, ensuring that the equivalent parameters accurately reflect the dominant dynamic characteristics of the recovery curve. If the linear combination coefficient of a certain basis function has an extremely high proportion, it indicates that this basis function dominates the voltage recovery process, and the equivalent recovery time constant will be close to its time constant.

[0246] The theoretical recovery time refers to the time required for the overall voltage recovery curve to rise back to the preset voltage recovery threshold from the initial moment of a voltage dip. It is a fundamental indicator for evaluating the recovery speed. The preset voltage recovery threshold is set according to the requirements for safe operation of the power grid, and is usually a set percentage of the steady-state voltage target value to ensure that the voltage recovers to a safe operating range.

[0247] First, establish the equation for the theoretical recovery time, and set the overall voltage recovery curve value equal to the preset voltage recovery threshold: ,in, Indicates the preset voltage recovery threshold. Indicates the theoretical recovery time. Indicates the target value of steady-state voltage. This indicates the start time of the voltage sag.

[0248] Solve the above equation by transformation:

[0249] Rearranging the terms, we obtain the expression for summing the exponent terms:

[0250] Based on the simplified calculation using the equivalent recovery time constant, since the equivalent recovery time constant incorporates the dynamic characteristics of each basis function, it can be solved using a numerical iteration method. The initial value of the iteration is set to The convergence condition for the iteration is that the values ​​calculated in two consecutive calculations are equal to the sum of the sums ... The difference is less than the set precision threshold.

[0251] Numerical iterative methods approximate the equation step by step until the difference between the left and right sides meets the accuracy requirements, avoiding the complexity of directly solving multi-exponential equations. During the iteration process, each update... Then calculate the summation term on the left. If the summation term is equal to... If the absolute difference is greater than the precision threshold, then adjust. Continue iterating until the conditions are met.

[0252] The theoretical recovery time is the time required for the voltage to rise from a sag to a safe threshold during an ideal, oscillation-free recovery process, providing a benchmark for subsequent corrections that take into account the effects of actual oscillations.

[0253] S524. Based on the residual between the weighted average voltage recovery sequence and the overall voltage recovery curve, calculate the oscillation deviation index of the voltage recovery process, and then fuse and correct the theoretical recovery time to obtain the overall recovery time index.

[0254] Optionally, the oscillation deviation index can be used to quantify the fluctuation difference between the actual weighted average voltage recovery sequence and the ideal overall voltage recovery curve, reflecting the stability of the voltage recovery process. During the actual recovery process, voltage oscillations may occur due to factors such as unit control strategies and changes in grid impedance. The theoretical recovery time alone cannot fully reflect the recovery efficiency; therefore, oscillation deviation needs to be introduced for correction. The expression for the oscillation deviation index can be: ,in, This represents the oscillation deviation index. The larger the value, the more violent the oscillation and the worse the stability during the actual voltage recovery process. Indicates the time of the wind farm station The weighted average voltage, i.e., the actual recovered sequence data; This indicates the overall voltage recovery curve at time [time]. The voltage value, i.e., the ideal recovery trajectory; Indicates the start time of the voltage sag; This indicates the duration of the preset time window, which is consistent with the time window used for data acquisition and integration calculation in the previous steps, ensuring the entire process of coverage voltage restoration.

[0255] If the oscillation deviation index exceeds the set stability threshold, it indicates that the oscillation during the recovery process is too severe and may affect the stable operation of the power grid. It needs to be adjusted as a key focus in the subsequent optimization of control strategies.

[0256] The overall recovery time index can be obtained by fusing and correcting the oscillation deviation index with the theoretical recovery time. It comprehensively reflects recovery speed and stability, and is a core time-domain indicator for evaluating synergistic effectiveness. The fusion correction formula can be: ,in, This indicates the overall recovery time index; the smaller the value, the better the overall efficiency of voltage recovery at the wind farm. Indicates the theoretical recovery time; This represents the preset oscillation correction coefficient, used to adjust the weight of the oscillation deviation on the total recovery time; This indicates the oscillation deviation index.

[0257] The preset oscillation correction coefficient needs to be determined based on power grid operation specifications and engineering experience, and its value must balance the importance of recovery speed and stability. If the power grid has a low tolerance for voltage oscillations, a larger correction coefficient will result in a longer overall recovery time for more severe oscillations; conversely, if more emphasis is placed on recovery speed, a smaller correction coefficient will mitigate the impact of oscillation deviations. The rationality of the correction coefficient needs to be verified through multiple sets of historical data to ensure that recovery processes with different oscillation levels can be effectively distinguished by the overall recovery time index.

[0258] To verify the effectiveness of the overall recovery time indicator, its correlation with the actual recovery completion time needs to be calculated. The actual recovery completion time is defined as the moment when the weighted average voltage first reaches the preset voltage recovery threshold and the oscillation deviation does not exceed the set range within the subsequent set time period. If the correlation coefficient between the overall recovery time indicator and the actual recovery completion time is higher than the set threshold, it indicates that the indicator can accurately reflect the actual recovery performance; otherwise, the preset voltage recovery threshold or oscillation correction coefficient needs to be adjusted, and the indicator calculation logic needs to be optimized.

[0259] The aforementioned collaborative response management method for low-voltage ride-through in wind farms collects real-time operating status data from multiple wind turbines within the wind farm. It extracts features from the timing of voltage changes at the turbine terminals within a preset time window to obtain a voltage dynamic response feature sequence. Based on this feature sequence, it performs distributed consistency synchronization and alignment of the voltage support action timing of each turbine, generating a farm-level collaborative timing matrix. This unifies the originally asynchronous and discrete reactive power response timing of each turbine under a global reference timing, eliminating the oscillations and support cancellation problems in the voltage recovery process caused by asynchronous responses between turbines. Furthermore, through analysis... The contribution characteristics of each unit to grid voltage recovery are characterized and the collaborative contribution index is calculated to quantify the actual supporting role of each unit in the station-level collaborative operation, avoiding the imbalance of distribution where some units contribute excessively while others provide insufficient support. Finally, the collaborative contribution index is input into the efficiency evaluation model to quantitatively evaluate the overall low-voltage ride-through collaborative operation efficiency of the station, enabling closed-loop measurement of collaborative quality. This improves the voltage recovery speed and stability of wind farms during low-voltage ride-through, reduces the voltage oscillation amplitude during the recovery process, reduces redundancy and conflicts in reactive power support among units, and enables precise synchronous collaboration and balanced contribution distribution of multi-unit responses.

[0260] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0261] Based on the same inventive concept, this application also provides a collaborative response management system for low-voltage ride-through of wind farms, used to implement the aforementioned collaborative response management method for low-voltage ride-through of wind farms. The solution provided by this system is similar to the implementation described in the above method. Therefore, the specific limitations of one or more embodiments of the collaborative response management system for low-voltage ride-through of wind farms provided below can be found in the limitations of the collaborative response management method for low-voltage ride-through of wind farms described above, and will not be repeated here.

[0262] In one exemplary embodiment, such as Figure 2 As shown, a schematic diagram of a collaborative response management system 10 for low-voltage ride-through at wind farms is provided, including:

[0263] The data acquisition and monitoring module 11 is used to acquire real-time operating status data of multiple wind turbine units in the wind farm.

[0264] The unit dynamic feature extraction module 12 is used to extract features of the terminal voltage change sequence of each wind turbine within a preset time window based on real-time operating status data, so as to obtain the voltage dynamic response feature sequence of the wind turbine.

[0265] The station-level timing coordination and alignment module 13 is used to perform distributed consistency synchronization and alignment of the voltage support action timing of each wind turbine based on the voltage dynamic response characteristic sequence, and generate a station-level coordination timing matrix.

[0266] Unit Coordination Contribution Assessment Module 14 is used to analyze the contribution characteristics of different wind turbine units to grid voltage recovery based on the site-level coordination time series matrix, and obtain the coordination contribution index of each wind turbine unit.

[0267] The collaborative performance evaluation module 15 is used to input the collaborative contribution index of each wind turbine into the performance evaluation model, to quantitatively evaluate the overall low-voltage ride-through collaborative action of the wind farm, and to obtain collaborative performance evaluation data.

[0268] In one embodiment, such as Figure 3 A computer device 300 is provided, comprising:

[0269] At least one processor 301, and at least one memory 302 communicatively connected to said processor 301; said memory stores application code executable by said processor, said application code being executed by said processor to enable said processor to perform the steps of a collaborative response management method for low-voltage ride-through of wind farms as described above;

[0270] The computer device may also include: sensor 303;

[0271] The processor 301, memory 302, and sensor 303 can be connected via bus 304 or other means. The figure shows an example of connection via bus 304. Figure 3 The character is represented by a single thick line, but this does not mean that there is only one bus or a type of bus.

[0272] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0273] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0274] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A collaborative response management method for low-voltage ride-through at wind farms, characterized in that, The method includes: S1. Obtain real-time operating status data of multiple wind turbine units within the wind farm; S2. Based on the real-time operating status data, feature extraction is performed on the timing of the terminal voltage change of each wind turbine within a preset time window to obtain the voltage dynamic response feature sequence of the wind turbine. S3. Based on the voltage dynamic response feature sequence, perform distributed consistency synchronization and alignment of the voltage support action timing of each wind turbine to generate a station-level collaborative timing matrix. S4. Based on the site-level collaborative time series matrix, analyze the contribution characteristics of different wind turbine units to grid voltage recovery, and obtain the collaborative contribution index of each wind turbine unit. S5. Input the collaborative contribution index of each wind turbine into the efficiency evaluation model to quantitatively evaluate the efficiency of the overall low-voltage ride-through collaborative action of the wind farm and obtain collaborative efficiency evaluation data.

2. The method according to claim 1, characterized in that, S2 includes: S21. Based on the real-time operating status data, extract the terminal voltage timing data of each wind turbine within the preset time window; S22. Perform time-domain noise reduction on the terminal voltage timing data to obtain smoothed voltage timing data for each wind turbine unit; S23. Based on the smoothed voltage time series data, calculate the instantaneous voltage drop depth and recovery rate of each wind turbine during the voltage sag event to obtain a preliminary dynamic feature set; S24. Input the preliminary dynamic feature set into the feature fusion model, perform feature splicing and feature mapping on the preliminary dynamic feature set, and obtain the comprehensive dynamic response vector of each wind turbine unit; S25. Arrange the comprehensive dynamic response vector in chronological order to obtain the voltage dynamic response characteristic sequence of each wind turbine.

3. The method according to claim 2, characterized in that, S3 includes: S31. Based on the voltage dynamic response feature sequence, extract the starting point and key turning point of the dynamic response time sequence of each wind turbine, and generate a fault time sequence event tag set for each wind turbine. S32. Based on the electrical connection topology of each wind turbine in the wind farm, a distributed information interaction network is constructed. S33. Using the fault time-series event tag set of each wind turbine as the initial state, perform multiple rounds of iterative consensus in the distributed information interaction network to update the time-series event estimates of each wind turbine; wherein, the update formula for the time-series event estimates is: in, Indicates the first During the first iteration Timing event estimates for the unit, For the first Taiwanese unit in Estimated time series events after rounds of iteration. In order to be with the first The set of neighboring units directly connected to the unit. For the first Taiwanese unit to the first Weighting coefficients for the Taiwanese unit; S34. When the change in the estimated time series event values ​​of all the wind turbine units is less than the preset convergence threshold, the iteration stops, and the finally agreed-upon estimated time series event values ​​are used as the global baseline time series. S35. Based on the global reference timing sequence, perform time axis alignment and interpolation resampling on the original voltage dynamic response feature sequences of each wind turbine to obtain aligned voltage dynamic response feature sequences, and arrange all the aligned voltage dynamic response feature sequences according to the sequence number of the wind turbine to obtain the site-level collaborative timing matrix.

4. The method according to claim 1, characterized in that, S4 includes: S41. Based on the station-level collaborative time sequence matrix, calculate the product of the real-time reactive power output increment and voltage deviation change of each wind turbine during the voltage recovery process to obtain the instantaneous power-voltage product sequence of each wind turbine. S42. Integrate the instantaneous power-voltage product sequence within the preset time window to obtain the total energy contribution value of each wind turbine during the entire low-voltage ride-through event. S43. Input the total contribution energy value into the contribution normalization model, calculate the normalized weight of each unit relative to the total contribution of all units in the station, and obtain the initial weight; the contribution normalization model includes a summation layer and a division layer, the summation layer is used to calculate the sum of the total contribution energy values ​​of all the wind turbine units, and the division layer is used to divide the total contribution energy value of each wind turbine unit by the sum of the total contribution energy values ​​to obtain the initial weight; S44. The initial weights and the response speed factors of each wind turbine in the critical stage of voltage recovery are weighted and fused to obtain the comprehensive contribution factor. S45. Normalize the comprehensive contribution factor to obtain the collaborative contribution index of each wind turbine.

5. The method according to claim 1, characterized in that, The performance evaluation model includes an input layer, a hidden layer, and an output layer, and S5 includes: S51. Based on the collaborative contribution index of each wind turbine, the overall collaborative index of the wind farm is calculated; wherein, the collaborative index is used to quantify the balance of the contribution distribution of each turbine, and the expression of the collaborative index can be: in, As a synergy indicator, For the first The collaborative contribution index of the Taiwanese unit. This represents the total number of wind turbine units within the wind farm. S52. Based on the site-level collaborative time series matrix, extract the overall voltage recovery curve of the wind farm during the voltage recovery phase, and calculate the time required for the average voltage of the site to recover to the preset voltage, thereby obtaining the overall recovery time index. S53. Input the multi-dimensional feature vector composed of the coordination index, the overall recovery time index, and the maximum voltage drop depth of the wind farm during the voltage sag into the input layer; S54. Input the multidimensional feature vector into the hidden layer, and perform nonlinear transformation and feature fusion on the multidimensional feature vector through the hidden layer to obtain the hidden state representation; S55. Input the hidden state representation to the output layer, perform linear weighting and activation function mapping on the hidden state representation through the output layer, and output a collaborative performance score. S56. Use the collaborative effectiveness score as the collaborative effectiveness evaluation data.

6. The method according to claim 5, characterized in that, S52 includes: S521. Based on the site-level collaborative time series matrix, extract the voltage time series data segments of each wind turbine in the voltage recovery stage, and perform weighted aggregation according to the rated capacity ratio of each wind turbine to obtain the weighted average voltage recovery sequence of the wind farm. S522. Input the weighted average voltage recovery sequence into the parametric recovery trajectory fitting model, and perform parametric curve fitting on the weighted average voltage recovery sequence to obtain the overall voltage recovery curve and its fitting parameters; wherein, the expression of the overall voltage recovery curve is: in, This indicates that the overall voltage recovery curve at time [time missing] voltage value, Indicates the target value of steady-state voltage. Indicates the first The linear combination coefficients of exponentially decaying basis functions, Indicates the first The time constant of an exponentially decaying basis function Indicates the start time of the voltage sag. This indicates the number of exponentially decaying basis functions; S523. Based on the fitting parameters of the overall voltage recovery curve, calculate the equivalent recovery time constant, and solve for the time corresponding to when the overall voltage recovery curve reaches the preset voltage recovery threshold according to the equivalent recovery time constant, to obtain the theoretical recovery time; S524. Based on the residual between the weighted average voltage recovery sequence and the overall voltage recovery curve, calculate the oscillation deviation index of the voltage recovery process, and fuse and correct the theoretical recovery time using the oscillation deviation index to obtain the overall recovery time index; wherein, the expression for the oscillation deviation index is: in, This indicates the oscillation deviation index. Indicates the time of the wind farm station The weighted average voltage, This indicates the overall voltage recovery curve at time [time]. voltage value, Indicates the start time of the voltage sag. This indicates the duration of the preset time window. This indicates the preset oscillation correction coefficient.

7. A collaborative response management system for low-voltage ride-through at wind farms, characterized in that, The system includes: The data acquisition and monitoring module is used to acquire real-time operating status data of multiple wind turbine units within the wind farm. The unit dynamic feature extraction module is used to extract features from the timing sequence of the terminal voltage change of each wind turbine within a preset time window based on the real-time operating status data, so as to obtain the voltage dynamic response feature sequence of the wind turbine. The station-level timing coordination and alignment module is used to perform distributed consistency synchronization and alignment of the voltage support action timing of each wind turbine based on the voltage dynamic response feature sequence, and generate a station-level coordination timing matrix. The unit collaborative contribution evaluation module is used to analyze the contribution characteristics of different wind turbine units to grid voltage recovery based on the site-level collaborative time series matrix, and obtain the collaborative contribution index of each wind turbine unit. The collaborative performance evaluation module is used to input the collaborative contribution index of each wind turbine into the performance evaluation model, to quantitatively evaluate the overall low-voltage ride-through collaborative performance of the wind farm, and to obtain collaborative performance evaluation data.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

Images