Method and system for monitoring running state of wind driven generator

Abstract

The invention discloses a wind driven generator operation state monitoring method and system, and relates to the technical field of wind driven generators and data analysis, and the method comprises the steps: collecting electric energy operation parameter data, carrying out the preprocessing, extracting feature values, and splicing a high-dimensional feature vector; performing similarity analysis and dimension reduction on the basis of the high-dimensional feature vectors to obtain initial positions of the high-dimensional feature vectors in a two-dimensional space, calculating two-dimensional similarity probabilities among the high-dimensional feature vectors, defining a target function to perform iterative updating, outputting a two-dimensional position set, constructing a topological node set, and calculating a neighborhood relationship and a topological energy value of nodes. Generating a topological topographic map, and carrying out node clustering to obtain a healthy cluster and an abnormal cluster; according to the invention, the dynamic clustering identification of the operation state of the wind driven generator can be realized, the high-precision, visual monitoring and early fault early warning of the operation state of the wind driven generator can be realized, and the high-precision, visual monitoring and early fault early warning of the operation state of the wind driven generator can be realized.

Description

Technical Field

[0001] This invention relates to the field of wind turbine and data analysis technology, and in particular to a method and system for monitoring the operating status of a wind turbine. Background Technology

[0002] With the global energy structure shifting towards low-carbon transformation, wind power, as an important form of renewable energy, is experiencing a continuous increase in installed capacity and operational complexity. Modern large-scale wind turbine systems typically employ a distributed sensing and monitoring architecture, using turbine monitoring and data acquisition systems to record key operating parameters in real time, such as wind speed, wind direction, active power, reactive power, grid frequency, and power factor. Traditional operational status monitoring methods are mostly based on single-parameter threshold judgments or empirical feature statistical analysis, such as judging output anomalies through active power-wind speed characteristic curves, judging mechanical faults through bearing vibration characteristics, and judging inverter unit imbalances through current waveforms. However, the operating status of wind turbines exhibits significant time-varying and multi-source coupling characteristics, and is affected by wind field disturbances, grid load fluctuations, and the dynamic response of the control system. Its operating data displays high-dimensional nonlinearity and dynamic correlation. Traditional methods based on single features or low-dimensional statistics often fail to accurately identify early abnormal signals and struggle to reveal the nonlinear evolution characteristics in complex energy transfer processes, leading to delayed fault prediction, high false alarm rates, and reduced energy efficiency.

[0003] Existing technologies mostly only focus on feature dimensionality reduction or cluster visualization, lacking the ability to couple and model with the characteristics of power distribution. They cannot establish a quantitative mapping relationship between high-dimensional energy state and operational health. In addition, traditional clustering algorithms are prone to blurred cluster boundaries and decreased anomaly identification accuracy when dealing with the sparse and unbalanced characteristics of wind turbine operation data. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method and system for monitoring the operating status of wind turbines, which solves the problem that existing technologies mostly only focus on feature dimensionality reduction or cluster visualization, lack the ability to couple and model with the characteristics of power distribution, and cannot establish a quantitative mapping relationship between high-dimensional energy state and operating health. In addition, traditional clustering algorithms are prone to problems such as blurred cluster boundaries and decreased accuracy of anomaly identification when dealing with the sparse and unbalanced characteristics of wind turbine operating data.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for monitoring the operating status of a wind turbine generator, comprising, Collect power operation parameter data, preprocess it, extract feature values ​​and concatenate them into a high-dimensional feature vector; Similarity analysis and dimensionality reduction are performed based on high-dimensional feature vectors to obtain the initial position of the high-dimensional feature vectors in two-dimensional space. The two-dimensional similarity probability between high-dimensional feature vectors is calculated, an objective function is defined for iterative update, and a two-dimensional position set is output. A topological node set is constructed, the neighborhood relationship and topological energy value of the nodes are calculated, and a topological topographic map is generated for node clustering to obtain healthy clusters and abnormal clusters. Based on healthy and abnormal clusters, the center and center ratio within each cluster are calculated to obtain the point-level power health index of the wind turbine, and then the operating status is determined and corresponding measures are implemented.

[0007] As a preferred embodiment of the wind turbine operation status monitoring method of the present invention, the method involves: performing similarity analysis and dimensionality reduction based on high-dimensional feature vectors to obtain the initial positions of the high-dimensional feature vectors in two-dimensional space, calculating the two-dimensional similarity probability between the high-dimensional feature vectors, defining an objective function for iterative updating, and outputting a two-dimensional position set, including: Based on the high-dimensional feature vectors in the high-dimensional feature vector sequence, the distance between the current high-dimensional feature vector and all high-dimensional feature vectors is calculated using Euclidean distance. Then, the standard deviation of the distance is taken, and the similarity between any two high-dimensional feature vectors in the high-dimensional space is calculated by combining the distance. This similarity weight is defined as the similarity weight. The similarity weights of all high-dimensional feature vectors are combined to generate a weight matrix. Then, the probability of each similarity weight is normalized based on the weight matrix to obtain the probability normalized value. Based on high-dimensional feature vectors, principal component analysis is used to obtain the initial two-dimensional position of each high-dimensional feature vector, and based on the initial two-dimensional position, the Gaussian kernel function is used to calculate the two-dimensional similarity probability between each high-dimensional feature vector and all high-dimensional feature vectors. Based on the two-dimensional similarity probability and the obtained probability normalization value, an objective function is defined, and the objective function value is minimized. The objective function value is calculated, and the gradient descent method is used to iteratively update the two-dimensional position of each high-dimensional feature vector. After the maximum number of iterations is reached, the two-dimensional position set is output, which contains the high-dimensional feature vector corresponding to the two-dimensional position.

[0008] As a preferred embodiment of the wind turbine operation status monitoring method of the present invention, the step of constructing a topological node set, calculating the neighborhood relationship and topological energy value of the nodes, generating a topological topographic map for node clustering, and obtaining healthy clusters and abnormal clusters includes: Based on the two-dimensional location set, a topological node set is established. After calculating the distance between each high-dimensional feature vector and all nodes using Euclidean distance, the distances are sorted in descending order. The high-dimensional feature vector corresponding to the minimum distance is assigned to the node. After obtaining the best matching node for each high-dimensional feature vector, a new node set is generated, which contains the best matching node for each high-dimensional feature vector. Based on the new node set, calculate the Euclidean distance between the current node and all nodes, and take the average value as the neighborhood radius. If the Euclidean distance between nodes is less than or equal to the neighborhood radius, then the node is marked as a neighborhood node; otherwise, it is removed. Based on neighboring nodes, a set of neighboring nodes is generated, and then the average distance of the nodes and the contribution of high-dimensional feature vectors to the node density are calculated in parallel. Standardization is performed based on density contribution. After calculating the topological energy value of each node by combining the average distance, interpolation is performed based on the two-dimensional location in the two-dimensional location set to form a topological topographic map. The topographic hierarchy is generated by contour lines and pseudo-color rendering, and then the hierarchy is marked, including blue to represent valley areas and marked as healthy energy areas, and black to represent peak areas and marked as abnormal energy boundary areas. Based on the topological map, the healthy energy region and abnormal energy boundary region in the map are used as the initial clusters. The distance from any node to the two initial clusters is calculated, and the initial cluster corresponding to the minimum distance is selected as the candidate cluster to which the node belongs. By statistically analyzing the distance between nodes and setting an execution threshold, if the minimum distance is less than the execution threshold, the node is merged into the candidate cluster; otherwise, the node is removed. This process is repeated to generate binary clusters, including healthy clusters and abnormal clusters.

[0009] As a preferred embodiment of the wind turbine operating status monitoring method of the present invention, the step of calculating the center and center ratio within a cluster based on healthy clusters and abnormal clusters to obtain the point-level power health index of the wind turbine includes: Based on healthy and abnormal clusters, the average high-dimensional feature vector within the cluster is used as the center. Based on the center, the distance from any node to the center is calculated using Euclidean distance, and then the normalized ratio of the distance is taken as the power health index of the wind turbine at that node.

[0010] As a preferred embodiment of the wind turbine operating status monitoring method of the present invention, the step of further determining the operating status and implementing corresponding measures includes: The mean health index of healthy and abnormal clusters is obtained by statistical method, and the midpoint value of the intersection area is used as the judgment threshold. When the electrical health index of a wind turbine node is greater than or equal to the judgment threshold, it indicates that the node is in a healthy state and the wind turbine is operating stably. Otherwise, it indicates that the node is in an abnormal state, the wind turbine is operating unstablely, or there is a fault or energy imbalance. In this case, a reminder message is sent to the monitoring personnel via wireless transmission technology.

[0011] As a preferred embodiment of the wind turbine operating status monitoring method of the present invention, the step of implementing corresponding measures includes: Based on the power health index, the rule engine maps abnormal states to distribution network risk level signals and generates decision control instructions based on the risk level signals.

[0012] As a preferred embodiment of the wind turbine operation status monitoring method of the present invention, the step of preprocessing the collected power operation parameter data and extracting feature values ​​to concatenate a high-dimensional feature vector includes: Real-time data collection of power operation parameters, including active power, reactive power, power factor, and grid frequency, is obtained from the SCADA system of the wind turbine. Interpolate and standardize electrical energy operation parameter data; The sliding window technique is used to divide the power operation parameter data, and the mean value formula is used to obtain the mean value of each data in the current window, resulting in the mean set of the current window, including the mean value of active power, the mean value of reactive power, the mean value of power factor and the mean value of grid frequency. Based on the mean set, the feature values ​​of each data point are calculated, including the active power deviation ratio, reactive power disturbance coefficient, power factor fluctuation amplitude, and grid coupling fluctuation index. The feature values ​​are concatenated to obtain the high-dimensional feature vector of the current window. The operation is repeated to obtain the high-dimensional feature vector of each window and combine them into a high-dimensional feature vector sequence.

[0013] Secondly, the present invention provides a wind turbine operating status monitoring system, comprising, The data acquisition and extraction module is used to collect power operation parameter data, perform preprocessing, extract feature values, and concatenate high-dimensional feature vectors. The clustering analysis module is used for similarity analysis and dimensionality reduction based on high-dimensional feature vectors. It calculates the two-dimensional similarity probability between high-dimensional feature vectors, defines an objective function for iterative updates, outputs a two-dimensional location set, constructs a topological node set, calculates the neighborhood relationship and topological energy value of the nodes, generates a topological topographic map for node clustering, and obtains healthy clusters and abnormal clusters. The calculation and judgment module is used to calculate the center and center ratio within a cluster based on healthy clusters and abnormal clusters, obtain the point-level power health index of the wind turbine, then determine the operating status and implement corresponding measures.

[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the wind turbine operating status monitoring method as described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the wind turbine operating status monitoring method as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: by calculating the similarity weights between high-dimensional features and performing probability normalization, and by combining the minimization of the Gaussian kernel function and the objective function, feature dimensionality reduction and topological position optimization are achieved. Then, based on the node density contribution and energy distribution, the topological energy value is calculated to form healthy energy zones and abnormal energy boundary zones, thereby realizing dynamic clustering identification of the operating status of wind turbines. Therefore, this invention can achieve high-precision, visual monitoring and early fault warning of the operating status of wind turbines. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the wind turbine operation status monitoring method in Example 1.

[0019] Figure 2 This is a structural diagram of the wind turbine operation status monitoring system in Example 1.

[0020] Figure 3 This is a flowchart of the running status determination in Example 1. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Example 1, referring to Figures 1-3 This is the first embodiment of the present invention, which provides a method for monitoring the operating status of a wind turbine generator, including the following steps: S1. Collect power operation parameter data, preprocess it, extract feature values ​​and concatenate them into a high-dimensional feature vector; Specifically, the collected power operation parameter data undergoes preprocessing, including: Real-time data collection of power operation parameters, including active power, reactive power, power factor, and grid frequency, is obtained from the SCADA system of the wind turbine. Interpolate and standardize electrical energy operation parameter data; The power operation parameter data is divided using the sliding window technique (each window contains data from multiple time points), and the mean value formula is used to obtain the mean value of each data point within the current window, resulting in the set of mean values ​​for the current window, including the mean value of active power, the mean value of reactive power, the mean value of power factor, and the mean value of grid frequency. Based on the mean set, the feature values ​​of each data point are calculated, including the active power deviation ratio, reactive power disturbance coefficient, power factor fluctuation amplitude value and grid coupling fluctuation index. The feature values ​​are concatenated to obtain the high-dimensional feature vector of the current window. The operation is repeated to obtain the high-dimensional feature vector of each window and combine them into a high-dimensional feature vector sequence. The active power deviation ratio is calculated using the following formula: In the formula, Indicates time Power deviation ratio at time Display window The average active power, Indicates wind speed The active power below (can be set according to IEC 61400-12-1 standard). Represents absolute value; The reactive power disturbance coefficient is formulated as follows: In the formula, Indicates at time The reactive power disturbance coefficient at that time. Display window The average reactive power, Indicates at time Reactive power at that time; The formula for the power factor fluctuation amplitude is: In the formula, Indicates at time The power factor fluctuation amplitude value at that time Indicates at time Power factor at time Display window The average power factor; The power grid coupling fluctuation index is formulated as follows: In the formula, Indicates at time Power grid coupling fluctuation index at that time Display window The average frequency of the power grid, Display window The average frequency of the power grid, Indicates the sampling time interval.

[0025] By interpolating and standardizing the electrical parameters collected from the wind turbine's SCADA system, continuous, comparable, and structured processing of operational data was achieved. Real-time SCADA acquisition technology ensures the high timeliness and completeness of multi-dimensional parameters such as active power, reactive power, power factor, and grid frequency, accurately reflecting the wind turbine's power conversion and grid-connected operation status. The interpolation repair method effectively eliminates the discontinuity caused by sensor distortion and data gaps, improving data continuity and repeatability, and providing smooth signal input for subsequent feature extraction. Standardization further unifies the dimensions and numerical ranges of different physical quantities, eliminating scale bias between features and providing a unified metric basis for high-dimensional feature analysis. Through sliding window and mean extraction techniques, time-series smoothing and dynamic feature extraction of wind turbine electrical operation data were achieved, significantly reducing the noise impact of wind speed disturbances and grid fluctuations, and improving data stability and monitoring sensitivity. By constructing a high-dimensional feature vector containing active power deviation ratio, reactive power disturbance coefficient, power factor fluctuation amplitude, and grid coupling fluctuation index, multi-dimensional correlation modeling between electromagnetic energy, mechanical power, and the control system was achieved, revealing the nonlinear coupling relationship in the wind turbine operating state. The calculation process of each feature parameter combines standard power curves and window statistical models, enabling the monitoring results to have physical interpretability and time resolution. A regularization correction term was introduced. This ensures the numerical stability of the algorithm under low-power conditions and enhances its engineering robustness. The resulting high-dimensional feature vector sequence provides structured input for subsequent similarity analysis and energy topological clustering.

[0026] S2. Based on high-dimensional feature vectors, perform similarity analysis and dimensionality reduction to obtain the initial position of the high-dimensional feature vectors in two-dimensional space, calculate the two-dimensional similarity probability between high-dimensional feature vectors, define an objective function for iterative update, output a two-dimensional position set, construct a topological node set, calculate the neighborhood relationship and topological energy value of the nodes, generate a topological terrain map for node clustering, and obtain healthy clusters and abnormal clusters. Specifically, similarity analysis and dimensionality reduction are performed based on high-dimensional feature vectors to obtain the initial positions of the high-dimensional feature vectors in two-dimensional space. The two-dimensional similarity probability between the high-dimensional feature vectors is calculated, an objective function is defined for iterative updates, and a set of two-dimensional positions is output, including: Based on the high-dimensional feature vectors in the high-dimensional feature vector sequence, the distance between the current high-dimensional feature vector and all high-dimensional feature vectors is calculated using Euclidean distance. Then, the standard deviation of the distance is taken, and the similarity between any two high-dimensional feature vectors in the high-dimensional space is calculated by combining the distance. This similarity weight is defined as the similarity weight. The similarity weights of all high-dimensional feature vectors are combined to generate a weight matrix. Then, the probability of each similarity weight is normalized based on the weight matrix to obtain the probability normalized value. The similarity between any two high-dimensional feature vectors calculated by combining distances in the high-dimensional space is defined as the similarity weight, and the formula is as follows: In the formula, Representing high-dimensional feature vectors and Similarity weights between Represents an exponential function. Representing high-dimensional feature vectors and The square of the distance between them Indicates the distance from the standard deviation; The probability normalization of each similarity weight is then performed based on the weight matrix, using the following formula: In the formula, Representing high-dimensional feature vectors and The probability normalization value of the similarity weight between the two. Representing high-dimensional feature vectors and Similarity weights between This represents the total number of high-dimensional feature vectors; Based on high-dimensional feature vectors, principal component analysis is used to obtain the initial two-dimensional position of each high-dimensional feature vector. In the formula, Representing high-dimensional feature vectors Two-dimensional position, and They represent high-dimensional feature vectors respectively. In a two-dimensional plane shaft and The coordinates of the axes are used, and based on the initial two-dimensional position, the two-dimensional similarity probability between each high-dimensional feature vector and all high-dimensional feature vectors is calculated using the Gaussian kernel function, as follows: In the formula, Representing high-dimensional feature vectors and Two-dimensional similarity probability between them Representing high-dimensional feature vectors Two-dimensional position, This represents the average distance between all high-dimensional feature vectors. Representing high-dimensional feature vectors Two-dimensional position, Indicates Euclidean distance; Among them, the average distance between the two-dimensional positions of all high-dimensional feature vectors The formula for obtaining it is: In the formula, This represents the total number of high-dimensional feature vectors; Based on the two-dimensional similarity probability and the obtained probability normalization value, an objective function is defined, and the objective function value is minimized. The formula is: In the formula, Represents the objective function value. Indicates the weights (which can be set via cross-validation); The objective function value is calculated, and the gradient descent method is used to iteratively update the two-dimensional position of each high-dimensional feature vector. After the maximum number of iterations is reached, the two-dimensional position set is output, which contains the high-dimensional feature vector corresponding to the two-dimensional position.

[0027] By modeling and normalizing the similarity between high-dimensional feature vectors, the intrinsic correlation between power operation parameters can be accurately reflected, significantly improving the stability and robustness of data feature representation. Secondly, the high-dimensional to low-dimensional mapping achieved by combining Euclidean distance and Gaussian kernel function effectively maintains the topological consistency of energy features in spatial structure, enabling wind turbine operating states to form clear healthy and abnormal zones in a two-dimensional topological map, enhancing the interpretability and visualization of operating states. Thirdly, through objective function minimization and gradient descent optimization mechanisms, this step enables dynamic adaptive adjustment of the mapping space, maintaining high recognition accuracy even when wind turbine operating conditions change, and possessing real-time learning and updating capabilities. Furthermore, by utilizing distance variance adjustment and probability constraint mechanisms, this invention can adapt to data scale and noise levels under different operating environments, effectively reducing false positives and false negatives.

[0028] Furthermore, a set of topological nodes is constructed, the neighborhood relationships and topological energy values ​​of the nodes are calculated, a topological topographic map is generated, and node clustering is performed to obtain healthy clusters and abnormal clusters, including: Based on the two-dimensional location set, a topological node set is established. After calculating the distance between each high-dimensional feature vector and all nodes using Euclidean distance, the distances are sorted in descending order. The high-dimensional feature vector corresponding to the minimum distance is assigned to the node. After obtaining the best matching node for each high-dimensional feature vector, a new node set is generated, which contains the best matching node for each high-dimensional feature vector. Based on the new node set, calculate the Euclidean distance between the current node and all nodes, and take the average value as the neighborhood radius. If the Euclidean distance between nodes is less than or equal to the neighborhood radius, then the node is marked as a neighborhood node; otherwise, it is removed. Based on neighboring nodes, a set of neighboring nodes is generated, and then the average distance of the nodes and the contribution of the high-dimensional feature vector to the node density are calculated in parallel. The formula is as follows: In the formula, Represents a node average distance, Represents a node The cardinality of the neighborhood set. Indicates belonging to, Represents a node and distance, Representing high-dimensional feature vectors For nodes density contribution, Representing high-dimensional feature vectors With nodes The distance; Standardization is performed based on density contribution. After calculating the topological energy value of each node by combining the average distance, interpolation is performed based on the two-dimensional location in the two-dimensional location set to form a topological topographic map. The topographic hierarchy is generated by contour lines and pseudo-color rendering, and then the hierarchy is marked, including blue to represent valley areas and marked as healthy energy areas, and black to represent peak areas and marked as abnormal energy boundary areas. The formula for calculating the topological energy value of each node is as follows: In the formula, Represents a node The topological energy value, The base of the natural logarithm. This represents the density adjustment coefficient (which can be set after obtaining the standard deviation of density contribution within all nodes using the standard deviation formula; the specific calculation formula is as follows:) , (This represents the standard deviation of density contribution within all nodes). This represents the density contribution after standardization.

[0029] Based on the topological map, the healthy energy region and abnormal energy boundary region in the map are used as the initial clusters. The distance from any node to the two initial clusters is calculated (using Euclidean distance). The initial cluster corresponding to the minimum distance is selected as the candidate cluster to which the node belongs. By statistically analyzing the distance between nodes and setting an execution threshold, if the minimum distance is less than the execution threshold, the node is merged into the candidate cluster; otherwise, the node is removed. This process is repeated to generate binary clusters, including healthy clusters and abnormal clusters.

[0030] By achieving optimal matching between high-dimensional features and topological nodes, a geometric representation of energy characteristics is realized, allowing for a direct visualization of energy distribution features in two-dimensional space. This solves the problems of blurred cluster boundaries and severe feature overlap in traditional methods. Secondly, an adaptive neighborhood radius is used to determine node connectivity, dynamically adjusting the clustering scale in different regions. This effectively avoids over- or under-clustering caused by fixed parameters, ensuring the rationality of the topological structure and the stability of the clustering results. Thirdly, by introducing a density contribution index, the local node density is combined with the high-dimensional energy distribution, significantly improving the distinguishability between healthy and abnormal regions and overcoming the limitations of traditional point density-based clustering. Subsequently, by combining local distance and density normalization factors in topological energy calculation, a continuous and smooth energy potential field is formed, enabling a visual representation of the energy state and facilitating the identification of healthy and abnormal energy boundaries in wind turbine operation. Finally, through an energy gradient-constrained clustering strategy, the node clustering results remain consistent with the actual energy distribution characteristics, improving the accuracy and interpretability of operational status identification.

[0031] S3. Based on healthy and abnormal clusters, calculate the center and center ratio within each cluster to obtain the point-level power health index of the wind turbine, then determine its operating status and implement corresponding measures. Specifically, based on healthy and abnormal clusters, the center and center ratio within each cluster are calculated to obtain the point-level power health index of the wind turbine, including: Based on healthy and abnormal clusters, the average high-dimensional feature vector within each cluster is used as the center, as shown in the formula: In the formula, Indicates the first The center within a cluster, Indicates the first The cardinality of a cluster, Represents a node The corresponding high-dimensional feature vector; Based on the center, the distance from any node to the center is calculated using Euclidean distance. Then, the normalized ratio of the distances is taken as the power health index of the wind turbine at that node. The formula is as follows: In the formula, Indicates the wind turbine at the node The power health index Represents a node To the The distance between the centers of each cluster Represents a node To the The distance between the centers within each cluster.

[0032] By constructing high-dimensional feature centers for healthy and abnormal clusters and calculating the power health index using the Euclidean distance ratio, an adaptive quantitative assessment of the wind turbine's operating status was achieved. This step establishes a geometrically symmetric structure of healthy and abnormal energy patterns in a high-dimensional feature space, significantly improving the accuracy and robustness of status identification. Statistical calculation of the mean values ​​of features within clusters effectively eliminates errors caused by random fluctuations in single nodes, enabling the healthy and abnormal feature centers to possess global representativeness and dynamic update capabilities. The distance ratio normalization mechanism continuously maps the health index to the [0,1] interval, avoiding the limitations of the fixed threshold method, and enabling adaptation to different operating conditions, thus improving sensitivity and generalization.

[0033] Further, the operational status is determined, including: The mean health index of healthy and abnormal clusters is obtained using statistical methods, and the midpoint value of the intersection region is used as the judgment threshold. The formula is as follows: In the formula, Indicates the judgment threshold. Indicates the first Mean health index within each cluster Indicates the first The average health index within each cluster; When the electrical health index of a wind turbine node is greater than or equal to the judgment threshold, it indicates that the node is in a healthy state and the wind turbine is operating stably. Otherwise, it indicates that the node is in an abnormal state, the wind turbine is operating unstablely, or there is a fault or energy imbalance. In this case, a reminder message is sent to the monitoring personnel via wireless transmission technology.

[0034] By introducing a threshold determination mechanism for healthy and abnormal clusters, dynamic identification and health measurement of wind turbine operating status are achieved. This step can self-learn and adjust the judgment criteria based on operating data, and by integrating multi-dimensional features (active power, reactive power, power factor, and grid frequency), the accuracy of anomaly detection and early identification capability are improved. Combined with a wireless communication module, real-time status feedback and remote intelligent early warning are realized, significantly improving the response efficiency of wind turbine operating status monitoring. Furthermore, implement corresponding measures, including: Based on the power health index, the rule engine maps abnormal states to distribution network risk level signals and generates decision control instructions based on the risk level signals.

[0035] The rules in the rule engine can be set by experts, for example: Rule 1: If Output low-risk "Level I"; Rule 2: If The output indicates a medium risk level of "Level II"; Rule 3: If The output is a high-risk "Level III" warning; After the rules engine outputs the risk level, following the principle of "power grid safety first," decision control instructions are generated based on the risk level signal, for example: When the risk level is "Level I", no instructions are generated, but the power operation parameters of the wind turbine are continuously monitored. When the risk level is "Level II" (medium risk), a "parameter fine-tuning instruction" is generated, for example, sending an instruction to the wind turbine main control system to "set an active power limit". When the risk level is "Level III", an "emergency stop command" is sent to the SCADA system; All instructions are executed only after monitoring personnel have confirmed, verified, and set the parameters.

[0036] By introducing a rules engine, abnormal power conditions are mapped to distribution network risk level signals. This enables the dynamic projection of localized wind turbine operational anomalies into grid-level risks, enhancing the overall intelligence of wind farm dispatching. Furthermore, the risk level signals guide the dispatch system to adjust reactive power compensation capacity or reallocate loads in advance, thereby preventing grid oscillations caused by power imbalances. When the risk level reaches a warning threshold, the rules engine triggers automatic control logic, outputting decision-making control commands to achieve preventative maintenance.

[0037] This embodiment also provides a wind turbine operating status monitoring system, including: The data acquisition and extraction module is used to collect power operation parameter data, perform preprocessing, extract feature values, and concatenate high-dimensional feature vectors. The clustering analysis module is used for similarity analysis and dimensionality reduction based on high-dimensional feature vectors. It calculates the two-dimensional similarity probability between high-dimensional feature vectors, defines an objective function for iterative updates, outputs a two-dimensional location set, constructs a topological node set, calculates the neighborhood relationship and topological energy value of the nodes, generates a topological topographic map for node clustering, and obtains healthy clusters and abnormal clusters. The calculation and judgment module is used to calculate the center and center ratio within a cluster based on healthy clusters and abnormal clusters, obtain the point-level power health index of the wind turbine, then determine the operating status and implement corresponding measures.

[0038] This embodiment also provides a computer device applicable to the wind turbine generator operating status monitoring method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the wind turbine generator operating status monitoring method proposed in the above embodiment. The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0039] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the wind turbine operating status monitoring method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0040] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for monitoring the operating status of a wind turbine generator, characterized in that: include, Collect power operation parameter data, preprocess it, extract feature values ​​and concatenate them into a high-dimensional feature vector; Similarity analysis and dimensionality reduction are performed based on high-dimensional feature vectors to obtain the initial position of the high-dimensional feature vectors in two-dimensional space. The two-dimensional similarity probability between high-dimensional feature vectors is calculated, an objective function is defined for iterative update, and a two-dimensional position set is output. A topological node set is constructed, the neighborhood relationship and topological energy value of the nodes are calculated, and a topological topographic map is generated for node clustering to obtain healthy clusters and abnormal clusters. Based on healthy and abnormal clusters, the center and center ratio within each cluster are calculated to obtain the point-level power health index of the wind turbine, and then the operating status is determined and corresponding measures are implemented.

2. The wind turbine generator operating status monitoring method as described in claim 1, characterized in that: The method involves similarity analysis and dimensionality reduction based on high-dimensional feature vectors to obtain the initial positions of the high-dimensional feature vectors in two-dimensional space, calculating the two-dimensional similarity probability between high-dimensional feature vectors, defining an objective function for iterative updates, and outputting a set of two-dimensional positions, including: Based on the high-dimensional feature vectors in the high-dimensional feature vector sequence, the distance between the current high-dimensional feature vector and all high-dimensional feature vectors is calculated using Euclidean distance. Then, the standard deviation of the distance is taken, and the similarity between any two high-dimensional feature vectors in the high-dimensional space is calculated by combining the distance. This similarity weight is defined as the similarity weight. The similarity weights of all high-dimensional feature vectors are combined to generate a weight matrix. Then, the probability of each similarity weight is normalized based on the weight matrix to obtain the probability normalized value. Based on high-dimensional feature vectors, principal component analysis is used to obtain the initial two-dimensional position of each high-dimensional feature vector, and based on the initial two-dimensional position, the Gaussian kernel function is used to calculate the two-dimensional similarity probability between each high-dimensional feature vector and all high-dimensional feature vectors. Based on the two-dimensional similarity probability and the obtained probability normalization value, an objective function is defined, and the objective function value is minimized. The objective function value is calculated, and the gradient descent method is used to iteratively update the two-dimensional position of each high-dimensional feature vector. After the maximum number of iterations is reached, the two-dimensional position set is output, which contains the high-dimensional feature vector corresponding to the two-dimensional position.

3. The wind turbine operating status monitoring method as described in claim 2, characterized in that: The process involves constructing a set of topological nodes, calculating the neighborhood relationships and topological energy values ​​of the nodes, generating a topological topographic map, and performing node clustering to obtain healthy clusters and abnormal clusters, including: Based on the two-dimensional location set, a topological node set is established. After calculating the distance between each high-dimensional feature vector and all nodes using Euclidean distance, the distances are sorted in descending order. The high-dimensional feature vector corresponding to the minimum distance is assigned to the node. After obtaining the best matching node for each high-dimensional feature vector, a new node set is generated, which contains the best matching node for each high-dimensional feature vector. Based on the new node set, calculate the Euclidean distance between the current node and all nodes, and take the average value as the neighborhood radius. If the Euclidean distance between nodes is less than or equal to the neighborhood radius, then the node is marked as a neighborhood node; otherwise, it is removed. Based on neighboring nodes, a set of neighboring nodes is generated, and then the average distance of the nodes and the contribution of high-dimensional feature vectors to the node density are calculated in parallel. Standardization is performed based on density contribution. After calculating the topological energy value of each node by combining the average distance, interpolation is performed based on the two-dimensional location in the two-dimensional location set to form a topological topographic map. The topographic hierarchy is generated by contour lines and pseudo-color rendering, and then the hierarchy is marked, including blue to represent valley areas and marked as healthy energy areas, and black to represent peak areas and marked as abnormal energy boundary areas. Based on the topological map, the healthy energy region and abnormal energy boundary region in the map are used as the initial clusters. The distance from any node to the two initial clusters is calculated, and the initial cluster corresponding to the minimum distance is selected as the candidate cluster to which the node belongs. By statistically analyzing the distance between nodes and setting an execution threshold, if the minimum distance is less than the execution threshold, the node is merged into the candidate cluster; otherwise, the node is removed. This process is repeated to generate binary clusters, including healthy clusters and abnormal clusters.

4. The wind turbine generator operating status monitoring method as described in claim 3, characterized in that: The point-level power health index of the wind turbine is obtained by calculating the center and center ratio within the cluster based on healthy and abnormal clusters, including: Based on healthy and abnormal clusters, the average high-dimensional feature vector within the cluster is used as the center. Based on the center, the distance from any node to the center is calculated using Euclidean distance, and then the normalized ratio of the distance is taken as the power health index of the wind turbine at that node.

5. The wind turbine generator operating status monitoring method as described in claim 4, characterized in that: The subsequent determination of the operating status includes: The mean health index of healthy and abnormal clusters is obtained by statistical method, and the midpoint value of the intersection area is used as the judgment threshold. When the electrical health index of a wind turbine node is greater than or equal to the judgment threshold, it indicates that the node is in a healthy state and the wind turbine is operating stably. Otherwise, it indicates that the node is in an abnormal state, the wind turbine is operating unstablely, or there is a fault or energy imbalance. In this case, a reminder message is sent to the monitoring personnel via wireless transmission technology.

6. The wind turbine generator operating status monitoring method as described in claim 5, characterized in that: The implementation of the corresponding measures includes: Based on the power health index, the rule engine maps abnormal states to distribution network risk level signals and generates decision control instructions based on the risk level signals.

7. The wind turbine generator operating status monitoring method as described in claim 6, characterized in that: The collected power operation parameter data is preprocessed, and feature values ​​are extracted and concatenated into a high-dimensional feature vector, including: Real-time data collection of power operation parameters, including active power, reactive power, power factor, and grid frequency, is obtained from the SCADA system of the wind turbine. Interpolate and standardize electrical energy operation parameter data; The sliding window technique is used to divide the power operation parameter data, and the mean value formula is used to obtain the mean value of each data in the current window, resulting in the mean set of the current window, including the mean value of active power, the mean value of reactive power, the mean value of power factor and the mean value of grid frequency. Based on the mean set, the feature values ​​of each data point are calculated, including the active power deviation ratio, reactive power disturbance coefficient, power factor fluctuation amplitude, and grid coupling fluctuation index. The feature values ​​are concatenated to obtain the high-dimensional feature vector of the current window. The operation is repeated to obtain the high-dimensional feature vector of each window and combine them into a high-dimensional feature vector sequence.

8. A wind turbine operating status monitoring system, based on the wind turbine operating status monitoring method according to any one of claims 1 to 7, characterized in that: include, The data acquisition and extraction module is used to collect power operation parameter data, perform preprocessing, extract feature values, and concatenate high-dimensional feature vectors. The clustering analysis module is used for similarity analysis and dimensionality reduction based on high-dimensional feature vectors. It calculates the two-dimensional similarity probability between high-dimensional feature vectors, defines an objective function for iterative updates, outputs a two-dimensional location set, constructs a topological node set, calculates the neighborhood relationship and topological energy value of the nodes, generates a topological topographic map for node clustering, and obtains healthy clusters and abnormal clusters. The calculation and judgment module is used to calculate the center and center ratio within a cluster based on healthy clusters and abnormal clusters, obtain the point-level power health index of the wind turbine, then determine the operating status and implement corresponding measures.

9. 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 steps of the wind turbine operating status monitoring method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the wind turbine operating status monitoring method according to any one of claims 1 to 7.

Images