Wind turbine fault diagnosis method and system based on mechanism data fusion

Abstract

The invention belongs to the technical field of wind turbine fault diagnosis. According to the mechanism data fusion-based wind turbine fault diagnosis method and system, a feature sequence, a time sequence feature, a residual error and a differential residual error are spliced and fused to obtain a fused feature vector; linearly mapping the fused feature vector to a high-dimensional hidden space, performing matrix multiplication among the features in the high-dimensional hidden space to obtain similarity distribution among the features, and normalizing to obtain an adaptive adjacent matrix; and carrying out matrix multiplication on the adaptive adjacency matrix and the features in the high-dimensional hidden space again, carrying out normalization to obtain a coupling feature matrix, adding the adaptive adjacency matrix and the coupling feature matrix to obtain an interactive feature vector, and obtaining a wind turbine fault diagnosis result according to the interactive feature vector. According to the method, adaptive modeling of different feature importance is realized, and misjudgment caused by environmental noise and working condition fluctuation is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind turbine fault diagnosis, and particularly relates to a wind turbine fault diagnosis method and system based on mechanism data fusion. BACKGROUND

[0002] The statements in this section merely provide background technology related to the present application and do not necessarily constitute prior art.

[0003] Since the wind turbine is an all-weather electromechanical integrated device, the main components of the wind turbine set, such as the blade, the gear box and the generator, bear alternating loads and various extreme working conditions for a long time, which is prone to a series of faults such as bearing damage, insulation aging and blade cracking. Therefore, it is necessary to perform fault diagnosis of the wind turbine. Traditional fault diagnosis methods, such as vibration analysis method and signal processing method, usually rely on the experience of experts and the extraction of artificial features, which makes these methods have obvious limitations in practical application. On the one hand, artificial feature extraction is difficult to fully capture the running characteristics of complex wind turbine sets under varying working conditions, and the diagnosis result is affected by subjectivity; on the other hand, the data source of a single signal is not enough to fully reflect the overall health status of the wind turbine set, and in the actual running process, environmental noise, load fluctuation and working condition switching and other factors will cause the characteristic signal to be blurred, thereby reducing the accuracy and reliability of the diagnosis.

[0004] In recent years, with the improvement of the intelligent level of the wind turbine set, the data-driven fault diagnosis method based on the Supervisory Control And Data Acquisition (SCADA) system has gradually become a research hotspot. The SCADA system can long-term and continuously collect running parameters such as wind speed, rotating speed, current, voltage, power and temperature, which provides rich resources for fault diagnosis. On this basis, deep learning methods are widely introduced, and through the construction of convolutional neural network, recurrent neural network or time series graph model, automatic identification and classification of the running state of the wind turbine set are realized. To some extent, this kind of method alleviates the limitations of artificial feature extraction, so that the model can learn the potential law directly from a large amount of historical data.

[0005] However, existing researches are mostly focused on modeling the time and space characteristics of SCADA data, that is, to realize fault diagnosis by capturing the time sequence patterns of signals or the spatial correlation between different sensor characteristics. Although such methods have made significant progress, there are still some deficiencies: the "space-time" modeling relying only on statistical correlation is easy to ignore the complex physical coupling relationship inside the wind turbine, lacks the combination with mechanism constraints, and leads to insufficient physical interpretability of the diagnosis results; under complex working conditions and multi-source interference, the simple data-driven model is easy to have insufficient generalization ability and misjudgment; some model structures are complex and have large number of parameters, which are strongly dependent on the number of samples and computing resources, leading to difficulties in deployment and real-time performance in actual wind farm applications. SUMMARY

[0006] In order to solve the problems of the prior art, the application provides a wind turbine fault diagnosis method and system based on mechanism data fusion, which introduces a power mechanism model into a deep learning model, and performs residual and differential residual calculation with the actual monitoring power in the form of residual signals. The method can not only rely on data-driven learning characteristics, but also combine mechanism constraints for anomaly detection. In the feature modeling process, the relationship graph between features is constructed through dynamic graph feature interaction, realizing adaptive modeling of different feature importance and reducing misjudgment caused by environmental noise and working condition fluctuations.

[0007] In order to achieve the above purpose, the application adopts the following technical solutions: In a first aspect, the application provides a wind turbine fault diagnosis method based on mechanism data fusion.

[0008] A wind turbine fault diagnosis method based on mechanism data fusion includes the following processes: The operation condition data of the wind turbine are preprocessed, and the preprocessing includes sliding window processing and standardization processing; The preprocessed feature sequence is time series extracted, and the time series features are obtained through a pre-trained neural network model; A power mechanism model is constructed using the preprocessed data, the theoretical power is determined according to the power mechanism model, and the residual and differential residual of the theoretical power and the real power are calculated; The feature sequence, time series feature, residual and differential residual are spliced and fused to obtain a fusion feature vector; The fusion feature vector is linearly mapped to a high-dimensional hidden space, the features in the high-dimensional hidden space are subjected to matrix multiplication operation between features, the similarity distribution between features is obtained, and the adaptive adjacency matrix is obtained after normalization; The adaptive adjacency matrix and the features in the high-dimensional hidden space are subjected to matrix multiplication operation again between the features, and a coupling feature matrix is obtained after normalization; the adaptive adjacency matrix and the coupling feature matrix are added to obtain an interaction feature vector, and a wind turbine fault diagnosis result is obtained according to the interaction feature vector.

[0009] In a second aspect, the present application provides a wind turbine fault diagnosis system based on mechanism data fusion.

[0010] A wind turbine fault diagnosis system based on mechanism data fusion comprises: A data preprocessing unit configured to preprocess the operating condition data of the wind turbine, the preprocessing comprising sliding window processing and standardization processing; A time sequence feature extraction unit configured to extract time sequence features from the preprocessed feature sequence by using a pre-trained neural network model; A mechanism data processing unit configured to construct a power mechanism model using the preprocessed data, determine a theoretical power according to the power mechanism model, and calculate residual error and differential residual error between the theoretical power and the real power; A multi-dimensional feature fusion unit configured to splice and fuse the feature sequence, the time sequence features, the residual error and the differential residual error to obtain a fusion feature vector; An adjacency matrix generation unit configured to linearly map the fusion feature vector to a high-dimensional hidden space, perform matrix multiplication operation between the features in the high-dimensional hidden space to obtain a similarity distribution between the features, and obtain an adaptive adjacency matrix after normalization; A diagnosis result generation unit configured to perform matrix multiplication operation again between the adaptive adjacency matrix and the features in the high-dimensional hidden space, obtain a coupling feature matrix after normalization, add the adaptive adjacency matrix and the coupling feature matrix to obtain an interaction feature vector, and obtain a wind turbine fault diagnosis result according to the interaction feature vector.

[0011] In a third aspect, the present application provides a computer device comprising a processor and a computer readable storage medium. The processor is adapted to execute a computer program. The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the wind turbine fault diagnosis method based on mechanism data fusion of the first aspect of the present application.

[0012] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, and the computer program is adapted to be loaded and executed by the processor to implement the wind turbine fault diagnosis method based on mechanism data fusion of the first aspect of the present application.

[0013] In the fifth aspect, the present application provides a computer program product, which comprises a computer program, and the computer program, when executed by a processor, implements the wind turbine fault diagnosis method based on mechanism data fusion of the first aspect of the present application.

[0014] Compared with the prior art, the present application has the following beneficial effects: The present application effectively combines data-driven feature learning and mechanism constraint anomaly detection by introducing a power mechanism model and calculating residual error and differential residual error with actual monitoring power, solves the problem of insufficient physical interpretability in traditional methods, uses dynamic graph feature interaction to construct a feature relationship graph, realizes adaptive modeling of different feature importance, greatly reduces misjudgment caused by environmental noise and working condition fluctuations, and overcomes the defects of insufficient generalization ability of previous methods under variable working conditions and noise interference, adopts a lightweight design strategy, significantly reduces training and reasoning time under the premise of ensuring diagnostic accuracy, solves the problem of large model structure of existing models that is difficult to meet the real-time application requirements of wind farms, and overall improves the stability, robustness and real-time performance of wind turbine fault diagnosis.

[0015] In view of the problem that most existing deep learning methods rely on statistical feature spatiotemporal modeling and are difficult to depict the complex dynamics mechanism inside the wind turbine, the present application introduces a power mechanism model in the network structure, calculates the residual error and differential residual error of the theoretical power and the measured power, forms information input with physical constraints, and ensures that the diagnostic result can reflect the true mechanism characteristics; in view of the problem that models relying purely on data-driven have insufficient generalization ability under variable working conditions and noise interference environment and are prone to misjudgment, the present application uses LSTM (Long Short-Term Memory, long short-term memory neural network) to model the time series evolution law, and combines DGI to construct a dynamic relationship graph between features, adaptively highlights key features, and suppresses redundant information, thereby improving the stability and robustness of diagnosis; in view of the problem that the existing model structure is large and the training and reasoning time is long, which is difficult to meet the real-time application requirements of wind farms, the PIA network (Physics-Integrated Analysis Network, mechanism fusion analysis network) proposed in the present application adopts a lightweight design strategy, can maintain the shallow structure of LSTM, and replaces the complex attention mechanism with DGI (Dynamic Graph Interaction, dynamic graph feature interaction), which significantly reduces the training and reasoning time under the premise of ensuring the diagnostic accuracy.

[0016] The operating condition data of the application is derived from the SCADA system, covering key parameters such as wind speed and rotating speed, ensuring the comprehensiveness and real-time of the data source, providing rich and accurate information basis for fault diagnosis, and through the use of these multi-dimensional and high-quality data, the running state change of the wind turbine can be more accurately captured, thereby improving the sensitivity and specificity of fault diagnosis, effectively solving the misdiagnosis and missed diagnosis problems caused by incomplete or inaccurate data; the specific form of the power mechanism model is defined, including parameters such as cut-in wind speed and full-load wind speed, and the output power characteristics of the wind turbine are described by an accurate physical model, so that the fault diagnosis not only relies on data-driven pattern recognition, but also combines physical laws for in-depth analysis, enhancing the physical interpretability of the diagnosis result and improving the accuracy of fault identification under complex working conditions.

[0017] The calculation method of the theoretical power of the application considers the influence of temperature, introduces the ambient temperature variable to make the power model more close to the actual operating conditions, improves the accuracy of the theoretical power calculation, and helps to more accurately capture the power fluctuation caused by environmental changes, thereby improving the sensitivity and reliability of fault diagnosis; the application defines the calculation method of residual error and differential residual error, quantifies the deviation between the theoretical power and the actual power, provides an intuitive and effective abnormal detection index for fault diagnosis, helps to quickly locate potential faults, reduces misjudgment and omission, and improves the diagnosis efficiency.

[0018] The application describes the process of obtaining the wind turbine fault diagnosis result according to the interaction feature vector, comprehensively integrates multi-dimensional feature information, and uses linear layers and activation modules for in-depth analysis, realizes accurate evaluation of the health state of the wind turbine, not only improves the accuracy of fault diagnosis, but also enhances the robustness of the model, so that it can maintain stable diagnosis performance in complex and variable operating environments.

[0019] The advantages of the additional aspects of the application will be partially given in the following description, partially will become obvious from the following description, or will be known by the practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0020] The drawings accompanying the specification of the application form part of the application and serve to provide further understanding of the application, the illustrative embodiments of the application and their description serve to explain the application without constituting an inappropriate limitation thereof.

[0021] Figure 1 The architecture diagram of the mechanism fusion analysis network provided for an exemplary embodiment of the application is shown in the figure; Figure 2 The power schematic diagram of the power overrun of the gear box oil temperature control system provided for an exemplary embodiment of the application is shown in the figure, wherein, Figure 2 (A) in the figure is a schematic diagram of the relationship between active power and environmental wind speed, Figure 2(B) of (A) is a schematic diagram of the relationship between the power difference and the ambient wind speed; Figure 3 The power diagram provided by an exemplary embodiment of the present application is a short-time blade pitch rate average power diagram, wherein, Figure 3 (A) of (A) is a schematic diagram of the relationship between the active power and the ambient wind speed, Figure 3 (B) of (A) is a schematic diagram of the relationship between the power difference and the ambient wind speed; Figure 4 The power diagram provided by an exemplary embodiment of the present application is a generator speed and impeller speed mismatch power diagram, wherein, Figure 4 (A) of (A) is a schematic diagram of the relationship between the active power and the ambient wind speed, Figure 4 (B) of (A) is a schematic diagram of the relationship between the power difference and the ambient wind speed; Figure 5 The power diagram provided by an exemplary embodiment of the present application is a cabin vibration fault power diagram, wherein, Figure 5 (A) of (A) is a schematic diagram of the relationship between the active power and the ambient wind speed, Figure 5 (B) of (A) is a schematic diagram of the relationship between the power difference and the ambient wind speed; Figure 6 The principle diagram provided by an exemplary embodiment of the present application is a dynamic graph feature interaction principle diagram; Figure 7 The schematic diagram provided by an exemplary embodiment of the present application is a comparative model loss curve; Figure 8 The schematic diagram provided by an exemplary embodiment of the present application is a confusion matrix Figure 1 , wherein, Figure 8 (A) of (A) is a confusion matrix diagram of the XGBoost model; Figure 8 (B) of (A) is a confusion matrix diagram of the Light GBM model; Figure 8 (C) of (A) is a confusion matrix diagram of the HGB model; Figure 8 (D) of (A) is a confusion matrix diagram of the CNN model; Figure 8 (E) of (A) is a confusion matrix diagram of the CNN-LSTM model; Figure 8 (F) of (A) is a confusion matrix diagram of the TCN model; Figure 9 The schematic diagram provided by an exemplary embodiment of the present application is a confusion matrix Figure 2 , wherein, Figure 9 (A) of (A) is a confusion matrix diagram of the RNN model; Figure 9 (B) of (A) is a confusion matrix diagram of the GCN model; Figure 9 (C) of (A) is a confusion matrix diagram of the EfficientNet model; Figure 9(D) in the diagram is a schematic representation of the confusion matrix of the Transformer model; Figure 9 (E) in the diagram is a schematic representation of the confusion matrix of the Swin-Transformer model; Figure 9 (F) in the diagram is a schematic representation of the confusion matrix of the NODE model; Figure 10 A confusion matrix illustration provided for an exemplary embodiment of the present invention Figure 3 ,in, Figure 10 (A) in the diagram is a schematic diagram of the confusion matrix of the Mamba model; Figure 10 (B) in the diagram is a schematic diagram of the confusion matrix of the ResNet model; Figure 10 (C) in the diagram is a schematic diagram of the confusion matrix of the PIA model; Figure 11 A schematic diagram of a wind turbine fault diagnosis system based on mechanism data fusion, provided as an exemplary embodiment of the present invention; Figure 12 A schematic diagram of a computer device provided for an exemplary embodiment of the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0024] This implementation proposes a wind turbine fault diagnosis method based on mechanism-data fusion. A physical mechanism model of segmented wind turbine power is introduced into the deep learning model, and residual and differential residual calculations are performed between the residual signal and the actual monitored power. This allows the model to not only learn features based on data but also detect anomalies by incorporating mechanism constraints. During feature modeling, an LSTM structure is used to extract temporal dynamic features, and a relationship graph between features is constructed through a Dynamic Graph Interaction (DGI) module. This enables adaptive modeling of the importance of different features, thereby reducing misjudgments caused by environmental noise and operating condition fluctuations. To balance accuracy and practical application requirements, a lightweight approach is adopted in the overall model design, proposing a Physics-Integrated Analysis Network (PIA).

[0025] like Figure 1 As shown, the wind turbine fault diagnosis method based on mechanism data fusion of the present invention includes the following process: Step S101: In the data processing stage, the data processing of each SCADA sensor is performed using a sliding window process; Step S102: In the temporal feature extraction stage, the data after the sliding window is extracted temporally, and temporal features are obtained through the LSTM network; Step S103: In the physical constraint construction stage, the mechanism power mechanism model is constructed using the data after sliding windowing, the theoretical power is calculated, and the residual and difference residual between the theoretical power and the actual power are obtained; Step S104: In the spatial matrix generation stage, the eigenvalues ​​obtained from the temporal features and physical constraints are fused into vectors for linear mapping and “graphical” interaction to obtain the correlation between each eigenvector, while suppressing redundant correlations.

[0026] In step S101 of the present invention, the data preprocessing includes sliding window construction and normalization transformation, specifically including: First, considering the temporal characteristics of the monitored signals, a sliding window mechanism is used for sample partitioning. Let the original signal sequence be as shown in equation (1): (1); in, For the length of time, For feature dimension, These represent the signal values ​​in the original signal sequence, Represents the first in the original signal sequence Each signal value It represents the real number field.

[0027] By setting the window length and step length The sequence can be divided into several training samples, as shown in equation (2): (2); in, Indicates the first Samples from one window, Representing the first The, the The, ..., the One window; The running status label at the end of the window. This approach, which uses labels, not only preserves the dynamic evolution of signals within local time periods but also enhances the number of samples and improves the robustness of temporal feature learning by constructing overlapping windows.

[0028] Secondly, considering the differences in dimensions and numerical ranges of features collected by different sensors, this invention employs standardization processing to map each feature to a unified numerical space, for any feature component. Its standardized form is as shown in equation (3): (3); in, and Features of the training set The mean and standard deviation, The transformation represents the total number of feature components, and the processed features satisfy equations (4) and (5): (4); (5); in, represent Expectations represent The variance.

[0029] During the testing phase, the data obtained during the training phase will still be used. and Normalization is performed to ensure the consistency of training and testing data distribution and avoid information leakage.

[0030] In step S102 of the present invention, after completing the preprocessing of the feature data, the dynamic features of the time series are further modeled to capture the dependence of the wind turbine's operating state between consecutive time points. A Long Short Term Memory (LSTM) network is used as the time series modeling unit to process the samples constructed by the sliding window.

[0031] Specifically, assuming the first window samples If equation (2) is satisfied, then the input corresponding to the LSTM is the length. The feature sequence is used as the feature sequence, and the output is the hidden state sequence at each time step. The hidden state at the final moment Used as a temporal characteristic representation for this window.

[0032] Inside the LSTM unit, the state update is jointly determined by the gating mechanism of equations (6)–(11): (6); (7); (8); (9); (10); (11); in, respectively represent input gate, forget gate and output gate, is a Sigmoid function, is an element-wise multiplication; represents an input vector at a time step; represents an input vector at a time step; represents a hidden state at a time step; represents a hidden state at a time step; represents a hidden state at a time step; represents a memory cell state at a time step; represents a memory cell state at a time step; represents a memory cell state at a time step; represents a memory cell state at a time step; respectively represent input weight matrices; and respectively represent recurrent weight matrices; and respectively represent bias vectors; tanh(•) represents a hyperbolic tangent function.

[0033] Through the above structure, the LSTM can keep key information in a long time span and forget unimportant historical signals, thereby effectively extracting the time sequence dependency in the sequence. The final obtained time sequence feature representation of the window is as shown in formula (12): (12); The above time sequence feature representation can comprehensively reflect the dynamic pattern of the input sequence in the time dimension within the window, and provide time sequence feature support for the subsequent fusion of mechanism residual signals and spatial interaction modules.

[0034] In step S103 of the application, according to the operation mechanism of the wind turbine, the output power of the wind turbine shows obvious segmented characteristics in different wind speed intervals. Specifically, a segmented function as shown in formula (13) is used to describe the mechanism power mechanism model: (13); wherein, are respectively the cut-in wind speed, the full-load wind speed and the cut-out wind speed, which are respectively 3 m / s, 10.8 m / s and 25 m / s in the wind turbine; is the rated power of the wind turbine, which is 2050 kW; is the air density;​​​​​​​ is the swept area of the wind wheel, is the wind speed; is the power coefficient, usually ranging from 0 to 0.59.

[0035] From a physical point of view: when the wind speed is lower than the cut-in wind speed, the kinetic energy of the airflow is not enough to drive the impeller to rotate, the unit does not generate electricity, and the power is zero; when the wind speed is higher than the cut-out wind speed, the unit is automatically shut down to protect the equipment, and the power is zero; between the full-load wind speed and the cut-out wind speed, the power is limited to the rated value ; while between the cut-in wind speed and the full-load wind speed, the power increases with the cube of the wind speed, and this interval is the most sensitive area of the difference between the mechanism model power and the actual power.

[0036] As shown in Figure 2 , Figure 3 , Figure 4 and Figure 5 , the power curves of normal working conditions and fault working conditions are less different in the low wind speed section and the full-load wind speed section; while between the cut-in wind speed and the full-load wind speed, the power curve of the fault working condition is obviously different from that of the normal working condition, reflecting the influence of different types of faults on the energy conversion efficiency of the wind turbine, and this difference is consistent with the core interval of the mechanism segmented model, so through the residual modeling of the mechanism constraint, the fault sensitivity of this interval can be significantly improved.

[0037] More specifically, Figure 2 (A) shows the change of active power of the wind turbine under different environmental wind speeds, it can be seen that between the cut-in wind speed and the full-load wind speed, the power increases significantly with the increase of wind speed, and after reaching the full-load wind speed, the power remains stable; Figure 2 (B) shows the difference between actual power and theoretical power with the change of environmental wind speed, under the fault condition of power overrun of the gear box oil temperature control system, the power difference of some wind speed sections is obviously large, indicating that the fault causes abnormal power output; Figure 3 (A) shows the normal output of active power of the wind turbine under different wind speed conditions; Figure 3 (B) reflects the change of the difference between actual power and theoretical power when the average of short-time blade pitch rate is too low, under the fault state, the power difference of the low wind speed section is particularly significant, indicating that the low pitch rate affects the power capture; Figure 4 (A) depicts the trend of active power of the wind turbine with the change of wind speed under normal working conditions; Figure 4 (B) shows the change of power difference when the generator speed and the impeller speed do not match, under the fault condition, the power difference of multiple wind speed sections exceeds the normal range, indicating that the speed mismatch seriously affects the power conversion efficiency.

[0038] According to formula (13), the output power between the cut-in wind speed and the full-load wind speed is obtained Further, the swept area is obtained by formula (14): (14); wherein, is the wind wheel radius.

[0039] The air density is obtained by formula (15): (15); wherein, is the gas constant, is the Kelvin temperature, is the atmospheric pressure.

[0040] The atmospheric pressure varies with the altitude, as formula (16): (16); wherein, is the standard atmospheric pressure at sea level; is the temperature decrement rate; is the altitude; is the standard temperature at sea level, is the gravitational acceleration; is the molar mass of air.

[0041] According to formula (13) to formula (16), the wind turbine at a fixed location has only the temperature and the wind speed as variables, and the rest are constants, so the theoretical power of the wind turbine between the cut-in wind speed and the full-load wind speed can be obtained, as shown in formula (17): (17); wherein, the constant is determined by the structural parameters of the unit, the altitude and the climate conditions.

[0042] In order to depict the difference between the theoretical power and the actual power, the present application defines the residual signal as shown in formula (18): (18); wherein, is the actual monitored power, is the theoretical power, that is, the output power that the mechanism model should have under normal operation.

[0043] At the same time, further construct the differential residual as formula (19) to represent the sensitivity of the residual to the wind speed and the temperature: (19); Through residuals With difference residual By constructing the structure, this invention can quantify the deviation between theoretical power and measured power. In particular, when entering the full-power wind speed range, energy conversion anomalies caused by different types of faults can be effectively captured. At the same time, the residual and differential residual characteristics are fused with the previous time series characteristics and a spatial matrix is ​​constructed on the DGI module.

[0044] In step S104 of the present invention, after completing the extraction of time-series features and the construction of residual signals, a DGI module is further proposed to model the dynamic coupling relationship between different features. By constructing an adaptive adjacency matrix, cross-feature dimension information fusion is achieved, thereby enhancing the ability to distinguish fault features.

[0045] like Figure 6 As shown, the input to the DGI module includes: raw features (i.e., the preprocessed data feature sequence), residual signal , differential residuals and temporal features extracted by LSTM The above multi-source features are concatenated to obtain a fusion vector, as shown in equation (20): (20); The concatenated input vector is linearly mapped to the high-dimensional latent space through equation (21): (twenty one); in, For trainable weight matrix, This is the hidden space dimension.

[0046] In the latent space, matrix multiplication (i.e. dot product similarity calculation) is first performed between features to obtain the similarity distribution between features. Then, after Softmax normalization, an adaptive adjacency matrix A is generated, as shown in Equation (22). This adjacency matrix can reflect the coupling strength between different features and is dynamically adjusted according to the changes in the input samples.

[0047] (twenty two); Among them, in the adaptive adjacency matrix A Indicates the first The first feature and the second The adjacency matrix, which assigns correlation weights to each feature, can be dynamically adjusted based on the input samples, exhibiting strong adaptability.

[0048] The adaptive adjacency matrix and the features in the high-dimensional hidden space are subjected to matrix multiplication again, and a coupling feature matrix is obtained after normalization; an interactive feature vector is obtained by adding the adaptive adjacency matrix and the coupling feature matrix; and a wind turbine fault diagnosis result is obtained according to the interactive feature vector.

[0049] The constructed adjacency matrix is subjected to matrix multiplication again with the hidden space features, as shown in formula (23): (23); The result of formula (23) After normalization, a coupling feature matrix is obtained, realizing information propagation and aggregation across features, enhancing the correlation expression between features, that is, each dimension feature is combined by other features related to it, thereby explicitly capturing the cross-feature coupling relationship; meanwhile, in order to maintain numerical stability and prevent overfitting, the application introduces a residual connection (that is, an interactive feature vector is obtained by adding the adaptive adjacency matrix and the coupling feature matrix) based on the aggregated output, and the interactive feature vector is subjected to linear layer and layer normalization processing, and then is subjected to processing of multiple same activation modules in sequence, and finally a wind turbine fault diagnosis result is obtained through a linear layer, wherein the activation module comprises a linear layer and a rectified linear unit activation function arranged in sequence.

[0050] A specific example is given below. The data set selected is SCADA system data of a certain wind farm, the resolution of the SCADA system data is 1 minute, and 12 feature variables of sensors are selected, which are environmental temperature, environmental wind speed, active power, gearbox cooling system pump internal temperature, generator speed, converter grid voltage, first blade variable pitch motor temperature, second blade variable pitch motor temperature, third blade variable pitch motor temperature, impeller speed, impeller speed and grid side C-phase current.

[0051] The fault types selected are four kinds of faults, namely, gearbox oil temperature control system power overlimit, short-time blade variable pitch rate average too low, generator speed and impeller speed mismatch and cabin vibration fault, and the constructed data set is shown in Table 1.

[0052] Table 1: Data set setting

[0053] In order to comprehensively evaluate the effectiveness of the wind turbine fault diagnosis method proposed in the application, the application selects a plurality of macro-averaging evaluation indexes to comprehensively investigate the classification ability of the model under different categories. Let the true positive number of the i-th category be , the true negative number be , the false positive number be , the false negative number be , and the total number of categories be , the overall accuracy of the model is , the index definitions are shown in formulas (24)-(30): (24); (25); (26); (27); (28); (29); (30); wherein, is used to measure the proportion of samples predicted as positive classes that actually belong to positive classes. Its significance lies in reflecting the credibility of the model's fault identification results. The higher the precision rate, the lower the false positive rate; represents the average classification accuracy of each category, avoiding the dominance of the overall accuracy by large categories; is used to measure the proportion of all true fault samples that are correctly identified, which can reflect the coverage ability of the model. In wind turbine diagnosis, high recall rate means less false negatives, ensuring the safety of equipment operation; is the geometric mean of recall rate and specificity, which can evaluate the balance ability of the model among different categories; is the harmonic mean of precision and recall, which takes into account both precision and recall, and is suitable for evaluating the overall performance of the model. reflects the proportion of normal samples that are incorrectly identified as faults. The lower the value, the smaller the false positive risk of the system. In actual wind farm applications, low FAR can reduce unnecessary repair and maintenance costs; reflects the proportion of fault samples that are misjudged as normal. The lower the value, the smaller the false negative risk of the system. In the wind power operation scenario, low MAR means that the model can effectively reduce potential safety hazards.

[0054] To verify the effectiveness of the fault diagnosis method proposed in the present application, the present application is trained and tested on a real wind turbine operation data set, and compared with a variety of existing deep learning models.

[0055] In the training phase, the present application uses cross-entropy loss function as the optimization target, and uses Adam optimizer for parameter update, with learning rate set to 0.01, batch size set to 64, and training rounds set to 100. To ensure the time series dependency, the input samples are constructed by sliding window, and the scale uniformity across feature dimensions is realized by standardization transformation. In the training process, macro-Accuracy is used as the preferred standard, the performance is monitored in real time on the test set, and the optimal model weight is saved for subsequent evaluation.

[0056] In order to comprehensively evaluate the diagnostic effect of the present application, a plurality of representative deep learning and traditional machine learning methods including Transformer, Swin-Transformer, Mamba, NODE, EfficientNet, ResNet, 1D-CNN, CNN-LSTM, GCN, RNN, TCN, HGB, GBDT-Light, GBM and XGBoost are selected as comparative models, which cover a plurality of mainstream technical systems such as convolution structure, recursive structure, graph neural network, time series modeling network and ensemble learning method.

[0057] As shown in Table 2, in macro-Accuracy, macro-Precision and macro-Recall and other multi-dimensional indicators, the PIA model proposed in the present application achieves the optimal result, although the inference speed of the model does not reach the optimal, but 2.73us is also enough to support the model to realize real-time detection of wind turbine faults, and the lightweight of the model is realized. Among other models, large models such as Transformer, Swin-Transformer and ResNet also achieve good results, but their inference speed increases by an order of magnitude compared with the PIA model, which is difficult to meet the requirements of real-time detection.

[0058] Table 2: Comparison of fault diagnosis of different models

[0059] As Figure 7 shown is the comparison of the Loss curve (i.e. loss curve) of different models in the training process. The PIA model proposed in the present application quickly decreases in the initial stage and tends to converge within 10 iterations, and the overall Loss value is significantly lower than that of the comparative model. This indicates that the PIA model has higher convergence speed and more stable optimization performance in feature extraction and fault pattern discrimination.

[0060] In addition, Figure 8 , Figure 9 and Figure 10 are the confusion matrices (representing the relationship between the predicted label and the true label, all corresponding to label 0, label 1, label 2, label 3 and label 4) of each model on the test set; Figure 8 , Figure 9 and Figure 10 each subgraph shows the classification performance of different models on the test set, including the distribution of true positives, false positives, true negatives and false negatives. By comparing the confusion matrices of each model, the recognition accuracy and misjudgment rate of each model on different categories can be directly observed.

[0061] The PIA model proposed in the application has high recognition accuracy on all fault categories, especially on complex fault categories such as 'gear box oil temperature control system power out of limit' and 'generator speed and impeller speed mismatch', and can effectively reduce the misjudgment rate and the false negative rate. Compared with other models, PIA is more prominent in category differentiation, which verifies the significant advantages of introducing the physical constraint and dynamic feature interaction mechanism.

[0062] Figure 11 A wind turbine fault diagnosis system based on mechanism data fusion is shown, comprising: A data preprocessing unit configured to preprocess the operating condition data of the wind turbine, the preprocessing including sliding window processing and standardization processing; A time series feature extraction unit configured to extract time series from the preprocessed feature sequence and obtain time series features through a pre-trained neural network model; A mechanism data processing unit configured to construct a power mechanism model using the preprocessed data, determine a theoretical power according to the power mechanism model, and calculate the residual error and the differential residual error between the theoretical power and the real power; A multi-dimensional feature fusion unit configured to concatenate and fuse the feature sequence, the time series features, the residual error and the differential residual error to obtain a fusion feature vector; An adjacency matrix generation unit configured to linearly map the fusion feature vector to a high-dimensional hidden space, perform matrix multiplication operation between the features in the high-dimensional hidden space to obtain the similarity distribution between the features, and obtain an adaptive adjacency matrix after normalization; A diagnosis result generation unit configured to perform matrix multiplication operation between the adaptive adjacency matrix and the features in the high-dimensional hidden space again, obtain a coupling feature matrix after normalization, add the adaptive adjacency matrix and the coupling feature matrix to obtain an interaction feature vector, and obtain the wind turbine fault diagnosis result according to the interaction feature vector.

[0063] It can be understood that the above-mentioned units can be combined into one or several other units respectively or all, or some of the units can be further divided into a plurality of units with smaller functions to constitute, which can realize the same operation without affecting the realization of the technical effects of the embodiments of the application. The above-mentioned units are divided based on logical functions, and the function of one unit can also be realized by multiple units, or the functions of multiple units can be realized by one unit. In other embodiments of the application, the system can also include other units, and these functions can also be realized by other units in actual application, and can be realized by multiple units.

[0064] According to another embodiment of the present application, the system of the present embodiment can be constructed by running a computer program (including program codes) capable of performing each step involved in the corresponding method of the present application on a general computing device such as a computer including processing elements and storage elements such as a Central Processing Unit (CPU), a Random Access Memory (RAM), a Read Only Memory (ROM), etc., the computer program can be recorded on a computer readable recording medium such as a computer readable recording medium and loaded into the above-mentioned computing device through the computer readable recording medium and run therein.

[0065] Figure 12 A computer device is shown, which includes a processor, a communication interface and a computer readable storage medium. Wherein, the processor, the communication interface and the computer readable storage medium can be connected through a bus or other means.

[0066] Wherein, the communication interface is used for receiving and sending data, the computer readable storage medium can be stored in the memory of the electronic device, the computer readable storage medium is used for storing a computer program, the computer program includes program instructions, and the processor is used for executing the program instructions stored in the computer readable storage medium.

[0067] The processor is the computing core and control core of the electronic device, which is suitable for implementing one or more instructions, and is particularly suitable for loading and executing one or more instructions to realize a corresponding method process or a corresponding function.

[0068] The processor is configured to perform the following process: The operating condition data of the wind turbine is preprocessed, and the preprocessing includes sliding window processing and standardization processing; The preprocessed feature sequence is time series extracted, and the time series features are obtained through the pre-trained neural network model; The power mechanism model is constructed by using the preprocessed data, the theoretical power is determined according to the power mechanism model, and the residual error and the differential residual error between the theoretical power and the real power are calculated; The feature sequence, the time series feature, the residual error and the differential residual error are spliced and fused to obtain a fusion feature vector; The fusion feature vector is linearly mapped to a high-dimensional hidden space, the features in the high-dimensional hidden space are subjected to matrix multiplication operation between features to obtain the similarity distribution between features, and the adaptive adjacency matrix is obtained after normalization; The adaptive adjacency matrix and the features in the high-dimensional hidden space are subjected to matrix multiplication again, and a coupling feature matrix is obtained after normalization; the interactive feature vector is obtained by adding the adaptive adjacency matrix and the coupling feature matrix; and the wind turbine fault diagnosis result is obtained according to the interactive feature vector.

[0069] The application further provides a computer readable storage medium, which is a memory device in an electronic device and is used for storing programs and data.

[0070] In the storage space, one or more instructions suitable for being loaded and executed by the processor are further stored, and the instructions can be one or more computer programs (including program codes).

[0071] In an embodiment, the computer readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer readable storage medium to implement the following process: The operation condition data of the wind turbine are preprocessed, and the preprocessing includes sliding window processing and standardization processing; The preprocessed feature sequence is subjected to time sequence extraction, and the time sequence features are obtained through a pre-trained neural network model; The power mechanism model is constructed by using the preprocessed data, the theoretical power is determined according to the power mechanism model, and the residual error and the differential residual error between the theoretical power and the real power are calculated; The feature sequence, the time sequence features, the residual error and the differential residual error are spliced and fused to obtain a fusion feature vector; The fusion feature vector is linearly mapped to a high-dimensional hidden space, the features in the high-dimensional hidden space are subjected to matrix multiplication again, and an adaptive adjacency matrix is obtained after normalization; The adaptive adjacency matrix and the features in the high-dimensional hidden space are subjected to matrix multiplication again, and a coupling feature matrix is obtained after normalization; the interactive feature vector is obtained by adding the adaptive adjacency matrix and the coupling feature matrix; and the wind turbine fault diagnosis result is obtained according to the interactive feature vector.

[0072] The application further provides a computer program product or computer program, which comprises computer instructions stored in a computer readable storage medium. The operation condition data of the wind turbine is preprocessed, and the preprocessing comprises sliding window processing and standardization processing. The preprocessed feature sequence is time series extracted, and the time series features are obtained through the pre-trained neural network model. The power mechanism model is constructed by using the preprocessed data, the theoretical power is determined according to the power mechanism model, and the residual error and the differential residual error between the theoretical power and the real power are calculated. The feature sequence, the time series features, the residual error and the differential residual error are spliced and fused to obtain a fusion feature vector. The fusion feature vector is linearly mapped to a high-dimensional hidden space, the features in the high-dimensional hidden space are subjected to matrix multiplication operation between features to obtain a similarity distribution between features, and an adaptive adjacency matrix is obtained after normalization. The adaptive adjacency matrix and the features in the high-dimensional hidden space are subjected to matrix multiplication operation between features again, a coupling feature matrix is obtained after normalization, an interaction feature vector is obtained by adding the adaptive adjacency matrix and the coupling feature matrix, and a wind turbine fault diagnosis result is obtained according to the interaction feature vector.

[0073] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the application can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software mode depends on the specific application and design constraints of the technical solution. Those of ordinary skill in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the application.

[0074] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present application are generated. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in or transmitted by a computer readable storage medium. The computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired (for example, coaxial cable, optical fiber, digital line) or wireless (for example, infrared, wireless, microwave, etc.) manner. The computer readable storage medium can be any available medium that can be accessed by a computer or a data processing device such as a server, data center, etc. integrated with one or more available media. The available media can be magnetic media (for example, floppy disk, hard disk, magnetic tape), optical media (for example, DVD), or semiconductor media (for example, solid state disk) and the like.

[0075] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A wind turbine fault diagnosis method based on mechanism data fusion, characterized in that, Includes the following processes: The operating condition data of the wind turbine is preprocessed, including sliding window processing and standardization processing. Temporal features are extracted from the preprocessed feature sequences and obtained through a pre-trained neural network model. A power mechanism model is constructed using the preprocessed data. The theoretical power is determined based on the power mechanism model, and the residuals and difference residuals between the theoretical power and the actual power are calculated. The feature sequence, temporal features, residuals, and difference residuals are concatenated and fused to obtain a fused feature vector; The fused feature vectors are linearly mapped to a high-dimensional latent space. The features in the high-dimensional latent space are then subjected to matrix multiplication to obtain the similarity distribution between features. After normalization, the adaptive adjacency matrix is ​​obtained. The adaptive adjacency matrix and the features in the high-dimensional latent space are multiplied again by matrix multiplication. After normalization, the coupled feature matrix is ​​obtained. The adaptive adjacency matrix and the coupled feature matrix are added to obtain the interactive feature vector. The wind turbine fault diagnosis result is obtained based on the interactive feature vector.

2. The wind turbine fault diagnosis method based on mechanism data fusion as described in claim 1, characterized in that, The operating condition data refers to the data collected by the data acquisition and monitoring system, including: wind speed, rotational speed, current, voltage, power, and temperature.

3. The wind turbine fault diagnosis method based on mechanism data fusion as described in claim 1, characterized in that, The power mechanism model is as follows: ,in, These are the cut-in wind speed, full-load wind speed, and cut-out wind speed, respectively. The rated power of the wind turbine unit. air density, The sweeping area of ​​the wind turbine. For wind speed, This is the power coefficient.

4. The wind turbine fault diagnosis method based on mechanism data fusion as described in claim 3, characterized in that, The theoretical power is determined based on the power mechanism model. ,include: ,in, Represents temperature.

5. The wind turbine fault diagnosis method based on mechanism data fusion as described in claim 1, characterized in that, The residual is: The differential residual is: , This represents the actual power. Theoretical power, For residuals, It represents the difference residual.

6. The wind turbine fault diagnosis method based on mechanism data fusion as described in any one of claims 1-5, characterized in that, The wind turbine fault diagnosis results are obtained based on the interaction feature vectors, including: The interactive feature vector is processed by a linear layer and layer normalization, and then processed by multiple identical activation modules in sequence. Finally, the wind turbine fault diagnosis result is obtained through a linear layer. The activation module includes a linear layer and a modified linear unit activation function set in sequence.

7. A wind turbine fault diagnosis system based on mechanism data fusion, characterized in that, include: The data preprocessing unit is configured to preprocess the wind turbine's operating condition data, including sliding window processing and standardization processing. The temporal feature extraction unit is configured to: extract temporal features from the preprocessed feature sequence and obtain temporal features through a pre-trained neural network model; The mechanism data processing unit is configured to: construct a power mechanism model using the preprocessed data, determine the theoretical power based on the power mechanism model, and calculate the residual and difference residual between the theoretical power and the actual power; The multidimensional feature fusion unit is configured to concatenate and fuse feature sequences, temporal features, residuals, and difference residuals to obtain a fused feature vector. The adjacency matrix generation unit is configured to: linearly map the fused feature vector to a high-dimensional latent space, perform matrix multiplication operations between features in the high-dimensional latent space to obtain the similarity distribution between features, and obtain an adaptive adjacency matrix after normalization. The diagnostic result generation unit is configured to: perform matrix multiplication between the adaptive adjacency matrix and the features in the high-dimensional latent space again, and obtain a coupled feature matrix after normalization; add the adaptive adjacency matrix and the coupled feature matrix to obtain an interactive feature vector; and obtain the wind turbine fault diagnosis result based on the interactive feature vector.

8. A computer device, characterized in that, include: Processor and computer-readable storage media; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the wind turbine fault diagnosis method based on mechanism data fusion as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1 to 6, which is a wind turbine fault diagnosis method based on mechanism data fusion.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the wind turbine fault diagnosis method based on mechanism data fusion as described in any one of claims 1 to 6.

Images