Fan blade icing diagnosis method considering sample imbalance condition

Abstract

The invention provides a fan blade icing diagnosis method considering a sample imbalance condition, and the method comprises the following steps: S1, obtaining the operation data of a wind turbine generator set as initial sample data, and carrying out the preprocessing of the initial sample data; s2, processing the unbalanced initial sample data to form a balanced sample; s3, extracting icing identification features from the balance sample; s4, the icing identification features are input into the Light GBM model, and the Light GBM model is trained; s5, operation data of a to-be-tested wind turbine generator set are obtained to serve as real-time samples, the real-time samples are preprocessed and then subjected to feature extraction, then the extracted features are input into the trained LightGBM model to obtain the icing grade, by means of the method, when a small number of samples exist in the wind turbine generator set, unbalance processing is carried out on the sample data, and the icing grade of the to-be-tested wind turbine generator set is obtained. Therefore, the final balance data is obtained, the accuracy and reliability of fan blade icing diagnosis are effectively improved, and accurate data support is provided for subsequent deicing operation and maintenance.

Description

Technical Field

[0001] The invention relates to an icing diagnosis method, and in particular to an icing diagnosis method for fan blades taking into account sample imbalance conditions. Background Art

[0002] Wind power generation, as an important component of clean energy, has been widely used. However, in cold regions, ice on wind turbine blades is a common problem, which seriously affects the operating efficiency and safety of wind turbines.

[0003] Blade icing not only reduces power generation efficiency and increases blade load, but may also cause equipment damage due to imbalance. The current mainstream blade icing diagnosis methods mainly rely on manual experience, physical modeling, or machine learning techniques based on balanced data sets. However, in actual applications, the number of samples of blade icing events is usually far less than that of non-icing samples, resulting in a serious imbalance in data samples. This data characteristic makes it impossible for traditional algorithms to effectively capture the characteristics of minority samples, resulting in reduced reliability of diagnostic results.

[0004] Therefore, in order to solve the above technical problems, it is urgent to propose a new technical means. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a wind turbine blade icing diagnosis method taking into account the sample imbalance condition, and to perform imbalance processing on the sample data when there are a small number of samples in the wind turbine set, so as to obtain the final balanced data, thereby effectively improving the accuracy and reliability of the wind turbine blade icing diagnosis, and providing accurate data support for subsequent deicing operation and maintenance.

[0006] The present invention provides a fan blade icing diagnosis method considering sample imbalance conditions, comprising the following steps:

[0007] S1. Obtaining wind turbine operation data as initial sample data, and preprocessing the initial sample data;

[0008] S2. Process the unbalanced initial sample data to form a balanced sample;

[0009] S3. Extract ice cover identification features from equilibrium samples;

[0010] S4. Input the ice cover recognition features into the LightGBM model to train the LightGBM model;

[0011] S5. Obtain the operating data of the wind turbine to be tested as real-time samples, perform feature extraction after preprocessing the real-time samples, and then input the extracted features into the trained LightGBM model to obtain the ice coverage level.

[0012] Further, in step S2, processing the unbalanced sample data to form a balanced sample specifically includes:

[0013] Select m samples from the initial sample data as benchmark samples X m :

[0014] X m =(x1,x2, … ,x i ,…,x m );

[0015] Calculate the Euclidean distance between the benchmark sample and the remaining sample data in the initial sample data, and select the k samples with the smallest distance in the remaining sample data as the nearest neighbor samples;

[0016] Randomly select M neighbor samples from the k neighbor samples as auxiliary samples Y mk :

[0017] Y mk =(y m1 ,y m2 , … ,y mj ,…,y mk );

[0018] Reconstruct the new sample as a balanced sample:

[0019] X new =x i +γ(y mj -x i ), where γ is a random coefficient between [0,1].

[0020] Furthermore, the initial sample data and the real-time sample data include temperature, humidity, wind speed and vibration signals of wind turbine blades.

[0021] Further, preprocessing the initial sample and the real-time sample includes:

[0022] The box plot detection method is used to remove outliers from sample data;

[0023] The sample data after removing outliers is processed by linear interpolation;

[0024] Perform Z-score normalization on the interpolated sample data:

[0025]

[0026] Among them, z represents the sample data after normalization, x represents the sample data before normalization, μ is the mean of the sample data, and σ represents the standard deviation of the sample data.

[0027] Further, the extracting of ice cover identification features includes:

[0028] Determine the mean, standard deviation, and peak value of the sample data;

[0029] The principal component analysis method is used to analyze the mean, standard deviation and peak value of the sample data to obtain the final ice cover characteristics.

[0030] The beneficial effects of the present invention are as follows: Through the present invention, when there are a small number of samples in the wind turbine, the sample data is unbalancedly processed to obtain the final balanced data, thereby effectively improving the accuracy and reliability of the wind turbine blade icing diagnosis and providing accurate data support for subsequent deicing operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0032] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0033] The present invention is further described in detail below:

[0034] The present invention provides a fan blade icing diagnosis method considering sample imbalance conditions, comprising the following steps:

[0035] S1. Obtaining wind turbine operation data as initial sample data, and preprocessing the initial sample data;

[0036] S2. Process the unbalanced initial sample data to form a balanced sample;

[0037] S3. Extract ice cover identification features from equilibrium samples;

[0038] S4. Input the ice cover recognition features into the LightGBM model to train the LightGBM model;

[0039] S5. Implement the acquisition of the operating data of the wind turbine to be tested as real-time samples, and perform feature extraction after preprocessing the real-time samples. Then, the extracted features are input into the trained LightGBM model to obtain the icing level. Through the above method, when there are a small number of samples in the wind turbine, the sample data is unbalanced to obtain the final balanced data, thereby effectively improving the accuracy and reliability of wind turbine blade icing diagnosis, and providing accurate data support for subsequent de-icing operation and maintenance.

[0040] In this embodiment, in step S2, processing the unbalanced sample data to form a balanced sample specifically includes:

[0041] Select m samples from the initial sample data as benchmark samples X m :

[0042] X m =(x1,x2,…,x i ,…,x m );

[0043] Calculate the Euclidean distance between the benchmark sample and the remaining sample data in the initial sample data, and select the k samples with the smallest distance in the remaining sample data as the nearest neighbor samples;

[0044] Randomly select M neighbor samples from the k neighbor samples as auxiliary samples Y mk :

[0045] Y mk =(y m1 ,y m2 ,…,y mj ,…,y mk );

[0046] Reconstruct the new sample as a balanced sample:

[0047] X new =x i +γ(y mj -x i ), where γ is a random coefficient between [0,1]. The above method can effectively ensure the balance of sample data, thereby ensuring the accuracy of the model parameter update during the prediction model LightGBM training, and further ensuring the accuracy of the final prediction results.

[0048] In this embodiment, the initial sample data and the real-time sample data include temperature, humidity, wind speed, and vibration signals of wind turbine blades.

[0049] In this embodiment, preprocessing the initial sample and the real-time sample includes:

[0050] The box plot detection method is used to remove outliers from sample data;

[0051] The sample data after removing outliers is processed by linear interpolation;

[0052] Perform Z-score normalization on the interpolated sample data:

[0053]

[0054] Among them, z represents the sample data after normalization, x represents the sample data before normalization, μ is the mean of the sample data, and σ represents the standard deviation of the sample data; among them, the box plot detection method and the linear interpolation method both adopt the existing technology, and their details are not repeated here.

[0055] In this embodiment, extracting ice cover identification features includes:

[0056] Determine the mean, standard deviation, and peak of the sample data; where:

[0057] The formulas for mean and standard deviation are:

[0058]

[0059] The principal component analysis method is used to obtain the final ice cover characteristics for the mean, standard deviation and peak value of the sample data. The principal component analysis method is a prior art and will not be described in detail here.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A wind turbine blade icing diagnosis method considering sample imbalance conditions, characterized by: The following steps are involved: S1. Obtaining wind turbine operation data as initial sample data, and preprocessing the initial sample data; S2. Process the unbalanced initial sample data to form a balanced sample; S3. Extract ice cover identification features from equilibrium samples; S4. Input the ice cover recognition features into the LightGBM model to train the LightGBM model; S5. Obtain the operating data of the wind turbine to be tested as real-time samples, perform feature extraction after preprocessing the real-time samples, and then input the extracted features into the trained LightGBM model to obtain the ice coverage level.

2. The wind turbine blade icing diagnosis method considering sample imbalance conditions according to claim 1 is characterized in that: In step S2, processing the unbalanced sample data to form a balanced sample specifically includes: Select m samples from the initial sample data as benchmark samples X m : X m =(x1,x2,…,x i ,…,x m ): Calculate the Euclidean distance between the benchmark sample and the remaining sample data in the initial sample data, and select the k samples with the smallest distance in the remaining sample data as the nearest neighbor samples; Randomly select M neighbor samples from the k neighbor samples as auxiliary samples Y mk : AND mk (and m1 ,and m2 ,…,and mj ,…,and mk ); Reconstruct the new sample as a balanced sample: X new =x i +γ(y mj -x i ), where γ is a random coefficient between [0,1].

3. The wind turbine blade icing diagnosis method considering sample imbalance conditions according to claim 1, characterized in that: The initial sample data and the real-time sample data include temperature, humidity, wind speed and vibration signals of wind turbine blades.

4. The wind turbine blade icing diagnosis method considering sample imbalance conditions according to claim 3 is characterized by: Preprocessing of initial samples and real-time samples includes: The box plot detection method is used to remove outliers from sample data; The sample data after removing outliers is processed by linear interpolation; Perform Z-score normalization on the interpolated sample data: Among them, z represents the sample data after normalization, x represents the sample data before normalization, μ is the mean of the sample data, and σ represents the standard deviation of the sample data.

5. The wind turbine blade icing diagnosis method considering sample imbalance conditions according to claim 1 is characterized in that: The extracting of ice cover identification features comprises: Determine the mean, standard deviation, and peak value of the sample data; The principal component analysis method is used to analyze the mean, standard deviation and peak value of the sample data to obtain the final ice cover characteristics.

Images