CN118669281A - Online state evaluation method, device and equipment of wind motor and storage medium - Google Patents

Abstract

The present invention relates to the technical field of state assessment of new energy equipment, and discloses an online state assessment method, device, equipment and storage medium for a wind turbine. The method includes collecting the operation data of the wind turbine in real time through the equipment, and performing abnormal value detection processing on the collected operation data; extracting target state features from the operation data after abnormal value detection processing through a relief algorithm Relief, and selecting the extracted target state features according to the correlation index of the features; using the operation data containing the target state features for training to establish an online state assessment model for the wind turbine; using the established online state assessment model to predict and judge the real-time collected data, and evaluate the online state of the wind turbine; the device includes a data processing module, a model building module, and a state assessment module. The present invention improves the accuracy and reliability of online state assessment of the wind turbine, and realizes real-time monitoring and prediction of the operation state of the wind turbine.

Description

A method, device, equipment and storage medium for online status assessment of wind turbine

Technical Field

The present invention relates to the technical field of new energy equipment status assessment, and in particular to an online status assessment method, device, equipment and storage medium for a wind turbine.

Background Art

As a green, pollution-free, and renewable energy source with mature technology, wind energy has become an indispensable force in solving energy pollution problems. The rapid development of the wind power industry and the widespread installation of wind turbines have brought huge challenges to the safe and stable operation of wind turbines. For example, due to the influence of wind resource distribution, wind turbines are mostly installed in windy places such as mountainous areas, seaports, and islands. They work in harsh environments such as severe cold and heat, extreme temperature differences, sandstorms, thunderstorms, and irregular wind changes for a long time, resulting in frequent failures of wind turbines, which seriously affects the safe and stable operation of wind turbines. In addition, wind turbines are complex, nonlinear, and strongly coupled control systems. The various components of the unit are coupled with each other. If any component has a minor failure, it will continue to deteriorate under the relatively lagging precise detection technology, and eventually cause a fatal failure, which increases the difficulty of wind turbine maintenance and overhaul. In addition, most wind turbines are located in remote areas, and it is often necessary to pay huge operating costs from the occurrence of failure to the restoration of normal operation, and also to bear the economic losses caused by downtime. According to the literature, my country's research and development capabilities and operation and maintenance technologies for wind turbines are insufficient, and the existing wind turbines in operation will gradually reach the warranty period, causing the operating costs of wind turbines to account for a higher proportion of total expenditures. The wind turbine control system has the highest failure rate, followed by blades, pitch systems, etc., and the average troubleshooting time is the longest for blades, followed by gearboxes and generators. Therefore, fault detection and early warning research is carried out for wind turbine blades with long troubleshooting time and high failure rate, to detect the occurrence of faults and issue early warnings, so that operation and maintenance personnel can respond in time to avoid greater economic losses. However, wind turbine blade fault detection and early warning face the following difficulties: 1) Since the monitoring data of the SCADA system of the wind turbine under normal working conditions contains a lot of noise information and there is partial data loss, the data collected by the SCADA system is discontinuous, which has an important impact on the fault detection results; 2) The wind turbine is a complex, nonlinear, and strongly coupled control system. The various state parameters are coupled with each other, have extremely complex correlations, and have a large number of features that are not related to the fault. If feature selection is not performed, the generalization ability of the model will be affected, and the accuracy of fault detection and early warning will be further affected; 3) In actual operation, due to environmental interference and changes in operation mode, the fault mode of the wind turbine will also change, so the blade cracking fault detection model needs to have a certain generalization ability. Therefore, how to utilize the value of the historical data of the wind turbine SCADA system itself, explore the relationship between various state parameters and wind turbine blade cracking failures, and realize blade cracking fault detection and early warning during wind turbine operation has an important impact on optimizing wind turbine operation and maintenance strategies, promoting safe and stable operation of power systems, and improving the economic effects of wind power generation.

Prior art 1, application number: CN202211548233.6 discloses a performance evaluation and energy efficiency diagnosis method and system for a threshold wind turbine, the method comprising: based on the active power of the target wind turbine to be diagnosed, online evaluation of the power generation performance of the target wind turbine; combining the results of the online evaluation to construct an energy efficiency status index system for the target wind turbine, and determining the benchmark value of each energy efficiency status index; constructing a wind turbine energy efficiency diagnosis ontology knowledge base for the energy efficiency status index system; obtaining real-time operation data of the target wind turbine, identifying energy efficiency anomalies according to the real-time operation data and the benchmark value, and performing energy efficiency fault diagnosis based on the energy efficiency anomaly identification result through the energy efficiency diagnosis ontology knowledge base. Although it is possible to evaluate the power generation performance of the wind turbine in real time, and accurately diagnose the energy efficiency fault mode and energy efficiency fault cause, the accuracy and pertinence of energy efficiency diagnosis are improved. However, the construction of the knowledge base requires a lot of data, which increases the data processing time to a certain extent, making the timeliness of fault diagnosis poor.

Prior art 2, application number: CN202111313388.7 discloses a method for evaluating the real-time output performance of large wind turbines. Based on the data set recorded by the data acquisition and monitoring control system under normal operation of the wind turbine, an outlier detection method using sliding window technology is designed, and a feature extraction model with time information extraction is designed. The XGBoost model is selected to fit the equivalent mathematical model of the wind energy utilization coefficient and the input feature, and the degree of overrun in the training set is statistically analyzed for real-time evaluation of output performance during online application. In the real-time output performance evaluation method, each data point can be judged multiple times, and its degree of outlier is statistically analyzed for outlier detection, so as to obtain more robust and flexible detection results; the time dependency of variables is considered in feature extraction, although the problem that the machine learning XGBoost model does not use time information is solved, the degree of overrun in the training set is used as the basis for online real-time evaluation, which avoids the interference of human subjective factors and ensures the accuracy of the evaluation. However, the function is relatively single, and it can only evaluate real-time processing, and cannot comprehensively evaluate the online status of the wind turbine, resulting in inaccurate evaluation results.

Prior art three, application number: CN202111589047.2 discloses a health assessment method for wind turbine generators based on improved stacked autoencoders, including the following steps: obtaining a training data set: cleaning the wind turbine generator operating data and then performing linear normalization processing to obtain effective training data and test data, and then training and testing the stacked autoencoder model; constructing multiple stacked autoencoder models; training each stacked autoencoder model; integrating and extracting the deep features of the training data set; using the trained benchmark model as the generator online status detector, and inputting the volume data into the benchmark model to obtain and output the health of the wind turbine generator in each time period. Although the accuracy is improved under the premise of ensuring the objectivity of the health assessment results, it is more sensitive and intuitive, and can detect fault trends before the fault occurs. However, it depends on all data points, which makes the model more affected by noise and outliers.

At present, the existing technologies 1, 2 and 3 have the problem that the intelligence level of the evaluation system is low and the basic data collection accuracy is poor. Therefore, the present invention provides an online status evaluation method, device, equipment and storage medium for a wind turbine, which supports real-time information collection of the wind turbine and improves the reliability of equipment operation.

Summary of the invention

The main purpose of the present invention is to provide a method, device, equipment and storage medium for online status evaluation of a wind turbine, so as to solve the problems in the prior art that the evaluation system has a low intelligence level and poor basic data collection accuracy.

To achieve the above object, the present invention provides the following technical solutions:

A method for evaluating the online status of a wind turbine, comprising:

Through sensor equipment, the vibration data, temperature data, current data and power data of the wind turbine are collected in real time, and abnormal value detection and processing are performed on the collected operation data;

Extract target state features from the operating data after outlier detection processing, the target state features are used to describe the operating state of the wind turbine; select the extracted target state features; use the operating data containing the target state features for training to establish an online state assessment model for the wind turbine;

Through cross-validation, the established online status assessment model is evaluated. The established online status assessment model is used to predict and judge the real-time collected data to evaluate the online status of the wind turbine. The assessment results are visualized and corresponding operations and maintenance are performed based on the assessment results.

As a further improvement of the present invention, the process of performing outlier detection processing on the collected operating data includes:

Determine the size of the running data window, start from the beginning of the running data sequence, put the data points in the window into an array, sort the data points in the window, and take the value in the middle position as the median; if the window size is an odd number, the median is the middle value after sorting; if the window size is an even number, the median is the average of the two middle values after sorting;

The median of the current window is used as the value of the current data point to replace the value of the corresponding position in the original running data sequence; the window is moved backward by one position until all data points are traversed to obtain the denoised running data;

Use statistical methods to perform descriptive statistical analysis on the denoised operating data to detect whether there are outliers that deviate significantly from other operating data; use box plots to identify outliers by judging the degree of deviation between data points and other data points, and delete data points identified as outliers from the data set.

As a further improvement of the present invention, the process of identifying outliers includes:

Calculate the mean and standard deviation of the denoised running data, which represent the central tendency and dispersion of the data respectively; draw a box plot to show the distribution of the running data. The box plot includes the upper edge, upper quartile, median, lower quartile and lower edge. The outlier is defined as a data point that exceeds the interquartile range of 1.5 times the upper and lower quartiles.

Calculate the Z-score value of each data point, which indicates the degree of deviation between the data point and the mean value. The formula is: Z = (X-μ)/σ, where X is the data point, μ is the mean, and σ is the standard deviation. Data points with a Z-score greater than 3 or less than -3 are considered outliers.

Grubbs’ Test is performed to identify single outliers. The difference between each data point and the mean is calculated to find data points that deviate significantly from other data points.

As a further improvement of the present invention, the construction process of the online status evaluation model includes:

Use the relief algorithm Relief to extract target state features from the running data processed by outlier detection, and calculate the correlation between the features and the target state to evaluate the importance of the features;

The extracted target state features are selected according to the correlation index, which determines which features are most important for describing the operating state of the wind turbine;

The online state assessment model of the wind turbine is established by training with the operating data containing the target state characteristics. During the training process, the model is constructed using the machine learning algorithm to assess the online state of the wind turbine according to the target state characteristics.

As a further improvement of the present invention, the process of the relief algorithm Relief includes:

Initialize feature weights. For each feature, initialize its weight to 0. Randomly select a sample as a reference sample. Calculate the similarity between other samples and the reference sample based on the Euclidean distance, and sort other samples according to the similarity.

For each feature, calculate the difference between the reference sample and the nearest neighbor sample. If the nearest neighbor sample belongs to the target state category, increase the feature weight; if the nearest neighbor sample does not belong to the target state category, decrease the feature weight;

Update the feature weight according to the difference value and sample weight, and repeat until all samples are traversed; perform feature selection based on the feature weight, and select the features that reach the weight threshold as the target state features.

As a further improvement of the present invention, the process of selecting the extracted target state features according to the correlation index includes:

Calculate the correlation between the features and the target state. For each feature, calculate the correlation index between it and the target state through mutual information;

According to the mutual information value, determine which features are most important for describing the operating status of the wind turbine, set a mutual information threshold, and only select features whose mutual information value exceeds the threshold as target state features; the larger the mutual information value, the higher the correlation between the feature and the target state, and it is considered to be an important feature;

According to the relevance index, sort according to the size of the relevance index and select the top-ranked features as the target state features;

The calculation process of mutual information includes:

Calculate the probability distribution of each feature and target state, that is, the probability distribution of the feature and the probability distribution of the target state; calculate the joint probability distribution of the feature and the target state, that is, the probability of the feature and the target state occurring at the same time;

Step 2: According to the marginal probability distribution and the joint probability distribution, the mutual information between the feature and the target state is calculated. The calculation formula of the mutual information is as follows:

I(X,Y)＝∑P(x,y)*log(P(x,y)/(P(x)*P(y)))

Where X represents the feature, Y represents the target state, P(x,y) represents the probability of the feature and the target state occurring at the same time, and P(x) and P(y) represent the marginal probability distribution of the feature and the target state respectively;

The calculated result of mutual information is compared with the threshold of mutual information, and the features whose mutual information value exceeds the threshold are selected as the target state features.

As a further improvement of the present invention, the process of building a model using a machine learning algorithm includes:

Normalize and preprocess the running data containing target state features, and select the random forest machine learning algorithm for modeling;

The preprocessed data set is divided into a training set and a test set. The training set is used for training and parameter tuning of the online state assessment model, and the test set is used to evaluate the performance of the online state assessment model.

Using the training set to train the selected machine learning algorithm, and adjusting the parameters of the online state assessment model; inputting the training set into the machine learning algorithm for training, the training set including the preprocessed operating data and the corresponding target state;

During the training process, the machine learning algorithm learns the parameters of the online state assessment model based on the relationship between the input features and the target state; the weights and biases of the online state assessment model are adjusted based on the corresponding relationship between the features of the training data and the target state;

During the training process, the depth of the decision tree, the number of trees in the random forest and other parameters are adjusted. After the training is completed, a trained online state evaluation model is obtained;

Among them, by trying different parameter values, the corresponding loss function value is calculated, and the parameter combination that minimizes the loss function is selected as the final model parameter setting.

To achieve the above object, the present invention also provides the following technical solutions:

An online state assessment device for a wind turbine, which is applied to the online state assessment method for a wind turbine, comprises:

The data processing module is used to collect the vibration data, temperature data, current data and power data of the wind turbine in real time through sensors and other equipment, and perform abnormal value detection and processing on the collected operating data;

The model building module is used to extract target state features from the operating data after outlier detection processing through the relief algorithm Relief. The target state features are used to describe the operating state of the wind turbine. The extracted target state features are selected according to the correlation index of the features.

The status assessment module is used to use the operating data containing the target status characteristics for training to establish an online status assessment model for the wind turbine; evaluate the established online status assessment model through cross-validation, use the established online status assessment model to predict and judge the real-time collected data, and evaluate the online status of the wind turbine; visualize the evaluation results, and perform corresponding operations and maintenance based on the evaluation results.

To achieve the above object, the present invention also provides the following technical solutions:

An electronic device, characterized in that it includes a processor and a memory coupled to the processor, wherein the memory stores program instructions executable by the processor; when the processor executes the program instructions stored in the memory, the online status assessment method of the wind turbine is implemented.

To achieve the above object, the present invention also provides the following technical solutions:

A storage medium, characterized in that program instructions are stored in the storage medium, and when the program instructions are executed by a processor, the online status assessment method of the wind turbine can be implemented.

The present invention collects the vibration data, temperature data, current data, power data and other operating data of the wind turbine in real time, and performs abnormal value detection processing, so as to eliminate the interference of abnormal data, ensure the accuracy of subsequent analysis and evaluation, and improve the reliability and accuracy of the online state evaluation of the wind turbine. Through the relief algorithm Relief, the target state features are extracted from the operating data after the abnormal value detection processing, and the features that have an important impact on the operating state of the wind turbine can be identified; the most representative features are selected through the correlation index of the features, which can reduce redundant information and improve the efficiency of model training; the online state evaluation model of the wind turbine is established, and the accurate prediction and judgment of the operating state of the wind turbine can be achieved. The established online state evaluation model is evaluated through cross-validation, and the performance and accuracy of the model can be verified. The established online state evaluation model is used to predict and judge the real-time collected data, and the operating state of the wind turbine can be monitored in real time, potential problems and faults can be discovered in time, and corresponding operations and maintenance can be performed; the evaluation results are visualized, and the operating state of the wind turbine can be intuitively understood, and a decision-making basis and guidance can be provided, so as to improve the operation and maintenance efficiency and reliability of the wind turbine.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a step flow chart of an embodiment of an online state assessment method for a wind turbine according to the present invention;

FIG2 is a schematic flow chart of the steps of performing abnormal value detection processing on the collected operating data according to an embodiment of the online state assessment method of a wind turbine according to the present invention;

3 is a schematic flow chart of steps for constructing an online status assessment model of an embodiment of an online status assessment method for a wind turbine according to the present invention;

FIG4 is a schematic flow chart of steps for visually displaying evaluation results according to an embodiment of an online status evaluation method for a wind turbine according to the present invention;

FIG5 is a schematic diagram of functional modules of an embodiment of an online status assessment device for a wind turbine according to the present invention;

FIG6 is a schematic structural diagram of an electronic device according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of the structure of a storage medium according to an embodiment of the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

The terms "first", "second" and "third" in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present invention are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

As shown in FIG. 1 , this embodiment provides an embodiment of an online state assessment method for a wind turbine. In this embodiment, the online state assessment method for a wind turbine specifically includes the following steps:

Step S1: using sensors and other equipment to collect vibration data, temperature data, current data, power data and other operating data of the wind turbine in real time, and performing abnormal value detection processing on the collected operating data;

Step S2: extracting target state features from the operating data after outlier detection processing, the target state features are used to describe the operating state of the wind turbine; selecting the extracted target state features; using the operating data containing the target state features for training to establish an online state assessment model for the wind turbine;

Step S3: Evaluate the established online status evaluation model through cross-validation, use the established online status evaluation model to predict and judge the real-time collected data, and evaluate the online status of the wind turbine; visualize the evaluation results, and perform corresponding operations and maintenance based on the evaluation results.

Preferably, step S1 of this embodiment collects the vibration data, temperature data, current data, power data and other operating data of the wind turbine in real time, and performs abnormal value detection processing, which can eliminate the interference of abnormal data, ensure the accuracy of subsequent analysis and evaluation, and improve the reliability and accuracy of the online state evaluation of the wind turbine. Step S2 extracts the target state features from the operating data after abnormal value detection processing through the relief algorithm Relief, and can identify the features that have an important impact on the operating state of the wind turbine; selects the most representative features through the correlation index of the features, which can reduce redundant information and improve the efficiency of model training; establishes an online state evaluation model for the wind turbine, and can realize accurate prediction and judgment of the operating state of the wind turbine. Step S3 evaluates the established online state evaluation model through cross-validation, which can verify the performance and accuracy of the model, and uses the established online state evaluation model to predict and judge the real-time collected data, so as to monitor the operating state of the wind turbine in real time, timely discover potential problems and faults, and perform corresponding operations and maintenance; the evaluation results are visualized, so as to intuitively understand the operating state of the wind turbine, provide decision-making basis and guidance, and improve the operation and maintenance efficiency and reliability of the wind turbine.

In summary, this embodiment improves the accuracy and reliability of online status assessment of wind turbines, realizes real-time monitoring and prediction of the operating status of wind turbines, promptly discovers and resolves potential problems and faults, and improves the operation and maintenance efficiency and reliability of wind turbines; ensures the safe operation of wind turbines, reduces failure rates and maintenance costs, improves wind power generation efficiency, and promotes the sustainable development of renewable energy.

Further, as shown in FIG2 , the process of performing outlier detection processing on the collected operating data in step S1 specifically includes the following steps:

Step S11: Determine the size of the running data window, start from the beginning of the running data sequence, put the data points in the window into an array in turn, sort the data points in the window, and take the value in the middle position as the median; if the window size is an odd number, the median is the middle value after sorting; if the window size is an even number, the median is the average of the two middle values after sorting;

Step S12: taking the median of the current window as the value of the current data point, replacing the value of the corresponding position in the original running data sequence; moving the window backward by one position until all data points are traversed to obtain the denoised running data;

Step S13: Use statistical methods to perform descriptive statistical analysis on the denoised operating data to detect whether there are outliers that are significantly deviated from other operating data; use box plots to identify outliers by judging the degree of deviation between data points and other data points, and delete data points identified as outliers from the data set.

Preferably, step S11 of this embodiment determines the size of the operation data window, sorts the data points in the window, and takes the median; by calculating the median, smoothes the data and eliminates noise; reduces the impact of random noise in the data on subsequent analysis, making the data more reliable and stable. Step S12 uses the median of the current window as the value of the current data point to replace the value of the corresponding position in the original operation data sequence; applies the denoised median value to the original data, further smoothes the data and reduces the impact of noise, obtains smoother and more reliable operation data, and provides a better data basis for subsequent outlier detection. Step S13 uses statistical methods, such as box plots, to perform descriptive statistical analysis on the denoised operation data, and detects whether there are outliers that are significantly deviated from other data; detects outliers by statistical analysis methods, that is, identifies data points that are significantly different from other data, identifies and excludes outliers to ensure the accuracy and reliability of subsequent data analysis and decision-making; by deleting outliers, the impact of abnormal data on the overall data distribution and analysis results can be reduced, and the quality and accuracy of the data can be improved.

Furthermore, the process of identifying outliers in step S13 specifically includes the following steps:

Step S131: Calculate the mean and standard deviation of the denoised operating data, which respectively represent the central tendency and dispersion of the data; draw a box plot to display the distribution of the operating data, where the box plot includes the upper edge, upper quartile, median, lower quartile and lower edge, and an outlier is defined as a data point that exceeds the interquartile range of 1.5 times the upper and lower quartiles;

Step S132: Calculate the Z-score value of each data point, which indicates the degree of deviation between the data point and the mean value. The formula is: Z = (X-μ)/σ, where X is the data point, μ is the mean value, and σ is the standard deviation. Data points with a Z-score greater than 3 or less than -3 are considered outliers.

Step S133: Perform Grubbs’ Test to identify single outliers, and find data points that deviate significantly from other data points by calculating the difference between each data point and the mean.

Preferably, step S131 of this embodiment describes the central tendency and dispersion of the data, and intuitively displays the distribution of the data; through the box plot, it can be identified whether the data point deviates from the normal range, and the data point that exceeds the interquartile range of 1.5 times the upper and lower quartiles is defined as an outlier; the data is visualized for analysis so as to intuitively find outliers. Step S132 calculates the Z-score. By calculating the degree of deviation between each data point and the mean, it is measured by the Z-score. If the Z-score is greater than 3 or less than -3, it means that the deviation between the data point and the mean is very large and is regarded as an outlier; the abnormal data points that deviate significantly from the mean are identified by statistical methods. Step S133 uses Grubbs'Test to identify single outliers, and by calculating the difference between each data point and the mean, finds out the data points that deviate significantly from other data points; by performing statistical tests, it can be determined which data points are significantly deviated outliers; and single abnormal data points are identified by statistical methods.

In summary, this embodiment performs outlier detection and processing on the collected operating data. By identifying and deleting outliers, the quality and accuracy of the data can be improved, making subsequent data analysis and modeling work more reliable and effective.

Further, as shown in FIG3 , the process of constructing the online status evaluation model in step S2 specifically includes the following steps:

Step S21: extracting target state features from the operating data processed by outlier detection using the relief algorithm Relief, and calculating the correlation between the features and the target state to evaluate the importance of the features;

Step S22: selecting the extracted target state features according to the correlation index, the correlation index determining which features are most important for describing the operating state of the wind turbine;

Step S23: using the operating data containing the target state characteristics for training to establish an online state assessment model for the wind turbine; during the training process, using a machine learning algorithm to construct a model to assess the online state of the wind turbine according to the target state characteristics.

Preferably, step S21 of this embodiment uses the relief algorithm Relief to extract target state features from the operating data processed by outlier detection, and calculates the correlation between the features and the target state to evaluate the importance of the features; through data analysis and feature extraction, features related to the operating state of the wind turbine are obtained, thereby improving the accuracy and reliability of the model; providing an important data basis for subsequent feature selection and model establishment. Step S22 selects the extracted target state features according to the correlation index, and the correlation index determines which features are most important for describing the operating state of the wind turbine; through correlation analysis, features that have an important impact on the operating state of the wind turbine are screened out, the dimension of the features is reduced, and the simplicity and interpretability of the model are improved; the efficiency and accuracy of the model are improved, while reducing the complexity of modeling. Step S23 uses the operating data containing the target state characteristics for training to establish an online state evaluation model for the wind turbine. During the training process, a machine learning algorithm is used to build a model to evaluate the online state of the wind turbine according to the target state characteristics; through model training, the association between the characteristics and the target state is learned, so as to achieve accurate evaluation of the online state of the wind turbine; a fast and accurate method is provided to evaluate the state of the wind turbine, timely discover potential problems and take corresponding operation and maintenance measures, so as to improve the stability and reliability of the wind turbine.

Furthermore, the process of the relief algorithm Relief in step S21 specifically includes the following steps:

Step S211: Initialize feature weights. For each feature, initialize its weight to 0; randomly select a sample as a reference sample; calculate the similarity between other samples and the reference sample according to the Euclidean distance, and sort other samples according to the similarity;

Step S212: For each feature, calculate the difference between the reference sample and the nearest neighbor sample. If the nearest neighbor sample belongs to the target state category, increase the feature weight; if the nearest neighbor sample does not belong to the target state category, reduce the feature weight;

Step S213: Update the feature weight according to the difference value and the sample weight, and repeat until all samples are traversed; perform feature selection according to the feature weight, and select the feature that reaches the weight threshold as the target state feature.

Preferably, step S211 of this embodiment initializes the feature weights and selects a reference sample, and calculates the similarity between other samples and the reference sample; by calculating the similarity, the degree of similarity between each sample and the reference sample can be understood, providing basic data for subsequent steps. Step S212 calculates the difference value between the reference sample and the nearest neighbor sample, and adjusts the feature weight according to the category of the nearest neighbor sample; by calculating the difference value and adjusting the weight, the importance of each feature to the target state can be evaluated, and the weight can be adjusted according to the category of the nearest neighbor sample to distinguish the contribution to the target state. Step S213 updates the feature weight according to the difference value and sample weight; by iteratively updating the feature weight, the ability to discriminate the target state can be gradually improved to more accurately select the target state feature.

In summary, the mitigation algorithm Relief of this embodiment improves the accuracy and reliability of the wind turbine online status assessment model by evaluating the correlation between the features and the target state. By accurately evaluating the online status of the wind turbine, potential problems can be detected in time, improving the operating efficiency and safety of the wind turbine.

Furthermore, the process of selecting the extracted target state features according to the correlation index in step S22 specifically includes the following steps:

Step S221: Calculate the correlation between the feature and the target state. For each feature, calculate the correlation index between it and the target state through mutual information.

Step S222: according to the mutual information value, determine which features are most important for describing the operating state of the wind turbine, set a mutual information threshold, and only select features whose mutual information value exceeds the threshold as target state features; the larger the mutual information value, the higher the correlation between the feature and the target state, and is considered to be an important feature;

Step S223: sort according to the relevance index and select the top-ranked features as the target state features.

Preferably, step S221 of this embodiment calculates the correlation between the feature and the target state, calculates the correlation index between it and the target state through mutual information, quantifies the degree of mutual dependence between the feature and the target state, and measures the contribution of the feature to the target state through mutual information; the calculated correlation index can be used to evaluate the correlation between the feature and the target state. Step S222 determines which features are most important for describing the target state according to the value of the mutual information, and only selects features whose mutual information values exceed the threshold as target state features by setting a threshold of the mutual information; the larger the mutual information value, the higher the correlation between the feature and the target state, which is considered to be an important feature; screen out features with higher correlation to the target state to retain features useful for target state prediction. Step S223 sorts according to the correlation index according to the size of the correlation index, and selects the top-ranked features as target state features; sorts according to the importance of the correlation index, and selects the features with the highest correlation as target state features; by selecting the top-ranked features, features that contribute more to target state prediction can be retained.

In summary, this embodiment selects features with high correlation with the target state, thereby improving the prediction accuracy and interpretability of the target state; feature selection can reduce data dimensions, reduce the impact of redundant features, and improve the generalization ability and efficiency of the model. At the same time, feature selection can also reduce the complexity and computational cost of the model, making the model easier to understand and explain.

Furthermore, the mutual information calculation process in step S221 specifically includes the following steps:

Step S2211: Calculate the probability distribution of each feature and target state, that is, the probability distribution of the feature and the probability distribution of the target state; calculate the joint probability distribution of the feature and the target state, that is, the probability of the feature and the target state occurring at the same time;

Step S2212: Calculate the mutual information between the feature and the target state according to the marginal probability distribution and the joint probability distribution. The calculation formula of the mutual information is as follows:

I(X,Y)＝∑P(x,y)*log(P(x,y)/(P(x)*P(y)))

Where X represents the feature, Y represents the target state, P(x,y) represents the probability of the feature and the target state occurring at the same time, and P(x) and P(y) represent the marginal probability distribution of the feature and the target state respectively;

Step S2213: Compare the calculation result of the mutual information with the threshold of the mutual information, and select the feature whose mutual information value exceeds the threshold as the target state feature.

Preferably, step S2211 of this embodiment calculates the probability distribution of features and target states, which can help understand the distribution of features and target states and provide a basis for the subsequent calculation of mutual information. Step S2212 can measure the correlation between features and target states by calculating mutual information. The larger the mutual information value, the higher the correlation between the feature and the target state, which is considered to be an important feature. Step S2213 selects features whose mutual information values exceed the threshold as target state features based on the threshold of the mutual information. By setting the threshold to control the number of selected features, only features with high correlation with the target state are selected, reducing redundant information.

In summary, this embodiment selects the feature that best describes the target state based on the correlation between the feature and the target state; it is of great significance for analyzing and predicting the target state, and can help to deeply understand the changing law of the target state, thereby making effective decisions and controls.

Furthermore, the process of using the machine learning algorithm to build a model in step S23 specifically includes the following steps:

Step S231: normalize and preprocess the operation data containing the target state characteristics, and select the random forest machine learning algorithm for modeling;

Step S232: Divide the preprocessed data set into a training set and a test set, the training set is used for training and parameter tuning of the online state assessment model, and the test set is used for evaluating the performance of the online state assessment model;

Step S233: Use the training set to train the selected machine learning algorithm and adjust the parameters of the online state assessment model.

Preferably, the normalization preprocessing in step S231 of this embodiment can unify the value ranges of different features and avoid excessive influence of certain features on model training; the random forest algorithm is selected for modeling because random forest has good feature selection ability and robustness. Step S232 divides the data set into a training set and a test set, which can be used for model training and evaluation. The training set is used to train the model, optimize the parameters and algorithms of the model so that it can fully learn the characteristics of the target state, and the test set is used to evaluate the performance of the model and verify the generalization ability and prediction accuracy of the model. Step S233 can optimize the performance of the model and improve the prediction accuracy of the model by training the model and adjusting the parameters. By optimizing the online status evaluation model, the accuracy and reliability of the online status evaluation of the wind turbine can be improved, providing support for the operation and maintenance management of the wind turbine and reducing the risk of failure and maintenance costs.

Furthermore, the process of using the training set to train the selected machine learning algorithm in step S233 specifically includes the following steps:

Step S2331: inputting a training set into a machine learning algorithm for training, wherein the training set includes preprocessed operating data and corresponding target states;

Step S2332: During the training process, the machine learning algorithm learns the parameters of the online state assessment model according to the relationship between the input features and the target state; and adjusts the weights and biases of the online state assessment model according to the corresponding relationship between the features of the training data and the target state;

The parameters of the online state evaluation model are initialized, and the initial values are random values; the data samples in the training set are used to input the online state evaluation model, and the prediction results of the online state evaluation model are calculated and compared with the actual target state; according to the comparison results, the value of the loss function is calculated to measure the gap between the prediction results of the online state evaluation model and the target state; the loss function is derived to obtain the gradient of the parameters of the online state evaluation model. The gradient represents the rate of change of the loss function at the current parameter value and indicates the direction of parameter update; according to the direction and size of the gradient, the parameters of the online state evaluation model are updated using gradient descent. The updating process is to subtract the learning rate from the current parameter value and multiply it by the gradient. The learning rate determines the step size of each parameter update; the iteration is repeated until the specified number of iterations is reached or the change in the loss function is small enough to meet the preset convergence condition;

Step S2333: During the training process, parameters such as the depth of the decision tree and the number of trees in the random forest are adjusted. After the training is completed, a trained online state evaluation model is obtained;

Among them, by trying different parameter values, the corresponding loss function value is calculated, and the parameter combination that minimizes the loss function is selected as the final model parameter setting.

Preferably, step S2331 of this embodiment inputs the training set into the machine learning algorithm for training, and trains the model through the data samples in the training set, including the preprocessed operation data and the corresponding target state; in order to enable the machine learning algorithm to learn the relationship between the input features and the target state, so as to be able to make accurate predictions and classifications. Step S2332 During the training process, the machine learning algorithm learns the parameters of the online state evaluation model according to the relationship between the input features and the target state, and adjusts the weights and biases of the model so that the model can better fit the training data; in order to minimize the loss function, the prediction results of the model are as close to the target state as possible. Step S2333 During the training process, the depth of the decision tree, the number of trees in the random forest and other parameters are adjusted, and the corresponding loss function values are calculated by trying different parameter values, and the parameter combination that minimizes the loss function is selected as the final model parameter setting, in order to optimize the performance of the model and improve the accuracy and generalization ability of the model.

In summary, this embodiment trains the model through the training set so that the model can learn the relationship between the input features and the target state, and minimizes the loss function by adjusting the model parameters so that the prediction results of the model are as close as possible to the target state; the ultimate goal is to obtain a trained online state evaluation model that can predict and classify new input features, thereby providing accurate results for subsequent applications. At the same time, through parameter tuning, the performance and generalization ability of the model can also be improved, so that the model has better adaptability and promotion ability.

Further, as shown in FIG4 , the process of visually displaying the evaluation results in step S3 specifically includes the following steps:

Step S31: Compare the prediction result of the online state evaluation model with the actual target state, calculate the evaluation index, and evaluate the performance of the online state evaluation model;

Step S32: input the real-time collected data into the established online status assessment model to obtain the prediction result of the model;

Step S33: Display the prediction results of the online status evaluation model in the form of a chart, and take timely maintenance measures based on the prediction results of the online status evaluation.

Preferably, step S31 of this embodiment can calculate evaluation indicators such as accuracy, recall rate, F1 value, etc. by comparing the prediction results of the online state evaluation model with the actual target state to evaluate the performance of the online state evaluation model; help judge the accuracy and reliability of the model, and also provide a reference for subsequent model improvement. Step S32 inputs the real-time collected data into the online state evaluation model to obtain the prediction results of the model; uses the established model to predict the real-time data to obtain the current online state of the wind turbine. The prediction result can be used to determine whether the wind turbine is operating normally or there is an abnormality. Step S33 displays the prediction results of the online state evaluation model in the form of a chart, which can intuitively display the online state of the wind turbine. Through visual display, operators can understand the operation of the wind turbine more quickly and take timely measures for maintenance and repair; it helps to improve the operating efficiency and reliability of the wind turbine, reduce downtime and maintenance costs, and improve the stability and economy of wind power generation.

In summary, by implementing the above steps, this embodiment can evaluate the online status of the wind turbine and take corresponding operation and maintenance measures in time, thereby ensuring the normal operation of the wind turbine and improving the reliability and efficiency of the wind power generation system.

Furthermore, the process of displaying the prediction results in the form of a chart in step S33 specifically includes the following steps:

Step S331: converting the prediction results into time series data;

Step S332: Select a line chart to display the prediction results. The horizontal axis is the time axis, indicating the continuity of time, and the data points are arranged in chronological order. The vertical axis is the value of the prediction result, which is the evaluation score of the state. The title is the change trend of the online state evaluation result. The horizontal axis label is time, and the vertical axis label is the evaluation result.

Step S333: Use data visualization tools to present the prediction results in the form of charts.

Preferably, step S331 of this embodiment converts the prediction results into time series data by arranging the prediction results in chronological order to facilitate subsequent visual display; in order to associate the prediction results with time, the changing trend of the prediction results can be observed more clearly. Step S332 selects a line chart to display the prediction results. A line chart is used to display the changing trend of the prediction results. The horizontal axis represents the continuity of time, and the vertical axis represents the value of the evaluation result. The setting of the title and label can provide a description and explanation of the content of the chart; through visual display, the changing trend of the prediction results can be intuitively observed to help users understand and analyze the performance of the evaluation model. Step S333 uses a data visualization tool to present the prediction results in the form of a chart, and converts the processed data into a chart through a data visualization tool, such as using Matplotlib, Seaborn and other libraries in Python; converts the data into an intuitive chart form, providing a more intuitive and easy-to-understand display method, so that users can more easily understand and analyze the prediction results and take corresponding maintenance measures in time.

As shown in FIG5 , this embodiment further provides an embodiment of an online state assessment device for a wind turbine. In this embodiment, the online state assessment device for a wind turbine is applied to the online state assessment method for a wind turbine in the above embodiment. The online state assessment device for a wind turbine comprises a data processing module 1, a model building module 2, and a state assessment module 3 which are electrically connected in sequence;

Among them, the data processing module 1 is used to collect the vibration data, temperature data, current data, power data and other operating data of the wind turbine in real time through sensors and other equipment, and perform outlier detection processing on the collected operating data; the model building module 2 is used to extract the target state characteristics from the operating data after outlier detection processing through the relief algorithm Relief, and the target state characteristics are used to describe the operating state of the wind turbine; according to the correlation index of the characteristics, the extracted target state characteristics are selected; the state evaluation module 3 is used to use the operating data containing the target state characteristics for training to establish an online state evaluation model of the wind turbine; through cross-validation, the established online state evaluation model is evaluated, and the real-time collected data is predicted and judged using the established online state evaluation model to evaluate the online state of the wind turbine; the evaluation results are visualized and corresponding operations and maintenance are performed according to the evaluation results.

Preferably, the data processing module 1 of this embodiment can identify and eliminate possible abnormal data by performing abnormal value detection processing on the real-time collected wind turbine operation data, thereby ensuring the accuracy and reliability of subsequent model construction and state evaluation. The model construction module 2 uses the relief algorithm Relief to extract target state features from the operation data after abnormal value detection processing. These features can effectively describe the operation state of the wind turbine. The quality and effect of the model can be improved by selecting features through correlation indicators. The state evaluation module 3 uses the operation data containing the target state features for training to establish an online state evaluation model for the wind turbine. The model is evaluated through cross-validation to evaluate the accuracy and generalization ability of the model; the established online state evaluation model is used to predict and judge the real-time collected data, so that the abnormal state of the wind turbine can be discovered in time, providing a basis for operation and maintenance.

In summary, this embodiment includes improving the reliability and accuracy of data, extracting effective features, establishing an accurate status assessment model, and timely discovering and handling abnormal situations; improving the operating efficiency and reliability of wind turbines, reducing failures and downtime, reducing maintenance costs, and improving power generation efficiency.

Furthermore, the data processing module 1 specifically includes:

The data sorting submodule is used to determine the size of the running data window, starting from the beginning of the running data sequence, and sequentially put the data points in the window into an array, sort the data points in the window, and take the value in the middle position as the median; if the window size is an odd number, the median is the middle value after sorting; if the window size is an even number, the median is the average of the two middle values after sorting;

The data denoising submodule is used to use the median of the current window as the value of the current data point to replace the value of the corresponding position in the original running data sequence; the window is moved backward by one position until all data points are traversed to obtain the denoised running data;

The outlier judgment submodule is used to use statistical methods to perform descriptive statistical analysis on the denoised operating data to detect whether there are outliers that are significantly deviated from other operating data; use box plots to identify outliers by judging the degree of deviation between data points and other data points, and delete data points identified as outliers from the data set.

Preferably, the data sorting submodule of this embodiment sorts the data points in the window and takes the median as the representative value of the window; determines the size of the window, and selects a suitable window representative value to balance the change of the data. The data denoising submodule uses the median of the current window as the value of the current data point, replacing the value of the corresponding position in the original data sequence; removes the noise in the original data by using the median, thereby obtaining a smoother data sequence. The outlier judgment submodule uses statistical methods and box-and-whisker plots to detect whether there are obviously deviated outliers in the denoised data; identifies outliers and deletes them from the data set based on the distribution of the data and the definition of outliers; ensures the accuracy and reliability of the data and avoids the influence of outliers on the results.

Furthermore, the model building module 2 specifically includes:

The feature evaluation submodule is used to extract the target state features from the running data processed by outlier detection using the relief algorithm Relief, and calculate the correlation between the features and the target state to evaluate the importance of the features;

A feature determination submodule is used to select the extracted target state features according to a correlation index, which determines which features are most important for describing the operating state of the wind turbine;

The model training submodule uses the operating data containing the target state characteristics for training to establish an online state assessment model for the wind turbine. During the training process, a machine learning algorithm is used to build a model to evaluate the online state of the wind turbine based on the target state characteristics.

Preferably, the feature evaluation submodule of this embodiment uses the Relief algorithm to extract target state features from the operating data processed by outlier detection, and evaluates the correlation between the features and the target state; determines which features are most important for describing the operating state of the wind turbine, thereby providing valuable feature information for subsequent feature selection and model training. The feature determination submodule selects the extracted target state features according to the correlation index; by screening out features with high correlation with the target state, the accuracy and effect of the model are improved, the feature dimension is reduced, and the interpretability and computational efficiency of the model are improved. The model training submodule uses the operating data containing the target state features for training to establish an online state evaluation model for the wind turbine; constructs a model through a machine learning algorithm, uses the target state features to evaluate the online state of the wind turbine, provides real-time prediction and monitoring of the wind turbine state, and helps improve the operating efficiency and reliability of the wind turbine.

Furthermore, the status assessment module 3 specifically includes:

The performance evaluation submodule is used to compare the prediction results of the online state evaluation model with the actual target state, calculate the evaluation index, and evaluate the performance of the online state evaluation model;

The result output submodule is used to input the real-time collected data into the established online status evaluation model to obtain the prediction results of the model;

The result display submodule is used to display the prediction results of the online status evaluation model in the form of charts and graphs, and take timely maintenance measures based on the prediction results of the online status evaluation.

Preferably, the performance evaluation submodule of this embodiment evaluates the performance of the model by comparing the prediction results of the online state evaluation model with the actual target state and calculating evaluation indicators such as accuracy, precision, recall, etc.; helps evaluate the accuracy and reliability of the online state evaluation model, understands the prediction ability of the model, and provides a basis for model improvement and optimization. The result output submodule inputs the real-time collected data into the online state evaluation model to obtain the prediction results of the model; realizes real-time state evaluation, judges the operating state of the wind turbine according to the prediction results of the model, and provides a basis for taking timely measures for maintenance and adjustment. The result display submodule displays the prediction results of the online state evaluation model in the form of charts, such as curve charts, scatter plots, etc.; through intuitive chart display, it helps users understand the operating state of the wind turbine, discover abnormal conditions or trends in time, make early warnings and decisions, and improve the operating efficiency and reliability of the wind turbine.

In summary, the purpose of the entire status assessment module of this embodiment is to achieve real-time monitoring and assessment of the operating status of the wind turbine by establishing a model, evaluating model performance and displaying results, so as to improve the operating efficiency of the wind turbine, reduce the failure rate, and provide a scientific basis for maintenance and adjustment.

As shown in FIG. 6 , this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 4 includes a processor 41 and a memory 42 coupled to the processor 41 .

The memory 42 stores program instructions for implementing the online state assessment method for a wind turbine according to any of the above embodiments.

The processor 41 is used to execute program instructions stored in the memory 42 to perform online status assessment of the wind turbine.

The processor 41 may also be referred to as a CPU (Central Processing Unit). The processor 41 may be an integrated circuit chip having signal processing capabilities. The processor 41 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

Further, FIG. 7 is a schematic diagram of the structure of a storage medium of an embodiment of the present application, wherein the storage medium 5 of the embodiment of the present application stores program instructions 51 capable of implementing all the above methods, wherein the program instructions 51 can be stored in the above storage medium in the form of a software product, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, a server, a mobile phone, and a tablet.

In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units. The above is only an implementation mode of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the specification and drawings of the present invention, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present invention.

The specific implementation methods of the invention are described in detail above, but they are only examples, and the invention is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the invention, and therefore, the equalization, modification, improvement, etc. made without departing from the spirit and principle of the invention should be included in the scope of the invention.

Claims (10)

1. A method for online status assessment of a wind turbine, characterized in that the method for online status assessment of a wind turbine comprises: Through sensor equipment, the vibration data, temperature data, current data and power data of the wind turbine are collected in real time, and abnormal value detection and processing are performed on the collected operation data; Extract target state features from the operating data after outlier detection processing, the target state features are used to describe the operating state of the wind turbine; select the extracted target state features; use the operating data containing the target state features for training to establish an online state assessment model for the wind turbine; Through cross-validation, the established online status assessment model is evaluated. The established online status assessment model is used to predict and judge the real-time collected data to evaluate the online status of the wind turbine. The assessment results are visualized and corresponding operations and maintenance are performed based on the assessment results. 2. The online state assessment method of a wind turbine according to claim 1 is characterized in that the process of performing abnormal value detection processing on the collected operating data comprises: Determine the size of the running data window, start from the beginning of the running data sequence, put the data points in the window into an array, sort the data points in the window, and take the value in the middle position as the median; if the window size is an odd number, the median is the middle value after sorting; if the window size is an even number, the median is the average of the two middle values after sorting; The median of the current window is used as the value of the current data point to replace the value of the corresponding position in the original running data sequence; the window is moved backward by one position until all data points are traversed to obtain the denoised running data; Use statistical methods to perform descriptive statistical analysis on the denoised operating data to detect whether there are outliers that deviate significantly from other operating data; use box plots to identify outliers by judging the degree of deviation between data points and other data points, and delete data points identified as outliers from the data set. 3. The online state assessment method of a wind turbine according to claim 2, characterized in that the process of identifying abnormal values comprises: Calculate the mean and standard deviation of the denoised running data, which represent the central tendency and dispersion of the data respectively; draw a box plot to show the distribution of the running data. The box plot includes the upper edge, upper quartile, median, lower quartile and lower edge. The outlier is defined as a data point that exceeds the interquartile range of 1.5 times the upper and lower quartiles. Calculate the Z-score value of each data point, which indicates the degree of deviation between the data point and the mean value. The formula is: Z = (X-μ)/σ, where X is the data point, μ is the mean, and σ is the standard deviation. Data points with a Z-score greater than 3 or less than -3 are considered outliers. Grubbs’ Test is performed to identify single outliers. The difference between each data point and the mean is calculated to find data points that deviate significantly from other data points. 4. The online state assessment method of a wind turbine according to claim 1, characterized in that the process of constructing an online state assessment model comprises: Use the relief algorithm Relief to extract target state features from the running data processed by outlier detection, and calculate the correlation between the features and the target state to evaluate the importance of the features; The extracted target state features are selected according to the correlation index, which determines which features are most important for describing the operating state of the wind turbine; The online state assessment model of the wind turbine is established by training with the operating data containing the target state characteristics. During the training process, the model is constructed using the machine learning algorithm to assess the online state of the wind turbine according to the target state characteristics. 5. The online state assessment method of a wind turbine according to claim 4, characterized in that the relief algorithm relief process comprises: Initialize feature weights. For each feature, initialize its weight to 0. Randomly select a sample as a reference sample. Calculate the similarity between other samples and the reference sample based on the Euclidean distance, and sort other samples according to the similarity. For each feature, calculate the difference between the reference sample and the nearest neighbor sample; if the nearest neighbor sample belongs to the target state category, increase the feature weight; if the nearest neighbor sample does not belong to the target state category, reduce the feature weight; Update the feature weight according to the difference value and sample weight, and repeat until all samples are traversed; perform feature selection based on the feature weight, and select the features that reach the weight threshold as the target state features. 6. The online state assessment method of a wind turbine according to claim 4 is characterized in that the process of selecting the extracted target state features according to the correlation index comprises: Calculate the correlation between the features and the target state. For each feature, calculate the correlation index between it and the target state through mutual information; According to the mutual information value, determine which features are most important for describing the operating status of the wind turbine, set a mutual information threshold, and only select features whose mutual information value exceeds the threshold as target state features; the larger the mutual information value, the higher the correlation between the feature and the target state, and it is considered to be an important feature; According to the relevance index, sort according to the size of the relevance index and select the top-ranked features as the target state features; The calculation process of mutual information includes: Calculate the probability distribution of each feature and target state, that is, the probability distribution of the feature and the probability distribution of the target state; calculate the joint probability distribution of the feature and the target state, that is, the probability of the feature and the target state occurring at the same time; Step 2: According to the marginal probability distribution and the joint probability distribution, the mutual information between the feature and the target state is calculated. The calculation formula of the mutual information is as follows: I(X,Y)＝∑P(x,y)*log(P(x,y)/(P(x)*P(y))) Where X represents the feature, Y represents the target state, P(x,y) represents the probability of the feature and the target state occurring at the same time, and P(x) and P(y) represent the marginal probability distribution of the feature and the target state respectively; The calculated result of mutual information is compared with the threshold of mutual information, and the features whose mutual information value exceeds the threshold are selected as the target state features. 7. The online state assessment method of a wind turbine according to claim 4, characterized in that the process of constructing a model using a machine learning algorithm comprises: Normalize and preprocess the running data containing target state features, and select the random forest machine learning algorithm for modeling; The preprocessed data set is divided into a training set and a test set. The training set is used for training and parameter tuning of the online state assessment model, and the test set is used to evaluate the performance of the online state assessment model. Using the training set to train the selected machine learning algorithm, and adjusting the parameters of the online state assessment model; inputting the training set into the machine learning algorithm for training, the training set including the preprocessed operating data and the corresponding target state; During the training process, the machine learning algorithm learns the parameters of the online state assessment model based on the relationship between the input features and the target state; the weights and biases of the online state assessment model are adjusted based on the corresponding relationship between the features of the training data and the target state; During the training process, the depth of the decision tree and the number of trees in the random forest are adjusted. After the training is completed, a trained online state evaluation model is obtained; Among them, by trying different parameter values, the corresponding loss function value is calculated, and the parameter combination that minimizes the loss function is selected as the final model parameter setting. 8. An online status assessment device for a wind turbine, applied to the online status assessment method for a wind turbine according to any one of claims 1 to 7, characterized in that the online status assessment device for a wind turbine comprises: The data processing module is used to collect the vibration data, temperature data, current data and power data of the wind turbine in real time through sensor equipment, and perform abnormal value detection and processing on the collected operating data; The model building module is used to extract target state features from the operating data after outlier detection processing through the relief algorithm Relief. The target state features are used to describe the operating state of the wind turbine. The extracted target state features are selected according to the correlation index of the features. The status assessment module is used to use the operating data containing the target status characteristics for training to establish an online status assessment model for the wind turbine; evaluate the established online status assessment model through cross-validation, use the established online status assessment model to predict and judge the real-time collected data, and evaluate the online status of the wind turbine; visualize the evaluation results, and perform corresponding operations and maintenance based on the evaluation results. 9. An electronic device, characterized in that it comprises a processor and a memory coupled to the processor, the memory storing program instructions executable by the processor; when the processor executes the program instructions stored in the memory, the online status assessment method of the wind turbine as described in any one of claims 1 to 7 is implemented. 10. A storage medium, characterized in that program instructions are stored in the storage medium, and when the program instructions are executed by a processor, the online state assessment method of a wind turbine according to any one of claims 1 to 7 can be implemented.