Fan blade icing monitoring method, device and equipment and medium

By constructing and training the icing monitoring neural network model, the problem of low recognition efficiency of fan blade icing is solved, and efficient automatic monitoring of icing of wind turbine blades is realized, which improves recognition accuracy and efficiency.

CN118934506AInactive Publication Date: 2024-11-12BOWANG (ZHANGBEI) NEW ENERGY CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411200589.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, the icing efficiency of wind turbine blades is low, which makes it difficult to accurately monitor the icing situation of wind turbine blades, which may lead to serious problems such as blade fracture.

Method used

By obtaining fan blade pictures, building and training an icing monitoring neural network model, adjusting neural network parameters to improve the accuracy of icing recognition, and achieving automated monitoring.

Benefits of technology

It improves the efficiency and automation of fan blade icing recognition, can accurately identify the icing condition of the blade in batches, and reduces the dependence on manual identification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118934506A_ABST
    Figure CN118934506A_ABST
Patent Text Reader

Abstract

The invention discloses a fan blade icing monitoring method, device and equipment and a medium, and the method comprises the steps: obtaining a plurality of fan blade pictures, and grouping the plurality of fan blade pictures to obtain a plurality of picture groups; a to-be-trained icing monitoring neural network model is constructed and initialized, the input of the to-be-trained icing monitoring neural network model is the fan blade picture, the output of the to-be-trained icing monitoring neural network model is an icing recognition result, and the icing recognition result comprises icing or non-icing; inputting the picture group into a to-be-trained icing monitoring neural network model, obtaining a plurality of icing identification results corresponding to the picture group, and adjusting neural network parameters of the to-be-trained icing monitoring neural network model according to the plurality of icing identification results; and when a preset number of iterations is reached, storing the latest neural network parameters, and obtaining a target icing monitoring neural network model. The invention belongs to the field of fan blade monitoring. According to the invention, the accuracy and efficiency of blade icing identification can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of fan blade monitoring, and in particular to a fan blade icing monitoring method, device, equipment and medium. Background Art

[0002] With the massive consumption of fossil fuels and serious environmental pollution, the penetration rate of clean and environmentally friendly renewable energy power generation continues to increase, and the scale of wind farm construction in cold and humid areas continues to expand, which will cause the blades of wind turbines to freeze, and in severe cases even cause blade breakage. Therefore, it is necessary to conduct in-depth research on the identification of icing conditions of wind turbine blades.

[0003] The research on the fault prediction of wind turbine blade icing is still in its infancy at home and abroad. Due to the complex structure of wind turbine blades, most of them are currently identified manually. However, relying solely on manual identification is not only inefficient, but also has low accuracy. Therefore, a wind turbine blade icing monitoring method is urgently needed. Summary of the invention

[0004] The embodiments of the present application solve the technical problem of low efficiency in identifying ice on fan blades in the prior art by providing a method, device, equipment and medium for monitoring ice on fan blades, thereby achieving the technical effect of improving the efficiency of identifying ice on fan blades.

[0005] In a first aspect, the present application provides a method for monitoring icing of wind turbine blades, the method comprising:

[0006] Acquire a number of fan blade images, and group the fan blade images to obtain a number of image groups, each group of images including at least two fan blade images, wherein the fan blade images include frozen blade images and unfrozen blade images;

[0007] Constructing and initializing a neural network model for ice monitoring to be trained, wherein the input of the neural network model for ice monitoring to be trained is a wind turbine blade image, and the output is an ice recognition result, wherein the ice recognition result includes ice or non-ice;

[0008] Inputting the picture group into the ice monitoring neural network model to be trained, obtaining a number of ice recognition results corresponding to the picture group, and adjusting the neural network parameters of the ice monitoring neural network model to be trained according to the number of ice recognition results;

[0009] When the preset number of iterations is reached, the latest neural network parameters are saved and the target icing monitoring neural network model is obtained.

[0010] Furthermore, the neural network parameters of the icing monitoring neural network model to be trained are adjusted according to several icing recognition results, including:

[0011] According to a number of ice recognition results, determine the accuracy parameter, positive ratio parameter and recall parameter corresponding to the image group;

[0012] The neural network parameters of the ice monitoring neural network model to be trained are adjusted according to the accuracy parameter, positive ratio parameter and recall rate parameter corresponding to each picture group.

[0013] Furthermore, the formula for the accuracy parameter includes:

[0014]

[0015] Among them, A is the accuracy parameter, P C is the number of correct icing identification results, P T To summarize the number of ice identification results.

[0016] Furthermore, the formulas for positive ratio parameters include:

[0017]

[0018] Among them, B is the positive ratio parameter, P Z The result of ice recognition is the amount of ice, P F The number of ice identification results that are not iced.

[0019] Furthermore, the formula for the recall parameter includes:

[0020]

[0021] Among them, C is the recall rate parameter, A is the accuracy parameter, and B is the positive ratio parameter.

[0022] Furthermore, after obtaining a number of wind turbine blade images, the method further includes:

[0023] Crop several wind turbine blade images so that the size of each wind turbine blade image remains consistent;

[0024] Label each wind turbine blade image, including real-time weather, icing degree, icing area, icing type, and wind turbine type;

[0025] According to the annotation of each fan blade image and the corresponding icing recognition result, determining the correlation between the annotation of the fan blade image and the icing recognition result;

[0026] According to a number of association relationships, the neural network parameters of the ice monitoring neural network model to be trained are adjusted.

[0027] Furthermore, the neural network parameters include at least:

[0028] Weights, biases, learning rate, batch size, optimizer, and momentum.

[0029] In a second aspect, the present application provides a wind turbine blade icing monitoring device, the device comprising:

[0030] An acquisition module is used to acquire a number of fan blade images, and group the fan blade images to obtain a number of image groups, each group of images includes at least two fan blade images, wherein the fan blade images include frozen blade images and unfrozen blade images;

[0031] A construction module is used to construct and initialize a neural network model for ice monitoring to be trained, wherein the input of the neural network model for ice monitoring to be trained is a wind turbine blade image, and the output is an ice recognition result, wherein the ice recognition result includes ice or non-ice;

[0032] An input module, used to input a picture group into the ice monitoring neural network model to be trained, obtain a number of ice recognition results corresponding to the picture group, and adjust the neural network parameters of the ice monitoring neural network model to be trained according to the number of ice recognition results;

[0033] The output module is used to save the latest neural network parameters and obtain the target icing monitoring neural network model when the preset number of iterations is reached.

[0034] In a third aspect, the present application provides an electronic device, including:

[0035] processor;

[0036] a memory for storing processor-executable instructions;

[0037] The processor is configured to execute to implement a wind turbine blade icing monitoring method as provided in the first aspect.

[0038] In a fourth aspect, the present application provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to implement a method for monitoring wind turbine blade icing as provided in the first aspect.

[0039] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0040] The present application obtains several fan blade images and groups them into several image groups, each of which includes at least two fan blade images, wherein the fan blade images include blade ice images and blade un-ice images; constructs a neural network model for ice monitoring to be trained and initializes it, wherein the input of the neural network model for ice monitoring to be trained is the fan blade image, and the output is the ice recognition result, wherein the ice recognition result includes ice or un-ice; inputs the image group into the neural network model for ice monitoring to be trained, obtains several ice recognition results corresponding to the image group, and adjusts the neural network parameters of the neural network model for ice monitoring to be trained according to the several ice recognition results; when the preset number of iterations is reached, the latest neural network parameters are saved, and the target ice monitoring neural network model is obtained. The present application can obtain the target ice monitoring neural network model by training the neural network, and the target ice monitoring neural network model can accurately identify the ice conditions of fan blade images in batches, thereby improving the recognition efficiency and automation of ice recognition of fan blade images. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0042] Figure 1 A schematic diagram of a flow chart of a method for monitoring icing of fan blades provided in this application;

[0043] Figure 2 A schematic diagram of the structure of a fan blade icing monitoring device provided in this application;

[0044] Figure 3 A schematic diagram of the structure of an electronic device provided in this application. DETAILED DESCRIPTION

[0045] The embodiment of the present application solves the technical problem of low efficiency in identifying ice on fan blades in the prior art by providing a method for monitoring ice on fan blades.

[0046] The technical solution of the embodiment of the present application is to solve the above technical problems, and the overall idea is as follows:

[0047] A method for monitoring icing of fan blades, the method comprising: obtaining a number of fan blade images, and grouping the number of fan blade images to obtain a number of image groups, each group of images including at least two fan blade images, wherein the fan blade images include blade icing images and blade unicing images; constructing a neural network model for icing monitoring to be trained and initializing it, wherein the input of the neural network model for icing monitoring to be trained is the fan blade image, and the output is an icing recognition result, wherein the icing recognition result includes icing or unicing; inputting the image group into the neural network model for icing monitoring to be trained, obtaining a number of icing recognition results corresponding to the image group, and adjusting the neural network parameters of the neural network model for icing monitoring to be trained according to the number of icing recognition results; when a preset number of iterations is reached, saving the latest neural network parameters, and obtaining a target icing monitoring neural network model.

[0048] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0049] First of all, the term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0050] This application provides Figure 1 A method for monitoring icing of fan blades is shown, comprising steps S11-S14:

[0051] Regarding step S11, a number of fan blade pictures are obtained, and the fan blade pictures are grouped to obtain a number of picture groups, each group of pictures includes at least two fan blade pictures, wherein the fan blade pictures include iced blade pictures and un-iced blade pictures.

[0052] Wind blades are components used in wind turbines or fans to capture wind energy. Wind blades are usually made of lightweight and strong materials (such as fiberglass, carbon fiber or plastic) and designed into a specific aerodynamic shape to maximize the efficiency of wind energy conversion.

[0053] In a wind turbine, the rotation of the blades drives the generator, which converts wind energy into electricity. The length, shape and number of wind turbine blades affect the efficiency of wind energy capture and power generation capacity.

[0054] The purpose of this application is to accurately monitor the icing condition of wind turbine blades in wind turbines. Therefore, the wind turbine blade images mainly include blade icing images and blade unicing images. After obtaining several wind turbine blade images, each wind turbine blade image can be cropped so that each wind turbine blade image is of the same size for easy identification.

[0055] After cropping each fan blade image, the fan blade images may be grouped, each group of images including at least two fan blade images, and each group of images may include blade ice images or blade non-ice images.

[0056] By grouping the wind turbine blade images, it is convenient to subsequently calculate the accuracy parameter, positive ratio parameter, and recall parameter of each group of images. Please refer to the following description for details.

[0057] Regarding step S12, a neural network model for ice monitoring to be trained is constructed and initialized. The input of the neural network model for ice monitoring to be trained is a picture of a wind turbine blade, and the output is an ice recognition result, wherein the ice recognition result includes ice or non-ice.

[0058] The ice monitoring neural network model to be trained constructed in this application may be a convolutional neural network. The input of the convolutional neural network may be RGB images, unprocessed and processed audio data, and the like.

[0059] The basic structure of a convolutional neural network consists of five main parts: input layer, convolution layer, downsampling layer, fully connected layer, and output layer. The convolutional neural network extracts the high-level semantic information contained in the original data in the input layer and performs a "feedforward operation", that is, convolution operation on the input data to obtain the global features of the data, and then downsamples the extracted features through pooling operation.

[0060] After that, operations such as connecting nonlinear activation functions are performed. The convolutional neural network relies on the last layer to process the target task (classification, regression, etc.) and finally obtain the target function. Generally speaking, convolutional layers and pooling layers are set alternately. The convolutional layers and pooling layers can be regarded as a convolutional block. The convolutional block appears several times, and the fully connected layer and the output layer are connected to it.

[0061]

[0062] The above is a continuous form of convolution, where s(n)=(f*w)(n) is a convolution operation, f is an input function, ω is a kernel function or a weighting function, and the output S obtained after the convolution operation is the feature map.

[0063]

[0064] The above is the continuous discrete form of convolution.

[0065] Regarding step S13, the picture group is input into the ice monitoring neural network model to be trained, a number of ice recognition results corresponding to the picture group are obtained, and the neural network parameters of the ice monitoring neural network model to be trained are adjusted according to the several ice recognition results.

[0066] When the fan blade images in the picture group are input into the ice monitoring neural network model to be trained, several ice recognition results corresponding to the picture group can be obtained. It can be understood that each fan blade image corresponds to an ice recognition result. In other words, the number of ice recognition results of the picture group is the number of fan blade images in the picture group.

[0067] According to a number of ice recognition results, the neural network parameters of the ice monitoring neural network model to be trained are adjusted, specifically including: according to a number of ice recognition results, the accuracy parameter, the positive ratio parameter and the recall rate parameter corresponding to the picture group are determined; according to the accuracy parameter, the positive ratio parameter and the recall rate parameter corresponding to each picture group, the neural network parameters of the ice monitoring neural network model to be trained are adjusted.

[0068] The formula for the accuracy parameter includes:

[0069]

[0070] Among them, A is the accuracy parameter, P C is the number of correct ice identification results, P T To summarize the number of ice identification results.

[0071] Formulas for positive ratio parameters include:

[0072]

[0073] Among them, B is the positive ratio parameter, P Z The result of ice recognition is the amount of ice, P F The number of ice-recognition results that are not iced.

[0074] The formula for the recall parameter includes:

[0075]

[0076] Among them, C is the recall rate parameter, A is the accuracy parameter, and B is the positive ratio parameter.

[0077] According to the actual icing conditions of the wind turbine blade images in the image group and the obtained icing recognition results, the accuracy parameter, the positive ratio parameter and the recall rate parameter can be determined.

[0078] For example, a certain image group contains 100 images of wind turbine blades, of which 50 are images of blade ice and 50 are images of blade non-ice. Among them, 96 images are accurately identified, 49 of the 96 images are accurately identified as iced, and 47 are accurately identified as non-iced. Based on the formula provided above, the accuracy parameter, positive ratio parameter, and recall parameter of the group can be accurately calculated.

[0079] The neural network parameters at least include: weight, bias, learning rate, batch size, optimizer, momentum, etc. It should be noted that when a certain group of images can be completely correctly identified, the neural network parameters do not need to be adjusted.

[0080] By adjusting the neural network parameters, the icing monitoring neural network model to be trained can be continuously enabled to more accurately and truly identify the icing results of each wind turbine blade image.

[0081] It should be noted that the positive ratio parameter reflects the ability of the trained icing monitoring neural network model to distinguish negative samples, the accuracy parameter reflects the icing monitoring neural network model to be trained, and the recall rate is a combination of the two. The higher the accuracy parameter, the positive ratio parameter and the recall rate parameter, the more accurate the recognition result of the icing monitoring neural network model to be trained.

[0082] Regarding step S14, when the preset number of iterations is reached, the latest neural network parameters are saved, and the target icing monitoring neural network model is obtained.

[0083] The preset number of iterations can be determined according to actual conditions. After the preset number of iterations, the target icing monitoring neural network model can be obtained, that is, the target icing monitoring neural network model can be used to accurately and batch identify the icing conditions of fan blade images, thereby improving the efficiency and automation of icing identification in fan blade images.

[0084] In addition, after obtaining a number of fan blade images, the method also includes: cropping the number of fan blade images so that the size of each fan blade image remains consistent; labeling each fan blade image, and the labeling content includes real-time weather, ice degree, ice area, ice type and fan type; determining the correlation between the labeling of the fan blade image and the ice recognition result based on the labeling of each fan blade image and the corresponding ice recognition result; and adjusting the neural network parameters of the ice monitoring neural network model to be trained based on the several correlation relationships.

[0085] The purpose of cropping a number of wind turbine blade images so that the size of each wind turbine blade image remains consistent is the same as described above and will not be elaborated here.

[0086] Each picture of the wind blade is labeled, and the labeled content includes the real-time weather, degree of icing, icing area, icing type and wind turbine type. It can be understood that the labeled content is the real-time weather, degree of icing, icing area, icing type and wind turbine type when the picture of the wind blade was taken.

[0087] R i =h(j=1∑ m α j w j +β·φ(C i )+γ·ψ(A i )+δ·θ(T i )+∈·χ(F i ))

[0088] Among them, α, β, γ, δ, ∈ are the preset weights for real-time weather, icing degree, icing area, icing type and fan type respectively, and h is the empirical value of the sensitivity coefficient, which can be adjusted according to the actual situation. H is a comprehensive function that maps all features to icing recognition results. m is a function for processing real-time weather (it can be encoded into a numerical feature such as one-hot encoding or an embedded vector). ψ is a function for processing icing area, which is used for feature standardization or normalization; θ is a function for processing icing type (it can be encoded into a numerical feature such as one-hot encoding or an embedded vector), and χ is a function for processing fan type (it can be encoded into a numerical feature such as one-hot encoding or an embedded vector).

[0089] Wj is the jth eigenvector of the real-time weather; Ci is the degree of icing, which can be set as discrete categories, such as Ci∈{c1,c2,…,ck}; Ai is the icing area; Ti is the icing type, F i is the fan type, R i is the recognition result of the i-th fan blade image (can be set to 0 for recognition and 1 for unrecognized). By determining the correlation between the annotation of each fan blade image and the corresponding icing recognition result, the influence of real-time weather, icing degree, icing area, icing type and fan type on accurate recognition can be obtained. By continuously adjusting the neural network parameters in the icing monitoring neural network model to be trained, the icing monitoring neural network model to be trained can overcome the influence of the above-mentioned real-time weather, icing degree, icing area, icing type and fan type on the icing monitoring neural network model to be trained, so that accurate recognition can be completed.

[0090] In summary, the present application obtains several fan blade images and groups the several fan blade images to obtain several image groups, each group of images includes at least two fan blade images, wherein the fan blade images include blade ice images and blade un-ice images; constructs a neural network model for ice monitoring to be trained and initializes it, wherein the input of the neural network model for ice monitoring to be trained is the fan blade image, and the output is the ice recognition result, wherein the ice recognition result includes ice or un-ice; inputs the image group into the neural network model for ice monitoring to be trained, obtains several ice recognition results corresponding to the image group, and adjusts the neural network parameters of the neural network model for ice monitoring to be trained according to the several ice recognition results; when the preset number of iterations is reached, the latest neural network parameters are saved, and the target ice monitoring neural network model is obtained. The present application can obtain the target ice monitoring neural network model by training the neural network, and the target ice monitoring neural network model can accurately and batch identify the ice conditions of fan blade images, thereby improving the recognition efficiency and automation of ice recognition of fan blade images.

[0091] Based on the same inventive concept, the present application provides Figure 2 A fan blade icing monitoring device is shown, the device comprising:

[0092] An acquisition module 21 is used to acquire a number of fan blade images, and group the fan blade images to obtain a number of image groups, each group of images includes at least two fan blade images, wherein the fan blade images include frozen blade images and unfrozen blade images;

[0093] A construction module 22 is used to construct and initialize a neural network model for ice monitoring to be trained, wherein the input of the neural network model for ice monitoring to be trained is a wind turbine blade image, and the output is an ice recognition result, wherein the ice recognition result includes ice or non-ice;

[0094] An input module 23, used to input the picture group into the ice monitoring neural network model to be trained, obtain a number of ice recognition results corresponding to the picture group, and adjust the neural network parameters of the ice monitoring neural network model to be trained according to the number of ice recognition results;

[0095] The output module 24 is used to save the latest neural network parameters and obtain the target icing monitoring neural network model when a preset number of iterations is reached.

[0096] Based on the same inventive concept, the present application also provides Figure 3 An electronic device as shown includes:

[0097] Processor 31;

[0098] A memory 32 for storing instructions executable by the processor 31;

[0099] The processor 31 is configured to execute to implement a wind turbine blade icing monitoring method as provided above.

[0100] Based on the same inventive concept, the present application also provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by the processor 31 of the electronic device, the electronic device can execute and implement a wind turbine blade icing monitoring device as provided above.

[0101] Since the electronic device introduced in this embodiment is an electronic device used to implement the information processing method in the embodiment of the present application, based on the information processing method introduced in the embodiment of the present application, a person skilled in the art can understand the specific implementation of the electronic device of the present embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is not described in detail here. As long as a person skilled in the art implements the electronic device used by the information processing method in the embodiment of the present application, it belongs to the scope of protection of this application.

[0102] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0103] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0104] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0105] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0106] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0107] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for monitoring icing of fan blades, It is characterized in that The method comprises: Acquire a number of fan blade images, and group the fan blade images to obtain a number of image groups, each group of images including at least two fan blade images, wherein the fan blade images include frozen blade images and unfrozen blade images; Constructing and initializing a neural network model for ice monitoring to be trained, wherein the input of the neural network model for ice monitoring to be trained is a picture of a wind turbine blade, and the output is an ice recognition result, wherein the ice recognition result includes ice or non-ice; Inputting the picture group into the ice monitoring neural network model to be trained, obtaining a number of ice recognition results corresponding to the picture group, and adjusting the neural network parameters of the ice monitoring neural network model to be trained according to the number of ice recognition results; When the preset number of iterations is reached, the latest neural network parameters are saved, and the target icing monitoring neural network model is obtained.

2. A method for monitoring icing of fan blades according to claim 1, It is characterized in that The neural network parameters of the to-be-trained icing monitoring neural network model are adjusted according to a number of icing recognition results, including: According to a number of ice recognition results, determine the accuracy parameter, positive ratio parameter and recall parameter corresponding to the image group; The neural network parameters of the ice monitoring neural network model to be trained are adjusted according to the accuracy parameter, positive ratio parameter and recall rate parameter corresponding to each picture group.

3. A method for monitoring icing of fan blades according to claim 2, It is characterized in that The formula for the accuracy parameter includes: Among them, A is the accuracy parameter, P C is the number of correct ice identification results, P T To summarize the number of ice identification results.

4. A method for monitoring icing of fan blades according to claim 2, It is characterized in that Formulas for positive ratio parameters include: Among them, B is the positive ratio parameter, P Z The result of ice recognition is the amount of ice, P F The number of ice identification results that are not iced.

5. A method for monitoring icing of fan blades according to claim 2, It is characterized in that The formula for the recall parameter includes: Among them, C is the recall rate parameter, A is the accuracy parameter, and B is the positive ratio parameter.

6. A method for monitoring icing of fan blades according to claim 1, It is characterized in that After obtaining a number of wind turbine blade images, the method further includes: Crop several wind turbine blade images so that the size of each wind turbine blade image remains consistent; Annotate each wind turbine blade image, wherein the annotation content includes real-time weather, icing degree, icing area, icing type and wind turbine type; According to the annotation of each fan blade image and the corresponding icing recognition result, determining the correlation between the annotation of the fan blade image and the icing recognition result; According to a number of association relationships, the neural network parameters of the ice monitoring neural network model to be trained are adjusted.

7. A method for monitoring icing of fan blades according to claim 1, It is characterized in that The neural network parameters include at least: Weights, biases, learning rate, batch size, optimizer, and momentum.

8. A fan blade icing monitoring device, It is characterized in that The device comprises: An acquisition module is used to acquire a number of fan blade images, and group the fan blade images to obtain a number of image groups, each group of images includes at least two fan blade images, wherein the fan blade images include frozen blade images and unfrozen blade images; A construction module is used to construct and initialize a neural network model for ice monitoring to be trained, wherein the input of the neural network model for ice monitoring to be trained is a picture of a wind turbine blade, and the output is an ice recognition result, wherein the ice recognition result includes ice or non-ice; An input module, used for inputting a picture group into the ice monitoring neural network model to be trained, obtaining a number of ice recognition results corresponding to the picture group, and adjusting the neural network parameters of the ice monitoring neural network model to be trained according to the number of ice recognition results; The output module is used to save the latest neural network parameters and obtain the target icing monitoring neural network model when the preset number of iterations is reached.

9. An electronic device, It is characterized in that include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute to implement a wind turbine blade icing monitoring method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, It is characterized in that When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to implement a method for monitoring icing of wind turbine blades as described in any one of claims 1 to 7.