A method and system for monitoring faults of a wind power plant based on operational data

Abstract

The application discloses a kind of wind power equipment's fault monitoring method and system based on operation data, first obtain the historical operation data containing fault type, utilize the correlation algorithm of pre-set and quantify the correlation between each operating parameter and fault, and mapping to two-dimensional plane, construct structured point set, enhance the visibility and handleability of multidimensional data.In the key influence factor screening process, select the operating parameter with the largest correlation as the reference point, has clear physical fault characterization significance, can focus on the core parameter highly relevant to the target fault mechanism, to improve the accuracy of candidate factor set from the source, to further screen key influence factor.And accordingly construct exclusive fault monitoring submodel, can ensure that model input has high pertinence, to greatly improve the identification accuracy and classification clarity of each type of fault, improve the fault detection accuracy of wind power equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind power equipment monitoring, and in particular to a wind power equipment fault monitoring method and system based on operation data. BACKGROUND

[0002] Wind power equipment is a comprehensive system that converts wind energy into electrical energy. The core components include a wind wheel (blade), a mechanical energy conversion device (such as a gearbox and bearing), a generator, a support structure (tower), and an intelligent control system. According to the direction of the main shaft, wind power equipment can be divided into horizontal axis wind turbines and vertical axis wind turbines. Horizontal axis wind turbines account for more than 95% of the global market share and are widely used in large-scale onshore and offshore wind farms. Once a fault occurs in wind power equipment, it can have a serious impact on operational safety and economic benefits, including power generation loss, reduced equipment life, increased safety hazards, rising maintenance costs, and environmental pollution risks. For example, gearbox failure can cause the wind turbine to shut down for 7-15 days, resulting in a daily power generation loss of 30-50 thousand yuan. The maintenance cost of offshore wind farms is more than three times that of onshore wind farms. At the same time, equipment failures such as bearing micro-pitting or electrical system overheating can cause fires or structural damage, endangering personal and environmental safety.

[0003] However, in existing fault monitoring methods, hundreds of parameters are involved during the operation of wind power equipment, and the relevance of each parameter to a specific fault is complex and variable. Using artificial experience or full-featured input models results in unclear fault representation and high feature redundancy. At the same time, wind power equipment often experiences multiple faults coexisting (such as the superposition of gear wear and bearing damage), making it difficult for traditional single models to distinguish coupled features, leading to missed detection or misjudgment. SUMMARY

[0004] The present application provides a wind power equipment fault monitoring method and system based on operation data to improve the accuracy of wind power equipment fault monitoring.

[0005] To solve the above technical problems, the present application provides a wind power equipment fault monitoring method based on operation data, comprising:

[0006] Obtaining historical operation data of wind power equipment, the historical operation data including fault types and corresponding fault operation data under the fault types;

[0007] For any fault type, calculating the relevance of each operation parameter in the fault operation data to the fault type based on a pre-set relevance algorithm, and mapping the relevance of each fault type to a two-dimensional plane to generate a point set of the two-dimensional plane;

[0008] In the point set of the two-dimensional plane, the feature vector with the largest correlation degree is determined as a reference point, a candidate circle is generated with the reference point as the center and a preset length threshold as the radius, and all feature vectors in the candidate circle are screened out as a candidate key factor set;

[0009] Based on a preset sliding window, each operating parameter in the candidate key factor set is secondarily verified, and the operating parameter that passes the verification is taken as a key influence factor of the fault type;

[0010] The key influence factor corresponding to each fault type is obtained, and a fault monitoring sub-model is respectively constructed based on the key influence factor;

[0011] Based on the fault monitoring sub-model, each fault working condition is monitored, and the fault monitoring of the wind power equipment is completed.

[0012] Firstly, the historical operation data containing the fault type are obtained, the correlation degree between each operating parameter and the fault is quantified by using a preset correlation algorithm, and is mapped to a two-dimensional plane to construct a structured point set, thereby enhancing the visibility and processability of multi-dimensional data. In the key influence factor screening process, the operating parameter with the largest correlation degree is selected as the reference point, which has a clear physical fault characterization meaning, can focus on the core parameters highly related to the target fault mechanism, and improves the accuracy of the candidate factor set from the source. At the same time, the parameters with low correlation with the fault and susceptible to noise interference are excluded, and the robustness of fault feature extraction is improved. The candidate circle adopts a dynamic radius adjustment mechanism, and the radius size is adaptively adjusted in combination with the data distribution characteristics, so that the balance between noise resistance and generalization ability is achieved, and the risk of misjudgment or omission caused by the traditional fixed threshold is avoided. In addition, the operating parameters present clustering characteristics in the two-dimensional space under the driving of correlation, and the candidate circle can naturally capture the local parameter clustering area, thereby further enhancing the screening efficiency and accuracy. For multiple concurrent fault scenarios, the relevant factors of various faults are in a spatial separation state in the two-dimensional plane, and the fault decoupling identification of key parameters can be realized by generating candidate circles respectively, thereby improving the response ability to fault linkage under complex working conditions. Finally, compared with the traditional global clustering algorithm, the time complexity of the candidate circle strategy is reduced from O(n log n) to O(n), which significantly improves the online processing efficiency and is more suitable for the needs of high-frequency monitoring and real-time early warning of wind power equipment. By extracting the key influence factors most representative of different fault types and constructing exclusive fault monitoring sub-models accordingly, it can be ensured that the model input is highly targeted, thereby greatly improving the identification accuracy and classification clarity of various faults and improving the fault detection accuracy of wind power equipment.

[0013] ​Further, the correlation between each operation parameter and the target fault label is calculated based on a preset correlation algorithm for any fault type, and the correlation of each operation parameter is mapped to a two-dimensional plane to generate a point set of the two-dimensional plane, including:

[0014] A joint probability model is constructed based on each operation parameter in the fault operation data;

[0015] The correlation between each operation parameter and the target fault label is evaluated based on a maximum correlation entropy criterion and the joint probability model;

[0016] The correlation between each operation parameter and the target fault label is mapped to a two-dimensional plane based on a preset dimension reduction algorithm to generate a point set representing the correlation between each operation parameter and the target fault label.

[0017] The present application accurately quantifies the correlation between operation parameters and target fault labels by using a joint probability model and a maximum correlation entropy criterion, and maps the correlation result to a two-dimensional plane by a dimension reduction algorithm to form a visual point set. On the one hand, the joint probability model can depict the statistical dependence relationship between multiple parameters, and the maximum correlation entropy criterion ensures that the selected parameters and fault labels have the maximum information gain; on the other hand, the dimension reduction mapping can intuitively present the correlation strength between each parameter and the fault in the two-dimensional space, providing a more reliable mathematical foundation and visual support for subsequent key factor screening, thereby greatly improving the accuracy and interpretability of preliminary feature extraction.

[0018] Further, the feature vector with the maximum correlation is determined as a reference point, the distribution characteristics of the point set are calculated based on the reference point, and a screening threshold is determined based on the distribution characteristics, including:

[0019] The feature vector with the maximum correlation is determined as a reference point, and the Euclidean distance distribution of the point set is calculated based on the reference point;

[0020] The standard deviation of the point set is calculated based on the Euclidean distance distribution, and the product of a preset scale factor and the standard deviation is taken as the screening threshold.

[0021] The present application determines the feature vector with the maximum correlation as a reference point, and calculates the Euclidean distance distribution of the point set based on the reference point. The screening threshold is adaptively generated by the product of the standard deviation and the preset scale factor. It can dynamically adjust the screening radius according to the actual "convergence" degree of the data, taking into account noise resistance and coverage, and can avoid the problem of loose or tight fixed threshold, accurately capturing high correlation parameters and naturally excluding noise parameters, significantly improving the robustness and universality of key factor screening.

[0022] Further, the preset sliding window is used for secondary verification on each operation parameter in the candidate key factor set, and an operation parameter passing the verification is taken as a key influence factor of the fault type, and the method comprises the following steps of:

[0023] Based on the candidate key factor set, a candidate parameter sequence of each candidate key factor is screened out from the historical operation data.

[0024] Based on a preset window length and a preset step, overlapping sliding windows are generated on each candidate parameter sequence.

[0025] For any candidate parameter sequence, a local correlation value of the candidate key factor in the sliding window and a target fault label is calculated, a mean value and a standard deviation of the local correlation values of all sliding windows on the candidate parameter sequence are counted, when the mean value is greater than or equal to a preset minimum correlation threshold value and the standard deviation is less than or equal to a preset maximum fluctuation threshold value, it is determined that the candidate key factor passes the verification, and the candidate key factor is taken as a key influence factor of the fault type.

[0026] The sliding window is introduced for secondary verification on the candidate key factor after the candidate key factor is screened out, the local correlation in multiple overlapping windows is calculated, the mean value and the standard deviation are counted, the minimum correlation threshold value and the maximum fluctuation threshold value are set for strict determination, the false correlation parameter with short-term occurrence or large fluctuation can be eliminated, only the influence factor with stable correlation can be reserved, and therefore the false positive and false negative risks can be reduced, and the time sequence stability and the diagnosis reliability of the fault influence factor are enhanced.

[0027] Further, the key influence factor corresponding to each fault type is acquired, and a fault monitoring sub-model is constructed based on the key influence factor, and the method comprises the following steps of:

[0028] The key influence factor corresponding to each fault type is acquired, the time sequence characteristics of the key influence factor are extracted from the corresponding fault operation data, and a multi-dimensional feature vector is constructed.

[0029] A target prediction algorithm is matched based on the multi-dimensional feature vector, and an initial fault monitoring sub-model is constructed based on the target prediction algorithm.

[0030] Training data is generated based on the multi-dimensional feature vector, the initial fault monitoring sub-model is trained based on the training data, and a fault monitoring sub-model is acquired.

[0031] The application extracts the time sequence characteristics of the key influence factors passing the secondary check for each type of fault, constructs a multi-dimensional feature vector, and constructs an initial monitoring sub-model based on a target prediction algorithm matching the fault mechanism, so that each sub-model takes the most representative features as input, ensuring the specificity and diagnostic accuracy of the model, and laying a highly targeted foundation for subsequent model training and optimization, improving the coverage and recognition performance of multiple fault types.

[0032] Further, the training data is generated based on the multi-dimensional feature vector, and the initial fault monitoring sub-model is trained based on the training data to obtain a fault monitoring sub-model, including:

[0033] The multi-dimensional feature vector is cut into fixed-length sequence segments in a sliding window manner according to time sequence, and each sequence segment is labeled with a corresponding fault type or normal working condition label to generate training data;

[0034] The training data is divided into a training subset and a validation subset according to a predetermined ratio, and the initial fault monitoring sub-model is trained based on the training subset and the validation subset to obtain a fault monitoring sub-model.

[0035] The multi-dimensional feature vector is cut into fixed-length sequence segments in a sliding window manner and labeled, and then the training set and the validation set are divided for model training. The use of time sequence slices enhances sample diversity, effectively alleviating the problems of fault sample scarcity and class imbalance; at the same time, the separation of the training / verification set and the cross-validation mechanism further ensure the stability of the obtained fault monitoring sub-model.

[0036] Further, the fault monitoring sub-model is used to monitor each fault condition respectively to complete the fault monitoring of the wind power equipment, including:

[0037] The running data of the key influence factors corresponding to each fault type is collected in real time, and the running data is denoised, interpolated and normalized for pretreatment to generate an input vector for each fault detection sub-model;

[0038] The input vector is input into the corresponding fault monitoring sub-model to obtain a prediction score of the fault type;

[0039] The fault type is alarmed based on the prediction score and a preset alarm threshold to complete the fault monitoring of the wind power equipment.

[0040] The application collects and pre-processes key influence factor data in real time, inputs the key influence factor data into respective fault monitoring sub-models for online prediction, triggers an alarm based on a prediction score and a preset threshold, supports logical fusion and dynamic incremental updating of multiple model outputs, thereby constructing a full-process closed loop from data collection, pre-processing, model reasoning to alarm output, which can not only realize real-time and accurate monitoring of various faults, but also ensure that the system continuously adapts to environmental changes and fault evolution during operation through dynamic adjustment and incremental training, and comprehensively improves the timeliness of early warning and operation and maintenance efficiency of wind power equipment.

[0041] In a second aspect, the application provides a wind power equipment fault monitoring system based on operation data, comprising a data collection module, a correlation calculation module, a distribution determination module, a candidate factor screening module, a key factor screening module, a model construction module and a fault detection module.

[0042] The data collection module is configured to obtain historical operation data of the wind power equipment, wherein the historical operation data comprises a fault type and corresponding fault operation data under the fault type.

[0043] The correlation calculation module is configured to calculate, for any fault type, a correlation between each operation parameter in the fault operation data and a target fault label based on a preset correlation algorithm, and map the correlation of each operation parameter to a two-dimensional plane to generate a point set of the two-dimensional plane.

[0044] The distribution determination module is configured to determine a feature vector with the largest correlation as a reference point, calculate a distribution feature of the point set based on the reference point, and determine a screening threshold based on the distribution feature.

[0045] The candidate factor screening module is configured to generate a candidate circle with the reference point as the center and the screening threshold as the radius, and screen all feature vectors within the candidate circle as a candidate key factor set.

[0046] The key factor screening module is configured to perform secondary verification on each operation parameter in the candidate key factor set based on a preset sliding window, and take the operation parameters that pass the verification as key influence factors of the fault type.

[0047] The model construction module is configured to obtain key influence factors corresponding to each fault type, and construct fault monitoring sub-models based on the key influence factors.

[0048] The fault detection module is configured to monitor each fault condition based on the fault monitoring sub-models to complete fault monitoring of the wind power equipment.

[0049] Further, the distribution determination module is configured to determine a feature vector with the largest correlation degree as a reference point, calculate a distribution feature of the point set based on the reference point, and determine a screening threshold based on the distribution feature, including:

[0050] determining a feature vector with the largest correlation degree as a reference point, and calculating a Euclidean distance distribution of the point set based on the reference point;

[0051] calculating a standard deviation of the point set based on the Euclidean distance distribution, and taking a product of a preset scale factor and the standard deviation as the screening threshold.

[0052] Further, the candidate factor screening module is configured to perform secondary verification on each operation parameter in the candidate key factor set based on a preset sliding window, and take an operation parameter that passes the verification as a key impact factor of the fault type, including:

[0053] screening a candidate parameter sequence of each candidate key factor from the candidate key factor set in the historical operation data;

[0054] generating overlapping sliding windows on each candidate parameter sequence based on a preset window length and a preset step length;

[0055] for any candidate parameter sequence, calculating a local correlation degree value of the candidate key factor and the target fault label in the sliding window, and calculating a mean value and a standard deviation of the local correlation degree values of all sliding windows on the candidate parameter sequence; when the mean value is greater than or equal to a preset minimum correlation degree threshold and the standard deviation is less than or equal to a preset maximum fluctuation threshold, determining that the candidate key factor passes the verification, and taking it as a key impact factor of the fault type. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 a flowchart of a fault monitoring method of a wind power equipment based on operation data provided by an embodiment of the present application;

[0057] Figure 2 a structural diagram of a fault monitoring system of a wind power equipment based on operation data. DETAILED DESCRIPTION

[0058] The specific embodiments of the present application will be further described in detail below with reference to the accompanying drawings and embodiments. The following embodiments are used to illustrate the present application, but are not used to limit the scope of the present application.

[0059] The terms "first" and "second" and the like in the description and in the claims of the present application and in the drawings of the application are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. Furthermore, the terms "comprising", "including", "containing", and "having" and their conjugates, as used herein, are inclusive and are intended to permit the specification of features that would otherwise be excluded in some manner by virtue of their being limiting. For example, a process, method, article, or apparatus that comprises a list of steps or elements is not necessarily limited to the listed steps or elements but can include additional or other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

[0060] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. It is expressly understood that any of the embodiments described herein can be incorporated in to other embodiments.

[0061] Embodiment 1

[0062] Reference is made to Figure 1 , Figure 1 A flowchart of a fault monitoring method for wind power equipment based on operation data is provided for an embodiment of the application. The embodiment of the application provides a fault monitoring method for wind power equipment based on operation data, which comprises steps 101 to 107, and specifically as follows:

[0063] Step 101: Obtain historical operation data of the wind power equipment, which includes fault types and corresponding fault operation data under the fault types;

[0064] In this embodiment, during the daily operation of the wind farm, the operation parameters of the wind turbine are collected in real time by devices such as the SCADA system, the vibration monitoring system, the oil analyzer, and the temperature sensor and stored in the historical database. The historical operation data at least includes fault type labels and fault operation data under the corresponding labels, and the fault operation data includes different operation parameters.

[0065] In this embodiment, every time a shutdown maintenance or automatic alarm event occurs, the operation and maintenance system generates corresponding fault type labels such as "gearbox gear wear", "bearing early crack", "generator stator winding overheat", etc. according to the fault phenomenon and the maintenance report. The fault type labels can be confirmed by manual review or an automatic classification module based on the fault knowledge base to ensure the accuracy and consistency of the labels.

[0066] In the embodiment, for each fault type, a plurality of operating parameter sequences of the unit are collected and recorded within a preset time window (e.g. 24 hours before and after the fault) before and after the fault, which include but are not limited to: mechanical vibration signals (peak acceleration, spectral energy distribution, 0.5 times meshing frequency sideband amplitude, etc.); oil monitoring indicators (oil viscosity, water content, metal debris concentration, etc.); temperature data (gearbox oil temperature, bearing temperature, generator stator winding temperature, etc.); electrical parameters (stator current harmonic content, power factor, generator power output, etc.); environment and operating conditions (wind speed, wind direction, unit speed, yaw angle, etc.).

[0067] In the embodiment, each operating parameter is continuously recorded at a fixed sampling frequency (e.g. 1 Hz or 10 Hz) and is marked with a timestamp to form complete time series data.

[0068] In the embodiment, the synchronous storage of the fault type label and the fault operating data can obtain a structured historical operating data table. Each record contains fields such as "fault type", "collection time period", "operating parameter name", and "parameter value sequence", which lays a data foundation for subsequent correlation calculation and feature extraction based on the correlation algorithm.

[0069] Step 102: For any fault type, calculate the correlation between each operating parameter in the fault operating data and the target fault label based on a preset correlation algorithm, and map the correlation of each operating parameter to a two-dimensional plane to generate a point set of the two-dimensional plane;

[0070] In the embodiment, the calculation of the correlation between each operating parameter in the fault operating data and the target fault label based on a preset correlation algorithm and the mapping of the correlation of each operating parameter to a two-dimensional plane to generate a point set of the two-dimensional plane for any fault type include:

[0071] Construct a joint probability model based on each operating parameter in the fault operating data;

[0072] Evaluate the correlation between each operating parameter and the target fault label based on the maximum correlation entropy criterion and the joint probability model;

[0073] Map the correlation between each operating parameter and the target fault label to a two-dimensional plane based on a preset dimension reduction algorithm to generate a point set representing the correlation between each operating parameter and the target fault label.

[0074] In the embodiment, taking the "gearbox gear wear" fault type of a wind turbine as an example, the system first extracts a multi-source sensor data sequence containing a fault label from a historical operation database, and the data includes several (M) operating parameters such as cabin temperature, gearbox oil temperature, 0.5 times meshing frequency sideband amplitude in the main shaft vibration envelope spectrum, lubricating oil metal particle concentration, etc. For any fault type, the joint probability model is constructed by using kernel density estimation (KDE) or multivariate Gaussian distribution assumption with the operating parameters and the fault label as inputs , and the marginal distribution is calculated ; then based on the maximum correlation entropy criterion, the correlation between each operating parameter and the fault label is quantified by using the mutual information formula.

[0075] In the embodiment, the mutual information formula is:

[0076] (1)

[0077] In the embodiment, after the obtained M-dimensional mutual information vector is processed by dimension reduction algorithms such as principal component analysis (PCA) or t-SNE, a visualization distribution map composed of a point set can be obtained in a two-dimensional plane, where the spatial position of each coordinate point directly reflects the correlation strength of the parameter to the fault type and the joint feature aggregation of the parameter with other parameters.

[0078] In the embodiment, the given quantity not only describes the nonlinear statistical dependence, but also provides a clear geometric basis for subsequent reference point selection and candidate circle screening through dimension reduction mapping.

[0079] In the embodiment, the correlation between the operating parameters and the target fault label is accurately quantified by using the joint probability model and the maximum correlation entropy criterion, and the correlation result is mapped to a two-dimensional plane by using the dimension reduction algorithm to form a visualization point set. On the one hand, the joint probability model can describe the statistical dependence relationship among multiple parameters, and the maximum correlation entropy criterion ensures that the selected parameters have the maximum information gain with the fault label; on the other hand, the dimension reduction mapping can intuitively present the correlation strength of each parameter to the fault in a two-dimensional space, and provide a more reliable mathematical basis and visualization support for subsequent key factor screening, thereby greatly improving the accuracy and interpretability of preliminary feature extraction.

[0080] Step 103: determining a feature vector with the maximum correlation as a reference point, calculating distribution features of the point set based on the reference point, and determining a screening threshold based on the distribution features;

[0081] In the embodiment, the feature vector with the largest correlation degree is determined as a reference point, distribution characteristics of the point set are calculated based on the reference point, and a screening threshold is determined based on the distribution characteristics, including:

[0082] The feature vector with the largest correlation degree is determined as a reference point, and the Euclidean distance distribution of the point set is counted based on the reference point.

[0083] The standard deviation of the point set is calculated based on the Euclidean distance distribution, and the product of a preset scale factor and the standard deviation is taken as the screening threshold.

[0084] In the embodiment, after the two-dimensional plane point set mapping is completed, the point with the largest mutual information value or other correlation degree metric in the point set is first identified as a reference point. Then, the Euclidean distances of all other points in the point set to the reference point are calculated, and the distribution characteristics of the obtained distance values are counted to determine the standard deviation σ of the Euclidean distance. Finally, a scale factor k is preset according to experience or engineering requirements, and the product r = k·σ of the scale factor and the standard deviation is taken as the screening threshold.

[0085] In the embodiment, the threshold value can be adaptively adjusted according to the overall "concentration" degree of the point set, and can ensure that the candidate circle can cover parameters highly related to the target fault mechanism and effectively exclude discrete noise points, thereby providing reliable geometric and statistical basis for subsequent key influence factor screening based on the candidate circle.

[0086] In the embodiment, the feature vector with the largest correlation degree is determined as a reference point, the Euclidean distance distribution of the point set is counted based on the reference point, and the product of the standard deviation and a preset scale factor is adaptively generated as a screening threshold. The screening radius can be dynamically adjusted according to the actual "concentration" degree of the data, and both noise resistance and coverage are considered. In addition, the problem of loose or tight caused by a fixed threshold value can be avoided, the high correlation degree parameters can be accurately captured, and the noise parameters can be naturally excluded, thereby significantly improving the robustness and universality of the key factor screening.

[0087] Step 104: A candidate circle is generated with the reference point as the center and the screening threshold as the radius, and all feature vectors within the candidate circle are screened out as a candidate key factor set.

[0088] In this embodiment, after the two-dimensional plane point set mapping and threshold calculation are completed, a candidate circle is constructed on the two-dimensional plane with the reference point corresponding to the maximum mutual information as the center and the threshold r=k·σ as the radius. Specifically, each feature vector coordinate in the mapped point set is traversed, and the Euclidean distance between the feature vector coordinate and the reference point is calculated. When the distance is less than or equal to the preset radius r, it is considered that the corresponding operating parameter has sufficient statistical correlation and physical representation significance with the target fault label, and thus the feature vector corresponding to the parameter is included in the candidate key factor set.

[0089] In this embodiment, through the candidate circle screening, a group of operating parameters highly related to the core fault mechanism can be effectively covered, and at the same time, discrete points with weak fault correlation or large noise influence are automatically excluded, thereby providing a refined and information-rich candidate factor set for subsequent sliding window secondary verification and sub-model construction.

[0090] Step 105: performing secondary verification on each operating parameter in the candidate key factor set based on a preset sliding window, and taking the operating parameter passing the verification as a key influence factor of the fault type;

[0091] In this embodiment, the secondary verification on each operating parameter in the candidate key factor set based on a preset sliding window, and taking the operating parameter passing the verification as a key influence factor of the fault type, includes:

[0092] filtering out, based on the candidate key factor set, a candidate parameter sequence of each candidate key factor in the historical operation data;

[0093] generating, based on a preset window length and a preset step length, overlapping sliding windows on each candidate parameter sequence;

[0094] For any candidate parameter sequence, calculating a local correlation value of the candidate key factor in the sliding window and the target fault label; calculating a mean value and a standard deviation of the local correlation values of all sliding windows on the candidate parameter sequence; when the mean value is greater than or equal to a preset minimum correlation threshold and the standard deviation is less than or equal to a preset maximum fluctuation threshold, determining that the candidate key factor passes the verification, and taking the candidate key factor as a key influence factor of the fault type.

[0095] In this embodiment, after the preliminary screening of the candidate key factor is completed, the continuous time sequence value sequence of each candidate key factor is extracted from the historical operation data, which is denoted as a candidate parameter sequence. Subsequently, based on a preset sliding window length L (for example, 100 sampling points) and a step length Δ (for example, 20 sampling points), a plurality of overlapping sliding windows are generated on each candidate parameter sequence in turn, and each window corresponds to a subsequence of adjacent operation data.

[0096] In the embodiment, for each sliding window, the local correlation value between the candidate key factor and the target fault label in the window is calculated using the same correlation algorithm as the preliminary screening, and the sequence of local correlation values of all windows is summarized. Then, the mean μ and the standard deviation σ of the sequence are calculated. When and only when μ ≥ τ_min (τ_min is a preset minimum correlation threshold, such as 0.6) and σ ≤ τ_max (τ_max is a preset maximum fluctuation threshold, such as 0.1), it is determined that the correlation between the candidate factor and the fault label in time is both strong and stable, that is, the check passes, and finally it is confirmed as the key influencing factor of the fault type.

[0097] In the embodiment, the sliding window secondary verification process can eliminate incidental noise interference or short-term abnormal correlation, and ensure that the selected key factor has persistent and reliable fault indication ability in different running stages.

[0098] In the embodiment, after the candidate key factor is screened, a sliding window is introduced for secondary verification. By calculating the local correlation in multiple overlapping windows and calculating the mean and standard deviation, setting a minimum correlation threshold and a maximum fluctuation threshold for strict determination, short-term incidental or fluctuation pseudo-correlation parameters can be eliminated, and only the influencing factors with persistent and stable correlation are retained, thereby significantly reducing the false positive and false negative risk, and enhancing the time stability and diagnostic reliability of the fault influencing factor.

[0099] Step 106: Obtain the key influencing factor corresponding to each fault type, and construct a fault monitoring sub-model based on the key influencing factor;

[0100] In the embodiment, the key influencing factor corresponding to each fault type is obtained, and a fault monitoring sub-model is constructed based on the key influencing factor, which includes:

[0101] Obtain the key influencing factor corresponding to each fault type, extract the time sequence characteristics of the key influencing factor in the corresponding fault operation data, and construct a multi-dimensional feature vector;

[0102] Match a target prediction algorithm based on the multi-dimensional feature vector, and construct an initial fault monitoring sub-model based on the target prediction algorithm;

[0103] Generate training data based on the multi-dimensional feature vector, and train the initial fault monitoring sub-model based on the training data to obtain a fault monitoring sub-model.

[0104] In this embodiment, taking the "gearbox gear wear" fault type as an example, first, according to the key influence factor list obtained by the foregoing sliding window verification, such as the 0.5 times meshing frequency sideband amplitude in the main shaft vibration envelope spectrum, the lubricating oil metal particle concentration and the gear box oil temperature M' term parameter, the continuous time sequence of each parameter is extracted in the corresponding fault operation data with a unified sampling frequency; then, the time sequence of each key influence factor is cut into sliding window segments of the same length, and the statistical features (such as mean, variance, maximum, minimum) and frequency domain features (such as power spectrum density peak value, harmonic content) of each segment sequence are calculated, and the obtained features are spliced in a predetermined order to construct a multi-dimensional feature vector representing the fault condition . According to the dimension of the feature vector and the nature of the fault mode, the system automatically matches the applicable target prediction algorithm, such as support vector machine (SVM) for small sample high-dimensional features, or long short-term memory network (LSTM) for time series dependent features, and accordingly constructs an initial fault monitoring sub-model. Then, in the same sliding window manner, the historical multi-dimensional feature vector sequence and its corresponding fault label are packaged to form a training sample set, and are divided into a training set and a validation set according to a certain proportion, and the selected prediction algorithm is used to iteratively train on the training set, and finally the fault monitoring sub-model corresponding to each fault type is obtained, which is used for real-time fault identification and early warning.

[0105] In this embodiment, based on the multi-dimensional feature vector of each fault type, a modeling algorithm suitable for the fault mechanism is selected or customized, including but not limited to: a support vector machine (SVM) based classification algorithm for small sample high-dimensional feature fault identification; a long short-term memory network (LSTM) based time series prediction algorithm for capturing fault development trend; a Gaussian mixture model (GMM) or isolation forest based on anomaly detection for anomaly alarm in unsupervised environment.

[0106] In this embodiment, by extracting the time sequence features of the key influence factors that pass the secondary verification for each type of fault, a multi-dimensional feature vector is constructed, and an initial monitoring sub-model is constructed based on the target prediction algorithm matching the fault mechanism, so that each sub-model takes the most representative features as input, which not only ensures the specificity and diagnostic accuracy of the model, but also lays a highly targeted foundation for subsequent model training and optimization, and improves the coverage ability and recognition performance for multiple fault types.

[0107] In this embodiment, the training data is generated based on the multi-dimensional feature vector, and the initial fault monitoring sub-model is trained based on the training data to obtain a fault monitoring sub-model, including:

[0108] The multi-dimensional feature vector is cut into fixed-length sequence segments in a sliding window manner in chronological order, and each sequence segment is labeled with a corresponding fault type or normal condition label to generate training data;

[0109] The training data is divided into a training subset and a validation subset according to a preset ratio, and the initial fault monitoring sub-model is trained based on the training subset and the validation subset to obtain a fault monitoring sub-model.

[0110] In this embodiment, for the extracted multi-dimensional feature vector sequence , first, the multi-dimensional feature vector sequence is cut into several overlapping sequence segments in a sliding window manner on the time axis with a fixed length L and a step size Each sequence segment contains L multi-dimensional feature vectors of adjacent time points. Then, each sequence segment is labeled with a fault type or a "normal" label according to the corresponding fault occurrence time or normal operation time interval, thereby constructing an input-output paired training data set. Next, the training data is randomly divided into a training subset and a validation subset according to a preset ratio (for example, 70% training and 30% validation), and iterative learning (such as gradient descent, Adam optimization) is performed on the training subset with the initial fault monitoring sub-model as the skeleton, while the model performance indicators (such as accuracy, recall rate, loss function value) are periodically evaluated on the validation subset. By adjusting the learning rate or early stopping strategy, the model performance on the validation set is stable and convergent, and finally a fault monitoring sub-model with high recognition accuracy and generalization ability is obtained.

[0111] In this embodiment, the multi-dimensional feature vector is cut into fixed-length sequence segments in a sliding window manner and labeled, and then the training set and the validation set are divided for model training. The use of time series slicing enhances sample diversity, effectively alleviating the problems of scarce fault samples and class imbalance; at the same time, the separation and cross-validation mechanism of the training / verification set ensures the stability of the obtained fault monitoring sub-model in this embodiment.

[0112] Step 107: monitoring each fault condition based on the fault monitoring sub-model to complete the fault monitoring of the wind power equipment.

[0113] In this embodiment, the fault monitoring of the wind power equipment based on the fault monitoring sub-model includes:

[0114] Real-time acquisition of the operating data of the key influence factors corresponding to each fault type, and denoising, interpolation and normalization preprocessing of the operating data to generate input vectors for each fault detection sub-model;

[0115] input the input vector into a corresponding fault monitoring sub-model to obtain a prediction score of a fault type;

[0116] alarm the fault type based on the prediction score and a preset alarm threshold, and complete the fault monitoring of the wind power equipment.

[0117] In the embodiment, during the online operation stage of the wind turbine, the system collects real-time time series signals of key influence factors corresponding to each fault type through a high-speed data bus, such as the vibration amplitude of the main shaft, the oil temperature of the gearbox, and the metal particle concentration of the oil. Then, for the collected original data, first, high-frequency noise is removed by using wavelet denoising or low-pass filtering algorithm, and then missing data points caused by sensor packet loss or network jitter are repaired by using linear interpolation or spline interpolation. Then, the key factor data is normalized according to the mean and standard deviation in the training stage to generate an input vector with consistent numerical distribution and dimensionless. The input vector is sent to the monitoring sub-model of the corresponding fault type in real time, and the model outputs a prediction score reflecting the severity of the fault. When the score is greater than or equal to the pre-labeled alarm threshold, the alarm signal of the corresponding fault type is triggered and pushed to the monitoring center through the operation and maintenance platform, realizing online identification and early warning of the fault of the wind power equipment.

[0118] In the embodiment, in the multi-fault concurrent scenario, the alarm signals of each sub-model are logically ORed or weightedly fused to comprehensively generate the overall fault alarm state of the wind power equipment.

[0119] In the embodiment, according to the subsequent repair results and new fault samples, the sub-models triggered by the alarm are periodically or online incrementally trained and parameter-adjusted to maintain the monitoring accuracy and adaptability;

[0120] In the embodiment, the key influence factor data is collected and preprocessed in real time, input into each fault monitoring sub-model for online prediction, and an alarm is triggered based on the prediction score and the preset threshold, while supporting logical fusion and dynamic incremental update of multiple model outputs, thereby constructing a full-process closed loop from data collection, preprocessing, model inference to alarm output. Not only can the real-time and accurate monitoring of various faults be realized, but also the system can continuously adapt to environmental changes and fault evolution during operation through dynamic adjustment and incremental training, thereby comprehensively improving the timeliness of early warning and operation and maintenance efficiency of the wind power equipment.

[0121] In this embodiment, historical operational data containing fault types is first acquired. A pre-defined correlation algorithm is used to quantify the correlation between each operational parameter and the fault, and this data is mapped to a two-dimensional plane to construct a structured point set, enhancing the visibility and processability of the multi-dimensional data. During the screening of key influencing factors, the operational parameter with the highest correlation is selected as the benchmark point. This parameter has clear physical fault characterization significance and allows focus on core parameters highly correlated with the target fault mechanism, improving the accuracy of the candidate factor set from the source. Simultaneously, parameters with low correlation to faults and susceptible to noise interference are excluded, improving the robustness of fault feature extraction. The construction of the candidate circle employs a dynamic radius adjustment mechanism, adaptively adjusting the radius size based on data distribution characteristics. This achieves a balance between noise resistance and generalization ability, avoiding the risk of misjudgment or missed judgment caused by traditional fixed thresholds. Furthermore, operational parameters exhibit clustering characteristics in two-dimensional space driven by correlation. The candidate circle can naturally capture local parameter clustering regions, thereby enhancing the screening efficiency and accuracy in this embodiment. For scenarios with multiple concurrent faults, various fault-related factors are spatially separated in a two-dimensional plane. By generating candidate circles separately, fault decoupling identification of key parameters can be achieved, improving the response capability to fault linkage under complex operating conditions. Finally, compared with traditional global clustering algorithms, this candidate circle strategy only needs to calculate the distance between the point and the circle center, reducing the time complexity by […]. The computational complexity has been reduced to O(n), significantly improving online processing efficiency and better meeting the needs of high-frequency monitoring and real-time early warning for wind power equipment. By extracting the most representative key influencing factors for different fault types and constructing dedicated fault monitoring sub-models accordingly, the model input can be ensured to be highly targeted, thereby greatly improving the accuracy of fault identification and classification clarity, and enhancing the fault detection accuracy of wind power equipment.

[0122] Please refer to Figure 2 , Figure 2 A schematic diagram of a fault monitoring system for wind power equipment based on operational data is provided, including: a data acquisition module 201, a correlation calculation module 202, a distribution determination module 203, a candidate factor screening module 204, a key factor screening module 205, a model building module 206, and a fault detection module 207.

[0123] The data acquisition module 201 is used to acquire historical operating data of wind power equipment, the historical operating data including fault types and corresponding fault operating data under the fault types;

[0124] The correlation calculation module 202 is used to calculate the correlation between each operating parameter in the fault operation data and the target fault label based on a preset correlation algorithm for any fault type, and to map the correlation of each operating parameter to a two-dimensional plane to generate a point set of the two-dimensional plane.

[0125] The distribution determining module 203 is configured to determine a feature vector with the largest correlation degree as a reference point, calculate a distribution feature of the point set based on the reference point, and determine a screening threshold based on the distribution feature.

[0126] The candidate factor screening module 204 is configured to generate a candidate circle with the reference point as the center and the screening threshold as the radius, and screen all feature vectors within the candidate circle as a candidate key factor set.

[0127] The key factor screening module 205 is configured to perform secondary verification on each operating parameter in the candidate key factor set based on a preset sliding window, and take an operating parameter that passes the verification as a key influence factor of the fault type.

[0128] The model constructing module 206 is configured to obtain key influence factors corresponding to each fault type, and construct a fault monitoring sub-model based on the key influence factors.

[0129] The fault detecting module 207 is configured to monitor each fault condition based on the fault monitoring sub-model, and complete fault monitoring of the wind power equipment.

[0130] In this embodiment, the correlation calculating module is configured to, for any fault type, calculate the correlation degree between each operating parameter in the fault operating data and a target fault label based on a preset relevance algorithm, map the correlation degrees of the operating parameters to a two-dimensional plane, and generate a point set of the two-dimensional plane, including:

[0131] Constructing a joint probability model based on each operating parameter in the fault operating data;

[0132] Evaluating the correlation degree between each operating parameter and the target fault label based on a maximum correlation entropy criterion and the joint probability model;

[0133] Mapping the correlation degree between each operating parameter and the target fault label to a two-dimensional plane based on a preset dimension reduction algorithm, and generating a point set representing the correlation between each operating parameter and the target fault label.

[0134] In this embodiment, the distribution determining module is configured to determine a feature vector with the largest correlation degree as a reference point, calculate a distribution feature of the point set based on the reference point, and determine a screening threshold based on the distribution feature, including:

[0135] Determining a feature vector with the largest correlation degree as a reference point, and statistically calculating the Euclidean distance distribution of the point set based on the reference point;

[0136] Calculating the standard deviation of the point set based on the Euclidean distance distribution, and taking the product of a preset proportion factor and the standard deviation as the screening threshold.

[0137] In the embodiment, the key factor screening module is configured to perform secondary verification on each operation parameter in the candidate key factor set based on a preset sliding window, and take an operation parameter that passes the verification as a key impact factor of the fault type, and includes the following steps.

[0138] Screening a candidate parameter sequence of each candidate key factor from the candidate key factor set in the historical operation data.

[0139] Generating overlapping sliding windows on each candidate parameter sequence based on a preset window length and a preset step length.

[0140] For any candidate parameter sequence, calculating a local correlation value of the candidate key factor in the sliding window and the target fault label; calculating a mean value and a standard deviation of the local correlation values of all sliding windows on the candidate parameter sequence; when the mean value is greater than or equal to a preset minimum correlation threshold value and the standard deviation is less than or equal to a preset maximum fluctuation threshold value, determining that the candidate key factor passes the verification, and taking it as a key impact factor of the fault type.

[0141] In the embodiment, the model construction module is configured to obtain the key impact factor corresponding to each fault type, and construct a fault monitoring sub-model based on the key impact factor, and includes the following steps.

[0142] Obtaining the key impact factor corresponding to each fault type, extracting a time sequence feature of the key impact factor in the corresponding fault operation data, and constructing a multi-dimensional feature vector.

[0143] Matching a target prediction algorithm based on the multi-dimensional feature vector, and constructing an initial fault monitoring sub-model based on the target prediction algorithm.

[0144] Generating training data based on the multi-dimensional feature vector, and training the initial fault monitoring sub-model based on the training data to obtain a fault monitoring sub-model.

[0145] In the embodiment, the model construction module is configured to generate training data based on the multi-dimensional feature vector, and train the initial fault monitoring sub-model based on the training data to obtain a fault monitoring sub-model, and includes the following steps.

[0146] Dividing the multi-dimensional feature vector into fixed-length sequence segments in a sliding window manner in chronological order, labeling each sequence segment with a corresponding fault type or normal working condition label, and generating training data.

[0147] Dividing the training data into a training subset and a validation subset according to a preset proportion, and training the initial fault monitoring sub-model based on the training subset and the validation subset to obtain a fault monitoring sub-model.

[0148] In the embodiment, the fault detection module is configured to monitor each fault condition based on the fault monitoring sub-models, thereby completing the fault monitoring of the wind power equipment.

[0149] The running data of the key influence factors corresponding to each fault type is collected in real time, and the running data is subjected to denoising processing, interpolation processing and normalization preprocessing, thereby generating the input vectors of each fault detection sub-model.

[0150] The input vectors are input into the corresponding fault monitoring sub-models, thereby obtaining the predicted scores of the fault types.

[0151] The fault types are alarmed based on the predicted scores and the preset alarm threshold, thereby completing the fault monitoring of the wind power equipment.

[0152] In the embodiment, a terminal device is also provided, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the above-mentioned fault monitoring method of the wind power equipment based on the running data when executing the computer program.

[0153] In the embodiment, a computer readable storage medium is also provided, which includes a stored computer program, and when the computer program is running, the computer readable storage medium controls the device where the computer readable storage medium is located to execute the above-mentioned fault monitoring method of the wind power equipment based on the running data.

[0154] For example, the computer program can be divided into one or more modules, and the one or more modules are stored in the memory and executed by the processor to complete the present application. The one or more modules can be a series of computer program instructions capable of completing a specific function, which are used to describe the execution process of the computer program in the terminal device.

[0155] The terminal device can be a desktop computer, a notebook computer, a palm computer, a cloud server and other computing devices. The terminal device can include, but is not limited to, a processor, a memory, a display. Those skilled in the art can understand that the above components are only examples of the terminal device and do not constitute a limitation on the terminal device, and the terminal device can include more or fewer components than the components, or combine certain components, or different components, for example, the terminal device can also include an input / output device, a network access device, a bus, etc.

[0156] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is a control center of the terminal device, and connects all parts of the terminal device through various interfaces and lines.

[0157] The memory can be used to store computer programs and / or modules, and the processor realizes various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, a character conversion function, etc.), etc.; and the data storage area can store data created according to the use of the terminal device (such as audio data, character message data, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a nonvolatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory device.

[0158] If the module for monitoring the fault of the wind power equipment based on the operation data is implemented in the form of a software function unit and sold or used as an independent product, the module can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, an executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. Those of ordinary skill in the art can understand and implement it without creative labor.

[0159] The above-described specific embodiments further illustrate the purposes, technical solutions, and beneficial effects of the present application. It should be understood that the above-described specific embodiments are merely examples of the present application and are not intended to limit the protection scope of the present application. It is particularly pointed out that any modification, equivalent replacement, improvement, etc. made by those of ordinary skill in the art within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for failure monitoring of a wind power plant based on operational data, characterized in that, The method comprises the following steps: obtaining historical operation data of a wind power device, wherein the historical operation data comprises a fault type and corresponding fault operation data under the fault type; for any fault type, calculating the correlation degree of each operation parameter in the fault operation data with a target fault label based on a preset correlation algorithm, and mapping the correlation degree of each operation parameter to a two-dimensional plane to generate a point set of the two-dimensional plane; determining a feature vector with the maximum correlation degree as a reference point, calculating the distribution characteristics of the point set based on the reference point, and determining a screening threshold based on the distribution characteristics; generating a candidate circle with the reference point as the center and the screening threshold as the radius, and screening all feature vectors within the candidate circle as a candidate key factor set; based on a preset sliding window, performing secondary verification on each operation parameter in the candidate key factor set, and taking the operation parameters that pass the verification as key impact factors of the fault type; obtaining the key impact factors corresponding to each fault type, and respectively constructing a fault monitoring sub-model based on the key impact factors; respectively monitoring each fault condition based on the fault monitoring sub-model to complete the fault monitoring of the wind power device.

2. A method for monitoring faults in wind power equipment based on operational data according to claim 1, characterized in that, The method comprises the following steps: constructing a joint probability model based on each operation parameter in the fault operation data; evaluating the correlation degree between each operation parameter and the target fault label based on the maximum correlation entropy criterion and the joint probability model; mapping the correlation degree between each operation parameter and the target fault label to a two-dimensional plane based on a preset dimension reduction algorithm to generate a point set representing the correlation between each operation parameter and the target fault label.

3. A method for monitoring faults in wind power equipment based on operational data according to claim 2, characterized in that, The method comprises the following steps: determining a feature vector with the maximum correlation degree as a reference point, and calculating the distribution characteristics of the point set based on the reference point; calculating the standard deviation of the point set based on the Euclidean distance distribution, and taking the product of a preset proportion factor and the standard deviation as a screening threshold.

4. A method of fault monitoring of a wind power plant based on operational data according to claim 3, characterized in that, The method comprises the following steps: screening the candidate parameter sequence of each candidate key factor in the historical operation data based on the candidate key factor set; generating overlapping sliding windows on each candidate parameter sequence based on a preset window length and a preset step length; For any candidate parameter sequence, the local correlation value of the candidate key factor in the sliding window and the target fault label is calculated; the mean and standard deviation of the local correlation values of all sliding windows on the candidate parameter sequence are counted; when the mean is greater than or equal to a preset minimum correlation threshold, and the standard deviation is less than or equal to a preset maximum fluctuation threshold, it is determined that the candidate key factor passes the verification, and it is taken as a key influencing factor of the fault type.

5. A method of fault monitoring of a wind power plant based on operational data according to claim 4, characterized in that, The key influencing factor corresponding to each fault type is obtained, and a fault monitoring sub-model is constructed based on the key influencing factor, comprising: Obtaining the key influencing factor corresponding to each fault type, extracting the time sequence characteristics of the key influencing factor in the corresponding fault operation data, and constructing a multi-dimensional feature vector; Based on the multi-dimensional feature vector, a target prediction algorithm is matched, and an initial fault monitoring sub-model is constructed based on the target prediction algorithm; Based on the multi-dimensional feature vector, training data is generated, and the initial fault monitoring sub-model is trained based on the training data to obtain a fault monitoring sub-model.

6. A method of fault monitoring of a wind power plant based on operational data according to claim 5, characterized in that, The training data is generated by cutting the multi-dimensional feature vector into fixed-length sequence segments in a sliding window manner in time sequence, and each sequence segment is labeled with a corresponding fault type or normal working condition label, and the training data is generated; The training data is divided into a training subset and a validation subset according to a preset proportion, and the initial fault monitoring sub-model is trained based on the training subset and the validation subset to obtain a fault monitoring sub-model. The fault monitoring sub-model is used to monitor each fault condition respectively to complete the fault monitoring of the wind power equipment, comprising:

7. A method of fault monitoring of a wind power plant based on operational data according to claim 6, characterized in that, Real-time acquisition of the operation data of the key influencing factor corresponding to each fault type, and denoising, interpolation and normalization preprocessing of the operation data to generate an input vector of each fault detection sub-model; The input vector is input into the corresponding fault monitoring sub-model to obtain a prediction score of the fault type; Based on the prediction score and a preset alarm threshold, the fault type is alarmed to complete the fault monitoring of the wind power equipment. Comprising:

8. A fault monitoring system for wind power plants based on operational data, characterized in that Data acquisition module, correlation calculation module, distribution determination module, candidate factor screening module, key factor screening module, model construction module and fault detection module; The data acquisition module is used to acquire the historical operation data of the wind power equipment, and the historical operation data includes fault types and corresponding fault operation data under the fault types; The correlation calculation module is used to calculate the correlation of each operation parameter in the fault operation data and the target fault label based on a preset correlation algorithm for any fault type, and map the correlation of each operation parameter to a two-dimensional plane to generate a point set of the two-dimensional plane; The distribution determination module is used to determine a reference point as a feature vector with the maximum correlation, calculate the distribution characteristics of the point set based on the reference point, and determine a screening threshold based on the distribution characteristics; ​ The candidate factor screening module is configured to generate a candidate circle with the reference point as the center and the screening threshold as the radius, and screen all feature vectors within the candidate circle as a candidate key factor set; The key factor screening module is configured to perform secondary verification on each operating parameter in the candidate key factor set based on a preset sliding window, and determine an operating parameter that passes the verification as a key influence factor of the fault type; The model construction module is configured to obtain key influence factors corresponding to each fault type, and construct a fault monitoring sub-model based on the key influence factors. The fault detection module is configured to monitor each fault condition based on the fault monitoring sub-model, and complete fault monitoring of the wind power equipment.

9. A wind power plant failure monitoring system based on operational data as claimed in claim 8, wherein, The distribution determination module is configured to determine a feature vector with the largest correlation degree as a reference point, calculate distribution characteristics of the point set based on the reference point, and determine a screening threshold based on the distribution characteristics, including: determining a feature vector with the largest correlation degree as a reference point, and calculating a Euclidean distance distribution of the point set based on the reference point; calculating a standard deviation of the point set based on the Euclidean distance distribution, and determining a product of a preset proportion factor and the standard deviation as the screening threshold.

10. A wind power plant failure monitoring system based on operational data as claimed in claim 9, wherein, The candidate factor screening module is configured to perform secondary verification on each operating parameter in the candidate key factor set based on a preset sliding window, and determine an operating parameter that passes the verification as a key influence factor of the fault type, including: screening a candidate parameter sequence of each candidate key factor in the historical operation data based on the candidate key factor set; generating overlapping sliding windows on each candidate parameter sequence based on a preset window length and a preset step length; for any candidate parameter sequence, calculating a local correlation degree value of the candidate key factor in the sliding window and a target fault label, and calculating a mean value and a standard deviation of the local correlation degree values of all sliding windows on the candidate parameter sequence; when the mean value is greater than or equal to a preset minimum correlation degree threshold and the standard deviation is less than or equal to a preset maximum fluctuation threshold, determining that the candidate key factor passes the verification, and determining the candidate key factor as the key influence factor of the fault type.

Images