Modular monitoring system and monitoring method for a wind turbine

Abstract

This application relates to a modular monitoring system and method for wind turbine generator sets. The modular monitoring system includes a blade acceleration sensor, a stress sensor, a main shaft acceleration / moment of inertia sensor, a temperature sensor, a lightning monitoring sensor, and an icing monitoring sensor. The blade acceleration sensor is located on the inner side of the blade and is used to collect various data of the wind turbine rotor and blades. The stress sensor is located at the root of the blade and is used to monitor the strain value at the blade root. The main shaft acceleration / moment of inertia sensor is located in the main shaft bearing housing of the wind turbine generator set and is used to monitor the moment of inertia of the main shaft. The temperature sensor is located at a preset heating position of the wind turbine generator set and is used to monitor the temperature at the corresponding position. The lightning monitoring sensor is located on the lightning current discharge channel and is used to capture and record the passing lightning current. The icing monitoring sensor is located on the outer wall of the nacelle of the wind turbine generator set and is used to monitor whether the outer wall of the nacelle is icy.

Description

Technical Field

[0001] This application relates to the field of wind turbine technology, specifically to a modular monitoring system and method for wind turbine generator sets. Background Technology

[0002] As a major clean energy source, wind power is a crucial energy pillar for many countries and regions, and its development demand is increasing daily. To ensure the long-term stable operation of wind turbines, continuous and effective control and monitoring are essential. The rotor blades of wind turbines are critical components, and their long-term exposure to complex and changing environmental and load conditions makes them susceptible to problems such as icing, lightning strikes, and structural fatigue damage. These problems can directly lead to decreased power generation efficiency, unplanned outages, and even catastrophic failures. Currently, blade condition monitoring often employs single-function, independent systems (such as separate de-icing control systems or load monitoring systems), which suffer from drawbacks such as low system integration, high cost, fragmented data making comprehensive utilization difficult, and complex maintenance. Summary of the Invention

[0003] The first aspect of this application provides a modular monitoring system for wind turbine generator sets. This modular monitoring system can effectively reduce unplanned downtime caused by factors such as icing and malfunctions by monitoring the wind turbine generators in multiple dimensions, thereby increasing power generation.

[0004] The modular monitoring system for a wind turbine generator provided in the first aspect of this application includes at least one of a blade acceleration sensor, a stress sensor, a main shaft acceleration / moment of inertia sensor, a temperature sensor, a lightning monitoring sensor, and an icing monitoring sensor. The blade acceleration sensor is disposed on the inner side of the blade and close to the root of the blade, and is used to collect various data of the wind turbine rotor and the blade. The stress sensor is disposed at the root of the blade, and multiple stress sensors are disposed on one blade, and the stress sensor is used to monitor the strain value at the root of the blade. The main shaft acceleration / moment of inertia sensor is disposed on the main shaft bearing housing of the wind turbine generator and is used to monitor the moment of inertia of the main shaft. The temperature sensor is disposed at a preset heating position of the wind turbine generator and is used to monitor the temperature at the corresponding position. The lightning monitoring sensor is disposed on the lightning current discharge channel and is used to capture and record the passing lightning current. The icing monitoring sensor is disposed on the outer wall of the nacelle of the wind turbine generator and is used to monitor whether there is icing on the outer wall of the nacelle.

[0005] In one alternative embodiment, the modular monitoring system further includes a wireless communication module and a programmable safety module. The wireless communication module is used for data transmission, and the programmable safety module is used for receiving detection data and performing analysis and control functions, including: analyzing ice layer detection based on vibration data and generating de-icing control or shutdown / restart commands; implementing load optimization control based on load data and assessing structural damage risk; recording events based on lightning strike data and correlating them with turbine operating parameters for impact assessment; and conducting a comprehensive assessment and early warning of the structural health status of the blades based on continuous monitoring data.

[0006] The second aspect of this application provides a monitoring method for wind turbine generator sets, which can better achieve multi-dimensional real-time monitoring of wind turbine generator sets, thereby improving the operating efficiency, safety, and maintainability of wind turbine generator sets.

[0007] The monitoring method for wind turbine generators provided in the second aspect of this application includes the following steps: S1. Provide preset types of sensors according to monitoring needs to simultaneously collect vibration signals, strain signals and lightning strike event signals of rotor blades; S2. Perform localized preprocessing and data fusion on the collected multi-dimensional signals to generate a standardized monitoring data stream; S3. Perform integrated analysis on the monitoring data stream, and output the analysis results of each type of task through multi-algorithm collaboration and multi-task parallelism. The analysis results include icing state judgment, load optimization calculation, lightning strike event parameters, and structural health assessment. S4. Based on the integrated analysis results, generate control recommendations, maintenance warnings and status reports for wind turbine operation, and output them through an open interface.

[0008] In one optional scheme, in step S3, Kalman filtering is used to process the observation data during the analysis of the icing state judgment and the load optimization calculation; in the icing state judgment, Kalman filtering is used to filter noise and extract pure signal attenuation features to provide reliable input for subsequent neural network classification; in the load optimization calculation, Kalman filtering is used to predict the load change trend at the next moment to provide advance for dynamic optimization.

[0009] In one alternative approach, Kalman filtering evaluates the state of a linear system through a two-step prediction-update iteration, the mathematical expression of which is: X(k|k-1)=AX(k-1|k-1)+BU(k); In the formula, X(k|k-1) is the result predicted using the previous state, X(k-1|k-1) is the optimal result of the previous state, and U(k) is the control variable of the current state. When there is no control variable, U(k) = 0. P(k|k-1)=AP(k-1|k-1) A'+Q ; In the formula, P(k|k-1) is the covariance of X(k|k-1), P(k-1|k-1) is the covariance of X(k-1|k-1), A' represents the transpose of A, and Q is the covariance of the system process.

[0010] In one alternative approach, in step S3, a neural network is used for classification or regression prediction during the analysis of the icing state judgment and the structural health assessment; in the icing state judgment, the neural network outputs the icing state result based on the ultrasonic signal characteristics and vibration acceleration characteristics after Kalman filtering and noise reduction, through a neural network classification model; in the structural health assessment, the neural network identifies the location and extent of fatigue damage based on the temporal characteristics of blade vibration and stress.

[0011] In one optional embodiment, the neural network structure is an input layer-hidden layer-output layer, wherein the input layer comprises 3 neurons, the hidden layer consists of 2 layers with 12 neurons each, and the output layer comprises 4 neurons; the forward propagation formula of the neural network includes: Hidden layer output: h l =σ(W l h l−1 +b l ); Among them, h l For the output of layer l, W l Let b be the weight matrix of the l-th layer. l Here, σ is the bias vector, and σ is the activation function (ReLU: σ(x) = max(0,x)). Output layer output: ; Where y represents the probability values ​​of the four states, and W out b represents the output layer weights. out For output layer bias; The backpropagation formula includes: Loss function: ; Among them, t i For real labels, y i To predict probabilities; Weight update: , ; Where η is the learning rate. Let W be the gradient of the loss function with respect to the weights Wl.

[0012] In one optional approach, the training of the neural network is also included. The training process includes dataset preparation, data preprocessing, iterative training, and real-time inference. The dataset preparation includes collecting no less than 1000 samples under different icing conditions in the wind field, of which 70% are used for training and 30% for testing. The data preprocessing is used to standardize the data to avoid the influence of different units on the training. The iterative training is performed at least 500 times, updating the weights / biases through backpropagation until the classification accuracy of the test set is ≥95%. The real-time inference includes: during ensemble analysis, inputting the standardized data stream into the trained BP network and outputting the icing condition classification result.

[0013] In one optional embodiment, the control recommendations, maintenance alerts, and status reports generated in step S4 include: When the analysis result is no ice / light ice, the control recommendation is to maintain the current operating parameters, with no maintenance warnings, and the core content of the status report is that the blades have no risk of icing and are operating normally; When the analysis result is medium / heavy icing, the control recommendation is to start the de-icing system and adjust the pitch angle. The maintenance warning is that the ice layer affects the power generation efficiency and a re-inspection is required after de-icing. The core contents of the status report include the icing level, the de-icing execution status, and the assessment of power generation loss. When the analysis results indicate that the load is abnormally high, the control recommendations are to reduce the torque / adjust the speed and optimize the load distribution. The maintenance warning indicates an increased risk of structural fatigue and recommends stopping the machine for inspection. The core contents of the status report include the abnormal load value, duration, and structural stress distribution. The analysis results indicate that when a lightning strike occurs, the control recommendations are to briefly pause for self-testing, and restart if no damage is found. The maintenance warning includes recording lightning strike parameters, checking blades / electrical components, and the status report should include the core content of the lightning strike time / amplitude / location and the results of component damage investigation.

[0014] In an optional scheme, the control recommendations, maintenance warnings and status reports generated in step S4 also include: when the analysis result is that the structural health score is below the threshold, the control recommendation is to shut down for maintenance, the maintenance warning is that the blade has fatigue damage / cracks and to perform immediate maintenance, and the core content of the status report includes the location / degree of damage and the prediction of remaining service life. When the analysis results show that the structural health score is within the preset range, the control recommendation is to maintain operation and optimize the load. The maintenance warning is to regularly monitor the damage trend and plan preventive maintenance. The core content of the status report includes the health score and the damage development trend. When the analysis results indicate an abnormal temperature rise, the control recommendations are to reduce the load and enhance heat dissipation. The maintenance warnings are to identify overheating of critical components and troubleshoot the faults. The core content of the status report includes abnormal temperature points, heat-generating components, and correlation analysis of operating parameters.

[0015] The beneficial effects of this application are as follows: The modular monitoring system and method for wind turbine generators in this application minimize unplanned downtime caused by factors such as icing and malfunctions through real-time, multi-dimensional monitoring and rapid automatic response, thereby increasing power generation. Real-time monitoring enables early identification of structural damage and accurate recording of extreme events (such as lightning strikes), providing data support for preventing major failures and ensuring equipment and personnel safety. Furthermore, status-based preventative maintenance prompts replace traditional periodic or post-failure maintenance, reducing maintenance costs and downtime. The modular design allows users to flexibly select and combine monitoring functions according to specific needs (such as environmental threats in a particular wind farm), avoiding unnecessary initial investment and reducing system integration complexity and total cost through open interfaces. In addition, the system supports remote configuration and data access, facilitating centralized monitoring and operation and maintenance management.

[0016] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0017] Figure 1 This is a schematic diagram showing the installation locations of each sensor in a modular monitoring system. Figure 2 A flowchart illustrating the data acquisition and processing procedures for monitoring wind turbine generator sets; Figure 3 This is a schematic diagram of the wireless communication module in a modular monitoring system. Figure 4 This is a schematic diagram of the composition of a programmable security module in a modular monitoring system.

[0018] Figure reference numerals: 1. Accelerometer sensor; 2. Stress sensor; 3. Spindle acceleration / rotational inertia sensor; 4. Temperature sensor; 5. Lightning monitoring sensor; 6. Icing monitoring sensor.

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. Detailed Implementation

[0020] To better understand the technical solution of this application, the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0021] It should be understood that the described embodiments are merely some embodiments of this application, and not all embodiments. All other technical solutions obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0023] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0024] It should be noted that the directional terms such as "upper," "lower," "left," and "right" described in the embodiments of this application are used to describe the angles shown in the accompanying drawings and should not be construed as limiting the embodiments of this application. Furthermore, in the context, it should be understood that when it is mentioned that an element is connected "upper" or "lower" to another element, it can be directly connected to the other element "upper" or "lower," or indirectly connected to the other element "upper" or "lower" through an intermediate element.

[0025] like Figure 1-4 As shown, the first aspect of this application provides a modular monitoring system for a wind turbine generator set. This modular monitoring system mainly includes at least one of the following: a blade acceleration sensor 1, a stress sensor 2, a main shaft acceleration / moment of inertia sensor 3, a temperature sensor 4, a lightning monitoring sensor 5, and an icing monitoring sensor 6. The blade acceleration sensor 1 is disposed on the inner side of the blade and near the blade root, and is used to collect various data of the wind turbine rotor and blades. The stress sensor 2 is disposed at the blade root, and multiple stress sensors 2 are disposed on one blade, used to monitor the strain value at the blade root. The main shaft acceleration / moment of inertia sensor 3 is disposed on the main shaft bearing housing of the wind turbine generator set and is used to monitor the rotational inertia of the main shaft. The temperature sensor 4 is disposed at a preset heating position of the wind turbine generator set and is used to monitor the temperature at the corresponding position. The lightning monitoring sensor 5 is disposed on the lightning current discharge channel and is used to capture and record the passing lightning current. The icing monitoring sensor 6 is disposed on the outer wall of the nacelle of the wind turbine generator set and is used to monitor whether icing exists on the outer wall of the nacelle.

[0026] Specifically, blade acceleration sensors 1 are installed on the inner side of the blade (near the blade root), in an area approximately 1 / 3 to 1 / 2 of the blade length from the blade root. This area has a larger blade chord length, more pronounced vibration modes, and is easy to install and maintain. Stress sensors 2 are installed at the root of each blade, with 3 to 5 sensors installed, typically 4. Main shaft acceleration / moment of inertia sensors 3 are installed on the main shaft bearing housing (especially the bearing housing at the non-drive end) to measure the radial vibration (reflecting alignment and balance) and axial vibration of the main shaft. Temperature sensors 4 are installed in key heat-generating parts of the entire unit, such as the generator, gearbox, pitch system, yaw motor, yaw reducer, main bearing, power converter / inverter, and control cabinet. Lightning monitoring sensors 5 are installed on the lightning current discharge channel, with a Rogowski coil or current transformer installed on the lightning arrester lead at the root of each blade to capture and record the passing lightning current (amplitude, waveform, and charge). The icing monitoring sensor 6 can be an ultrasonic anemometer. By installing the ultrasonic anemometer on the top or side of the wind turbine nacelle, it is essential to ensure the probe faces the direction of the airflow and avoid obstructing the airflow due to the nacelle structure. Utilizing the signal characteristics of the ultrasonic anemometer (icing causes ultrasonic signal attenuation, changes in propagation speed, or complete malfunction), the presence of icing is determined by identifying this signal anomaly. The main functions of each sensor are as follows: like Figure 2 As shown, various data from the wind turbine rotor and blades are collected via blade acceleration sensor 1. The rotor data includes acceleration, angular velocity, azimuth angle, torque, rotational speed per minute, thrust, yaw, wind direction, and wind speed. The blade data includes acceleration, angular velocity, pitch angle, lightning strikes, temperature, and icing status. The initial collected data is analyzed using various algorithms in the data processing module to monitor for anomalies. The analyzed data is then transmitted to a programmable security module or uploaded to the cloud platform of the remote monitoring center via a wireless communication module.

[0027] Stress sensor 2: The core sensor detects the strain value (unit: Pa, including circumferential and axial strain) at the root of the blade, as well as the bending strain and torsional strain of the tower / main shaft; the main data acquisition object is the blade, while also covering the rotor; it is used to monitor the structural stress changes of the blade caused by wind load and icing, to determine the risk of fatigue damage (such as crack initiation), and can also assist in monitoring the strain distribution during the torque transmission process of the main shaft and assess the load status of the transmission system.

[0028] Spindle acceleration / rotational inertia sensor 3: Core detection of the spindle's radial vibration acceleration (unit: m / s²), axial vibration acceleration, and rotational inertia (reflecting rotational speed stability); the object of acquisition is the rotor; radial vibration data can detect spindle alignment deviation and bearing wear (frequency range 0.1~10000Hz), and rotational inertia can indirectly deduce spindle speed and torque fluctuations, and assist in load optimization calculations.

[0029] Temperature Sensor 4: Core detection of temperature values ​​(unit: °C) of key components, including generator winding temperature, gearbox oil temperature, pitch system temperature, and blade root temperature; the data collection covers the rotor and blades; used to monitor rotor-side gearbox oil temperature (normal range -50~150℃, refer to Newenergy oil temperature sensor parameters) and generator temperature to prevent overheating damage, and also to assist in judging blade icing (combined with icing sensor data when the ambient temperature is below 0℃) and abnormal structural heating (such as local high temperature after lightning strike).

[0030] Lightning monitoring sensor 5: The core of the sensor is to detect the amplitude (unit: A), waveform, and charge of the lightning current, as well as the instantaneous voltage / current at the lightning strike point on the blade; the sensor only collects data on the blade; the sensor captures the lightning strike signal through the Rogowski coil / current transformer on the down conductor of the lightning arrester at the blade root, records the event parameters, and correlates them with the blade stress mutation data to assess the damage impact of the lightning strike on the blade structure.

[0031] Icing monitoring sensor 6: The core of the sensor is to detect the signal strength attenuation of the ultrasonic anemometer, the change in signal propagation time, and the ambient humidity (as an auxiliary factor in the judgment); the sensor collects data on the blades (indirectly); it is installed on the outer shell of the nacelle and indirectly judges the icing status of the blades by the degree of ultrasonic signal attenuation (e.g., the signal strength drops from 42dB to below 20dB). The signal is stable when there is no ice, and the signal attenuation increases when there is light or heavy ice. The icing level needs to be confirmed in conjunction with temperature sensor 4 (ambient temperature < 0℃).

[0032] like Figure 3-4 As shown, in one specific embodiment, the modular monitoring system further includes a wireless communication module and a programmable safety module. The wireless communication module is used for data transmission, and the programmable safety module is used for receiving detection data and performing analysis and control functions, including: analyzing ice layer detection based on vibration data and generating de-icing control or shutdown / restart commands; implementing load optimization control based on load data and assessing structural damage risk; recording events based on lightning strike data and correlating them with turbine operating parameters for impact assessment; and conducting a comprehensive assessment and early warning of the structural health status of the blades based on continuous monitoring data.

[0033] Specifically, such as Figure 3 The wireless communication module shown is used to transmit the data collected by the sensor to the analysis and control unit. The analysis results, early warning information and control commands are transmitted to the main control system, remote monitoring center or maintenance management system of the wind turbine. It also supports remote configuration and parameter adjustment of the system.

[0034] Figure 4The programmable safety module shown can perform the following analysis and control functions based on the received detection data: analyze ice layer detection based on vibration data and generate de-icing control or shutdown / restart commands; implement load optimization control based on load data and assess structural damage risk; record events based on lightning strike data and correlate them with turbine operating parameters for impact assessment; and conduct comprehensive assessment and early warning of the structural health status of the blades based on continuous monitoring data. A second aspect of this application provides a monitoring method for a wind turbine generator set, the monitoring method comprising the following steps: S1. Provide preset types of sensors according to monitoring needs to simultaneously collect vibration signals, strain signals and lightning strike event signals of rotor blades.

[0035] S2. Localized preprocessing and data fusion are performed on the collected multi-dimensional signals. After standardizing the original multi-dimensional signals with unified format, unified precision, and unified semantics, a standardized monitoring data stream is generated for subsequent integrated analysis, providing a unified data foundation for performing analysis tasks such as ice layer judgment and load calculation.

[0036] S3. Integrate and analyze the monitoring data stream, and output the analysis results of each type of task through multi-algorithm collaboration and multi-task parallelism. The analysis results include icing status judgment, load optimization calculation, lightning strike event parameters, and structural health assessment. Its core logic is: take standardized data as input, match exclusive algorithm models for different analysis tasks, promote each task simultaneously through "parallel computing architecture", and finally output the analysis results of each type of task.

[0037] In this step, Kalman filtering is used to process the observation data during the analysis of icing state judgment and load optimization calculation. In the icing state judgment: the ultrasonic anemometer signal is easily affected by airflow turbulence and electromagnetic interference. Kalman filtering is used to filter noise and extract pure signal attenuation features to provide reliable input for subsequent neural network classification. In the load optimization calculation: based on historical load data (stress, torque), Kalman filtering is used to predict the load change trend at the next moment, providing a lead time for dynamic optimization.

[0038] It's important to explain that Kalman filtering is a recursive estimation algorithm that achieves optimal estimation of the state of a linear system through a two-step "prediction-update" iteration. Assuming the system model follows a linear Gaussian distribution, the core formula is as follows: X(k|k-1)=AX(k-1|k-1)+BU(k); In the formula, X(k|k-1) is the result predicted using the previous state, X(k-1|k-1) is the optimal result of the previous state, and U(k) is the control variable of the current state. When there is no control variable, U(k) = 0. P(k|k-1)=AP(k-1|k-1)A'+Q; In the formula, P(k|k-1) is the covariance of X(k|k-1), P(k-1|k-1) is the covariance of X(k-1|k-1), A' represents the transpose of A, and Q is the covariance of the system process.

[0039] Wind power scenario adaptation example (signal denoising in ice layer condition judgment): System state xk: actual signal strength of ultrasonic anemometer; Observation value zk: ultrasonic signal strength (including noise) in standardized data stream; Process noise Q, observation noise R: calibrated using measured wind field data (e.g., Q=10−4, R=0.01, adjusted according to actual environment); Iterative process: For each set of standardized data (10Hz sampling frequency) received, perform one "prediction-update" cycle and output the denoised signal strength. .

[0040] Furthermore, in this step, neural networks are used for classification or regression prediction during the analysis of icing state assessment and structural health evaluation. The core function of neural networks is to learn nonlinear data characteristics to achieve classification or regression prediction. In the integrated analysis, they are mainly used for two scenarios: ice state assessment and structural health evaluation. In ice state assessment: based on the ultrasonic signal characteristics and vibration acceleration characteristics after Kalman filtering and noise reduction, a neural network classification model is used to output the state results of "no ice / light ice / medium ice / heavy ice". In structural health evaluation: based on the temporal characteristics of blade vibration and stress, a neural network is used to identify the location and degree of fatigue damage (such as crack length and structural aging level).

[0041] Algorithm selection and network structure (taking ice layer state classification as an example): BP neural network (backpropagation neural network) is selected (suitable for small sample sizes and classification tasks). The network structure is "input layer - hidden layer - output layer": Input layer: 3 neurons (ultrasonic signal attenuation, leaf X-axis vibration acceleration, ambient temperature); Hidden layer: 2 layers, 12 neurons per layer (activation function: ReLU, to avoid gradient vanishing); Output layer: 4 neurons (corresponding to no ice / light ice / medium ice / heavy ice, activation function: Softmax, output probability distribution).

[0042] The forward propagation formula for a neural network includes: Hidden layer output: hl = σ(Wlhl−1 + bl); Where hl is the output of the l-th layer, Wl is the weight matrix of the l-th layer, bl is the bias vector, and σ is the activation function (ReLU: σ(x)=max(0,x)). Output layer output: ; Where y is the probability value of the four states, Wout is the output layer weight, and bout is the output layer bias; The backpropagation formula includes: Loss function: ; Where ti is the true label and yi is the predicted probability; Weight update: , ; Where η is the learning rate. Let W be the gradient of the loss function with respect to the weights Wl.

[0043] Training Dataset Preparation: 1000 sets of samples (including ultrasonic signals, vibration, and temperature data) were collected under different icing conditions in the wind field. 70% were used for training and 30% for testing. Data Preprocessing: Standardization (input data was mapped to the [0,1] interval) was implemented to avoid the influence of different units on training. Iterative Training: 500 iterations were performed, and weights / biases were updated through backpropagation until the classification accuracy of the test set was ≥95%. Real-time Inference: During ensemble analysis, the standardized data stream was input into the trained BP network, and the icing condition classification results were output.

[0044] Applications of neural networks in structural health assessment: A CNN-LSTM hybrid neural network (capturing spatial features + temporal features) was selected: Input: spectral graph of blade vibration signal (generated by Fourier transform, containing spatial features) and stress temporal data (containing temporal features); Core logic: CNN extracts damage features (such as vibration frequency shift) from the spectral graph, LSTM learns the trend change of stress temporal sequence, and outputs a structural health score (0-100 points) after fusion.

[0045] The overall collaborative logic of integrated analysis (including algorithm interaction): 1. Data Input: Standardized data streams are synchronously distributed to four parallel task units according to "time slices" (1 slice per 100ms); 2. Algorithm Collaboration: Ice Layer Judgment Unit: First, Kalman filter noise reduction → then input BP neural network for classification → output icing status; Load Optimization Unit: First, Kalman filter to predict load trends → then input model prediction and control algorithm → output optimization scheme; Structural Health Unit: First, extract vibration / stress features → then input CNN-LSTM network → output health score; 3. Results Integration: The analysis results from the four units are summarized in the "Results Integration Module", contradictory data is removed (such as when the lightning strike event is not related to the structural damage, it is recorded separately), and a unified analysis report is generated.

[0046] S4. Based on the integrated analysis results, generate control recommendations, maintenance warnings, and status reports for wind turbine operation, and output them through an open interface. The generated control recommendations, maintenance warnings, and status reports include: When the analysis result is no ice / light icing, the control recommendation is to maintain the current operating parameters, with no maintenance warning. The core content of the status report is that the blades have no risk of icing and are operating normally. When the analysis result is moderate / heavy icing, the control recommendation is to start the de-icing system and adjust the pitch angle. The maintenance warning is that the ice layer affects power generation efficiency and a re-inspection is required after de-icing. The core content of the status report includes the icing level, de-icing execution status, and power generation loss assessment. When the analysis result is abnormally high load, the control recommendation is to reduce torque / adjust speed and optimize load distribution. The maintenance warning is that the risk of structural fatigue is increased and a shutdown inspection is recommended. The core content of the status report includes the abnormal load value, duration, and structural stress distribution. When the analysis result is a lightning strike, the control recommendation is to briefly suspend the system for self-inspection and restart if no damage is found. The maintenance warning is to record lightning strike parameters and check the blades / electrical components. The core content of the status report includes the lightning strike time / amplitude / location and the results of component damage investigation.

[0047] When the analysis result shows that the structural health score is below the threshold, the control recommendation is to shut down for maintenance. The maintenance warning is that the blades have fatigue damage / cracks and require immediate maintenance. The core content of the status report includes the location / degree of damage and the prediction of remaining service life. When the analysis result shows that the structural health score is within the preset range, the control recommendation is to maintain operation and optimize the load. The maintenance warning is to regularly monitor the damage trend and plan preventive maintenance. The core content of the status report includes the health score and the damage development trend. When the analysis result shows an abnormal temperature increase, the control recommendation is to reduce the load and enhance heat dissipation. The maintenance warning is that key components are overheating and faults should be investigated. The core content of the status report includes the abnormal temperature point, the heat-generating component, and the correlation analysis of operating parameters.

[0048] As can be seen from the above embodiments, the modular monitoring system and method of this wind turbine generator unit, through real-time, multi-dimensional monitoring and rapid automatic response, minimizes unplanned downtime caused by factors such as icing and malfunctions, thereby increasing power generation. Real-time monitoring enables early identification of structural damage and accurate recording of extreme events (such as lightning strikes), providing data support for preventing major failures and ensuring equipment and personnel safety. Furthermore, status-based preventative maintenance prompts replace traditional periodic or post-failure maintenance, reducing maintenance costs and downtime. The modular design allows users to flexibly select and combine monitoring functions according to specific needs (such as environmental threats in a particular wind farm), avoiding unnecessary initial investment, and reducing system integration complexity and total cost through open interfaces. In addition, the system supports remote configuration and data access, facilitating centralized monitoring and operation and maintenance management.

[0049] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A modular monitoring system for wind turbine generator sets, characterized in that, The system includes at least one of the following: a blade acceleration sensor, a stress sensor, a main shaft acceleration / moment of inertia sensor, a temperature sensor, a lightning monitoring sensor, and an icing monitoring sensor. The blade acceleration sensor is disposed on the inner side of the blade and near the root of the blade, and is used to collect various data of the wind turbine rotor and the blade. The stress sensor is disposed at the root of the blade, and multiple stress sensors are disposed on one blade, and the stress sensor is used to monitor the strain value at the root of the blade. The main shaft acceleration / moment of inertia sensor is disposed on the main shaft bearing housing of the wind turbine generator set and is used to monitor the moment of inertia of the main shaft. The temperature sensor is disposed at a preset heating position of the wind turbine generator set and is used to monitor the temperature at the corresponding position. The lightning monitoring sensor is disposed on the lightning current discharge channel and is used to capture and record the passing lightning current. The icing monitoring sensor is disposed on the outer wall of the nacelle of the wind turbine generator set and is used to monitor whether there is icing on the outer wall of the nacelle.

2. The modular monitoring system for wind turbine generator sets according to claim 1, characterized in that, It also includes a wireless communication module and a programmable safety module. The wireless communication module is used for data transmission, and the programmable safety module is used for receiving detection data and performing analysis and control functions, including: analyzing ice layer detection based on vibration data and generating de-icing control or shutdown / restart commands; implementing load optimization control and assessing structural damage risk based on load data; recording events based on lightning strike data and correlating them with turbine operating parameters for impact assessment; and conducting comprehensive assessment and early warning of the structural health status of the blades based on continuous monitoring data.

3. A monitoring method for wind turbine generator sets, characterized in that, Includes the following steps: S1. Provide preset types of sensors according to monitoring needs to simultaneously collect vibration signals, strain signals and lightning strike event signals of rotor blades; S2. Perform localized preprocessing and data fusion on the collected multi-dimensional signals to generate a standardized monitoring data stream; S3. Perform integrated analysis on the monitoring data stream, and output the analysis results of each type of task through multi-algorithm collaboration and multi-task parallelism. The analysis results include icing state judgment, load optimization calculation, lightning strike event parameters, and structural health assessment. S4. Based on the integrated analysis results, generate control recommendations, maintenance warnings and status reports for wind turbine operation, and output them through an open interface.

4. The monitoring method for wind turbine generator sets according to claim 3, characterized in that, In step S3, Kalman filtering is used to process the observation data during the analysis of the icing state judgment and the load optimization calculation. In the icing state judgment, Kalman filtering is used to filter noise and extract pure signal attenuation features to provide reliable input for subsequent neural network classification. In the load optimization calculation, Kalman filtering is used to predict the load change trend at the next moment to provide advance for dynamic optimization.

5. The monitoring method for wind turbine generator sets according to claim 4, characterized in that, Kalman filtering evaluates the state of a linear system through a two-step prediction-update process. Its mathematical expression is: X(k|k-1)=AX(k-1|k-1)+BU(k); In the formula, X(k|k-1) is the result predicted using the previous state, X(k-1|k-1) is the optimal result of the previous state, and U(k) is the control variable of the current state. When there is no control variable, U(k) = 0. P(k|k-1)=AP(k-1|k-1) A'+Q ; In the formula, P(k|k-1) is the covariance of X(k|k-1), P(k-1|k-1) is the covariance of X(k-1|k-1), A' represents the transpose of A, and Q is the covariance of the system process.

6. The monitoring method for wind turbine generator sets according to claim 4, characterized in that, In step S3, a neural network is used for classification or regression prediction during the analysis of the icing state judgment and the structural health assessment. In the icing state judgment, the neural network outputs the icing state result based on the ultrasonic signal characteristics and vibration acceleration characteristics after Kalman filtering and noise reduction, through a neural network classification model. In the structural health assessment, the neural network identifies the location and extent of fatigue damage based on the temporal characteristics of blade vibration and stress.

7. The monitoring method for wind turbine generator sets according to claim 6, characterized in that, The neural network has an input layer, a hidden layer, and an output layer. The input layer has 3 neurons, the hidden layer has 2 layers with 12 neurons each, and the output layer has 4 neurons. The forward propagation formula of the neural network includes: Hidden layer output: h l =σ(W l h l−1 +b l ); Among them, h l For the output of layer l, W l Let b be the weight matrix of the l-th layer. l Here, σ is the bias vector, and σ is the activation function (ReLU: σ(x) = max(0,x)). Output layer output: ; Where y represents the probability values ​​of the four states, and W out b represents the output layer weights. out For output layer bias; The backpropagation formula includes: Loss function: ; Among them, t i For real labels, y i To predict probabilities; Weight update: , ; Where η is the learning rate. Let W be the gradient of the loss function with respect to the weights Wl.

8. The monitoring method for wind turbine generator sets according to claim 6 or 7, characterized in that, It also includes training the neural network. The training process includes dataset preparation, data preprocessing, iterative training, and real-time inference. The dataset preparation includes collecting no less than 1000 sets of samples under different icing conditions in the wind field, of which 70% are used for training and 30% for testing. The data preprocessing is used to standardize the data to avoid the influence of different units of measurement on the training. The iterative training is conducted at least 500 times, updating the weights / biases through backpropagation until the classification accuracy of the test set is ≥95%. The real-time inference includes: during ensemble analysis, inputting a standardized data stream into a trained BP network and outputting icing state classification results.

9. The monitoring method for a wind turbine generator set according to any one of claims 3-7, characterized in that, The control recommendations, maintenance warnings, and status reports generated in step S4 include: When the analysis result is no ice / light ice, the control recommendation is to maintain the current operating parameters, with no maintenance warnings, and the core content of the status report is that the blades have no risk of icing and are operating normally; When the analysis result is medium / heavy icing, the control recommendation is to start the de-icing system and adjust the pitch angle. The maintenance warning is that the ice layer affects the power generation efficiency and a re-inspection is required after de-icing. The core contents of the status report include the icing level, the de-icing execution status, and the assessment of power generation loss. When the analysis results indicate that the load is abnormally high, the control recommendations are to reduce the torque / adjust the speed and optimize the load distribution. The maintenance warning indicates an increased risk of structural fatigue and recommends stopping the machine for inspection. The core contents of the status report include the abnormal load value, duration, and structural stress distribution. The analysis results indicate that when a lightning strike occurs, the control recommendations are to briefly pause for self-testing, and restart if no damage is found. The maintenance warning includes recording lightning strike parameters, checking blades / electrical components, and the status report should include the core content of the lightning strike time / amplitude / location and the results of component damage investigation.

10. The monitoring method for wind turbine generator sets according to claim 9, characterized in that, The control recommendations, maintenance warnings, and status reports generated in step S4 also include: when the analysis result shows that the structural health score is below the threshold, the control recommendation is to shut down for maintenance; the maintenance warning is that the blades have fatigue damage / cracks and require immediate maintenance; and the core content of the status report includes the location / degree of damage and the prediction of remaining service life. When the analysis results show that the structural health score is within the preset range, the control recommendation is to maintain operation and optimize the load. The maintenance warning is to regularly monitor the damage trend and plan preventive maintenance. The core content of the status report includes the health score and the damage development trend. When the analysis results indicate an abnormal temperature rise, the control recommendations are to reduce the load and enhance heat dissipation. The maintenance warnings are to identify overheating of critical components and troubleshoot the faults. The core content of the status report includes abnormal temperature points, heat-generating components, and correlation analysis of operating parameters.

Images