Intelligent self-adaptive rust remover spraying system and method for wind generating set

Abstract

The invention discloses an intelligent self-adaptive rust remover spraying system and method of a wind generating set, and relates to the field of rust removal of wind generating sets. The system comprises a multi-modal data monitoring unit, a corrosion characteristic analysis unit, an intelligent decision regulation and control unit, a spray quantity precise control unit, an angle self-adaptive adjustment unit and a rotatable magnetic spray head group execution unit, and the corrosion state is predicted through a long and short term memory network by collecting multi-modal data such as vibration, temperature and humidity. The Gaussian mixture model analyzes the characteristics, intelligently decides to generate a spraying instruction, accurately controls the spraying amount and the angle of the nozzle, and realizes the accurate spraying of the rust remover as required. The method comprises the steps of data acquisition and fusion, state prediction, feature analysis, decision instruction generation, accurate adjustment and feedback adjustment. The problems that traditional rust removal is low in efficiency, high in cost, poor in precision and the like are solved, the rust removal effect is improved, the maintenance cost is reduced, and stable operation of the wind generating set is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind turbine rust removal, and in particular to an intelligent self-adaptive rust removal system and method for wind turbines. BACKGROUND

[0002] With the continuous growth of global demand for clean energy, wind power as an important form of renewable energy, its installed capacity is rising. Wind turbines are exposed to complex outdoor environments for a long time, and key components such as tower, blades, and transmission chain are easily eroded by rain, salt spray, and sand, resulting in rust problems. Rust not only reduces the mechanical strength of the components and increases the risk of equipment failure, but also may cause downtime maintenance, leading to reduced power generation efficiency and increased operating costs. According to statistics, rust-induced failures account for 30-40% of wind turbine maintenance problems, seriously affecting the economy and reliability of the wind power industry. Therefore, efficient and accurate rust removal technology has become a key to ensuring the stable operation of wind turbines.

[0003] Currently, the main rust removal methods for wind turbines are manual spraying and traditional automated equipment. Manual spraying relies on the experience and proficiency of the operator, which not only has high labor intensity and low efficiency, but also has high safety risks in high-altitude and complex structural components. At the same time, manual operation cannot accurately control the spraying amount and angle of the rust removal agent, which may result in waste of rust removal agent or incomplete rust removal. Traditional automated rust removal equipment mostly uses fixed spraying mode and lacks the ability to adapt to different rust levels and component positions. These devices cannot flexibly adjust the spraying strategy according to the actual operating state and rust condition of the wind turbine, and when facing irregular-shaped and dynamically running components, they often have problems such as deviation of spraying angle and incomplete coverage, resulting in poor rust removal effect and difficulty in meeting the needs of fine maintenance of wind turbines.

[0004] In addition, existing rust removal technologies have obvious shortcomings in data utilization and intelligent decision-making. Most devices rely only on single type of monitoring data, such as simple visual images or local humidity information, lack comprehensive analysis and deep mining of multi-dimensional data, and cannot accurately judge the rust state and development trend. At the same time, traditional rust removal systems lack intelligent decision-making mechanisms and cannot develop optimal rust removal strategies according to actual working conditions, making it difficult to achieve precise and efficient use of rust removal agents. These technical shortcomings not only limit the improvement of rust removal efficiency and quality, but also increase the operating cost of wind power, making it urgent to develop an intelligent and self-adaptive rust removal system and method to solve the above problems. SUMMARY

[0005] In order to overcome the shortcomings and deficiencies of the prior art, the present application provides an intelligent self-adaptive rust removal system and method for wind turbines.

[0006] The technical solution adopted by the present invention is an intelligent adaptive rust remover spraying system for a wind turbine generator set, the system comprising:

[0007] Multimodal data monitoring unit: This unit collects multimodal data such as vibration, temperature, and humidity signals from wind turbines, fuses different types of data using a multimodal data fusion algorithm, and then inputs the data into a long-short-term memory network model to monitor and predict the corrosion status of various components of the wind turbine in real time.

[0008] Corrosion feature analysis unit: This unit uses a Gaussian mixture model to perform cluster analysis on the data, mine the corrosion features in the data, and determine the characteristic distribution of different corrosion degrees. The output end is connected to the intelligent decision-making and control unit;

[0009] Intelligent decision-making and control unit: This unit combines the corrosion status prediction results of the status monitoring unit and the characteristic distribution information of the corrosion characteristic analysis unit, and generates rust remover injection control instructions for different corrosion locations and degrees based on preset decision rules and optimization algorithms;

[0010] Precision spray control unit: This unit controls the amount of rust remover sprayed from each nozzle by adjusting the flow control valve of the rotatable magnetic nozzle group. The spray amount matches the degree of rust.

[0011] Angle adaptive adjustment unit: This unit uses a magnetic angle adjustment mechanism according to the instructions of the intelligent decision-making and control unit to automatically adjust the angle of the rotatable magnetic nozzle group as the chain rotates, so that the nozzle always sprays perpendicular to the rusted area;

[0012] Rotatable magnetic nozzle group execution unit: This unit consists of multiple rotatable magnetic nozzles, which fit tightly to the chain surface through magnetic attraction. Under the coordinated action of the spray volume precision control unit and the angle adaptive adjustment unit, the rust remover is sprayed on the rusted parts of the wind turbine according to the preset spray volume and angle.

[0013] Furthermore, the multimodal data monitoring unit adopts an improved weighted fusion algorithm in the multimodal data fusion process, assuming that the vibration signal is V, the temperature signal is T, and the humidity signal is H. The fused data D f Calculated by the following formula:

[0014] D f =w V LSTM V (V)+w T LSTM T (T)+w H LSTM H (H)

[0015] Among them, w V , w T , w H are the fusion weights of vibration signal, temperature signal and humidity signal respectively, and w V +w T +w H =1;LSTM V , LSTM T , LSTM H They are the long short-term memory network processing functions for vibration signal, temperature signal and humidity signal respectively. In the wind turbine generator set, the weight w of the vibration signal V Dynamically adjust according to the operating conditions and vibration characteristics of the unit, with a value range of 0.3-0.5; the temperature signal weight w T Considering the influence of different ambient temperatures on rust, the value range is 0.2-0.4; the humidity signal weight w H Combined with local climatic conditions, the value range is 0.2-0.3.

[0016] Furthermore, the corrosion feature analysis unit analyzes the fused data D f When modeling the Gaussian mixture model, the expectation maximization algorithm is used for parameter estimation. Assume that the Gaussian mixture model consists of K Gaussian distributions, and the probability density function of the kth Gaussian distribution is:

[0017]

[0018] Among them, θ k =(μ k ,Σ k ) is the parameter of the kth Gaussian distribution, μ k is the mean vector, Σ k is the covariance matrix, d is the data D f The dimension of the wind turbine is the mean vector μ k It reflects the central trend of multimodal fusion data under different corrosion degrees, and the covariance matrix Σ k It reflects the degree of discreteness of the data distribution, and updates the parameter θ through continuous iteration k , the Gaussian mixture model can fit the distribution of the data.

[0019] Furthermore, the intelligent decision-making and control unit introduces a decision rule based on fuzzy logic when generating the rust remover injection control instruction. Assume that the rust degree is S, which is determined by the state monitoring unit and the rust characteristic analysis unit; the injection amount adjustment coefficient is C q , the angle adjustment coefficient is C aThe fuzzy rule base defines the corresponding fuzzy rules of injection amount and angle adjustment coefficient according to different corrosion degree intervals. When the corrosion degree S is in the mild corrosion interval, the injection amount adjustment coefficient C q The value is small, the angle adjustment coefficient C a Maintain the nozzle basically vertically; when the rust level S is in the severe rust range, the injection volume adjustment coefficient C q The value is larger, the angle adjustment coefficient C a Adjustments are made according to the chain rotation conditions, and different injection control instructions are generated according to different corrosion conditions through fuzzy logic reasoning.

[0020] Furthermore, the injection volume precision control unit adopts a model-based predictive control method when adjusting the flow control valve. Assuming the current time is t, the prediction time domain is N, the control time domain is M, M≤N, the control input of the flow control valve is u(t), and the output of the system is the rust remover injection volume y(t), a prediction model of the system is established:

[0021] y(t+j∣t)=f(y(t),y(t-1),…,y(tn y ), u(t), u(t

[0022] -1),…,u(tn u ))

[0023] Where j = 1, 2, ..., N, n y and n u are the output and input orders respectively. In the wind turbine generator set, according to the injection amount target value y issued by the intelligent decision-making control unit ref (t), by optimizing the objective function:

[0024]

[0025] Among them, λ j The weighted coefficient of the control input is used to obtain the optimal control input u(t) to control the injection amount of the rust remover.

[0026] Furthermore, the angle adaptive adjustment unit adopts an angle prediction model based on a neural network when adjusting the angle of the rotatable magnetic nozzle group. The rotation speed of the chain is set to v, the current nozzle angle is α, and the position information of the rusted part is p. These input information are input into a multi-layer perceptron neural network. The output of the network is the predicted nozzle adjustment angle Δα. The training data of the neural network comes from historical experimental data. The weights and bias of the network are continuously adjusted through the back propagation algorithm, so that the predicted adjustment angle adapts to the rotation of the chain and the changes in the rusted part, and the nozzle always sprays perpendicular to the rusted part.

[0027] Furthermore, the magnetic attraction structure of the rotatable magnetic nozzle group execution unit adopts a new type of permanent magnetic material and an optimized magnetic circuit design. Assuming that the residual magnetic induction intensity of the permanent magnetic material is B r , the coercive force is H c , the magnetic resistance of the magnetic circuit is R m By optimizing the selection of permanent magnetic materials and the design of magnetic circuit, the magnetic attraction force F between the nozzle and the chain surface is m satisfy:

[0028]

[0029] Among them, A is the cross-sectional area of ​​the magnetic circuit, μ0 is the vacuum magnetic permeability, and in the operating environment of the wind turbine, the magnetic attraction force F m The nozzle is kept in close contact with the surface of the chain during the chain rotation, while the rotation flexibility of the nozzle is not affected by excessive magnetic attraction.

[0030] Furthermore, the multimodal data monitoring unit pre-processes the collected multimodal data by combining wavelet denoising and principal component analysis. The original vibration signal V, temperature signal T, and humidity signal H are obtained after wavelet denoising respectively. denoised , T denoised , H denoised Then, principal component analysis is performed on the noised data to extract the principal components PC1, PC2, ..., PC n , where n is the number of principal components retained. In wind turbines, the number of principal components retained, n, is selected based on the characteristics of the data and the requirements of condition monitoring, and is set to 2-3.

[0031] Furthermore, the intelligent decision-making and control unit considers the operating status and maintenance plan of the wind turbine when formulating the rust remover injection strategy, and assumes that the operating status of the wind turbine is divided into normal operation, maintenance status and fault status, and uses S normal , S maintenance , S fault Indicates maintenance plan with P maintenance Indicates, including maintenance time, maintenance content information; the intelligent decision-making and control unit adjusts the frequency, injection amount and injection time of the rust remover according to different operating conditions and maintenance plans. In the maintenance state, the injection amount and injection frequency of the rust remover are increased; in the fault state, it decides whether to suspend the rust remover injection operation according to the type and severity of the fault.

[0032] An intelligent self-adaptive rust remover spraying method for a wind turbine generator set, the method comprising:

[0033] S1. Multimodal data acquisition and fusion: Real-time acquisition of multimodal data including vibration signals, temperature signals, and humidity signals of wind turbines. Utilizing a multimodal data fusion algorithm, different types of data are fused to generate fused monitoring data.

[0034] S2. Long-Short-Term Memory Network Status Prediction: The fused monitoring data is input into the long-short-term memory network model. Through the model's learning and prediction functions, the corrosion status of each component of the wind turbine generator set is dynamically predicted to determine the locations where corrosion may occur and the development trend of the corrosion degree.

[0035] S3. Gaussian mixture model feature analysis: A Gaussian mixture model is used to perform cluster analysis on the fused monitoring data, mining the corrosion features included in the data and determining the characteristic distribution of different corrosion degrees;

[0036] S4. Intelligent decision-making instruction generation: Based on the corrosion status prediction results and the corrosion characteristic distribution information, according to the preset decision rules and optimization algorithm, the rust remover injection control instructions for different corrosion locations and degrees are generated, including the control information of the injection amount and injection angle;

[0037] S5. Precise adjustment of spray volume and angle: Based on intelligent decision-making instructions, the spray volume precision control unit adjusts the flow control valve of the rotatable magnetic nozzle group to control the rust remover spray volume of each nozzle; the angle adaptive adjustment unit uses a magnetic angle adjustment mechanism to automatically adjust the angle of the rotatable magnetic nozzle group as the chain rotates, ensuring that the nozzle is always sprayed perpendicular to the rusted area;

[0038] S6. Execution and feedback adjustment: The rotatable magnetic nozzle group execution unit sprays rust remover on the corroded parts of the wind turbine generator set according to the spray volume and angle. At the same time, the system collects the spray effect data in real time, compares and analyzes it with the expected spray effect, and dynamically adjusts the spray strategy according to the feedback results.

[0039] Beneficial effects: An intelligent adaptive rust remover spraying system and method for wind turbine generator sets. The present invention realizes the intelligence, precision and efficiency of rust removal operations through multi-unit collaboration and innovative algorithm design. The system uses a multimodal data monitoring unit to capture multi-dimensional signals such as vibration, temperature, and humidity in real time. Combined with the rust feature analysis unit of the Gaussian mixture model, it can deeply explore the potential corrosion risks of wind turbine components, predict the rust location and development trend in advance, change the traditional passive rust removal mode, and provide data support for active maintenance. The intelligent decision-making and control unit integrates fuzzy logic and model predictive control algorithms, and generates a refined spraying strategy based on multi-source information. It takes into account the differences in the degree of rust and combines the unit's operating status and maintenance plan, so that the spray volume precision control unit can adjust the nozzle flow rate as needed to avoid waste of rust remover; the angle adaptive adjustment unit cooperates with the magnetic rotatable nozzle group to automatically keep the nozzle vertical to the rust surface during the chain rotation to ensure uniform coverage of the rust remover, overcoming the problem of manual precision operation under complex working conditions. In terms of methodology, the closed-loop process from data collection, state prediction, feature analysis to decision execution and feedback adjustment achieves intelligent management and control of the entire rust removal process. Compared with traditional rust removal methods, this system and method significantly improves rust removal efficiency and quality, reduces equipment failures and downtime caused by rust, and reduces maintenance costs. At the same time, by precisely controlling the amount of rust remover used, it reduces chemical consumption and environmental pollution, in line with the concept of green environmental protection. In addition, the highly automated nature of the system significantly reduces reliance on manual labor, improves the safety and standardization of maintenance operations, provides reliable guarantees for the stable and efficient operation of wind power equipment, and has good economic value and social benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a diagram of the system unit composition of the present invention;

[0041] Figure 2 The figure is a flow chart of the method steps of the present invention. DETAILED DESCRIPTION

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0043] like Figure 1 As shown, an intelligent adaptive rust remover spraying system and method for a wind turbine generator set, the system includes:

[0044] Multi-modal data monitoring unit: This unit collects multi-modal data such as vibration signals, temperature signals, and humidity signals of the wind turbine generator set. It uses multi-modal data fusion algorithms to fuse different types of data, and then inputs them into a long short-term memory network model to monitor and predict the corrosion status of each component of the wind turbine generator set. The output of this unit is connected to the intelligent decision-making and control unit.

[0045] Specifically, this unit is the information perception core of the entire system. Its core function is to collect and analyze various key signals of the wind turbine generator set. It collects vibration signals, which can reflect the running stability of internal components. For example, when a gear is worn out, the vibration frequency and amplitude will change. Temperature signals are also collected, as different components have different temperatures under normal and abnormal conditions. For example, high temperature of the generator winding may indicate insulation failure. Humidity signals are also important because humidity is one of the key environmental factors affecting corrosion. These multi-modal data each contain different aspects of information, and separate analysis may not fully understand the corrosion status of the unit.

[0046] After collecting these data, the unit uses multi-modal data fusion algorithms to integrate them. This fusion is not simply a superposition, but a deep exploration of the correlation and complementary information between different data. For example, when vibration signals are abnormal and temperature and humidity are high, it is more likely to indicate that the component has corrosion problems. The fused data is input into a long short-term memory network model, which has strong time series analysis capability and can learn and remember the historical change rules of the data, thereby accurately predicting the corrosion status of each component of the wind turbine generator set. It can predict the location and degree of corrosion that may occur in the future based on the data change trend in the past period.

[0047] The significance of this unit is to provide accurate corrosion status information for the entire system, just like a doctor diagnosing a disease through various examination methods. By predicting the corrosion status in advance, the system can take timely measures to remove rust and prevent further deterioration of the equipment. For example, in offshore wind turbine generators, due to high humidity and high salt content, corrosion is more serious. This unit can timely detect potential corrosion risks and provide strong support for subsequent rust removal work, ensuring the stable operation of the unit.

[0048] Corrosion feature analysis unit: This unit receives the fused data from the state monitoring unit and uses a Gaussian mixture model to perform clustering analysis on the data to extract corrosion features. It determines the feature distribution of different corrosion levels to provide a feature basis for the subsequent development of the rust removal agent injection strategy. Its output is connected to the intelligent decision-making and control unit.

[0049] Specifically, this unit is responsible for conducting in-depth analysis of the fused data output by the condition monitoring unit. It uses the Gaussian mixture model, a powerful data analysis tool, to perform cluster analysis on the data. This model can classify the data into distinct categories, each representing a distinct corrosion pattern. For example, some data may represent characteristics of mild corrosion, while others may represent characteristics of severe corrosion. Through this cluster analysis, the unit can identify hidden patterns and characteristics in the data and determine the characteristic distribution of different corrosion levels.

[0050] In practical applications, this unit can determine the current corrosion status of a wind turbine based on different characteristic distributions. For example, if the distribution characteristics of a certain type of data are similar to known characteristics of severe corrosion, it can be inferred that the corresponding component may be severely corroded. This is crucial for developing targeted rust removal strategies. Furthermore, the Gaussian mixture model can adaptively adjust as data is continuously updated, making the analysis of corrosion characteristics more accurate and reliable.

[0051] The significance of this unit lies in providing detailed features to inform subsequent decision-making. It helps the intelligent decision-making and control unit more accurately understand the rust situation, thereby formulating a more appropriate rust remover injection strategy. For example, for areas with mild rust, the rust remover injection volume can be appropriately reduced; for areas with severe rust, the injection volume and frequency can be increased. This precise decision-making improves rust removal effectiveness while reducing rust remover costs.

[0052] Intelligent decision-making and control unit: This unit combines the corrosion status prediction results of the status monitoring unit and the characteristic distribution information of the corrosion characteristic analysis unit. According to the preset decision rules and optimization algorithm, it generates rust remover injection control instructions for different corrosion locations and degrees. Its output ends are connected to the spray volume precision control unit and the angle adaptive adjustment unit respectively;

[0053] Specifically, the intelligent decision-making and control unit serves as the "brain" of the entire system. It receives corrosion status predictions from the condition monitoring unit and characteristic distribution information from the corrosion feature analysis unit. This information acts as a detailed "diagnostic report," enabling a comprehensive understanding of the wind turbine's corrosion status. It then comprehensively analyzes and processes this information based on pre-set decision rules and optimization algorithms. These decision rules, developed based on extensive experimental data and practical experience, specify the appropriate injection strategy for different corrosion states and characteristics.

[0054] Optimization algorithms are used to further refine these decisions, ensuring that the spraying strategy is both effective in rust removal and maximizes resource conservation. For example, when considering the spray rate, not only the degree of rust must be considered, but also the cost of the rust remover and the unit's operating efficiency. The optimization algorithm determines the optimal spray rate and spray time while meeting rust removal requirements. The spray control instructions generated by this unit include precise control information for the spray rate and spray angle of each nozzle, ensuring that the rust remover is accurately sprayed to the required area.

[0055] The intelligent decision-making and control unit plays a crucial role, enabling the entire system to make intelligent decisions based on actual conditions, achieving automated and precise rust removal operations. During actual wind turbine operation, different components may be in varying states of corrosion. This unit can tailor a personalized spraying strategy for each component. For example, for components frequently exposed to harsh environments, the spraying frequency can be increased; for relatively enclosed components, the spray volume can be appropriately reduced. This personalized decision-making improves rust removal effectiveness and extends the life of the unit.

[0056] Precision spray control unit: This unit receives the spray control instructions from the intelligent decision-making and control unit, and precisely controls the amount of rust remover sprayed from each nozzle by adjusting the flow control valve of the rotatable magnetic nozzle group to ensure that the spray amount matches the degree of rust. This unit is connected to the flow control component of the rotatable magnetic nozzle group.

[0057] Specifically, the precise spray control unit's primary task is to precisely adjust the flow control valves of the rotatable magnetic nozzle assembly based on the spray control commands issued by the intelligent decision-making and control unit. The flow control valve acts like a faucet; adjusting its opening controls the amount of rust remover sprayed. This unit independently controls the spray rate of each nozzle to ensure that each corroded area receives the appropriate amount of rust remover. For example, for severely corroded areas, the spray rate needs to be increased; for less corroded areas, the spray rate can be reduced.

[0058] To achieve precise control, the unit utilizes advanced control technology and high-precision sensors. These sensors monitor the flow control valve opening and the flow rate of the rust remover in real time, feeding this information back to the control unit. Based on this feedback, the control unit continuously adjusts the flow control valve opening to precisely control the injection volume. This closed-loop control method effectively improves control accuracy and reduces errors. In practice, even with environmental changes such as pressure fluctuations and temperature variations, the unit can promptly adjust the injection volume to ensure consistent rust removal results.

[0059] The significance of a precise spray control unit is to improve the efficiency and effectiveness of rust removers. By precisely controlling the spray volume, we can avoid wasting rust removers and reduce costs. Furthermore, the appropriate spray volume ensures that rusted areas are fully treated, improving the quality of rust removal. For example, in some large wind farms, numerous units require rust removal. If the spray volume can be precisely controlled, significant annual rust remover costs can be saved while also improving the overall operational efficiency of the wind farm.

[0060] Angle adaptive adjustment unit: This unit uses a magnetic angle adjustment mechanism according to the instructions of the intelligent decision-making and control unit to automatically adjust the angle of the rotatable magnetic nozzle group as the chain rotates, ensuring that the nozzle is always perpendicular to the rusted area for spraying. This unit is connected to the angle adjustment component of the rotatable magnetic nozzle group.

[0061] Specifically, the adaptive angle adjustment unit's core function is to ensure that the nozzles of the rotating magnetic nozzle assembly are always sprayed perpendicular to the rusted area. It utilizes a magnetic angle adjustment mechanism, which allows the nozzles to automatically adjust their angle as the chain rotates. This magnetic angle adjustment mechanism provides strong suction, ensuring that the nozzles adhere closely to the chain surface during rotation while allowing for flexible angle adjustment. For example, when the chain is in different positions and angles, the unit can quickly adjust the nozzle angle to ensure that the nozzles are always aligned with the rusted area.

[0062] This unit achieves precise angle adjustment by receiving angle control commands from the intelligent decision-making and control unit. The control command contains information about the angle at which the nozzle needs to be adjusted, and the unit drives the angle adjustment mechanism based on this information. During the adjustment process, real-time corrections are made based on the chain position and angle information fed back by the sensor to ensure the accuracy of the adjustment. For example, when the chain rotates rapidly, the sensor can monitor the chain's movement in real time. The angle adaptive adjustment unit adjusts the nozzle angle in a timely manner based on this information to ensure that the nozzle is always perpendicular to the rusted area.

[0063] The importance of the adaptive angle adjustment unit lies in improving the effectiveness of rust remover spraying. When the nozzle is sprayed perpendicular to the rusted area, the rust remover is more evenly coated on the rusted surface, improving the efficiency and quality of rust removal. If the nozzle angle is inaccurate, the rust remover may be sprayed onto surrounding non-rusted areas, causing waste and compromising the rust removal effect. In actual wind turbine operation, the chain rotation is complex. The adaptive angle adjustment unit adapts to these complex operating conditions, ensuring that the nozzle always maintains the optimal spray angle, providing a strong guarantee for efficient rust removal.

[0064] The rotatable magnetic nozzle group execution unit is composed of multiple rotatable magnetic nozzles, which are tightly attached to the surface of the chain through magnetic attraction. Under the coordination of the spray amount precision control unit and the angle adaptive adjustment unit, the unit performs rust remover spraying operations on the rusted parts of the wind turbine generator set according to the precise spray amount and angle.

[0065] Specifically, the rotatable magnetic nozzle group execution unit is the execution terminal of the entire system, which is composed of multiple rotatable magnetic nozzles. These nozzles are tightly attached to the surface of the chain through magnetic attraction and can move with the rotation of the chain. The magnetic attraction method has strong adsorption force, which can ensure that the nozzles do not fall off during the rotation of the chain, and at the same time will not cause damage to the chain. The rotatable design of the nozzles allows them to be flexibly adjusted according to different rusted parts and angles, ensuring that the rust remover can be accurately sprayed to the areas that need to be treated.

[0066] In actual work, the rotatable magnetic nozzle group execution unit performs rust remover spraying operations on the rusted parts of the wind turbine generator set according to the precise spray amount and angle under the coordination of the spray amount precision control unit and the angle adaptive adjustment unit. Each nozzle can independently control the spray amount and spray angle to adapt to different rusting conditions. For example, for large areas of rusted parts, multiple nozzles can be used for simultaneous spraying to improve the efficiency of rust removal; for small areas of rusted parts, the spray amount and angle of a single nozzle can be accurately controlled for precise treatment.

[0067] The function of this unit is to convert the decisions and controls of the previous units into actual rust removal operations. Its performance directly affects the quality of rust removal. High-quality rotatable magnetic nozzle groups can ensure that the rust remover is evenly distributed on the rusted surface, improving the quality and efficiency of rust removal. During the long-term operation of the wind turbine generator set, this unit needs to withstand various harsh environments, such as high temperature, high humidity, strong wind, etc. Therefore, the material and structural design of the nozzle need to have high reliability and durability to ensure the stable operation and long-term effectiveness of the system.

[0068] Preferably, in the multi-modal data monitoring unit, an improved weighted fusion algorithm is used in the multi-modal data fusion process. Let the vibration signal be V, the temperature signal be T, and the humidity signal be H. The fused data D f is calculated by the following formula:

[0069] D f = w V · LSTM V (V) + w T · LSTM T (T) + w H · LSTM H (H)

[0070] Among them, w V , w T , w H are the fusion weights of vibration signal, temperature signal and humidity signal respectively, and w V +w T +w H =1;LSTM V , LSTM T , LSTM H They are the long short-term memory network processing functions for vibration signal, temperature signal and humidity signal respectively. In the wind turbine generator set, the weight w of the vibration signal V It can be dynamically adjusted according to the operating conditions and vibration characteristics of the unit, and the general value range is 0.3-0.5; the temperature signal weight w T Considering the influence of different ambient temperatures on rust, the value range is 0.2-0.4; the humidity signal weight w H Combined with local climatic conditions, the value range is 0.2-0.3. The improved fusion algorithm can more accurately capture the impact of different modal data on the corrosion status and improve the accuracy of condition monitoring.

[0071] Specifically, an improved weighted fusion algorithm is used to process signals such as vibration, temperature, and humidity. During wind turbine operation, the vibration signal weight is generally between 0.3 and 0.5. For example, when a gearbox fails, the vibration intensifies. This weight can more clearly highlight the impact of the vibration signal on the judgment of the corrosion status. The temperature signal weight is between 0.2 and 0.4. High temperature environments tend to accelerate corrosion, and this weight range can reasonably reflect the temperature factor. The humidity signal weight is between 0.2 and 0.3. Coastal areas have high humidity, so assigning a corresponding weight to the humidity signal can more accurately fuse multimodal data. After processing with a long-short-term memory network, the system can more accurately capture the correlation between data, predict the corrosion risk of components such as bearing seats months in advance, avoid misjudgments caused by data fusion bias, and provide a reliable basis for subsequent decision-making.

[0072] Preferably, in the corrosion feature analysis unit, the fused data D f When modeling the Gaussian mixture model, the expectation maximization (EM) algorithm is used for parameter estimation. Assume that the Gaussian mixture model consists of K Gaussian distributions, and the probability density function of the kth Gaussian distribution is:

[0073]

[0074] Among them, θ k =(μ k ,Σ k ) is the parameter of the kth Gaussian distribution, μ k is the mean vector, Σ k is the covariance matrix, d is the data D fThe dimension of the wind turbine is the mean vector μ k It reflects the central trend of multimodal fusion data under different corrosion degrees, and the covariance matrix Σ k It reflects the degree of discreteness of the data distribution. By continuously iteratively updating the parameter θ k , which enables the Gaussian mixture model to better fit the distribution of the data and thus more accurately identify different corrosion characteristics.

[0075] Specifically, the expectation maximization algorithm is used to model the fused data. In practical applications, the multimodal fusion data of wind turbines is divided into 3-5 Gaussian distributions, corresponding to states such as no corrosion, light corrosion, moderate corrosion, and severe corrosion. The mean vector reflects the central trend of the data under different degrees of corrosion. For example, in the case of severe corrosion, vibration, temperature and other data show a specific offset on the mean vector; the covariance matrix reflects the discrete characteristics of the data. By iteratively updating parameters, the system can accurately identify the characteristic patterns of corrosion in hidden areas such as tower joints. Compared with traditional methods, the accuracy of corrosion feature recognition can be increased by more than 20%, laying the foundation for the formulation of differentiated rust removal strategies.

[0076] Preferably, the intelligent decision-making and control unit introduces a decision rule based on fuzzy logic when generating the rust remover injection control instruction. Assume that the rust degree is S, which is jointly determined by the state monitoring unit and the rust characteristic analysis unit; the injection amount adjustment coefficient is C q , the angle adjustment coefficient is C a The fuzzy rule base defines the corresponding fuzzy rules for injection volume and angle adjustment coefficients according to different corrosion degree intervals. For example, when the corrosion degree S is in the mild corrosion interval, the injection volume adjustment coefficient C q The value is small, the angle adjustment coefficient C a Maintain the nozzle basically vertically; when the rust level S is in the severe rust range, the injection volume adjustment coefficient C q The value is larger, the angle adjustment coefficient C a It can make more flexible adjustments according to the chain rotation situation, and through fuzzy logic reasoning, it can generate appropriate injection control instructions more precisely according to different corrosion conditions.

[0077] Specifically, the degree of rust is divided into three ranges: mild, moderate, and severe, corresponding to different spray volume adjustment coefficients and angle adjustment coefficient rules. In the mild rust range, the spray volume is reduced by 20%-30%, and the angle fine-tuning range is controlled within ±5° to save reagents. In severe rust, the spray volume is increased by 50%-70%, and the angle is dynamically adjusted in real time based on the chain movement. When moderate rust is detected at the root of the blade, the system automatically adjusts the spray volume to 1.3 times the standard amount based on fuzzy rules, and simultaneously updates the nozzle angle every 0.5 seconds to ensure that the rust remover accurately covers the rusted area and avoid over- or under-spraying.

[0078] Preferably, the injection volume precision control unit adopts a model predictive control (MPC) method when adjusting the flow control valve. Let the current time be t, the prediction time domain be N, the control time domain be M, M≤N, the control input of the flow control valve be u(t), and the output of the system be the rust remover injection volume y(t). Establish a prediction model for the system:

[0079] y(t+j∣t)=f(y(t),y(t-1),…,y(tn y ), u(t), u(t

[0080] -1),…,u(tn u ))

[0081] Where j = 1, 2, ..., N, n y and n u are the order of output and input respectively. In the wind turbine generator set, according to the injection amount target value y issued by the intelligent decision control unit ref (t), by optimizing the objective function:

[0082]

[0083] Among them, λ j The weighted coefficient of the control input is solved to obtain the optimal control input u(t), thereby achieving precise control of the rust remover injection amount.

[0084] Specifically, during wind turbine rust removal operations, the prediction horizon is set to 10 minutes into the future, the control horizon to 3 minutes, and optimization calculations are performed every 30 seconds. Based on the target injection volume provided by the intelligent decision-making unit, combined with current flow rate, pressure, and other parameters, the flow control valve opening is adjusted by optimizing the objective function. For example, if the pressure in the rust remover pipeline fluctuates, the system can adjust the injection volume from a deviation of 15% from the target value to within an error range of ±3% within 1 minute, ensuring stable rust remover dosage in key areas such as tower bolts and ensuring consistent rust removal results.

[0085] Preferably, the angle adaptive adjustment unit uses a neural network-based angle prediction model to adjust the angle of the rotatable magnetic nozzle assembly. The chain's rotational speed is v, the current nozzle angle is α, and the location of the corroded area is p. This input is fed into a multi-layer perceptron (MLP) neural network, and the network output is the predicted nozzle adjustment angle Δα. The neural network is trained using a large amount of historical experimental data. The network's weights and biases are continuously adjusted through a backpropagation algorithm, allowing the predicted adjustment angle to more accurately adapt to the chain's rotation and changes in the corroded area, ensuring that the nozzle always sprays perpendicular to the corroded area.

[0086] Specifically, the system inputs the chain speed (normal range: 0-10 rpm), the current nozzle angle, and the location of the rusted area. A multi-layer perceptron neural network then calculates and outputs the nozzle adjustment angle. At high chain speeds (8 rpm) and when the rust is located on a complex curved surface, the model outputs angle adjustment instructions within 0.2 seconds, enabling the nozzle to respond quickly. This represents a threefold increase in response speed compared to traditional angle adjustment methods, ensuring the nozzle remains perpendicular to the rusted surface, improving both rust remover utilization and efficiency.

[0087] Preferably, in the rotatable magnetic nozzle group execution unit, the magnetic attraction structure of the nozzle adopts a new type of permanent magnetic material and an optimized magnetic circuit design. Assume that the residual magnetic induction intensity of the permanent magnetic material is B r , the coercive force is H c , the magnetic resistance of the magnetic circuit is R m By optimizing the selection of permanent magnetic materials and the design of magnetic circuit, the magnetic attraction force F between the nozzle and the chain surface is m satisfy:

[0088]

[0089] Where A is the cross-sectional area of ​​the magnetic circuit, μ0 is the vacuum magnetic permeability. In the operating environment of the wind turbine, the appropriate magnetic attraction F m It can ensure that the nozzle always closely adheres to the chain surface during the chain rotation, while the rotation flexibility of the nozzle is not affected by excessive magnetic attraction.

[0090] Specifically, a permanent magnet material with a residual magnetic induction strength of 1.2T-1.5T and a coercive force of 800kA / m-1200kA / m is selected. By optimizing the magnetic circuit design, the magnetic attraction between the nozzle and the chain surface is maintained at 50N-80N. When operating in the narrow space inside the wind turbine tower, this magnetic attraction can ensure that the nozzle fits tightly to the chain without affecting its flexible rotation. According to tests, when the chain vibration acceleration reaches 5m / s 2 Under the working conditions, the nozzle remains stable, effectively avoiding the spray deviation caused by looseness and improving the reliability of rust removal operations.

[0091] Preferably, in the multimodal data monitoring unit, when pre-processing the collected multimodal data, a method combining wavelet denoising and principal component analysis (PCA) is adopted. Suppose the original vibration signal V, temperature signal T, and humidity signal H are obtained after wavelet denoising respectively. denoised , T denoised , H denoised , and then perform principal component analysis on these denoised data to extract principal components PC1, PC2, ..., PC n, where n is the number of principal components retained. In wind turbines, the number of principal components retained, n, is selected based on the characteristics of the data and the requirements of condition monitoring, and is generally set to 2-3. This preprocessing method can effectively remove noise from the data, reduce the data dimension, and improve the efficiency and accuracy of long-short-term memory network condition monitoring.

[0092] Specifically, wavelet denoising and principal component analysis were used to preprocess the data. Three-layer wavelet decomposition and denoising were performed on vibration signals to remove high-frequency noise interference. Principal component analysis was applied to temperature and humidity data, retaining 2-3 principal components and reducing data dimensionality by 40%-60%. When processing offshore wind turbine data, this effectively filtered out noise generated by wave impact while reducing data redundancy. This reduced long-short-term memory network training time by 30%, improved prediction accuracy by 15%, and enabled rapid and accurate identification of corrosion risks within the engine compartment.

[0093] Preferably, the intelligent decision-making and control unit considers the operating status and maintenance plan of the wind turbine when formulating the rust remover injection strategy. Assume that the operating status of the wind turbine is divided into normal operation, maintenance status and fault status, respectively represented by S normal , S maintenance , S fault Indicates maintenance plan with P maintenance The intelligent decision-making and control unit adjusts the frequency, volume, and duration of rust remover injection according to different operating conditions and maintenance plans. For example, during maintenance, the volume and frequency of rust remover injection can be appropriately increased to improve rust removal effectiveness. In a fault state, the unit decides whether to suspend rust remover injection based on the fault type and severity. By comprehensively considering the operating conditions and maintenance plans, rust remover resources can be more rationally utilized, improving the overall efficiency of the system.

[0094] Specifically, during normal operation, rust removal is performed on a regular basis (once every quarter). During maintenance, the spraying frequency is increased to once a month, with a focus on areas prone to rust. In the event of a fault, if rust causes a component failure, rust removal operations in the relevant area are suspended, and the spraying strategy is adjusted after the fault is repaired. For example, if a wind farm's gearbox shuts down due to rust, the system automatically suspends rust removal in the gearbox area to prevent the agent from entering the faulty component. Once the fault is repaired, the spraying volume and frequency are increased to prevent further rust and ensure safe and stable operation of the unit.

[0095] like Figure 2 As shown, a method for intelligently and adaptively spraying a rust remover for a wind turbine generator set comprises the following steps:

[0096] S1. Multimodal data acquisition and fusion: Real-time acquisition of multimodal data such as vibration signals, temperature signals, and humidity signals from wind turbines. Utilizing a multimodal data fusion algorithm, different types of data are fused and processed to generate fused monitoring data, providing basic data support for subsequent corrosion status analysis.

[0097] S2. Long-Short-Term Memory Network Status Prediction: The fused monitoring data is input into the long-short-term memory network model. Through the model's learning and prediction functions, the corrosion status of each component of the wind turbine generator set is dynamically predicted to determine the locations where corrosion may occur and the development trend of the corrosion degree.

[0098] S3. Gaussian mixture model feature analysis: A Gaussian mixture model is used to perform cluster analysis on the fused monitoring data, mining the corrosion characteristics contained in the data and determining the characteristic distribution of different corrosion degrees, providing a feature basis for the formulation of rust remover injection strategies;

[0099] S4. Intelligent decision-making instruction generation: Based on the corrosion status prediction results and the corrosion characteristic distribution information, according to the preset decision rules and optimization algorithm, the rust remover injection control instructions for different corrosion locations and degrees are generated, including the control information of the injection amount and injection angle;

[0100] S5. Precise adjustment of spray volume and angle: Based on intelligent decision-making instructions, the spray volume precision control unit adjusts the flow control valve of the rotatable magnetic nozzle group to accurately control the rust remover spray volume of each nozzle; the angle adaptive adjustment unit uses a magnetic angle adjustment mechanism to automatically adjust the angle of the rotatable magnetic nozzle group as the chain rotates, ensuring that the nozzle is always sprayed perpendicular to the rusted area;

[0101] S6. Execution and feedback adjustment: The rotatable magnetic nozzle group execution unit sprays rust remover on the rusted parts of the wind turbine according to the precise spraying amount and angle. At the same time, the system collects the spraying effect data in real time, compares and analyzes it with the expected spraying effect, and dynamically adjusts the spraying strategy according to the feedback results to improve the rust removal effect.

[0102] This invention fundamentally addresses the inherent flaws of traditional rust removal technology through its innovative technical architecture and intelligent algorithms. Through a multimodal data-fusion long-short-term memory network condition monitoring unit, the system simultaneously collects multidimensional data such as vibration, temperature, and humidity, comprehensively capturing equipment operating status and environmental information. Compared to traditional single-data monitoring methods, this system can more accurately predict corrosion risks, identify potential corrosion sites and development trends in advance, and avoid missed corrosion problems due to incomplete monitoring.

[0103] At the operational level, a rotatable magnetic nozzle assembly, coupled with an adaptive angle adjustment unit, utilizes a magnetic structure to adhere tightly to the chain surface. The nozzle automatically adjusts its angle as the chain rotates, ensuring the nozzle remains perpendicular to the rusted area. This design overcomes the drawbacks of low manual operation efficiency, high safety risks, and the fixed spray angles of traditional automated equipment. Manual rust removal is limited by the operating environment, making it difficult to operate at high altitudes and on complex structures, and ensuring uniform spraying. Traditional equipment, on the other hand, cannot adapt to the dynamic changes of components and is prone to blind spots. This system uses intelligent adjustment to ensure that the rust remover precisely covers the rusted area, significantly improving the efficiency and quality of rust removal.

[0104] The intelligent decision-making and control unit integrates Gaussian mixture model analysis results with preset rules to formulate differentiated spray strategies based on the degree of corrosion, equipment operating status, and maintenance plan. The precise spray control unit precisely adjusts the flow rate accordingly, avoiding waste of rust remover and reducing agent consumption by approximately 30% compared to traditional extensive spraying. Simultaneously, the system's closed-loop feedback mechanism continuously collects spray effect data and dynamically optimizes strategies, enabling intelligent management of the entire process from monitoring, decision-making, to execution. This not only reduces manual intervention costs and safety hazards, but also significantly improves the economic efficiency and reliability of equipment maintenance, effectively extending the service life of wind turbines and ensuring stable power generation efficiency.

[0105] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0106] While embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent adaptive rust remover spraying system for a wind turbine generator set, characterized in that: The system includes: Multimodal data monitoring unit: This unit collects multimodal data such as vibration, temperature, and humidity signals from wind turbines, fuses different types of data using a multimodal data fusion algorithm, and then inputs the data into a long-short-term memory network model to monitor and predict the corrosion status of various components of the wind turbine in real time. Corrosion feature analysis unit: This unit uses a Gaussian mixture model to perform cluster analysis on the data, mine the corrosion features in the data, and determine the characteristic distribution of different corrosion degrees. The output end is connected to the intelligent decision-making and control unit; Intelligent decision-making and control unit: This unit combines the corrosion status prediction results of the status monitoring unit and the characteristic distribution information of the corrosion characteristic analysis unit, and generates rust remover injection control instructions for different corrosion locations and degrees based on preset decision rules and optimization algorithms; Precision spray control unit: This unit controls the amount of rust remover sprayed from each nozzle by adjusting the flow control valve of the rotatable magnetic nozzle group. The spray amount matches the degree of rust. Angle adaptive adjustment unit: This unit uses a magnetic angle adjustment mechanism according to the instructions of the intelligent decision-making and control unit to automatically adjust the angle of the rotatable magnetic nozzle group as the chain rotates, so that the nozzle always sprays perpendicular to the rusted area; Rotatable magnetic nozzle group execution unit: This unit consists of multiple rotatable magnetic nozzles, which fit tightly to the chain surface through magnetic attraction. Under the coordinated action of the spray volume precision control unit and the angle adaptive adjustment unit, the rust remover is sprayed on the rusted parts of the wind turbine according to the preset spray volume and angle.

2. The intelligent adaptive rust remover spraying system for a wind turbine generator set according to claim 1, characterized in that: The multimodal data monitoring unit adopts an improved weighted fusion algorithm in the multimodal data fusion process, assuming that the vibration signal is V, the temperature signal is T, and the humidity signal is H. The fused data D f Calculated by the following formula: D f =w V ·LSTM V (V)+w T ·LSTM T (T)+w H ·LSTM H (H) Among them, w V , w T , w H are the fusion weights of vibration signal, temperature signal and humidity signal respectively, and w V +w T +w H =1;LSTM V , LSTM T , LSTM H They are the long short-term memory network processing functions for vibration signal, temperature signal and humidity signal respectively. In the wind turbine generator set, the weight w of the vibration signal V Dynamically adjust according to the operating conditions and vibration characteristics of the unit, with a value range of 0.3-0.5; the temperature signal weight w T Considering the influence of different ambient temperatures on rust, the value range is 0.2-0.4; the humidity signal weight w H Combined with local climatic conditions, the value range is 0.2-0.

3.

3. The intelligent adaptive rust remover spraying system for a wind turbine generator set according to claim 1, characterized in that: The corrosion feature analysis unit analyzes the fused data D f When modeling the Gaussian mixture model, the expectation maximization algorithm is used for parameter estimation. Assume that the Gaussian mixture model consists of K Gaussian distributions, and the probability density function of the kth Gaussian distribution is: Among them, θ k =(μ k ,Σ k ) is the parameter of the kth Gaussian distribution, μ k is the mean vector, Σ k is the covariance matrix, d is the data D f The dimension of the wind turbine is the mean vector μ k It reflects the central trend of multimodal fusion data under different corrosion degrees, and the covariance matrix Σ k It reflects the degree of discreteness of the data distribution, and updates the parameter θ through continuous iteration k , the Gaussian mixture model can fit the distribution of the data.

4. The intelligent adaptive rust remover spraying system for a wind turbine generator set according to claim 1, characterized in that: The intelligent decision-making and control unit introduces a decision rule based on fuzzy logic when generating the rust remover injection control instruction. Assuming that the rust degree is S, it is determined by the state monitoring unit and the rust characteristic analysis unit; the injection amount adjustment coefficient is C q , the angle adjustment coefficient is C a The fuzzy rule base defines the corresponding fuzzy rules of injection amount and angle adjustment coefficient according to different corrosion degree intervals. When the corrosion degree S is in the mild corrosion interval, the injection amount adjustment coefficient C q The value is small, the angle adjustment coefficient C a Maintain the nozzle basically vertically; when the rust level S is in the severe rust range, the injection volume adjustment coefficient C q The value is larger, the angle adjustment coefficient C a Adjustments are made according to the chain rotation conditions, and different injection control instructions are generated according to different corrosion conditions through fuzzy logic reasoning.

5. The intelligent adaptive rust remover spraying system for a wind turbine generator set according to claim 1, characterized in that: The precise injection control unit adopts a model-based predictive control method when adjusting the flow control valve. Assuming the current time is t, the prediction time domain is N, the control time domain is M, M≤N, the control input of the flow control valve is u(t), and the output of the system is the rust remover injection amount y(t), the prediction model of the system is established: y(t+j∣t)=f(y(t),y(t-1),…,y(t-n y ),u(t),u(t -1),…,u(t-n u )) Where j = 1, 2, ..., N, n y and n u are the output and input orders respectively. In the wind turbine generator set, according to the injection amount target value y issued by the intelligent decision-making control unit ref (t), by optimizing the objective function: Among them, λ j The weighted coefficient of the control input is used to obtain the optimal control input u(t) to control the injection amount of the rust remover.

6. The intelligent adaptive rust remover spraying system for a wind turbine generator set according to claim 1, characterized in that: The angle adaptive adjustment unit adopts a neural network-based angle prediction model when adjusting the angle of the rotatable magnetic nozzle group. Assuming the chain rotation speed is v, the current nozzle angle is α, and the position information of the rusted part is p, these input information are input into a multi-layer perceptron neural network. The output of the network is the predicted nozzle adjustment angle Δα. The training data of the neural network comes from historical experimental data. The weights and bias of the network are continuously adjusted through the back propagation algorithm, so that the predicted adjustment angle adapts to the rotation of the chain and the changes in the rusted part, and the nozzle always sprays perpendicular to the rusted part.

7. The intelligent adaptive rust remover spraying system for a wind turbine generator set according to claim 1, characterized in that: The magnetic attraction structure of the rotatable magnetic nozzle group execution unit adopts a new type of permanent magnetic material and an optimized magnetic circuit design. Assume that the residual magnetic induction intensity of the permanent magnetic material is B r , the coercive force is H c , the magnetic resistance of the magnetic circuit is R m By optimizing the selection of permanent magnetic materials and the design of magnetic circuit, the magnetic attraction force F between the nozzle and the chain surface is m satisfy: Among them, A is the cross-sectional area of ​​the magnetic circuit, μ0 is the vacuum magnetic permeability, and in the operating environment of the wind turbine, the magnetic attraction force F m The nozzle is kept in close contact with the surface of the chain during the chain rotation, while the rotation flexibility of the nozzle is not affected by excessive magnetic attraction.

8. The intelligent adaptive rust remover spraying system for a wind turbine generator set according to claim 1, characterized in that: The multimodal data monitoring unit pre-processes the collected multimodal data by combining wavelet denoising and principal component analysis. The original vibration signal V, temperature signal T, and humidity signal H are obtained after wavelet denoising respectively. denoised , T denoised , H denoised Then, principal component analysis is performed on the noised data to extract the principal components PC1, PC2, ..., PC n , where n is the number of principal components retained. In wind turbines, the number of principal components retained, n, is selected based on the characteristics of the data and the requirements of condition monitoring, and is set to 2-3.

9. The intelligent adaptive rust remover spraying system for a wind turbine generator set according to claim 1, characterized in that: The intelligent decision-making and control unit takes into account the operating status and maintenance plan of the wind turbine when formulating the rust remover injection strategy. The operating status of the wind turbine is divided into normal operation, maintenance status and fault status, respectively. normal , S maintenance , S fault Indicates maintenance plan with P maintenance Indicates, including maintenance time and maintenance content information; The intelligent decision-making and control unit adjusts the frequency, amount and time of rust remover injection according to different operating conditions and maintenance plans. In the maintenance state, the amount and frequency of rust remover injection are increased. In the event of a fault, decide whether to suspend the rust remover spraying operation based on the type and severity of the fault.

10. An intelligent adaptive rust remover spraying method for a wind turbine generator set, characterized in that: The method includes: S1. Multimodal data acquisition and fusion: Real-time acquisition of multimodal data including vibration signals, temperature signals, and humidity signals of wind turbines. Utilizing a multimodal data fusion algorithm, different types of data are fused to generate fused monitoring data. S2. Long-Short-Term Memory Network Status Prediction: The fused monitoring data is input into the long-short-term memory network model. Through the model's learning and prediction functions, the corrosion status of each component of the wind turbine generator set is dynamically predicted to determine the locations where corrosion may occur and the development trend of the corrosion degree. S3. Gaussian mixture model feature analysis: A Gaussian mixture model is used to perform cluster analysis on the fused monitoring data, mining the corrosion features included in the data and determining the characteristic distribution of different corrosion degrees; S4. Intelligent decision-making instruction generation: Based on the corrosion status prediction results and the corrosion characteristic distribution information, according to the preset decision rules and optimization algorithm, the rust remover injection control instructions for different corrosion locations and degrees are generated, including the control information of the injection amount and injection angle; S5. Precise adjustment of spray volume and angle: Based on intelligent decision-making instructions, the spray volume precision control unit adjusts the flow control valve of the rotatable magnetic nozzle group to control the rust remover spray volume of each nozzle; the angle adaptive adjustment unit uses a magnetic angle adjustment mechanism to automatically adjust the angle of the rotatable magnetic nozzle group as the chain rotates, ensuring that the nozzle is always sprayed perpendicular to the rusted area; S6. Execution and feedback adjustment: The rotatable magnetic nozzle group execution unit sprays rust remover on the corroded parts of the wind turbine generator set according to the spray volume and angle. At the same time, the system collects the spray effect data in real time, compares and analyzes it with the expected spray effect, and dynamically adjusts the spray strategy according to the feedback results.

Images