Ice covering monitoring method for wind driven generator

Abstract

The invention discloses an ice covering monitoring method for a wind driven generator, and belongs to the technical field of wind power generation. Constructing a data set through the wind power blade sample, and training the machine learning model to obtain a wind power generation blade ice covering monitoring model; then, aiming at a target wind power generation blade of which ice coating needs to be measured, firstly obtaining three indexes of the blade surface temperature, the blade surface reflectance and the effective peak value of the blade surface ultrasonic receiving waveform of the target wind power generation blade, standardizing the three indexes, and inputting the indexes into the wind power generation blade ice coating monitoring model; monitoring results of whether the blade is covered with ice and the ice covering thickness are obtained; compared with the prior art, for the target wind power generation blade for ice coating measurement, only three indexes of the blade surface temperature, the blade surface reflectance and the effective peak value of the blade surface ultrasonic receiving waveform of the target wind power generation blade need to be obtained and input into the wind power generation blade ice coating monitoring model; and the monitoring result of whether the blade is iced and the iced thickness can be obtained.

Description

Technical Field

[0001] The invention belongs to the technical field of wind power generation, and in particular relates to a method for monitoring ice walking on a wind turbine. Background Art

[0002] Wind power generation refers to the conversion of wind kinetic energy into electrical energy. Wind energy is a clean, pollution-free, renewable energy source. Wind power generation has the advantages of being renewable and inexhaustible; being clean and having good environmental benefits; having a short construction period; and having flexible installed capacity.

[0003] The future of wind power installed capacity will usher in faster development. However, one of the key challenges is the icing of wind turbines in areas with low temperatures and high humidity. Due to the accumulation of ice, the aerodynamic performance of wind turbine blades decreases, that is, the lift drops rapidly and the drag increases sharply, thereby reducing wind power generation. At the same time, the accumulation of ice increases loads and noise, affects the balance and stability of the wind turbine structure, and may cause blade breakage. Ice may fall from wind turbine blades, posing a threat to personal safety. Globally, annual production losses caused by icing are as high as 50%. Ice on wind turbines greatly affects their reliability.

[0004] Wind turbines generally need to be equipped with corresponding blade ice monitoring systems and de-icing control systems. When the blade ice monitoring equipment detects ice on the blades, it sends a signal to the de-icing control system to de-ice the fan blades and change the fan operation strategy. Therefore, for wind turbines installed in cold areas that often encounter icing environments, the accuracy of the blade ice monitoring equipment directly affects the effect of the subsequent de-icing system and the operating performance of the fan. Improving the accuracy of blade ice monitoring is very important for improving the power generation efficiency of the wind turbine.

[0005] For example, Chinese patent application number CN202310651323.6 discloses a method for monitoring wind turbine blade icing, which includes: denoising and segmented enhancement of collected wind turbine blade images; constructing a wind turbine blade icing recognition model, optimizing the wind turbine blade icing recognition model using an improved non-convex optimization method to obtain an optimal wind turbine blade icing recognition model; using the optimal wind turbine blade icing recognition model to extract ice edge images from the enhanced wind turbine blade images, and calculating the ice thickness on the surface of the wind turbine blades. The present invention performs denoising and enhancement processing on images collected in harsh environments, constructs a wind turbine blade icing recognition model based on image morphological features, selects the optimal edge detection operator template to perform edge detection on the image, and then calculates the ice thickness on the surface of the wind turbine blades, thereby achieving real-time monitoring of wind turbines. Summary of the invention

[0006] The purpose of the present invention is to address the problem of low prediction accuracy caused by incomplete consideration of existing attention mechanisms when making predictions, and propose a wind turbine ice monitoring method. A data set is constructed through wind blade samples, and a machine learning model is trained at the same time. An ice monitoring model for wind turbine blades is obtained through supervised learning; then, for the target wind turbine blades that need to be measured for ice, three indicators of blade surface temperature, blade surface reflectivity, and effective peak value of blade surface ultrasonic receiving waveform are first obtained, and the three indicators are standardized and input into the wind turbine blade ice monitoring model to obtain monitoring results of whether the blade is iced and the thickness of the ice; compared with the prior art, for the target wind turbine blades that need to be measured for ice, only three indicators of blade surface temperature, blade surface reflectivity, and effective peak value of blade surface ultrasonic receiving waveform need to be obtained, and the three indicators are input into the wind turbine blade ice monitoring model to obtain monitoring results of whether the blade is iced and the thickness of the ice, which is very convenient and easy to manage.

[0007] In order to solve the above technical problems, the present invention provides a wind turbine icing monitoring method, comprising: S1, obtaining a data set composed of a large number of wind blade samples, wherein each wind blade sample in the data set includes five indicators: blade surface temperature, blade surface reflectivity, effective peak value of ultrasonic receiving waveform on blade surface, whether the blade is icing, and ice thickness; S2. Using the data set to train a machine learning model, wherein the machine learning model uses three indicators as inputs, namely, the standardized blade surface temperature, the blade surface reflectivity, and the effective peak value of the ultrasonic receiving waveform on the blade surface, and uses whether the blade is on ice and the ice thickness as outputs, to obtain a wind turbine blade ice monitoring model through supervised learning; S3. For the target wind turbine blades that need to be tested for ice rubbing, first obtain three indicators of the target wind turbine blades, namely, blade surface temperature, blade surface reflectivity, and effective peak value of the ultrasonic receiving waveform on the blade surface. After standardization, input them into the wind turbine blade ice rubbing monitoring model to obtain the monitoring results of whether the blades are rubbing ice and the thickness of the ice rubbing.

[0008] As a further improvement of the technical solution: a large number of wind blade samples include wind blade ice-walking samples and wind blade non-ice-walking samples.

[0009] As a further improvement of the present technical solution: in S1, the data set is divided into a training set and a validation set, the training set is used to simultaneously train multiple machine learning models, and the validation set is used to evaluate the performance accuracy of different machine learning models, and the best performing machine learning model is selected to construct a wind turbine blade ice monitoring model.

[0010] As a further improvement of the present technical solution: in S2, the selected machine learning model is at least one of logistic regression, linear discriminant analysis, K nearest neighbor, naive Bayes, support vector machine, random forest, and neural network.

[0011] As a further improvement of this technical solution: the selected machine learning model is a random forest model.

[0012] As a further improvement of the technical solution: a temperature sensor is provided on the side wall of the wind power generation blade, and the temperature sensor is used to detect the blade surface temperature of the wind power generation blade.

[0013] As a further improvement of the technical solution: an ultrasonic sensor is installed on the side wall of the wind turbine blade through a bracket, and the ultrasonic sensor is used to detect the effective peak value of the ultrasonic receiving waveform on the blade surface of the wind turbine blade.

[0014] As a further improvement of the technical solution: a reflective photoelectric sensor is installed on the side wall of the wind turbine blade through a bracket, and the reflective photoelectric sensor is used to detect the reflectivity of the blade surface of the wind turbine blade.

[0015] Beneficial effects of the present invention: This solution constructs a wind turbine blade ice monitoring model. For the target wind turbine blades to be iced, it only needs to obtain three indicators: the blade surface temperature, blade surface reflectivity, and the effective peak value of the ultrasonic receiving waveform on the blade surface. Then, the three indicators are input into the wind turbine blade ice monitoring model to obtain the monitoring results of whether the blade is iced and the thickness of the ice, thereby realizing automated and intelligent ice monitoring of the blades. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0017] Figure 1 A schematic diagram of a wind turbine ice monitoring method according to an embodiment of the present invention; Figure 2 It is a schematic diagram of the surface structure of the wind power generation blade of the present invention.

[0018] In the accompanying drawings, the components represented by the reference numerals are listed as follows: 1. Temperature sensor; 2. Wind turbine blades; 3. Ultrasonic sensor; 4. Reflective photoelectric sensor. DETAILED DESCRIPTION

[0019] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0020] Example 1: Figure 1 As shown, a method for monitoring ice walking on a wind turbine includes: S1, obtaining a data set composed of a large number of wind blade samples, wherein the large number of wind blade samples include wind blade ice walking samples and wind blade non-ice walking samples, and each wind blade sample in the data set includes five indicators: blade surface temperature, blade surface reflectivity, effective peak value of ultrasonic receiving waveform on the blade surface, whether the blade is ice walking, and ice walking thickness; The data set is divided into a training set and a validation set. The training set is used to train multiple machine learning models at the same time, and the validation set is used to evaluate the performance accuracy of different machine learning models. The best performing machine learning model is selected to build a wind turbine blade ice monitoring model. S2. Use the data set to train the machine learning model, where the machine learning model uses the standardized blade surface temperature, blade surface reflectivity, and effective peak value of the blade surface ultrasonic receiving waveform as input, and uses whether the blade is on ice and the ice thickness as output, and obtains a wind turbine blade ice monitoring model through supervised learning; The selected machine learning model is at least one of logistic regression, linear discriminant analysis, K nearest neighbor, naive Bayes, support vector machine, random forest, and neural network. The machine learning model selected in the present invention is a random forest model. S3. For the target wind turbine blade 2 that needs to be tested for ice erosion, first obtain three indicators of the target wind turbine blade 2, namely, the blade surface temperature, the blade surface reflectivity, and the effective peak value of the blade surface ultrasonic receiving waveform. After standardization, input them into the wind turbine blade ice erosion monitoring model to obtain the monitoring results of whether the blade is eroding and the thickness of the ice erosion. If the blade is eroding, output 1; if the blade is not eroding, output 0.

[0021] A temperature sensor 1 is provided on the side wall of the wind turbine blade 2 , and the temperature sensor 1 is used to detect the blade surface temperature of the wind turbine blade 2 .

[0022] An ultrasonic sensor 3 is mounted on the side wall of the wind power blade 2 via a bracket. The ultrasonic sensor 3 is used to detect the effective peak value of the ultrasonic receiving waveform on the blade surface of the wind power blade 2 .

[0023] A reflective photoelectric sensor 4 is installed on the side wall of the wind turbine blade 2 through a bracket, and the reflective photoelectric sensor 4 is used to detect the reflectivity of the blade surface of the wind turbine blade 2. Specifically, the machine learning model selected in the present invention is a BP neural network model, which takes three characteristic indicators of blade surface temperature, blade surface reflectivity, and effective peak value of ultrasonic receiving waveform on the blade surface of each set of sample data as input, and takes the judgment result of whether it is a zombie course as output, wherein the number of input layer nodes is 3 and the number of output layer nodes is 2.

[0024] The number of hidden layer nodes in the BP neural network model is: ; Wherein, n corresponds to the number of nodes in the input layer, m corresponds to the number of nodes in the output layer, and a corresponds to a constant in the interval [1,10]. In this embodiment, the activation functions of the hidden layer and the output layer adopt the Tansig function.

[0025] During training, the input information is divided into a prediction set and a training set. After the training, the validity of the model is verified in the test set. If the verification fails (if the average AP value of the test set is <0.95, the verification fails; if the average AP value is ≥0.95, the verification passes), more labeled data is added; if the verification passes, the training ends.

[0026] Embodiment 2: Taking a wind power fan as an example, the following steps are included: First, determine the monitoring goals and needs, and clarify the purpose of monitoring, such as ensuring the safety of wind turbines in winter, the safety of ice operations, or scientific research. Determine the key areas of monitoring and the time scale of monitoring. Secondly, select appropriate monitoring methods. Select the most appropriate monitoring method based on the monitoring goals and site conditions, such as obtaining a data set composed of a large number of wind blade samples. A large number of wind blade samples include wind blade ice samples and wind blade non-ice samples, and each wind blade sample in the data set includes five indicators: blade surface temperature, blade surface reflectivity, effective peak value of blade surface ultrasonic receiving waveform, whether the blade is iced, and ice thickness.

[0027] Then, formulate a monitoring plan and develop a detailed monitoring plan, including monitoring frequency, monitoring point layout, and monitoring equipment selection. At the same time, formulate an emergency plan to deal with unexpected situations that may occur during the monitoring process, and prepare the required monitoring equipment and tools to ensure their accuracy and reliability. When conducting preliminary on-site surveys to obtain a large number of wind blade samples in the composition of the data set, set up monitoring points at predetermined locations to install monitoring equipment and regularly obtain information such as the color, cracks, and snow accumulation of wind blade samples. Use sonar, radar, drones and other equipment to collect data, and organize, verify and preprocess the collected data. Based on the collected data, professional software or models are used to further analyze the data to predict the trend of ice surface changes, such as using the data set to train a machine learning model, where the machine learning model uses the standardized blade surface temperature, blade surface reflectivity, and effective peak value of the blade surface ultrasonic receiving waveform as input, and whether the blade is on ice and the thickness of the ice as output. Through supervised learning, a wind turbine blade ice monitoring model is obtained, and then, for the target wind turbine blade 2 that needs to be measured for ice, the three indicators of the blade surface temperature, blade surface reflectivity, and effective peak value of the blade surface ultrasonic receiving waveform are first obtained, standardized, and input into the wind turbine blade ice monitoring model to obtain the monitoring results of whether the blade is on ice and the thickness of the ice.

[0028] Finally, prepare a monitoring report, including monitoring methods, monitoring results, data analysis, conclusions and recommendations. Then communicate the monitoring results and reports to relevant departments and personnel in a timely manner. Adjust the monitoring plan based on the feedback and conduct additional monitoring if necessary. Carry out continuous monitoring according to the monitoring plan. Regularly evaluate the monitoring results and adjust the monitoring strategy according to the actual situation.

[0029] This embodiment constructs a wind turbine blade ice monitoring model. For the target wind turbine blade to be iced, it only needs to obtain three indicators of the target wind turbine blade's blade surface temperature, blade surface reflectivity, and effective peak value of the blade surface ultrasonic receiving waveform, and input them into the wind turbine blade ice monitoring model. Then, the monitoring results of whether the blade is iced and the ice thickness can be obtained, thereby realizing automated and intelligent ice monitoring of the blade.

[0030] The above specific embodiments are preferred embodiments of the present invention, and are not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to the present specific embodiments. All equivalent changes made in accordance with the shape, structure, and method of the present invention are within the protection scope of the present invention.

Claims

1. A method for monitoring ice on a wind turbine. It is characterized in that The steps include: S1. Obtain a large number of wind blade samples and construct a data set; S2. Using the data set to train the machine learning model, and obtaining a wind turbine blade ice monitoring model through supervised learning; S3. For the target wind turbine blades that need to be tested for ice rubbing, first obtain three indicators of the target wind turbine blades, namely, blade surface temperature, blade surface reflectivity, and effective peak value of the ultrasonic receiving waveform on the blade surface. After standardization, input them into the wind turbine blade ice rubbing monitoring model to obtain the monitoring results of whether the blades are rubbing ice and the thickness of the ice rubbing.

2. A wind turbine ice monitoring method according to claim 1, It is characterized in that The data set includes five indicators for each wind blade sample in the data set, including blade surface temperature, blade surface reflectivity, effective peak value of ultrasonic receiving waveform on the blade surface, whether the blade is on ice, and ice thickness.

3. A wind turbine ice monitoring method according to claim 2, It is characterized in that The wind blade samples further include wind blade ice samples and wind blade non-ice samples.

4. A wind turbine ice monitoring method according to claim 3, It is characterized in that In S1, the data set is divided into a training set and a validation set. The training set is used to simultaneously train multiple machine learning models, and the validation set is used to evaluate the performance accuracy of different machine learning models. The best performing machine learning model is selected to construct a wind turbine blade ice monitoring model.

5. A wind turbine ice monitoring method according to claim 1, It is characterized in that In S2, the selected machine learning model is at least one of logistic regression, linear discriminant analysis, K nearest neighbor, naive Bayes, support vector machine, random forest, and neural network.

6. A wind turbine ice monitoring method according to claim 5, It is characterized in that The machine learning model uses three indicators as input: the standardized blade surface temperature, the blade surface reflectivity, and the effective peak value of the blade surface ultrasonic receiving waveform, and uses whether the blade is on ice and the thickness of the ice as output.

7. A wind turbine ice monitoring method according to claim 6, It is characterized in that The blade surface temperature is sensed by a temperature sensor arranged on the side wall of the wind power generation blade.

8. A wind turbine ice monitoring method according to claim 5, It is characterized in that The effective peak value of the ultrasonic wave receiving waveform on the blade surface is sensed by an ultrasonic sensor arranged on a bracket on the side wall of the wind power generation blade.

9. A wind turbine ice monitoring method according to claim 8, It is characterized in that The blade surface reflectivity is sensed by a reflective photoelectric sensor arranged on a bracket on the side wall of the wind power generation blade.

Images