Self-adaptive tower drum damage detection method based on inherent frequency

Abstract

The invention relates to the technical field of wind power generation, and discloses a tower drum damage self-adaptive detection method based on inherent frequency. For a tower drum vibration signal, the efficient detection method for tower drum damage is established based on the inherent frequency. According to the specific technical scheme, the method comprises the following steps that 1, a signal processing module collects original tower drum vibration signals as input signals, and the input signals are subjected to a signal decomposition algorithm to obtain a plurality of modal functions; step 2, a feature extraction and adaptive threshold detection module extracts feature parameters of each modal function, applies an adaptive threshold detection method and obtains the inherent frequency of the tower drum; and step 3, an intelligent classification and damage detection module automatically classifies the extracted characteristic parameters through an intelligent classification model, and constructs a tower drum damage detection model, thereby realizing on-line tower drum damage adaptive detection. The method provides important reference value for health management of the wind turbine generator.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power, and in particular to a tower damage adaptive detection method based on natural frequency. Background Art

[0002] With the rapid development of wind power generation technology, wind turbines have been widely used around the world. The tower of a wind turbine, as its core supporting structure, bears the heavy responsibility of supporting the wind turbine and bearing the external wind load. Due to long-term exposure to complex environments such as strong winds and vibrations, the tower may suffer varying degrees of damage, which will affect the normal operation of the wind turbine and may even cause the tower to collapse in severe cases, causing huge economic losses and safety hazards. Therefore, timely and accurate detection of tower damage is crucial to ensure the safe and stable operation of wind turbines.

[0003] According to the literature [“Adaptive Active Generator Torque Controller Design UsingMulti-Objective Optimization for Tower Lateral Load Reduction in MonopileOffshore Wind Turbines,” (IEEE Access, vol. 11, pp. 115894-115910, 2023.)], measures to alleviate tower fatigue and reduce damage through torque control are proposed, but it can only alleviate damage and cannot effectively detect it. Traditional tower damage detection methods still rely mainly on manual inspections or regular equipment inspections. These methods not only have problems such as high cost, manual errors, and long detection cycles, but are also difficult to adapt to changing environmental conditions. To this end, developing a damage detection method that can operate effectively under different environmental conditions and has adaptive detection capabilities has become a key technical problem in the current research on tower health monitoring.

[0004] The present invention proposes an adaptive tower damage detection method based on natural frequency. By combining the signal decomposition algorithm and the intelligent classification model, it can efficiently and accurately detect tower damage in complex environments. The method can automatically identify the changes in the tower natural frequency and perform real-time monitoring and evaluation based on the differences between healthy and damaged states, thereby providing strong technical support for the health management of wind turbines and significantly improving the detection accuracy and response speed. Summary of the invention

[0005] The technical problem to be solved by the present invention is to propose an adaptive method based on natural frequency for the tower damage detection problem. By constructing a signal processing module, a feature extraction and adaptive threshold detection module and an intelligent classification and damage detection module to work together, efficient and adaptive detection of tower damage can be achieved, and potential damage can be discovered in a timely and accurate manner to ensure the safety and reliability of wind turbines.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A tower damage adaptive detection method based on natural frequency comprises the following steps:

[0008] Step 1, the signal processing module collects input signals and obtains multiple modal functions through a signal decomposition algorithm;

[0009] Step 2: feature extraction and adaptive threshold detection module, by extracting the characteristic parameters of each modal function, applying the adaptive threshold detection method and obtaining the tower natural frequency;

[0010] Step 3: Intelligent classification and damage detection module, which automatically classifies the extracted feature parameters through the intelligent classification model and constructs a tower damage detection model, thereby realizing online adaptive tower damage detection;

[0011] In step 1, the main function of the signal processing module is to decompose the input signal into multiple different intrinsic mode functions while filtering out noise, extract the dynamic characteristics of tower vibration, and provide a more accurate basis for subsequent damage detection;

[0012] In the step 1, the signal processing module uses a signal decomposition method to decompose the signal into a series of modal functions, each modal function can accurately reflect the change pattern of the signal in a specific time period, and the signal decomposition method adopted needs to effectively overcome the situation of spectrum aliasing and reasonably calculate the number of decomposition layers, and finally obtain multiple modal functions after denoising;

[0013] In step 2, the feature extraction in the feature extraction and adaptive threshold detection module is to obtain the characteristic parameters of the normal signal and the fault signal in each modal function by continuously iterating through a method of effectively measuring the irregularity of the inherent mode and other effective features;

[0014] In step 2, the adaptive threshold detection in the feature extraction and adaptive threshold detection module effectively analyzes the extracted modal function feature parameters through an adaptive detection method, and realizes adaptive threshold detection through the obtained modal function feature parameters;

[0015] In the step 2, in the feature extraction and adaptive threshold detection module, after completing the adaptive threshold detection, the feature parameters extracted from each modal function are compared and screened with the detection threshold, and the modal function containing the tower natural frequency information is accurately identified. Through this process, the tower natural frequency and its amplitude characteristics corresponding to the signal segment are determined;

[0016] In step 3, the intelligent classification and damage detection module adopts an efficient intelligent classification algorithm, and by constructing an intelligent classification model, the obtained tower natural frequency and amplitude characteristics are trained offline and updated repeatedly to improve the classification accuracy, thereby achieving the role of efficient intelligent classification;

[0017] In the step 3, the damage detection types in the intelligent classification and damage detection module are divided into tower settlement and inclination, tower surface damage, tower flange damage, tower bolt loosening, etc., and according to the severity of the damage, the most serious case will cause the tower to collapse. This module can well detect various damage types and automatically alarm and shut down the wind turbine system before irreversible damage occurs to the tower. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Attached Figure 1 is a system block diagram of the present invention; DETAILED DESCRIPTION

[0019] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are used to explain the present invention, and should not be interpreted as limiting the present invention.

[0020] Example

[0021] Combination Figure 1 The tower damage adaptive detection method based on natural frequency of the present invention comprises the following steps:

[0022] Step 1, the signal processing module collects input signals and obtains multiple modal functions through a signal decomposition algorithm;

[0023] As an adaptive time-frequency analysis method, intrinsic time scale decomposition adaptively decouples non-stationary signals into a series of independent modal functions and a residual component according to their own local time scale characteristics (the obtained modal function is called the intrinsic rotation component in this case), effectively improving spectrum aliasing and signal mutation, and effectively playing the role of signal decomposition and denoising, as shown in formula (1):

[0024]

[0025] Among them, X t represents the signal to be decomposed, represents the (k+1)th layer baseline component, represents the monotonic trend and the lowest frequency limit extracted, H represents the intrinsic rotation factor, and p represents the number of decomposition levels.

[0026] By repeating the above decomposition process with the baseline signal as a new signal to be decomposed, a series of intrinsic rotation components can be obtained, and these multiple intrinsic rotation components are used as the input signal of the next module;

[0027] Step 2: feature extraction and adaptive threshold detection module, by extracting the characteristic parameters of each modal function, applying the adaptive threshold detection method and obtaining the tower natural frequency;

[0028] Fractal dimension is an important parameter to describe the characteristics of fractal theory. In Euclidean geometry, the dimension is an integer, while in fractal geometry, the dimension is extended to the fractional dimension. The tower vibration signal can be regarded as a two-dimensional curve formed by bending and extending a one-dimensional straight line. The fractal dimension method can be used to measure the complexity of the tower vibration signal, and then the effective features of each inherent rotation component can be extracted according to the fractal dimension value.

[0029] As a kind of fractal dimension, the box dimension can effectively reflect the fault characteristics of the signal. The box dimension model is established as shown in formula (2):

[0030]

[0031] Where D represents the box dimension, ρ represents the grid divided into multidimensional Euclidean space, and N(ρ) represents the grid count of the set with grid width ρ;

[0032] The selection of detection threshold has an important influence on the acquisition of natural frequency. Through the fuzzy C-means clustering method, the box dimension characteristic parameters of all natural rotation components are divided into two categories, namely, the box dimension characteristic parameters containing the tower natural frequency and the box dimension characteristic parameters excluding the tower natural frequency. By setting the number of clusters and iteratively calculating the membership of sample data, the center of each cluster is updated with the membership function until the membership is stable.

[0033] When the smallest cluster center satisfies formula (3), the adaptive detection threshold formula (4) can be obtained:

[0034] min(center)>α·mean(D noise ) (3)

[0035]

[0036] Among them, center represents the cluster center, α represents the adjustment coefficient, and D noise represents the box dimension without the tower natural frequency, Th high Indicates the highest threshold, Th lowIndicates the minimum threshold;

[0037] After the box dimension characteristic parameters extracted by fractal theory are adaptively determined through the fuzzy C-means clustering algorithm, the natural frequency and amplitude characteristics of the tower can be effectively obtained and input into the next module;

[0038] Step 3: Intelligent classification and damage detection module, which automatically classifies the extracted feature parameters through the intelligent classification model and constructs a tower damage detection model, thereby realizing online adaptive tower damage detection;

[0039] A quantum fuzzy naive Bayes classification method is adopted. Firstly, the traditional naive Bayes model is extended to a fuzzy naive Bayes model by introducing fuzzy prior probability and fuzzy conditional probability. Based on this model, a fuzzy Bayes classification framework suitable for complex data processing is constructed.

[0040] On this basis, the fuzzy data set is further quantized, and a dedicated quantum computing path is designed. A quantum fuzzy naive Bayes classification algorithm is proposed, which is divided into a training phase and a prediction phase.

[0041] Specifically, in the training stage, the tower natural frequency and its amplitude characteristics obtained in step 2 are used as training data sets and encoded into quantum states. Quantum counting technology is combined with auxiliary quantum states to perform phase estimation operations, and the quantum states are measured to calculate the fuzzy conditional probability. The fuzzy conditional probability is accurately encoded through the characteristic probability amplitude of the quantum state to meet the subsequent classification requirements.

[0042] Specifically, in the prediction stage, the fuzzy probability of the online input data is encoded into the quantum state, and the fuzzy conditional probability obtained in the training stage is used to assign values ​​to the quantum state. In order to improve the classification accuracy, the Grover algorithm is used to amplify the quantum state probability amplitude corresponding to the target category label, while reducing the probability amplitude interference of irrelevant categories, thereby ensuring that the classification result meets the requirements of the fuzzy conditional probability. Finally, according to the calculated quantum fuzzy conditional probability, the input data is classified in real time to realize the online intelligent classification function.

[0043] The damage to wind turbine tower components can usually be divided into tower settlement and tilt, tower skin damage, tower flange damage, tower bolt loosening, etc., and the most serious damage may cause tower collapse. To avoid such risks, a large amount of offline data from the wind farm is processed by the signal processing module, feature extraction and adaptive threshold detection module, and the intelligent classification algorithm can classify various damages into different categories. After the online data enters the system, it can be quickly processed by the corresponding module to achieve accurate classification.

[0044] Through the intelligent classification and damage detection module, all types of tower damage can be fully monitored. Before irreversible damage occurs to the tower, the system will automatically send out an alarm signal and trigger the shutdown of the wind turbine, thereby effectively ensuring equipment safety and reducing losses.

[0045] The present invention does not require the addition of additional sensors. It relies on the original data of the wind turbine set and designs a natural frequency method in which three modules, namely, a signal processing module, a feature extraction and adaptive threshold detection module, and an intelligent classification and damage detection module, work together to effectively detect damage to large components of the wind turbine tower. It can still maintain high-precision damage detection capabilities under complex conditions such as tower shadow effect, wind shear, and in harsh environments such as typhoons.

[0046] The above description of the present invention is only an example for clearly illustrating the present invention, and is not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A tower damage adaptive detection method based on natural frequency, the present invention sets up a signal processing module, a feature extraction and adaptive threshold detection module and an intelligent classification and damage detection module, characterized in that: The method comprises the following steps: Step 1, the signal processing module collects input signals and obtains multiple modal functions through a signal decomposition algorithm; Step 2: feature extraction and adaptive threshold detection module, by extracting the characteristic parameters of each modal function, applying the adaptive threshold detection method and obtaining the tower natural frequency; Step 3: Intelligent classification and damage detection module, which automatically classifies the extracted feature parameters through the intelligent classification model and constructs a tower damage detection model, thereby realizing online adaptive tower damage detection.

2. The tower damage adaptive detection method based on natural frequency according to claim 1 is characterized in that: In step 1, the signal decomposition algorithm includes intrinsic time scale decomposition method, local mean decomposition method, empirical mode decomposition method, and variational mode decomposition method.

3. The tower damage adaptive detection method based on natural frequency according to claim 1 is characterized in that: In step 2, the feature extraction methods include fuzzy entropy function method, principal component analysis method, and fractal dimension method.

4. The tower damage adaptive detection method based on natural frequency according to claim 1 is characterized in that: In step 2, the adaptive threshold detection method can be established according to the minimum mean square error filter model, the Kalman filter model, and the Bayesian maximum likelihood estimation model.

5. The tower damage adaptive detection method based on natural frequency according to claim 1 is characterized in that: In step 3, the intelligent classification model includes a logistic regression model, a decision tree model, a support vector machine model, and a random forest model, and also includes a combined application of the above intelligent classification models. The optimization algorithm includes a chaotic sparrow optimization algorithm, a convergence factor improved optimization algorithm, and a spectral optimization algorithm.

6. The tower damage adaptive detection method based on natural frequency according to claim 1 is characterized in that: In step 3, the types of tower damage include tower settlement and tilting, tower surface damage, tower flange damage, and tower bolt loosening.

Images