Wind turbine generator operation data processing method and system

Abstract

The invention relates to the technical field of data processing, in particular to a wind turbine generator operation data processing method and system, and the method comprises the steps: obtaining original wind turbine generator operation data; performing segmentation processing on the original wind turbine generator operation data to obtain a plurality of operation data blocks; performing vibration time domain feature extraction on the operation data block to obtain time domain feature information; performing temperature change feature extraction on the operation data block to obtain temperature feature information; and performing wind speed power associated feature extraction on the operation data block to obtain wind speed power feature information. According to the method, detailed time domain feature extraction, temperature change feature extraction and wind speed power correlation feature extraction are carried out on the operation data of the wind turbine generator, and multi-source data fusion is carried out in combination with the deep learning model, so that the distinguishing and predicting capability on the operation state of the equipment is remarkably enhanced; and improvement of accuracy and efficiency of prediction analysis results aiming at different states is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for processing wind turbine operation data. Background Art

[0002] Wind turbines are devices that use wind energy to generate electricity. They convert the kinetic energy of wind into electrical energy. During operation, the integrated multiple sensors generate massive, high-dimensional time-series operating data that constitutes a digital image of the equipment status. Effective computational analysis of this data is key to achieving predictive maintenance, improving operational efficiency, and ensuring safety.

[0003] Traditional processing methods usually rely on single-dimensional features or simple statistics, and only analyze the time domain or frequency domain characteristics of vibration data. It is difficult to capture the complex coupling relationship between different physical quantities, resulting in insufficient effective information density of subsequent analysis model input, affecting the accuracy of data classification and identification; the data processing of existing technologies uses the same set of analysis models for processing and executes complex deep diagnostic algorithms, resulting in huge waste of computing resources and processing delays; the use of simple models cannot conduct in-depth and accurate analysis of data that actually has fault risk precursors or anomalies. Due to the lack of prior identification of data status, it is impossible to dynamically adjust the processing strategy and implement lightweight trend prediction for healthy data; initiate medium-complexity early warning analysis for potential risk data; and only call the most resource-intensive deep analysis model for confirmed abnormal data.

[0004] Therefore, it is urgent to propose a wind turbine operation data processing method that can perform multi-dimensional feature fusion and state classification on the data, and then dynamically call different analysis models according to the classification results to improve the accuracy of trend prediction, potential risk prediction and risk evolution prediction, thereby achieving a dual improvement in the efficiency and accuracy of industrial time series data processing. Summary of the Invention

[0005] To achieve the above-mentioned object, the present invention provides the following technical solution: a method for processing wind turbine operation data, comprising: Acquire original wind turbine operating data; segment the original wind turbine operating data to obtain multiple operating data blocks; Extracting vibration time domain features from the operation data block to obtain time domain feature information; extracting temperature change features from the operation data block to obtain temperature feature information; extracting wind speed and power correlation features from the operation data block to obtain wind speed and power feature information; Classifying the original wind turbine operating data into a normal operating state, a fault risk precursor state, or an abnormal operating state according to the time domain characteristic information, the temperature characteristic information, and the wind speed and power characteristic information; In response to the original wind turbine operating data being in the normal operating state, inputting the original wind turbine operating data into a basic analysis model to perform trend prediction and health assessment; In response to the original wind turbine operating data being in the fault risk precursor state, inputting the original wind turbine operating data into an early warning analysis model to perform abnormal trend prediction and potential risk prediction; In response to the original wind turbine operating data being in the abnormal operating state, the original wind turbine operating data is input into a deep analysis model to perform multi-source data fusion and risk evolution prediction.

[0006] Preferably, performing vibration time domain feature extraction on the operation data block to obtain time domain feature information; performing temperature change feature extraction on the operation data block to obtain temperature feature information; performing wind speed power correlation feature extraction on the operation data block to obtain wind speed power feature information, including: Calculating time domain statistics of the vibration signal in the operation data block; wherein the time domain statistics include maximum value, minimum value, average value and root mean square value; obtaining the time domain feature information according to the time domain statistics; Calculating a temperature change curve of the temperature monitoring data in the operation data block, and determining a temperature change rate and a temperature fluctuation range according to the temperature change curve; Calculating the linear regression coefficient of the temperature monitoring data and the operating time to obtain the correlation coefficient between the temperature and the operating time; obtaining the temperature characteristic information according to the temperature change rate, the temperature fluctuation range, and the correlation coefficient between the temperature and the operating time; Calculating the Pearson correlation coefficient between the wind speed data and the power output data in the operation data block; Identify the moment of sudden wind speed change, and calculate the power response delay time corresponding to the moment of sudden wind speed change; The power output data above the rated wind speed is counted to obtain power saturation; and the wind speed power characteristic information is obtained according to the Pearson correlation coefficient, the power response delay time and the power saturation.

[0007] Preferably, classifying the original wind turbine operating data into a normal operating state, a fault risk precursor state, or an abnormal operating state according to the time domain characteristic information, the temperature characteristic information, and the wind speed and power characteristic information includes: Performing feature splicing on the time domain feature information, the temperature feature information, and the wind speed and power feature information to obtain a fused operation feature information set; generating a similarity metric corresponding to the original wind turbine operation data based on the fused operation feature information set, wherein the similarity metric is used to represent the degree of similarity between each operation data block in the original wind turbine operation data; Inputting the fusion operation feature information set into a pre-trained feature extraction network, wherein the feature extraction network outputs a state feature matrix; Determining at least one similar operating state set and an operating state summary corresponding to each similar operating state set in the original wind turbine operating data according to the state feature matrix and the similarity metric; wherein the state feature matrix is ​​used to represent the state distribution of each operating data block in the original wind turbine operating data in the entire original wind turbine operating data; Counting the number of similar operating state sets; determining the operating state type corresponding to each similar operating state set based on the operating state summary of each similar operating state set; Based on the number of the operating status types, the original wind turbine operating data is classified into a normal operating state, a fault risk precursor state or an abnormal operating state.

[0008] Preferably, trend prediction and health assessment are performed on the original wind turbine operating data to obtain a corresponding state assessment vector, including: Performing sliding window smoothing processing on the original wind turbine operating data to obtain smoothed data; inputting the smoothed data into a long short-term memory network model to extract time series features; The time series features are superimposed on the original wind turbine operating data, and fully connected layer mapping is performed after superposition to obtain the state assessment vector.

[0009] Preferably, abnormal trend prediction and potential risk prediction are performed on the original wind turbine operating data to obtain corresponding early warning event data, including: Performing wavelet transform on the original wind turbine operating data to obtain a multi-scale feature map; Inputting the multi-scale feature map into a convolutional neural network model to extract local abnormal features; The local abnormal features are matched with a preset fault risk pattern library for similarity to obtain the warning event data.

[0010] Preferably, multi-source data fusion and risk evolution prediction are performed on the original wind turbine operating data to obtain a corresponding data processing package, including: Performing feature standardization processing on the original wind turbine operating data to obtain standardized feature data; Inputting the standardized feature data into a graph neural network model to construct a data association graph; Performing a graph convolution operation on the data association graph to extract global fault risk features; The global fault risk feature is input into a classifier to obtain the data processing package.

[0011] Preferably, the joint training method of the basic analysis model, the early warning analysis model and the deep analysis model includes: Acquire multiple sets of training data sets; wherein the training data sets include sample operating data of wind turbines in different operating states and corresponding state labels; Perform input configuration processing on the initial basic analysis model, initial early warning analysis model, and initial in-depth analysis model respectively to obtain the input configuration strategy of each model; the input configuration strategy of any model is used to indicate the state labels that need to be processed during the training process of the model; Based on the input configuration strategy of each model, the sample running data is input into the corresponding initial model for training to obtain the output result of each model; Determine the model loss information of each model based on the state label and output results of each model; the model loss information of any model is obtained by fitting the trend prediction loss, abnormal trend prediction loss and risk prediction loss. The trend prediction loss is used to represent the category deviation between the operating state predicted by the model and the actual state, the abnormal trend prediction loss is used to represent the position deviation between the abnormal point detected by the model and the actual abnormal point, and the risk prediction loss is used to represent the category deviation between the fault risk type identified by the model and the actual fault risk type; According to the model loss information of each model, the parameters of the initial basic analysis model, the initial early warning analysis model and the initial deep analysis model are optimized respectively to obtain the basic analysis model, the early warning analysis model and the deep analysis model.

[0012] Preferably, the model loss information of each model is determined based on the state label and output result of each model, including: Obtaining a first weight ratio between the trend prediction loss and the abnormal trend prediction loss, and a second weight ratio between the trend prediction loss and the risk prediction loss; Using the first weight ratio and the second weight ratio, adjusting the contribution of the trend prediction loss in the fitting process; The third weight ratio between the abnormal trend prediction loss and the risk prediction loss is used to adjust the contribution of the abnormal trend prediction loss and the risk prediction loss in the fitting process.

[0013] Preferably, the parameter optimization of the initial basic analysis model, the initial early warning analysis model and the initial depth analysis model includes: In response to the trend prediction loss in the model loss information being greater than a first preset threshold, preferentially adjusting parameters related to basic operation feature extraction in the initial basic analysis model; In response to the abnormal trend predicted loss in the model loss information being greater than a second preset threshold, preferentially adjusting the parameters related to the abnormal trend in the initial early warning analysis model; In response to the risk prediction loss in the model loss information being greater than a third preset threshold, the parameters related to risk type identification in the initial depth analysis model are preferentially adjusted.

[0014] A wind turbine operating data processing system, applicable to the above-mentioned wind turbine operating data processing method, comprises: A data block unit is used to obtain original wind turbine operating data; segment the original wind turbine operating data to obtain multiple operating data blocks; a feature extraction unit configured to extract vibration time domain features from the operation data block to obtain time domain feature information; extract temperature change features from the operation data block to obtain temperature feature information; and extract wind speed and power correlation features from the operation data block to obtain wind speed and power feature information; an operation classification unit, configured to classify the original wind turbine operation data into a normal operation state, a fault risk precursor state, or an abnormal operation state according to the time domain characteristic information, the temperature characteristic information, and the wind speed and power characteristic information; a trend prediction unit, configured to input the original wind turbine operating data into a basic analysis model in response to the original wind turbine operating data being in the normal operating state, and perform trend prediction and health assessment; a fault early warning unit, configured to input the original wind turbine operating data into an early warning analysis model in response to the original wind turbine operating data being in the fault risk precursor state, and perform abnormal trend prediction and potential risk prediction; The fault diagnosis unit is configured to input the original wind turbine operating data into a deep analysis model in response to the original wind turbine operating data being in the abnormal operating state, and perform multi-source data fusion and risk evolution prediction.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention extracts detailed time-domain features, temperature change features, and wind speed-power correlation features from wind turbine operating data, and combines this with a deep learning model to perform multi-source data fusion, effectively learning and utilizing the inherent coupling relationship between different physical quantities. Furthermore, by comprehensively analyzing time-domain features, temperature features, and wind speed-power features, the present invention significantly enhances the ability to distinguish and predict the operating status of the equipment, thereby improving the accuracy of operating status classification and future trend prediction. (2) The present invention jointly trains the basic analysis model, the early warning analysis model, and the deep analysis model, and dynamically optimizes them according to the loss information of each model, thereby avoiding the computational redundancy of a single complex model in processing all data, realizing on-demand allocation and optimized scheduling of computing resources, significantly improving the timeliness of large-scale industrial time series data processing tasks, and ensuring the accuracy of prediction results for different states. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic flow chart of the steps of the overall method in one embodiment of the present invention; Figure 2 FIG. 1 is a schematic diagram of the system architecture of the overall system in one embodiment of the present invention.

[0017] In the figure: 1. Data segmentation unit; 2. Feature extraction unit; 3. Operation classification unit; 4. Trend prediction unit; 5. Fault warning unit; 6. Fault diagnosis unit. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] For example 1, please refer to Figure 1 The present invention provides a technical solution: a method for processing wind turbine operation data, comprising: S1. Obtaining original wind turbine operating data; segmenting the original wind turbine operating data to obtain multiple operating data blocks; S2. Extract vibration time domain features from the operation data block to obtain time domain feature information; extract temperature change features from the operation data block to obtain temperature feature information; extract wind speed and power correlation features from the operation data block to obtain wind speed and power feature information; S3. Classify the original wind turbine operating data into a normal operating state, a fault risk precursor state, or an abnormal operating state based on the time domain characteristic information, the temperature characteristic information, and the wind speed and power characteristic information; S4. In response to the original wind turbine operating data being in a normal operating state, inputting the original wind turbine operating data into a basic analysis model to perform trend prediction and health assessment; S5. In response to the original wind turbine operating data being in a fault risk precursor state, inputting the original wind turbine operating data into an early warning analysis model to perform abnormal trend prediction and potential risk prediction; S6. In response to the original wind turbine operating data being in an abnormal operating state, the original wind turbine operating data is input into a deep analysis model to perform multi-source data fusion and risk evolution prediction.

[0020] It should be noted that the original wind turbine operation data includes vibration signal data, temperature monitoring data, wind speed collection data and power output data; the time domain characteristic information includes the mean, variance, peak factor and kurtosis coefficient of the vibration signal; the temperature characteristic information includes the temperature change rate, temperature fluctuation range and the correlation coefficient between temperature and operating time; the wind speed and power characteristic information includes the Pearson correlation coefficient between wind speed and power, the power response delay time when the wind speed changes suddenly and the power saturation above the rated wind speed.

[0021] In an optional embodiment, vibration time domain feature extraction is performed on the operation data block to obtain time domain feature information; temperature change feature extraction is performed on the operation data block to obtain temperature feature information; and wind speed power correlation feature extraction is performed on the operation data block to obtain wind speed power feature information, including: Calculate the time domain statistics of the vibration signal in the running data block; wherein the time domain statistics include maximum value, minimum value, average value and root mean square value; obtain time domain feature information based on the time domain statistics; Calculate the temperature change curve of the temperature monitoring data in the operation data block, and determine the temperature change rate and temperature fluctuation range based on the temperature change curve; Calculate the linear regression coefficient of temperature monitoring data and operating time to obtain the correlation coefficient between temperature and operating time; obtain temperature characteristic information based on the temperature change rate, temperature fluctuation range and the correlation coefficient between temperature and operating time; Calculate the Pearson correlation coefficient between wind speed data and power output data in the running data block; Identify the moment of sudden wind speed change and calculate the power response delay time corresponding to the moment of sudden wind speed change; The power output data above the rated wind speed is counted to obtain the power saturation; the wind speed power characteristic information is obtained based on the Pearson correlation coefficient, power response delay time and power saturation.

[0022] It should be noted that the maximum value refers to the maximum amplitude in the vibration signal, representing the maximum intensity of the vibration; the minimum value refers to the minimum amplitude in the vibration signal, representing the minimum intensity of the vibration; the average value refers to the mean of the vibration signal, reflecting the overall average level of the signal; the root mean square value refers to the value obtained by squaring, averaging and then taking the square root of the vibration signal, representing the effective vibration intensity of the signal; the RMS value is often used to evaluate the energy or intensity of the signal; based on these time domain statistics, the comprehensive characteristics of the vibration signal can be obtained, thereby evaluating whether the unit has abnormal vibration, which helps to judge the health status of the equipment; the temperature change curve plots the temperature change curve over time based on the temperature monitoring data in the operation data block; the temperature change rate refers to the speed of temperature change per unit time; it can be obtained by calculating the slope of the temperature curve. A rate that is too fast or too slow may mean that there is a problem with the equipment or that it is being disturbed by the outside world; the temperature fluctuation range refers to the difference between the maximum and minimum values ​​of the temperature change over a period of time; a large fluctuation range may indicate that the temperature control system of the equipment is unstable; the regression coefficient is obtained by performing linear regression analysis on the temperature monitoring data and the operating time; this can reflect the trend and law of temperature change during the operation of the equipment; combined with the temperature change rate and fluctuation range, the temperature fluctuation range can be used to determine the temperature of the equipment. The Pearson correlation coefficient is a statistic that measures the strength of the linear relationship between two variables. Here, it is used to calculate the correlation between wind speed data and power output data. If the correlation between wind speed and power is strong, it means that the working efficiency of the wind turbine is highly correlated with the change in wind speed. When the wind speed changes drastically, it may cause the power output of the turbine to change. By identifying the moment of sudden change in wind speed, the reaction of the turbine to the change in wind speed can be observed. The power response delay time is the wind speed. The delay time for the unit power output to change after a sudden change in wind speed occurs; a long delay time may be a sign of a slow unit response, which may indicate a slow response problem with the equipment; power saturation means that when the wind speed reaches a certain value (i.e., above the rated wind speed), the unit's power output tends to saturate and no longer increases significantly with increasing wind speed; power saturation describes the stability of power output when the wind speed exceeds the rated value; combining the Pearson correlation coefficient, power response delay time, and power saturation, the unit's operating characteristics under different wind speed conditions can be extracted; for example, whether it can respond quickly to wind speed changes, whether power saturation is likely to occur, etc.

[0023] In an optional embodiment, the original wind turbine operating data is classified into a normal operating state, a fault risk precursor state, or an abnormal operating state based on the time domain characteristic information, the temperature characteristic information, and the wind speed and power characteristic information, including: Perform feature splicing on the time domain feature information, temperature feature information and wind speed and power feature information to obtain a fused operation feature information set; generating a similarity metric corresponding to the original wind turbine operation data based on the fused operation feature information set, wherein the similarity metric is used to represent the degree of similarity between each operation data block in the original wind turbine operation data; The fusion operation feature information set is input into the pre-trained feature extraction network, and the feature extraction network outputs the state feature matrix; Determining at least one similar operating state set and an operating state summary corresponding to each similar operating state set in the original wind turbine operating data according to a state feature matrix and a similarity metric; wherein the state feature matrix is ​​used to represent the state distribution of each operating data block in the original wind turbine operating data in the entire original wind turbine operating data; Counting the number of similar operating state sets; determining the operating state type corresponding to each similar operating state set based on the operating state summary of each similar operating state set; Based on the number of operating status types, the raw wind turbine operating data is classified into normal operating status, fault risk precursor status or abnormal operating status.

[0024] It should be noted that these three types of feature information are fused (spliced) to obtain a comprehensive "fused operation feature information set"; this set contains features extracted from multiple dimensions (vibration, temperature, wind speed and power, etc.), which helps to comprehensively describe the operating status of the wind turbine; by analyzing the fused operation feature information set, the similarity between each operation data block is calculated; the similarity metric is used to measure the similarity between wind turbine data in different time periods or different operating states; for example, Euclidean distance, cosine similarity and other methods can be used to measure the similarity of different data blocks; the similarity metric helps identify which data blocks exhibit similar features among multiple operating data blocks of the wind turbine; this is very important for classifying and clustering different operating states of wind turbines; the fused operation feature information set is input into a pre-trained feature extraction network; this network is a deep learning model that can extract higher-order, abstract features from the fused features after training; the role of the feature extraction network is to convert the input data into a state The state feature matrix represents the characteristics of the wind turbine under different operating states; the state feature matrix is ​​a high-dimensional matrix, in which each row represents the characteristics of a data block and the columns represent different feature dimensions; this matrix describes the state distribution of different data blocks during the entire operation of the wind turbine, helping to identify the operating state corresponding to each data block; based on the state feature matrix and similarity measurement, at least one "similar operating state set" can be identified in all operating data; these sets are composed of similar operating data blocks, reflecting the similar states of the wind turbine within a specific time period; each similar operating state set will have a corresponding "operating state summary", that is, the representative feature or average feature of the set; for example, the average temperature, wind speed, vibration, etc. of a set can be used as the representative state of the set; count the number of all identified similar operating state sets to understand the distribution of each state set; by counting the number of these sets, it can be determined whether the unit is in normal operating state, fault risk precursor state or abnormal operating state.

[0025] In an optional embodiment, trend prediction and health assessment are performed on the original wind turbine operating data to obtain a corresponding state assessment vector, including: Perform sliding window smoothing on the original wind turbine operating data to obtain smoothed data; The smoothed data is input into the long short-term memory network model to extract time series features; The time series features are superimposed on the original wind turbine operation data, and then fully connected layer mapping is performed to obtain the state assessment vector.

[0026] It should be noted that sliding window is a commonly used signal processing technology used to reduce noise and fluctuations in data and help extract smoother trends. It sets a window size, slides on the original data, and calculates the mean, weighted average, or other statistics of the data points in each window. In this way, short-term fluctuations can be smoothed out and long-term trends in the data can be retained. Smoothed data means that high-frequency noise and sudden fluctuations in the data have been removed, making it more stable and easier to analyze later. Generally, this smoothing process helps make data more reliable and consistent during analysis, avoiding the impact of short-term anomalies on the training of the overall model. Long Short-Term Memory Network (LSTM) LSTM (Long Short-Term Memory (STM) is a special recurrent neural network (RNN) used to process and predict time series data. It can learn long-term dependencies in sequence data, effectively avoiding the gradient vanishing problem that traditional RNNs may encounter when processing long sequences. LSTM is very suitable for processing time-continuous change patterns such as wind turbine operation data. When the smoothed wind turbine operation data is input into the LSTM model, the model automatically learns and extracts the time series features in the data. For example, LSTM can capture the dynamic changes in wind turbine operation in different time periods and find potential regularities and periodic changes. The time series features extracted by LSTM are superimposed with the original wind turbine operation data, that is, the two types of information are combined. Generally, the features extracted by LSTM are implicit features learned through time series, which can reflect the trends and patterns of the data. Combining these features with the original data can enable the model to better understand the specific operation status of the wind turbine and enhance the model's expressive power. The superposition operation not only retains the detailed information of the original data, but also incorporates the time series features extracted by LSTM, so that the subsequent model can more comprehensively evaluate the operating status of the wind turbine. In deep neural networks, the fully connected layer (FCL) is a kind of deep neural network. A fully connected layer is a layer that connects each input neuron to the output neuron. Here, the superimposed data is passed to the fully connected layer for processing. The fully connected layer is usually used to map the extracted features to a high-dimensional space to help further analyze the data characteristics and produce the final prediction or evaluation results. Through the mapping of the fully connected layer, the model will generate an output based on the input data (including time series features and raw data). This output is usually a state assessment vector. The output can be a classification result (such as normal, fault risk, abnormal) or a continuous value (such as fault risk probability, health score, etc.).

[0027] In an optional embodiment, abnormal trend prediction and potential risk prediction are performed on the original wind turbine operating data to obtain corresponding early warning event data, including: Perform wavelet transform on the original wind turbine operation data to obtain multi-scale feature maps; Input the multi-scale feature map into the convolutional neural network model to extract local abnormal features; The local abnormal features are matched with the preset fault risk pattern library for similarity to obtain early warning event data.

[0028] It should be noted that wavelet transform is an effective signal processing technology. It can extract the local features of the signal at different frequencies and time domains by decomposing the signal at multiple scales. Unlike the traditional Fourier transform, wavelet transform can provide local information of the signal in both the time domain and the frequency domain, which is very useful for analyzing signals with non-stationary characteristics (such as the operating data of wind turbines). Wavelet transform decomposes the signal by changing the scale (or frequency) to obtain features at multiple scales. In the operating data of wind turbines, there may be fluctuations of different frequencies and durations (such as short-term instantaneous failure risks or long-term systemic problems). Wavelet transform can help extract these features at different scales. The obtained multi-scale feature map is the result of summarizing the features of the signal at multiple scales, which can reflect the changes at different levels of the data. Convolutional neural network (CNN) is a powerful deep learning model widely used in tasks such as image processing and speech recognition. In this scenario, CNN is used to process the multi-scale feature map obtained by wavelet transform. Through convolutional layers, pooling layers and fully connected layers, CNN can automatically extract useful features from input data and perform further pattern recognition and classification. Through the convolution operation of CNN, the model can learn local features in the data, which may include abnormal fluctuation patterns in the operating data of wind turbines. For example, the vibration of wind turbines The wind turbine signal may have sudden abnormal fluctuations, or its operating parameters may have abnormal fluctuations. CNN can extract these local abnormal features from the multi-scale feature map; through the filtering operation of the convolution layer, CNN can efficiently identify these local patterns that are different from the normal operating state; the fault risk pattern library is a "database" containing the fault risks of different types of wind turbines, which contains feature descriptions of various possible fault risk patterns (such as mechanical failure risk, electrical failure risk, etc.); these fault risk patterns are usually obtained through historical data, expert knowledge or experimental research, which can help the system identify which patterns are related to a specific fault risk; after extracting local abnormalities, the system can identify which patterns are related to a specific fault risk. After finding the common features, the model will compare these features with the preset fault risk pattern library and calculate the similarity; this process is usually completed by calculating the distance between features (such as Euclidean distance, cosine similarity, etc.); if the extracted local abnormal features are highly similar to the features of a certain fault risk pattern in the library, then it can be determined that the wind turbine may have a corresponding fault risk; the results of similarity matching can provide a basis for subsequent fault risk diagnosis; through similarity matching, the system will generate early warning event data; if the detected local abnormal features are highly similar to the features of a certain fault risk pattern, the system will issue a warning signal, indicating that the wind turbine may have a specific fault risk.

[0029] In an optional embodiment, multi-source data fusion and risk evolution prediction are performed on the original wind turbine operating data to obtain a corresponding data processing package, including: Perform feature standardization processing on the original wind turbine operating data to obtain standardized feature data; Input the standardized feature data into the graph neural network model to build a data association graph; Perform graph convolution on the data association graph to extract global fault risk features; The global fault risk characteristics are input into the classifier to obtain a data processing package.

[0030] It should be noted that Graph Neural Network (GNN) is a deep learning method specifically used to process graph-structured data; graph-structured data includes nodes (data points) and edges (relationships between nodes); in the fault risk diagnosis of wind turbines, the collected data of each sensor can be regarded as a node, and the relationship between different sensors (such as position proximity, signal correlation, etc.) can be used as edges to form a graph; by establishing a connection between the standardized feature data and the relationship between sensors, we can construct a "data association graph"; in this graph, the nodes of the graph represent different sensors, and the edges represent the relationship between sensors (for example, the physical location proximity or signal correlation); this graph structure can capture the spatial and signal correlation between sensors, so that the model can learn the mutual influence between different sensors through the structure of the graph; Graph Convolution (Graph Convolution) is the core operation in GNN, which is used to propagate and learn features on graph data; traditional convolutional neural networks (CNN) operate on regular grid data, while graph convolution can work on unstructured graph data and can propagate features through the connection relationship between nodes; through graph convolution operation, the model can transmit information and fuse the features of multiple nodes based on the correlation between sensors to extract global fault risk features; these global features can include the overall operating status of the wind turbine, helping the model to identify abnormal patterns of the entire system, rather than just local fault risk information; through the graph convolution layer, the relationship between nodes is transmitted and updated layer by layer, thereby extracting global fault risk features (such as the overall fault risk signal pattern of the system); the classifier is a model used to make classification decisions based on input features; common classifiers include support vector machines (SVM), decision trees, neural networks, etc.; here, the global fault risk features will be passed to the classifier as input after graph convolution for prediction of the fault risk type; the output of the classifier is the data processing package of the wind turbine; by analyzing the global features extracted by the graph convolution network, the classifier can determine whether the wind turbine has a fault risk and determine the type of fault risk (such as motor failure risk, transmission system failure risk, etc.); the final result will help operation and maintenance personnel to promptly detect the fault risk of the wind turbine and take appropriate maintenance measures, thereby improving the reliability and maintenance efficiency of the wind turbine.

[0031] In an optional embodiment, a joint training method for the basic analysis model, the early warning analysis model, and the deep analysis model includes: Acquire multiple sets of training data sets; wherein the training data sets include sample operating data of wind turbines under different operating states and corresponding state labels; Perform input configuration processing on the initial basic analysis model, initial early warning analysis model, and initial in-depth analysis model respectively to obtain the input configuration strategy of each model; the input configuration strategy of any model is used to indicate the state labels that need to be processed during the training process of the model; Based on the input configuration strategy of each model, the sample running data is input into the corresponding initial model for training to obtain the output results of each model; Based on the state label and output results of each model, the model loss information of each model is determined. The model loss information of any model is obtained by fitting the trend prediction loss, abnormal trend prediction loss, and risk prediction loss. The trend prediction loss is used to represent the category deviation between the operating state predicted by the model and the actual state. The abnormal trend prediction loss is used to represent the position deviation between the abnormal points detected by the model and the actual abnormal points. The risk prediction loss is used to represent the category deviation between the fault risk type identified by the model and the actual fault risk type. According to the model loss information of each model, the parameters of the initial basic analysis model, the initial early warning analysis model and the initial deep analysis model are optimized respectively to obtain the basic analysis model, the early warning analysis model and the deep analysis model.

[0032] It should be noted that the initial basic analysis model, the initial early warning analysis model, and the initial deep analysis model are the initial versions of the model, which are used for basic analysis, early warning analysis, and deep diagnosis respectively. Each model is designed for different needs of wind turbines. The basic analysis model is usually used to perform a simple analysis of the operating status of the wind turbine. The early warning analysis model is used to predict whether the wind turbine is about to have a failure risk or anomaly, and issue an early warning. The deep analysis model is used to diagnose the failure risk more accurately and analyze the specific failure risk type and location. For each model, determine which input data needs to be received by the model, and configure the model training based on this data. The input configuration strategy of each model indicates the model needs to process during the training process. The state label of the analysis; for example, the basic analysis model may only care about the two state labels of normal and fault risk, while the deep analysis model may need more detailed labels, such as different types of fault risk types; according to the above input configuration strategy, the sample operation data is input into the corresponding initial model for training; at this stage, the model will learn based on the given data and labels, and adjust the internal parameters so that it can gradually fit the patterns and laws in the data; the loss function is an indicator to measure the difference between the model prediction results and the actual results; each model will calculate the loss value based on its output results and the corresponding state label; the trend prediction loss is used to measure the deviation between the model's predicted operation state and the actual state; it focuses on the model Whether the state change trend in the time series accurately reflects the actual operating state change; for example, whether the model can correctly predict the future operating state of the equipment (such as from "normal" to "fault risk"); the abnormal trend prediction loss is used to measure the position deviation between the model and the actual abnormal point when detecting the abnormal point; if the model can correctly detect the abnormality in the wind turbine (such as sudden changes in current, voltage, etc.), then the abnormal trend prediction loss is small; if the model fails to correctly identify the abnormality, the loss will increase; the risk prediction loss is used to measure the deviation between the fault risk type identified by the model and the actual fault risk type; risk prediction is the core of the process, and the model needs to correctly identify different fault risk types (such as motor fault wind turbine). If the model incorrectly classifies the failure risk type, the loss will increase. Based on the loss information of each model, the model parameters are adjusted through optimization algorithms (such as gradient descent) to make it perform better on the training set and reduce the loss value. Through optimization, the model can improve accuracy, reduce errors, and better complete the task. The basic analysis model is adjusted based on the state labels in the training data to enable it to better distinguish between normal and failure risk states. The early warning analysis model is optimized based on the trend and anomaly information in the training data to enable it to predict the potential failure risk of wind turbines in advance. The deep analysis model is optimized based on the risk prediction labels in the training data to enable it to accurately diagnose the failure risk type of wind turbines.After parameter optimization, the basic analysis model, early warning analysis model, and in-depth analysis model have become more accurate. The optimized models can perform more accurate analysis, early warning, and fault risk diagnosis based on different operating data. Each optimized model has different tasks: the basic analysis model is responsible for preliminarily determining the operating status of the equipment, the early warning analysis model can identify potential fault risks in advance, and the in-depth analysis model can conduct in-depth analysis and accurately diagnose specific fault risk types.

[0033] In an optional embodiment, determining the model loss information of each model based on the state label and output result of each model includes: Obtaining a first weight ratio between the trend prediction loss and the abnormal trend prediction loss, and a second weight ratio between the trend prediction loss and the risk prediction loss; Using the first weight ratio and the second weight ratio, the contribution of the trend prediction loss in the fitting process is adjusted; The third weight ratio between abnormal trend prediction loss and risk prediction loss is used to adjust the contribution of abnormal trend prediction loss and risk prediction loss in the fitting process.

[0034] It should be noted that the first weight ratio is the ratio between trend prediction loss and abnormal trend prediction loss, which is used to indicate the relative importance of these two loss terms in model training. For example, if the first weight ratio is 2, then the contribution of trend prediction loss will be twice that of abnormal trend prediction loss. This weight ratio is set by analyzing the relative importance of each task, the training requirements of the model, data characteristics, and other factors. The second weight ratio is the ratio between trend prediction loss and risk prediction loss, which determines the relative importance of the two in the total loss. For example, if the second weight ratio is 1.5, it means that the contribution of trend prediction loss in training is 1.5 times that of risk prediction loss. This ratio needs to be adjusted according to the training objectives and the focus of the task. For example, if diagnosing fault risk is more important than predicting trends, the weight of risk prediction loss may be higher. Based on the values ​​of the first and second weight ratios, the weight of trend prediction loss in the fitting process is adjusted. The fitting process is the process of model learning, in which the optimization algorithm attempts to minimize the total loss function. The weight of trend prediction loss may increase or decrease accordingly based on the first and second weight ratios. For example, if the first weight ratio is large (indicating that abnormal trend prediction is less important), the weight of trend prediction loss may increase or decrease accordingly. The third weight ratio represents the ratio between the anomaly trend prediction loss and the risk prediction loss, balancing the contributions of these two tasks in the loss function. For example, a third weight ratio of 2 indicates that the anomaly trend prediction task has a greater weight than the risk prediction task. Adjusting the contributions of these two losses is to ensure a balance between the model's performance in detecting anomalies and identifying fault risks, and should be determined based on the model's ultimate application scenario. For example, if the accuracy of anomaly detection is more critical than the identification of fault risk types, the weight of the anomaly trend prediction loss can be appropriately increased. During training, the final form of the loss function combines the various loss terms according to certain weights. For multi-task learning, a common practice is to assign a weight to each loss term, reflecting the relative importance of the tasks. These weights are gradually adjusted based on the desired task optimization objective, ensuring that the model correctly focuses on the most important tasks during training. For example, if the model performs poorly in identifying fault risks, the weight of the risk prediction loss can be increased. If trend prediction is more important than anomaly trend prediction, the weight of the trend prediction loss can be increased.

[0035] In an optional embodiment, the parameters of the initial basic analysis model, the initial early warning analysis model, and the initial in-depth analysis model are optimized, including: In response to the trend prediction loss in the model loss information being greater than a first preset threshold, preferentially adjusting parameters related to basic operation feature extraction in the initial basic analysis model; In response to the abnormal trend prediction loss in the model loss information being greater than a second preset threshold, preferentially adjusting parameters related to abnormal point detection in the initial early warning analysis model; In response to the risk prediction loss in the model loss information being greater than a third preset threshold, parameters related to fault risk type identification in the initial depth analysis model are preferentially adjusted.

[0036] It should be noted that the first preset threshold is a pre-set threshold, which is usually determined through experiments or experience; if the trend prediction loss exceeds this threshold, it means that the performance of the model in this regard is not up to standard; at this time, priority is given to adjusting parameters related to basic operation feature extraction, such as data preprocessing methods, feature selection, feature engineering methods, etc.; by optimizing these features, the trend prediction ability of the model is improved, thereby reducing losses; the second preset threshold is used to determine whether the abnormal trend prediction loss is too large; if the loss is greater than the second threshold, it means that the abnormal trend prediction function of the model is insufficient; anomaly point detection refers to identifying those points in the data that obviously do not conform to the normal pattern; possible reasons include equipment failure risk, environmental anomalies, etc.; when the abnormal trend prediction loss exceeds the preset threshold, priority is given to adjusting parameters related to anomaly point detection, such as selecting different abnormal trend prediction algorithms (such as statistical-based methods or machine learning methods), optimizing the sensitivity of abnormal trend prediction, etc. sensitivity, adjust the threshold of the abnormal trend prediction model, etc.; the risk prediction loss measures the performance of the model in identifying specific fault risk types; a larger loss means that the model has poor accuracy in the risk prediction task; the third preset threshold is a pre-set threshold used to determine whether the risk prediction task needs to be optimized; if the loss exceeds the third threshold, it means that the risk prediction effect is not ideal; fault risk type identification refers to identifying the type of fault risk that the equipment has (for example, battery failure risk, transmission system failure risk, etc.); each fault risk type has different symptoms, and the model needs to make accurate classifications based on the input data; when the risk prediction loss is too large, give priority to adjusting the parameters related to fault risk type identification, such as optimizing the structure of the classifier, adjusting the labels of the fault risk categories, using more appropriate loss functions, increasing the sample size of fault risk data, etc.; this can improve the model's ability to identify fault risk types, thereby reducing losses.

[0037] For example 2, please refer to Figure 2 The present invention provides a technical solution: a wind turbine operation data processing system, which is applicable to the above-mentioned wind turbine operation data processing method, comprising: The data block unit 1 is used to obtain the original wind turbine operating data; the original wind turbine operating data is segmented and processed to obtain multiple operating data blocks; Feature extraction unit 2 is used to extract vibration time domain features from the operation data block to obtain time domain feature information; extract temperature change features from the operation data block to obtain temperature feature information; and extract wind speed and power correlation features from the operation data block to obtain wind speed and power feature information; An operation classification unit 3 is used to classify the original wind turbine operation data into a normal operation state, a fault risk precursor state or an abnormal operation state according to the time domain characteristic information, the temperature characteristic information and the wind speed and power characteristic information; A trend prediction unit 4 is configured to input the original wind turbine operating data into a basic analysis model in response to the original wind turbine operating data being in a normal operating state, perform trend prediction and health assessment on the original wind turbine operating data, and obtain a corresponding state assessment vector; A fault warning unit 5 is configured to input the original wind turbine operating data into a warning analysis model to perform trend prediction and health assessment in response to the original wind turbine operating data being in a fault risk precursor state; The fault diagnosis unit 6 is configured to input the original wind turbine operating data into the deep analysis model in response to the original wind turbine operating data being in an abnormal operating state, and perform multi-source data fusion and risk evolution prediction.

[0038] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for processing wind turbine operation data, characterized in that ,include: Acquire original wind turbine operating data; segment the original wind turbine operating data to obtain multiple operating data blocks; Extracting vibration time domain features from the operation data block to obtain time domain feature information; extracting temperature change features from the operation data block to obtain temperature feature information; extracting wind speed and power correlation features from the operation data block to obtain wind speed and power feature information; Classifying the original wind turbine operating data into a normal operating state, a fault risk precursor state, or an abnormal operating state according to the time domain characteristic information, the temperature characteristic information, and the wind speed and power characteristic information; In response to the original wind turbine operating data being in the normal operating state, inputting the original wind turbine operating data into a basic analysis model to perform trend prediction and health assessment; In response to the original wind turbine operating data being in the fault risk precursor state, inputting the original wind turbine operating data into an early warning analysis model to perform abnormal trend prediction and potential risk prediction; In response to the original wind turbine operating data being in the abnormal operating state, the original wind turbine operating data is input into a deep analysis model to perform multi-source data fusion and risk evolution prediction.

2. A method for processing wind turbine operation data according to claim 1, characterized in that , extracting vibration time domain features from the operation data block to obtain time domain feature information; Extracting temperature variation characteristics of the operation data block to obtain temperature characteristic information; Extracting wind speed and power correlation features from the operating data block to obtain wind speed and power feature information includes: Calculating time domain statistics of the vibration signal in the operation data block; wherein the time domain statistics include maximum value, minimum value, average value and root mean square value; obtaining the time domain feature information according to the time domain statistics; Calculating a temperature change curve of the temperature monitoring data in the operation data block, and determining a temperature change rate and a temperature fluctuation range according to the temperature change curve; Calculating the linear regression coefficient of the temperature monitoring data and the operating time to obtain the correlation coefficient between the temperature and the operating time; obtaining the temperature characteristic information according to the temperature change rate, the temperature fluctuation range, and the correlation coefficient between the temperature and the operating time; Calculating the Pearson correlation coefficient between the wind speed data and the power output data in the operation data block; Identify the moment of sudden wind speed change, and calculate the power response delay time corresponding to the moment of sudden wind speed change; The power output data above the rated wind speed is counted to obtain power saturation; and the wind speed power characteristic information is obtained according to the Pearson correlation coefficient, the power response delay time and the power saturation.

3. A method for processing wind turbine operation data according to claim 2, characterized in that: Classifying the original wind turbine operating data into a normal operating state, a fault risk precursor state, or an abnormal operating state according to the time domain characteristic information, the temperature characteristic information, and the wind speed and power characteristic information includes: Performing feature splicing on the time domain feature information, the temperature feature information, and the wind speed and power feature information to obtain a fused operation feature information set; generating a similarity metric corresponding to the original wind turbine operation data based on the fused operation feature information set, wherein the similarity metric is used to represent the degree of similarity between each operation data block in the original wind turbine operation data; Inputting the fusion operation feature information set into a pre-trained feature extraction network, wherein the feature extraction network outputs a state feature matrix; Determining at least one similar operating state set and an operating state summary corresponding to each similar operating state set in the original wind turbine operating data according to the state feature matrix and the similarity metric; wherein the state feature matrix is ​​used to represent the state distribution of each operating data block in the original wind turbine operating data in the entire original wind turbine operating data; Counting the number of similar operating state sets; determining the operating state type corresponding to each similar operating state set based on the operating state summary of each similar operating state set; Based on the number of the operating status types, the original wind turbine operating data is classified into a normal operating state, a fault risk precursor state or an abnormal operating state.

4. A method for processing wind turbine operation data according to claim 3, characterized in that: Perform trend prediction and health assessment on the original wind turbine operating data to obtain a corresponding state assessment vector, including: Performing sliding window smoothing processing on the original wind turbine operating data to obtain smoothed data; inputting the smoothed data into a long short-term memory network model to extract time series features; The time series features are superimposed on the original wind turbine operating data, and fully connected layer mapping is performed after superposition to obtain the state assessment vector.

5. A method for processing wind turbine operation data according to claim 4, characterized in that: Perform abnormal trend prediction and potential risk prediction on the original wind turbine operation data to obtain corresponding early warning event data, including: Performing wavelet transform on the original wind turbine operating data to obtain a multi-scale feature map; Inputting the multi-scale feature map into a convolutional neural network model to extract local abnormal features; The local abnormal features are matched with a preset fault risk pattern library for similarity to obtain the warning event data.

6. A method for processing wind turbine operation data according to claim 5, characterized in that: Perform multi-source data fusion and risk evolution prediction on the original wind turbine operation data to obtain a corresponding data processing package, including: Performing feature standardization processing on the original wind turbine operating data to obtain standardized feature data; Inputting the standardized feature data into a graph neural network model to construct a data association graph; Performing a graph convolution operation on the data association graph to extract global fault risk features; The global fault risk feature is input into a classifier to obtain the data processing package.

7. A method for processing wind turbine operation data according to claim 6, characterized in that: The joint training method of the basic analysis model, the early warning analysis model and the deep analysis model includes: Acquire multiple sets of training data sets; wherein the training data sets include sample operating data of wind turbines in different operating states and corresponding state labels; Perform input configuration processing on the initial basic analysis model, initial early warning analysis model, and initial in-depth analysis model respectively to obtain the input configuration strategy of each model; the input configuration strategy of any model is used to indicate the state labels that need to be processed during the training process of the model; Based on the input configuration strategy of each model, the sample running data is input into the corresponding initial model for training to obtain the output result of each model; Determine the model loss information of each model based on the state label and output results of each model; the model loss information of any model is obtained by fitting the trend prediction loss, abnormal trend prediction loss and risk prediction loss. The trend prediction loss is used to represent the category deviation between the operating state predicted by the model and the actual state, the abnormal trend prediction loss is used to represent the position deviation between the abnormal point detected by the model and the actual abnormal point, and the risk prediction loss is used to represent the category deviation between the fault risk type identified by the model and the actual fault risk type; According to the model loss information of each model, the parameters of the initial basic analysis model, the initial early warning analysis model and the initial deep analysis model are optimized respectively to obtain the basic analysis model, the early warning analysis model and the deep analysis model.

8. A method for processing wind turbine operation data according to claim 7, characterized in that: Based on the status label and output results of each model, the model loss information of each model is determined, including: Obtaining a first weight ratio between the trend prediction loss and the abnormal trend prediction loss, and a second weight ratio between the trend prediction loss and the risk prediction loss; Using the first weight ratio and the second weight ratio, adjusting the contribution of the trend prediction loss in the fitting process; The third weight ratio between the abnormal trend prediction loss and the risk prediction loss is used to adjust the contribution of the abnormal trend prediction loss and the risk prediction loss in the fitting process.

9. A method for processing wind turbine operation data according to claim 8, characterized in that: Optimizing parameters of the initial basic analysis model, the initial early warning analysis model, and the initial in-depth analysis model includes: In response to the trend prediction loss in the model loss information being greater than a first preset threshold, preferentially adjusting parameters related to basic operation feature extraction in the initial basic analysis model; In response to the abnormal trend predicted loss in the model loss information being greater than a second preset threshold, preferentially adjusting the parameters related to the abnormal trend in the initial early warning analysis model; In response to the risk prediction loss in the model loss information being greater than a third preset threshold, the parameters related to risk type identification in the initial depth analysis model are preferentially adjusted.

10. A wind turbine operating data processing system, applicable to a wind turbine operating data processing method according to any one of claims 1 to 9, characterized in that: include: Data block unit, used to obtain original wind turbine operation data; Segment-processing the original wind turbine operating data to obtain multiple operating data blocks; a feature extraction unit, configured to extract vibration time domain features from the operation data block to obtain time domain feature information; Extracting temperature variation characteristics of the operation data block to obtain temperature characteristic information; Extracting wind speed and power correlation features from the operation data block to obtain wind speed and power feature information; an operation classification unit, configured to classify the original wind turbine operation data into a normal operation state, a fault risk precursor state, or an abnormal operation state according to the time domain characteristic information, the temperature characteristic information, and the wind speed and power characteristic information; a trend prediction unit, configured to input the original wind turbine operating data into a basic analysis model in response to the original wind turbine operating data being in the normal operating state, and perform trend prediction and health assessment; a fault early warning unit, configured to input the original wind turbine operating data into an early warning analysis model in response to the original wind turbine operating data being in the fault risk precursor state, and perform abnormal trend prediction and potential risk prediction; The fault diagnosis unit is configured to input the original wind turbine operating data into a deep analysis model in response to the original wind turbine operating data being in the abnormal operating state, and perform multi-source data fusion and risk evolution prediction.

Images