A method and system for early warning of a variable pitch system failure

Abstract

The present application relates to the technical field of variable pitch system fault early warning, and particularly provides a variable pitch system fault early warning method and system, comprising: acquiring characteristic variable data of a variable pitch system of a wind turbine generator; calculating equivalent equivalent torque of each blade in a preset time interval according to variable pitch drive motor torque data in the characteristic variable data; performing fault prediction on a drive motor temperature sensor, a cooling system, an encoder and a slip ring communication in the variable pitch system; preprocessing the characteristic variable data through a linear regression model, dividing the preprocessed characteristic variable data into a training data set and a test data set; inputting the test data set into a pre-trained auto-encoding network model to calculate monitoring index parameters of the test data under the auto-encoding network model, comparing the monitoring index parameters with preset threshold parameters, and outputting a variable pitch system fault early warning result. The present application is used for overcoming the technical problem that a conventional monitoring method of a variable pitch system can only perform fault early warning after mechanical components are damaged.

Description

Technical Field

[0001] This invention relates to the field of pitch system fault early warning technology, and more specifically, to a pitch system fault early warning method and system. Background Technology

[0002] Currently, there are over 110,000 megawatt-class wind turbines connected to the grid using variable frequency and pitch control. With the development of wind turbines, all manufacturers and equipment suppliers are focusing on technological advancements and iterative research into the economic aspects of the equipment itself. For example, regarding the pitch system, which controls turbine power regulation and is crucial to overall turbine safety, manufacturers are emphasizing the development of pitch systems with stronger drive capabilities and higher control precision, while ensuring equipment performance and reasonably reducing costs. However, as existing equipment operates for extended periods, its reliability increasingly impacts turbine safety and power generation hours. Therefore, researching early warning systems for mechanical failures in pitch systems is extremely important.

[0003] Currently, fault warnings for mechanical components in wind turbine pitch systems primarily rely on threshold settings. Based on the components' materials, manufacturing processes, and fatigue calculations, these components typically have maximum and rated values, with different fault levels assigned to different threshold levels. However, traditional mechanical component fault monitoring methods are overly simplistic and passive, only detecting faults when significant damage occurs, by which time the unit's reliability and power generation may already be severely compromised. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for early warning of pitch system faults, in order to overcome the technical problem that traditional monitoring methods for pitch systems can only provide early warning of faults after mechanical components are damaged, and to achieve reliable and accurate early warning of faults for pitch systems.

[0005] The first aspect of the present invention provides a method for early warning of faults in a pitch system, comprising the following steps:

[0006] Obtain characteristic variable data of the wind turbine pitch system;

[0007] Calculate the equivalent torque of each blade within a preset time interval based on the pitch drive motor torque data in the characteristic variable data;

[0008] Fault prediction is performed on the drive motor temperature sensor, cooling system, encoder, and slip ring communication in the pitch system, and the fault prediction results are obtained.

[0009] The feature variable data is preprocessed using a linear regression model, and the preprocessed feature variable data is divided into training dataset and test dataset.

[0010] The test dataset is input into a pre-trained autoencoder network model to calculate the test data. The monitoring index parameters under the autoencoder network model are then compared with the preset threshold parameters to output the pitch system fault warning result.

[0011] Furthermore, the calculation of the equivalent torque of each blade within a preset time interval based on the pitch drive motor torque data in the feature variable data specifically includes:

[0012] The torque data of the pitch drive motor of the propeller is input into the first calculation module. The first calculation module calculates the equivalent torque of the propeller within a preset time interval. The calculation formula of the first calculation module is as follows:

[0013]

[0014] Among them, T j-eq-l T represents the equivalent torque of the j-th blade during the first preset time interval, p represents the mechanical fatigue parameter of the blade, and T represents the equivalent torque of the j-th blade during the first preset time interval. j-1-i This represents the drive motor torque data of the j-th blade at the i-th time node within the first preset time interval, t. i This represents the time interval between the i-th time node and the (i-1)-th time node.

[0015] The correlation and significance parameters between adjacent blades are calculated based on the equivalent torque of the blades.

[0016] Furthermore, the method also includes:

[0017] The Mahalanobis distance, a mechanical fault detection and evaluation index, is calculated based on motor current data, motor temperature data, pitch angle data, speed data, and equivalent torque data from the characteristic variable data.

[0018] Mechanical faults are identified based on Mahalanobis distance, correlation hyperparameters, and significance parameters, and early warning results for mechanical faults are output.

[0019] Furthermore, for fault prediction of the drive motor temperature sensor in the pitch system, the specific fault prediction results include:

[0020] Drive motor temperature sensor prediction: Analyze the second-level characteristic variable data of the drive motor temperature sensor. If the number of times the temperature data difference with a time interval of 1 second is greater than the preset temperature difference is greater than the preset number, it is determined that there is an abnormality in the drive motor temperature sensor.

[0021] Furthermore, fault prediction is performed on the cooling system of the pitch system, and the fault prediction results specifically include:

[0022] Cooling system prediction: Analyze the second-level characteristic variable data of the cooling system, compare the second-level temperature data of multiple blades and obtain the switching information of the cooling fan. After the cooling fan is turned on, if the second-level temperature data change curve of multiple blades is a non-sinusoidal fluctuation, it is determined that there is an anomaly in the cooling system.

[0023] Furthermore, fault prediction is performed on the encoder in the pitch system, and the fault prediction results are obtained specifically including:

[0024] Encoder prediction: Obtain the blade angle position information and compare it with the position information fed back by the encoder. If they are inconsistent, it is determined that there is an error in the encoder.

[0025] Furthermore, fault prediction is performed on the slip ring communication in the pitch system, and the fault prediction results are obtained specifically including:

[0026] Slip ring communication prediction: Obtain the slip ring's transmitted and received data, compare the contents of the transmitted and received data, and calculate the proportion of the transmitted and received data that are consistent. If the proportion is lower than a preset threshold, it is determined that there is an anomaly in the slip ring communication.

[0027] Furthermore, the early warning method also includes confirming whether the blades are installed correctly, specifically including:

[0028] Acquire blade angle data of the pitch system under electromagnetic hovering and constant speed pitch conditions to determine whether there is any deviation in the blade angle;

[0029] When the wind turbine is running, the load data of the drive motor is acquired and combined with the blade phase data to generate a load change curve.

[0030] Determine whether the blade is under balanced stress based on the load change curve.

[0031] Furthermore, preprocessing the feature variable data using a linear regression model specifically includes:

[0032] Obtain outliers from the characteristic variable data of the pitch system, and remove status words and power limit status words from the outliers;

[0033] Based on wind speed conditions, filter the data between the cut-in wind speed and the cut-out wind speed from the characteristic variable data;

[0034] Based on the wind speed-power curve, a linear regression model is used to clean the data by removing status words, power-limited status words, and outliers after filtering.

[0035] Calculate the mean and standard deviation of each variable in the feature variable data, and perform standardization on the feature variable data.

[0036] The second aspect of the present invention provides a pitch system fault early warning system, comprising:

[0037] The acquisition module is configured to acquire characteristic variable data of the wind turbine pitch system.

[0038] The first calculation module is configured to calculate the equivalent torque of each blade within a preset time interval based on the pitch drive motor torque data in the feature variable data.

[0039] The second calculation module is configured to calculate the correlation parameters and significance parameters between adjacent blades based on the equivalent torque.

[0040] The third calculation module is configured to calculate the Mahalanobis distance based on the equivalent torque and the motor current data, motor temperature data, pitch angle data, and speed data in the characteristic variable data.

[0041] The prediction module is configured to perform fault prediction on the drive motor temperature sensor, cooling system, encoder and slip ring communication in the pitch system, and obtain the fault prediction results.

[0042] The first judgment module is configured to judge mechanical faults based on Mahalanobis distance, correlation parameters, and significance parameters, and output mechanical fault warning signals.

[0043] The second judgment module is configured to output the pitch system fault warning result based on the trained autoencoder network model.

[0044] The beneficial effects of this invention include:

[0045] 1. This invention calculates the equivalent torque of each blade within a preset time interval based on the motor torque data in the characteristic variable data; calculates the correlation parameters and significance parameters between adjacent blades based on the equivalent torque of each blade; calculates the Mahalanobis distance based on the equivalent torque and the motor current data, motor temperature data, pitch angle data, and speed data in the characteristic variable data; and judges mechanical faults based on the Mahalanobis distance, correlation parameters, and significance parameters, outputting a mechanical fault warning. This fault warning method can detect mechanical faults in a timely and accurate manner, and can also predict the timing of fault occurrence. It uses correlation parameter and significance parameter methods to better analyze the relationship between various parts of the pitch system, discover potential fault sources, and thus avoid the problem of existing technologies that can only notify of faults after mechanical components are damaged, leading to the failure of the generator's power generation function. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 A flowchart illustrating the pitch system fault early warning method provided in an embodiment of the present invention;

[0048] Figure 2 A linear relationship graph of pitch speed versus time provided in an embodiment of the present invention;

[0049] Figure 3 A linear relationship graph of torque versus time provided in an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of the fault device provided in an embodiment of the present invention. Detailed Implementation

[0051] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention.

[0052] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0053] Please refer to Figures 1 to 3 As shown, the first aspect of the present invention provides a method for early warning of faults in a pitch system, comprising the following steps:

[0054] Step S100: Obtain characteristic variable data of the wind turbine pitch system;

[0055] In step S100, the characteristic variable data of the pitch system are mainly obtained from the wind farm's data acquisition and monitoring control system (SCADA). The characteristic variable data related to the pitch system mainly include various mechanical parameter data used to monitor and control the operating status and performance of the wind turbine blades. The data includes at least: blade angle, blade speed, torque, vibration, temperature, oil pressure, etc.

[0056] Step S200: Calculate the equivalent torque of each blade within a preset time interval based on the pitch drive motor torque data in the feature variable data; the equivalent torque specifically refers to the torque generated when the wind turbine blades rotate under wind force.

[0057] Step S200 includes:

[0058] Step S210: Input the pitch drive motor torque data of the blade into the first calculation module, and calculate the equivalent torque of the blade within a preset time interval according to the first calculation module. The calculation formula of the first calculation module is as follows:

[0059]

[0060] Among them, T j-eq-1 T represents the equivalent torque of the j-th blade during the first preset time interval, p represents the mechanical fatigue parameters of the blade, which are various parameters affecting the fatigue life of the mechanical structure of wind turbine blades. Common mechanical fatigue parameters include the blade's natural frequency, vibration mode, tip velocity, vibration magnitude, moment of inertia, axial force, torque, etc. j-l-i This represents the drive motor torque data of the j-th blade at the i-th time node within the first preset time interval, t. i This represents the time interval between the i-th time node and the (i-1)-th time node.

[0061] Step S220: Calculate the correlation parameters and significance parameters between adjacent blades based on the equivalent torque of the blades. Specifically, by performing correlation analysis on the equivalent torque of each blade, the correlation parameters between adjacent blades can be calculated. Commonly used correlation analysis methods such as Pearson correlation coefficient and Spearman rank correlation coefficient can be used for calculation. Then, the significance parameters between adjacent blades are calculated using the t-test method to determine the significance of the correlation parameters, thereby identifying potential mechanical faults in the pitch system.

[0062] Step S300: Perform fault prediction on the drive motor temperature sensor, cooling system, encoder and slip ring communication in the pitch system, and obtain the fault prediction results;

[0063] Step S300 includes:

[0064] Step S310: Obtain the fault prediction result of the drive motor temperature sensor:

[0065] Analyzing the second-level characteristic variable data of the drive motor temperature sensor, if the number of times the temperature data difference with a time interval of 1 second is greater than a preset temperature difference is greater than a preset number within a preset time period, then it is determined that the drive motor temperature sensor is abnormal. In this embodiment, the preset number ranges from 1 to 20 times, and the preset temperature difference ranges from 0 to 20 degrees Celsius. For example, if the number of times the temperature data difference with a time interval of 1 second is greater than 5 degrees Celsius within a preset time period of 48 hours is greater than twice, then it is determined that the temperature sensor is abnormal.

[0066] Step S320: Obtain the cooling system fault prediction results:

[0067] Analyzing the second-level characteristic variable data of the cooling system, comparing the second-level temperature data of multiple blades, and obtaining the on / off information of the radiator fan, if the second-level temperature data change curves of multiple blades are non-sinusoidal fluctuations after the radiator fan is turned on, an anomaly is judged in the cooling system. The sinusoidal fluctuation of the blade temperature data represents the regular oscillation of wind turbine blade temperature under the influence of wind energy. The principle is that due to the periodic changes in wind energy, the wind turbine blades are affected by periodic temperature fluctuations during operation, producing a temperature change curve similar to a sine wave. Therefore, sinusoidal temperature fluctuations can be used to assess the fatigue damage and reliability of wind turbine blades. The greater the temperature fluctuation, the greater the change in internal stress of the wind turbine blade, the higher the degree of fatigue damage, and the shorter the lifespan. Therefore, real-time monitoring and analysis of temperature fluctuations are necessary to promptly identify and predict potential blade failures and damage. In this embodiment, when predicting the cooling system of the pitch system, during the continuous operation of the pitch motor, when it continues to run below the rated speed-torque curve, the pitch motor will eventually reach a thermal equilibrium state within a certain temperature range. During the process of motor temperature rise, when the motor temperature exceeds 40 degrees Celsius, the motor cooling fan speed will be turned on. At this time, the motor temperature will be significantly suppressed, and when the motor is cooled to below the hysteresis temperature, the pitch motor temperature will show a significant decrease.

[0068] Step S330: Obtain encoder fault prediction results:

[0069] The propeller blade angle position information is obtained and compared with the position information fed back by the encoder. If they are inconsistent, it is determined that the encoder is abnormal. In this embodiment, by judging the encoder fault, the fault type can be further determined, such as position deviation, data offset, signal interference, etc.

[0070] Step S340: Obtain the slip ring communication fault prediction result:

[0071] The system acquires the data sent and responded by the slip ring, compares the contents of the sent and responded data, and calculates the proportion of the content that matches the content of the sent and responded data. If the proportion is lower than the preset proportion threshold, it is determined that there is an abnormality in the slip ring communication. If an abnormality in the slip ring communication is found, the abnormality can be handled by reconnecting the slip ring, checking whether the connection of the structural parts is loose, or replacing the slip ring, so as to ensure the normal operation of the pitch system.

[0072] Step S300 also includes step S350: confirming that the blades are installed correctly, specifically including:

[0073] Step S351: Obtain blade angle data of the pitch system under electromagnetic hovering and constant speed pitch conditions, and determine whether there is a deviation in the blade angle;

[0074] Step S352: When the wind turbine is running, acquire the drive motor load data and combine it with the blade phase data to generate a load change curve;

[0075] Step S353: Determine whether the blade is under stress balance based on the load change curve; specifically, if there are fluctuations in the load change curve with amplitudes exceeding or falling below the average level, it is determined that the blade is under stress imbalance.

[0076] Step S300 also includes step S360: analyzing whether there are any abnormalities in the drivetrain through four dimensions: pitch speed, torque, start-up time, and blade temperature. (See also...) Figure 2 and Figure 3 As shown, by acquiring real-time data of pitch speed and torque, a graph showing the numerical changes of pitch speed and torque over time is plotted. The graph records the numerical changes of pitch speed and torque over time, respectively. Under normal operating conditions, the pitch system should be stable and periodic. Therefore, for... Figure 2 and Figure 3 By analyzing whether the changing trends of pitch speed and torque are consistent or whether the changing cycle is normal, it can be determined whether there are potential problems in the pitch drivetrain. The principle of analyzing anomalies from the start-up time and blade temperature is the same as above, and will not be repeated here.

[0077] Step S400: Preprocess the feature variable data using a linear regression model, and divide the preprocessed feature variable data into a training dataset and a test dataset;

[0078] In step S400, the preprocessing of the feature variable data using a linear regression model specifically includes:

[0079] Step S410: Obtain outliers in the characteristic variable data of the pitch system, and remove status words and power limit status words from the outliers; when the pitch system sensor data acquisition device writes data, there are preset status words and power limit status words. The purpose of removing status words and power limit status words is to avoid the problem of poor data quality.

[0080] Step S420: Based on the wind speed conditions, filter out the data between the cut-in wind speed and the cut-out wind speed from the characteristic variable data; since only the wind speed data between the cut-in wind speed and the cut-out wind speed is related to the corresponding power generation data, and other wind speed data will not affect the system output, it is only necessary to filter out the wind speed data between the cut-in wind speed and the cut-out wind speed.

[0081] Step S430: Based on the wind speed and power curve, a linear regression model is used to clean the data by removing status words, power limit status words, and outliers after filtering.

[0082] Step S440: Calculate the mean and standard deviation of each variable in the feature variable data, and perform standardization on the feature variable data. Considering the different dimensions of each feature variable, standardize the feature variable data using the mean and standard deviation of the feature variables, employing a standard normal distribution; specifically:

[0083]

[0084] Where x represents the feature variable, z represents the standardized value, mean(x) represents the mean of the feature variable x, and std(x) represents the standard deviation of the feature variable x;

[0085] Step S500: Input the test dataset into the pre-trained autoencoder network model to calculate the monitoring index parameters of the test data under the autoencoder network model, and compare the monitoring index parameters with the preset threshold parameters to output the pitch system fault warning result; wherein, the autoencoder network model is an unsupervised deep learning model. The autoencoder network model can reduce the dimensionality and reconstruct the original mechanical motion data, thereby realizing fault warning. Specifically, a pitch system autoencoder network is trained using the training dataset to learn the latent representation of the pitch system features. The trained autoencoder network model is used to encode the test dataset. The fault warning probability is calculated based on the monitoring index parameters of the model under the test dataset. By comparing the monitoring index with the preset threshold parameters, the fault warning result can be obtained.

[0086] Step S500 also includes:

[0087] Step S510: Calculate the Mahalanobis distance, a mechanical fault detection and evaluation index, based on the motor current data, motor temperature data, pitch angle data, speed data, and equivalent torque data in the feature variable data.

[0088] Step S520: Determine mechanical faults based on Mahalanobis distance, correlation hyperparameters, and significance parameters, and output mechanical fault warning results.

[0089] In summary, this invention calculates the equivalent torque of each blade within a preset time interval based on the motor torque data in the characteristic variable data; calculates the correlation and significance parameters between adjacent blades based on the equivalent torque of each blade; calculates the Mahalanobis distance based on the equivalent torque and the motor current, motor temperature, pitch angle, and speed data in the characteristic variable data; and identifies mechanical faults based on the Mahalanobis distance, correlation parameters, and significance parameters, outputting a mechanical fault warning. This fault warning method can detect mechanical faults in a timely and accurate manner, and can also predict the timing of fault occurrence. It uses correlation and significance parameter methods to better analyze the relationships between various parts of the pitch system, discover potential fault sources, and thus avoid the problem of existing technologies that can only notify of faults after mechanical components are damaged, leading to generator power generation function failure.

[0090] Please refer to Figure 4 The second aspect of the present invention provides a pitch system fault early warning system, comprising:

[0091] The acquisition module is configured to acquire characteristic variable data of the wind turbine pitch system.

[0092] The first calculation module is configured to calculate the equivalent torque of each blade within a preset time interval based on the pitch drive motor torque data in the feature variable data.

[0093] The second calculation module is configured to calculate the correlation parameters and significance parameters between adjacent blades based on the equivalent torque.

[0094] The third calculation module is configured to calculate the Mahalanobis distance based on the equivalent torque and the motor current data, motor temperature data, pitch angle data, and speed data in the characteristic variable data.

[0095] The prediction module is configured to perform fault prediction on the drive motor temperature sensor, cooling system, encoder and slip ring communication in the pitch system, and obtain the fault prediction results.

[0096] The first judgment module is configured to judge mechanical faults based on Mahalanobis distance, correlation parameters, and significance parameters, and output mechanical fault warning signals.

[0097] The second judgment module is configured to output a fault warning result for the pitch system based on a trained autoencoder network model. The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for early warning of faults in a pitch control system, characterized in that, Includes the following steps: Obtain characteristic variable data of the wind turbine pitch system; The equivalent torque of each blade within a preset time interval is calculated based on the pitch drive motor torque data in the characteristic variable data, specifically including: The torque data of the pitch drive motor of the propeller is input into the first calculation module. The first calculation module calculates the equivalent torque of the propeller within a preset time interval. The calculation formula of the first calculation module is as follows: in, Let represent the equivalent torque of the j-th blade during the first preset time interval, and p represent the mechanical fatigue parameters of the blade. This represents the drive motor torque data for the j-th blade at the i-th time node within the first preset time interval. This represents the time interval between the i-th time node and the (i-1)-th time node. Calculate the correlation and significance parameters between adjacent blades based on the equivalent torque of the blades; The Mahalanobis distance, a mechanical fault detection and evaluation index, is calculated based on motor current data, motor temperature data, pitch angle data, speed data, and equivalent torque data from the characteristic variable data. Mechanical faults are identified based on Mahalanobis distance, correlation parameters, and significance parameters, and mechanical fault warning results are output. Fault prediction is performed on the drive motor temperature sensor, cooling system, encoder, and slip ring communication in the pitch system, and the fault prediction results are obtained, specifically including: Drive motor temperature sensor prediction: Analyze the second-level characteristic variable data of the drive motor temperature sensor. If the number of times the temperature data difference with a time interval of 1 second is greater than the preset temperature difference is greater than the preset number within a preset time, it is determined that there is an abnormality in the drive motor temperature sensor. Cooling system prediction: Analyze the second-level characteristic variable data of the cooling system, compare the second-level temperature data of multiple blades and obtain the switching information of the cooling fan. After the cooling fan is turned on, if the second-level temperature data change curve of multiple blades is a non-sinusoidal fluctuation, it is determined that there is an anomaly in the cooling system. Slip ring communication prediction: Obtain the slip ring's transmitted and received data, compare the contents of the transmitted and received data, and calculate the proportion of the transmitted and received data that are consistent. If the proportion is lower than the preset proportion threshold, it is determined that there is an anomaly in the slip ring communication. The feature variable data is preprocessed using a linear regression model, and the preprocessed feature variable data is divided into training dataset and test dataset. The test dataset is input into a pre-trained autoencoder network model to calculate the test data. The monitoring index parameters under the autoencoder network model are then compared with the preset threshold parameters to output the pitch system fault warning result.

2. The pitch system fault early warning method according to claim 1, characterized in that, Fault prediction of the encoder in the pitch system and obtaining the fault prediction results specifically include: Encoder prediction: Obtain the blade pitch angle position information and compare it with the position information fed back by the encoder. If they are inconsistent, it is determined that there is an error in the encoder.

3. The pitch system fault early warning method according to claim 1, characterized in that, The early warning method also includes confirming whether the blades are installed correctly, specifically including: Acquire blade angle data of the pitch system under electromagnetic hovering and constant speed pitch conditions to determine whether there is any deviation in the blade angle; When the wind turbine is running, the load data of the drive motor is acquired and combined with the blade phase data to generate a load change curve. Determine whether the blade is under balanced stress based on the load change curve.

4. The pitch system fault early warning method according to any one of claims 1 to 3, characterized in that, Preprocessing of feature variable data using a linear regression model specifically includes: Obtain outliers from the characteristic variable data of the pitch system, and remove status words and power limit status words from the outliers; Based on wind speed conditions, filter the data between the cut-in wind speed and the cut-out wind speed from the feature variable data; Based on the wind speed-power curve, a linear regression model is used to clean the data by removing status words, power-limited status words, and outliers after filtering. Calculate the mean and standard deviation of each variable in the feature variable data, and perform standardization on the feature variable data.

5. A pitch system fault early warning system, characterized in that, The method for performing the pitch system fault early warning method according to any one of claims 1 to 4 includes: The acquisition module is configured to acquire characteristic variable data of the wind turbine pitch system; The first calculation module is configured to calculate the equivalent torque of each blade within a preset time interval based on the pitch drive motor torque data in the feature variable data. The second calculation module is configured to calculate the correlation parameters and significance parameters between adjacent blades based on the equivalent torque. The third calculation module is configured to calculate the Mahalanobis distance based on the equivalent torque and the motor current data, motor temperature data, pitch angle data, and speed data in the characteristic variable data. The prediction module is configured to perform fault prediction on the drive motor temperature sensor, cooling system, encoder and slip ring communication in the pitch system, and obtain the fault prediction results. The first judgment module is configured to judge mechanical faults based on Mahalanobis distance, correlation parameters, and significance parameters, and output mechanical fault warning results. The second judgment module is configured to output the pitch system fault warning result based on the trained autoencoder network model.

Images