Physical model and AI model combined blade clearance monitoring method
By combining the embedded structure of physical models and AI models, using CNN network and physical consistency constraints to train deep learning models, the accuracy of blade clearance prediction of wind turbine sets is solved, and high-precision blade status monitoring and protection is achieved.
Patent Information
- Application Number
- CN202510241599.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-03
- Publication Date
- 2025-07-18
AI Technical Summary
In the prior art, traditional wind turbine blade clearance calculation methods based on physical formulas are difficult to cope with the modeling challenges of complex environments, and a single AI model may violate physical laws, resulting in unreliable headroom prediction results.
Combining the physical model and AI model, the neural network is guided through an embedded structure, and the physical model of clearance calculation is used as the prior knowledge of the AI model. The deep learning model is trained using the CNN network structure and the loss function with physical consistency constraints to ensure that the prediction results are in line with the actual physical laws of the blade.
It realizes high-precision real-time monitoring and prediction of blade clearance of wind turbine units, provides an intelligent safety monitoring and protection system, and improves the accuracy of blade status evaluation and dynamic protection capabilities.
Smart Images
Figure CN120332099A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of blade condition monitoring of wind turbines, and particularly to a blade clearance monitoring method combining a physical model and an AI model. Background Art
[0002] During the operation of wind turbine blades, they are affected by complex aerodynamic forces and wind direction deflection angles, and their dynamic deformation affects clearance safety and structural fatigue life. Traditional calculation methods based on physical formulas have high physical consistency, but are limited by the modeling difficulty of complex environments. While a single AI model has powerful data learning ability, it may violate physical laws, resulting in unreliable prediction results.
[0003] Therefore, how to improve the accuracy of clearance prediction is a technical problem that needs to be solved urgently at present. Summary of the Invention
[0004] To solve the deficiencies of the prior art, the present invention provides a blade clearance monitoring method combining a physical model and an AI model. The physical formula based on the root wind direction deflection angle is embedded in the AI model. By combining the advantages of both, the physical model of clearance calculation is used as the prior knowledge of the AI model, and the neural network is guided through an embedded structure to enhance the physical rationality of the model, so as to achieve high-precision real-time monitoring and prediction of blade conditions.
[0005] The embodiments of the present invention provide the following solutions:
[0006] The embodiments of the present invention provide a blade clearance monitoring method combining a physical model and an AI model, and the method includes:
[0007] Step 1, using sensors to collect the operation data of the wind turbine, and using a clearance measurement device to measure the true value of the clearance;
[0008] Step 2, extracting features from the operation data of the wind turbine to obtain feature values;
[0009] Step 3, establishing a physical model based on the aerodynamic and mechanical characteristics of the blade, with the input being the sample data composed of feature values and the output being the physical calculation value of the clearance;
[0010] Step 4, establishing a clearance prediction model based on deep learning, with the input being the sample data composed of feature values and the output being the predicted value of the clearance;
[0011] Step 5, training the clearance prediction model based on deep learning;
[0012] Step 6, using the trained clearance prediction model based on deep learning to monitor the blade clearance in real time.
[0013] In an alternative embodiment, the wind turbine operation data in Step 1 includes wind speed, wind direction deviation angle, impeller speed, and pitch angle.
[0014] In an alternative embodiment, Step 2 further includes normalizing the wind turbine operation data.
[0015] In an alternative embodiment, the physical model based on blade aerodynamic and mechanical characteristics in Step 3 is represented by the following formula:
[0016]
[0017] where d physics represents the true value of the clearance, M(ξ) represents the bending moment distribution at blade position ξ, E represents the blade elastic modulus, I(ξ) represents the cross-sectional moment of inertia distribution at blade position ξ, R represents the total length of the blade, and r represents the outer integral variable used to calculate the superposition effect from the blade root to the tip.
[0018] In an alternative embodiment, the clearance prediction model based on deep learning in Step 4 adopts a CNN network structure.
[0019] In an alternative embodiment, Step 5 uses a loss function with physical consistency constraints to train the clearance prediction model based on deep learning.
[0020] In an alternative embodiment, the loss function is:
[0021]
[0022] where τ total represents the total loss function, represents the data fitting loss function, N is the number of samples, d model and d true represent the predicted clearance value and the true clearance value respectively, is a preset weight, represents the physical consistency loss function, and d physics represents the physically calculated clearance value.
[0023] In an alternative embodiment, the real-time monitoring of the blade clearance in Step 6 includes using the trained clearance prediction model based on deep learning to output the predicted clearance value at the current moment in real time, and using the historical predicted clearance values to estimate the predicted clearance values in the next T time periods.
[0024] In an alternative embodiment, the real-time monitoring of the blade clearance in Step 6 further includes performing a blade collision risk score based on the predicted clearance value.
[0025] The beneficial effects of the present invention based on its technical solution are as follows:
[0026] A blade clearance monitoring method combining a physical model and an AI model provided by the present invention breaks through the limitations of traditional single physical or AI methods. By introducing a physical consistency loss function, it ensures that the AI model conforms to the actual physical laws of the blade, combining physical rationality and data-driven flexibility, providing a new intelligent solution for the safe operation of wind turbine blades, enabling the construction of an intelligent monitoring and protection system, and achieving high-precision assessment and dynamic protection of blade states. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of this specification. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.
[0028] Figure 1 It is a schematic flow chart of a blade clearance monitoring method combining a physical model and an AI model provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art belong to the scope protected by the embodiments of the present invention.
[0030] This embodiment provides a blade clearance monitoring method combining a physical model and an AI model. Referring to Figure 1 , the method includes:
[0031] Step 1: Use sensors to collect operation data of the wind turbine, including wind speed, wind direction deviation angle, impeller rotation speed, and pitch angle; use a clearance measurement device to measure the true value of the clearance.
[0032] Step 2: Extract features from the operation data of the wind turbine to obtain feature values.
[0033] During this process, the collected data is normalized to adjust the scale of the features, so that different features have the same measurement range, thereby improving the training efficiency and prediction accuracy of the model. Ensure the quality of the input data. The feature variables are collected as follows:
[0034] Feature Unit Range <![CDATA[Wind speed (V hub )]]> m / s [0,50] <![CDATA[Wind direction deviation angle (θ wind )]]> ° [-180,180] Impeller rotational speed (ω) Rad / s [0.1,2.0] Pitch angle (β) ° [0,90]
[0035] Process the feature quantity through the Min-Max normalization formula:
[0036]
[0037] where x is the original feature value, x min and x max are the minimum and maximum values of the feature respectively, and x' is the normalized feature value.
[0038] Step 3: Establish a physical model based on the aerodynamic and mechanical characteristics of the blade. Its input is the sample data composed of feature values, and the output is the physical calculation value of the clearance. The physical model based on the aerodynamic and mechanical characteristics of the blade is represented by the following formula:
[0039]
[0040] where, d physics represents the true value of the clearance, M(ξ) represents the bending moment distribution at the blade position ξ, E represents the blade elastic modulus, I(ξ) represents the cross-sectional moment of inertia distribution at the blade position ξ, R represents the total length of the blade, and r represents the outer integral variable used to calculate the superposition effect from the blade root to the tip.
[0041] Step 4: Establish a clearance prediction model based on deep learning using the CNN network structure. Its input is the sample data composed of feature values, and at each moment, the input X t = [V hub , θ wind , ω, β] represents, and the output is the predicted clearance value.
[0042] Step 5: Train the clearance prediction model based on deep learning. The Adam (Adaptive Moment Estimation) or SGD (stochastic gradient descent) optimization method can be used to optimize the model parameters. And in order to make the model have strong prediction ability, a loss function with physical consistency constraints is used for training, combining physical laws and data prediction results for constraints:
[0043] The data fitting loss function is established using the mean square error calculation formula:
[0044]
[0045] where, N is the number of samples, d model is the predicted clearance value, and d true is the measured true value of the clearance.
[0046] Establish a physical consistency loss function:
[0047]
[0048] Where: d mode1 is the net clearance prediction value, and d physics is the physical calculation value of the net clearance obtained by physical model calculation.
[0049] A total loss function is established by combining the weight term φ:
[0050] τ total = τ data + φτ physics
[0051] Where φ is the weight term, which is used to adjust the proportion of data fitting and physical model, so that the model output not only meets the requirements of data-driven, but also is consistent with the results calculated by physical formulas.
[0052] Step 6, use the trained net clearance prediction model based on deep learning to monitor the blade net clearance in real time, including outputting the net clearance prediction value at the current moment by using the trained net clearance prediction model based on deep learning in real time, and predicting the net clearance prediction value in the future T time period by using the historical net clearance prediction value. It is also possible to perform a blade collision risk score based on the net clearance prediction value, assist the maintenance personnel to give risk information and optimization strategies, and execute adjustments (such as pitch angle optimization or emergency shutdown) to ensure the safety of the unit.
[0053] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can 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.
[0054] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (modules, systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.
[0055] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means that implement the function specified in one or more of the processes and / or blocks Figure 1 in the process Figure 1 or processes and / or blocks
[0056] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the function specified in one or more of the processes Figure 1 or processes and / or blocks Figure 1 or blocks.
[0057] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic inventive concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the present invention.
[0058] 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 equivalent technologies, the present invention is also intended to include these modifications and variations.
Claims
1. A method for monitoring blade clearance by combining a physical model and an AI model, characterized in that, The method includes: Step 1: Use sensors to collect the operation data of the wind turbine, and use a clearance measurement device to measure the true value of the clearance; Step 2: Extract features from the operation data of the wind turbine to obtain feature values; Step 3: Establish a physical model based on the aerodynamic and mechanical characteristics of the blade, with the input being the sample data composed of feature values and the output being the physically calculated value of the clearance; Step 4: Establish a clearance prediction model based on deep learning, with the input being the sample data composed of feature values and the output being the predicted value of the clearance; Step 5: Train the clearance prediction model based on deep learning; Step 6: Use the trained clearance prediction model based on deep learning to monitor the blade clearance in real time.
2. The blade clearance monitoring method combining a physical model and an AI model according to claim 1, characterized in that: The operation data of the wind turbine described in Step 1 includes wind speed, wind direction deviation angle, impeller speed, and pitch angle.
3. The blade clearance monitoring method combining a physical model and an AI model according to claim 1, characterized in that: Step 2 also includes normalizing the operation data of the wind turbine.
4. The blade clearance monitoring method combining a physical model and an AI model according to claim 1, characterized in that: The physical model based on the aerodynamic and mechanical characteristics of the blade described in Step 3 is represented by the following formula: where d physics represents the true value of the clearance, M(ξ) represents the bending moment distribution at the blade position ξ, E represents the blade elastic modulus, I(ξ) represents the sectional moment of inertia distribution at the blade position ξ, R represents the total length of the blade, and r represents the outer integral variable used to calculate the superposition effect from the blade root to the blade tip.
5. The blade clearance monitoring method combining a physical model and an AI model according to claim 1, characterized in that: The clearance prediction model based on deep learning described in Step 4 adopts a CNN network structure.
6. The blade clearance monitoring method combining a physical model and an AI model according to claim 1, wherein: Step 5 uses a loss function with physical consistency constraints to train the clearance prediction model based on deep learning.
7. The blade clearance monitoring method combining a physical model and an AI model according to claim 6, characterized in that: The loss function is: Among them, τ total represents the total loss function, represents the data fitting loss function, N is the number of samples, d model and d true represent the clearance prediction value and the clearance true value respectively, is a preset weight, represents the physical consistency loss function, d physics represents the clearance physical calculation value.
8. The blade clearance monitoring method combining a physical model and an AI model according to claim 1, characterized in that: The real-time monitoring of the blade clearance described in Step 6 includes using the trained clearance prediction model based on deep learning to output the predicted value of the clearance at the current moment in real time, and using the historical predicted values of the clearance to estimate the predicted value of the clearance in the next T time periods.
9. The blade clearance monitoring method combining a physical model and an AI model according to claim 1, characterized in that: The real-time monitoring of the blade clearance described in Step 6 also includes performing a blade collision risk score based on the predicted value of the clearance.