Fan fault diagnosis method based on SCADA data

Abstract

The invention specifically discloses a fan fault diagnosis method based on SCADA (Supervisory Control and Data Acquisition) data. The method comprises the following steps of S1, acquiring SCADA monitoring data in a fan running state and preprocessing the SCADA monitoring data; s2, X-direction vibration data and Y-direction vibration data of the fan cabin are extracted from the preprocessed SCADA monitoring data, and whether a fault exists or not is judged based on a spectral analysis method; and S3, inputting the corresponding data with the fault into a pre-constructed deep neural network model, and outputting the corresponding fault category of the fan blade. According to the method, the fault and the fault type in the fan operation process can be effectively identified based on the SCADA monitoring data, so that the maintenance cost of a wind power plant is reduced, the occurrence of accidental injury accidents caused by blade falling can be effectively prevented, and the method has a wide market application prospect.

Description

Technical Field

[0001] The present invention relates to the field of new energy technology, and in particular to a fan fault diagnosis method based on SCADA data. Background Art

[0002] In recent years, the field of wind power has been continuously developed, and the installed capacity of wind turbines has also been increasing. Generally, the design life of onshore wind turbines is 20 years, and the design life of offshore wind turbines is 25 years, with a warranty period of 5 years. However, during the service period, wind turbines will cause component wear and damage accumulation due to long-term operation, and the failure rate increases year by year. This means that behind the rapid development of wind power in the future, there will be a reality that a large number of installed wind turbines will be out of warranty one after another. At that time, wind power failure accidents will also usher in an outbreak period, and the rapid development of the wind power industry will face unprecedented severe challenges.

[0003] At present, the operation and maintenance of wind turbines are mainly divided into post-maintenance and regular maintenance. Post-maintenance is a more traditional maintenance method, that is, maintenance is carried out only when the blades have functional problems and cannot operate. Regular maintenance refers to maintenance after the blades have been used for a specified time, which is the current mainstream mode of wind turbine blade maintenance. For example, inspection personnel are sent out every month to check the damage of the blades using telescopes, high-altitude circling and descent visual inspections (little flying men), maintenance platform hanging baskets and other means. The above two maintenance methods require special shutdown for inspection, and there are problems such as poor safety, large workload, low efficiency, and being affected by the observation angle.

[0004] In recent years, more and more large wind farms use Supervisory Control and Data Acquisition (SCADA) systems to monitor the operating status of wind turbines because there is no need to install additional vibration sensors and better on-site implementation conditions. Wind power researchers generally believe that the operating status trends and potential fault information of the units are hidden in the massive multi-source data recorded by the SCADA system. In addition, with the rise and development of machine learning in recent years, it has become a new research hotspot to analyze the data features of fault data and normal data in various monitoring data of wind turbines through appropriate algorithms and implement fault diagnosis of wind turbines from the perspective of data feature mining. The present invention provides a wind turbine fault diagnosis method based on SCADA data. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a fan fault diagnosis method based on SCADA data, the method comprising the following steps:

[0006] S1. Obtain SCADA monitoring data under the operation state of the wind turbine, and pre-process the obtained SCADA monitoring data to remove abnormal data;

[0007] S2, extracting the X-direction vibration data and the Y-direction vibration data of the wind turbine cabin from the pre-processed SCADA monitoring data, analyzing the extracted vibration data based on the spectrum analysis method to determine whether there is a fault, if there is a fault, proceeding to step S3, if there is no fault, repeating step S2;

[0008] S3. Input the corresponding data of the fault in step S2 into a pre-built deep neural network model, and then output the corresponding fault category of the wind turbine blade.

[0009] Furthermore, the preprocessing of the acquired SCADA monitoring data in step S1 specifically includes:

[0010] S11, finding abnormal data under the operation state of the wind turbine from the acquired SCADA monitoring data and determining the category of the abnormal data;

[0011] S12. When the abnormal data is missing data, further determine whether the missing data is high-frequency data. If it is high-frequency data, use the fitting interpolation method to process the missing data. If it is not high-frequency data, use the method based on the average value of the previous and next data to replace the missing data.

[0012] S13. When the abnormal data is abnormal jump data, the abnormal jump data is processed by using a differential method, so that all abnormal jump data after the differential method fluctuate at a zero balance position.

[0013] Furthermore, in step S13, if the abnormal jump data still deviates greatly from the zero balance position after being processed by the differential method, the abnormal jump data is directly eliminated, and the data is supplemented by the fitting interpolation method or the average value method of the previous and next data.

[0014] Furthermore, in step S13, when the removed abnormal jump data is high-frequency data, a fitting interpolation method is used to complete it; if it is not high-frequency data, an average value of the previous and next data is used for replacement completion.

[0015] Furthermore, the deep neural network model pre-constructed in step S3 is obtained by training and optimization based on the fan blade simulation experiment data and the SCADA monitoring data during the actual operation of the fan blade, and the deep neural network model is used to output the fan blade fault category.

[0016] Furthermore, the wind turbine blade fault categories include blade surface crack fault, blade wear fault and blade deformation fault.

[0017] Furthermore, it also includes step S4, using the image acquisition unit to obtain in real time the surface image of the wind turbine blade in the operating state, and transmitting the acquired surface image of the wind turbine blade to a pre-built blade appearance deep neural network model to judge the blade appearance defects and damages.

[0018] Compared with the prior art, the wind turbine fault diagnosis method based on SCADA data provided by the present invention can effectively remove abnormal data by preprocessing and analyzing the collected SCADA monitoring data, providing a guarantee for subsequent fault diagnosis and analysis, and then analyzes the preprocessed SCADA data based on the spectrum analysis method to accurately determine whether there is a fault in the wind turbine blade. Finally, the fault category corresponding to the wind turbine blade can be output based on the pre-set deep neural network model. Therefore, the fault diagnosis method in this application can effectively identify the faults and their fault categories during the operation of the wind turbine, thereby reducing the maintenance cost of the wind power plant, and can effectively prevent the occurrence of accidental injuries caused by falling blades, and has broad market application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flow chart of a fan fault diagnosis method based on SCADA data in this application,

[0020] Figure 2 It is a spectrum analysis diagram of a selected section of wind turbine cabin vibration acceleration data in some implementation modes of the present application. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings.

[0022] like Figure 1 As shown, the present application provides a fan fault diagnosis method based on SCADA data, the method comprising the following steps:

[0023] S1. Obtain SCADA monitoring data under the operation state of the wind turbine, and pre-process the obtained SCADA monitoring data to remove abnormal data;

[0024] This step specifically includes:

[0025] S11, finding abnormal data under the operation state of the wind turbine from the acquired SCADA monitoring data and determining the category of the abnormal data;

[0026] S12. When the abnormal data is missing data, further determine whether the missing data is high-frequency data. If it is high-frequency data, use the fitting interpolation method to process the missing data. If it is not high-frequency data, use the method based on the average value of the previous and next data to replace the missing data.

[0027] S13. When the abnormal data is abnormal jump data, the abnormal jump data is processed by the difference method so that all abnormal jump data after the difference method fluctuate at the zero balance position. If the abnormal jump data still deviates greatly from the zero balance position after the difference method, the abnormal jump data is directly eliminated, and the fitting interpolation method or the average value of the previous and next data is used to complete the data. When the eliminated abnormal jump data is high-frequency data, the fitting interpolation method is used to complete the data. If it is not high-frequency data, the average value of the previous and next data is used to complete the data.

[0028] S2, extracting the X-direction vibration data and the Y-direction vibration data of the wind turbine cabin from the pre-processed SCADA monitoring data, analyzing the extracted vibration data based on the spectrum analysis method to determine whether there is a fault, if there is a fault, proceeding to step S3, if there is no fault, repeating step S2;

[0029] S3, inputting the corresponding data of the fault in step S2 into a pre-built deep neural network model, and then outputting the corresponding fault category of the fan blade;

[0030] Among them, the deep neural network model is trained and optimized based on the fan blade simulation experiment data and the SCADA monitoring data during the actual operation of the fan blade. The fan blade simulation experiment data includes the SCADA data monitored when the fan blade is running in the state of surface cracks, wear or deformation during the simulation experiment, and the SCADA data monitored when the fan blade is running normally during the simulation experiment. Based on the fan blade simulation experiment data, a trained pre-built deep neural network model is obtained, and then based on the deep neural network model, the fan blade fault category can be output.

[0031] In the above implementation, by preprocessing and analyzing the collected SCADA monitoring data, abnormal data can be effectively removed, and then real monitoring data can be obtained, which provides a guarantee for subsequent fault diagnosis and analysis; then, based on the spectrum analysis method, the preprocessed SCADA data is analyzed to accurately determine whether there is a fault in the wind turbine blade; finally, based on the pre-set deep neural network model, the fault category corresponding to the wind turbine blade can be output. Therefore, the fault diagnosis method in this application can effectively identify the faults and their fault categories during the operation of the wind turbine, thereby reducing the maintenance cost of the wind power plant, and can effectively prevent the occurrence of accidental injuries caused by falling blades, and has broad market application prospects.

[0032] It should be noted that the construction and training of deep neural network models belong to the existing technology and will not be elaborated here. At the same time, in order to better monitor the wind blades, during the construction and training of the deep neural network model, it is also possible to train according to the severity of the corresponding faults of the wind blades, so as to better judge whether the wind blades should be repaired or replaced.

[0033] In some embodiments, step S4 is also included, using an image acquisition unit to acquire in real time the surface image of the wind turbine blade in operation, and transmitting the acquired surface image of the wind turbine blade to a pre-built blade appearance deep neural network model to judge the blade appearance defects and damages.

[0034] In the above implementation, the surface image of the wind turbine blade in operation is collected by the image acquisition unit, and the appearance defects and damages of the blade are judged based on the surface image of the wind turbine blade. The wind turbine blade fault and its fault category are judged in combination with the SCADA monitoring data. The above two judgment results are combined to further improve the accuracy of the judgment, providing a strong guarantee for the accuracy of the final judgment result.

[0035] In order to further illustrate the working principle and beneficial technical effects of the present invention, in this embodiment, the fan cabin vibration acceleration data with a sampling frequency of 1 Hz and a sampling time of 10 minutes is selected for spectrum analysis. The analysis results are as follows: Figure 2 As shown, Figure 2 a shows the spectrum analysis of the X-direction vibration data of the wind turbine nacelle. Figure 2 b shows the spectrum analysis diagram of the Y-direction vibration data of the wind turbine nacelle.

[0036] from Figure 2 It can be seen that Figure 2 In a, the frequency was abnormal at 0.25-0.3 Hz; Figure 2 In b, the frequency was abnormal at 0.25-0.3Hz and the frequency was abnormal at 0.35-0.4Hz, indicating that the vibration acceleration data of the wind turbine cabin corresponds to the monitored part. Then, the wind turbine cabin vibration acceleration data is input into the pre-built deep neural network model, and the corresponding wind turbine blade fault category can be output. Therefore, the fault diagnosis method provided by the present application can effectively identify the faults and their fault categories during the operation of the wind turbine, thereby reducing the maintenance cost of the wind power plant and preventing accidental injuries caused by falling blades, and has broad market application prospects.

[0037] The above is a detailed introduction to a fan fault diagnosis method based on SCADA data provided by the present invention. This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the core idea of ​​the present invention. It should be pointed out that for ordinary technicians in this technical field, without departing from the principle of the present invention, the present invention can also be improved and modified, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A fan fault diagnosis method based on SCADA data, It is characterized in that The method comprises the following steps: S1. Obtain SCADA monitoring data under the operation state of the wind turbine, and pre-process the obtained SCADA monitoring data to remove abnormal data; S2, extracting the X-direction vibration data and the Y-direction vibration data of the wind turbine cabin from the pre-processed SCADA monitoring data, analyzing the extracted vibration data based on the spectrum analysis method to determine whether there is a fault, if there is a fault, proceeding to step S3, if there is no fault, repeating step S2; S3. Input the corresponding data of the fault in step S2 into a pre-built deep neural network model, and then output the corresponding fault category of the wind turbine blade.

2. The fan fault diagnosis method based on SCADA data as claimed in claim 1, It is characterized in that The preprocessing of the acquired SCADA monitoring data in step S1 specifically includes: S11, finding abnormal data under the operation state of the wind turbine from the acquired SCADA monitoring data and determining the category of the abnormal data; S12. When the abnormal data is missing data, further determine whether the missing data is high-frequency data. If it is high-frequency data, use the fitting interpolation method to process the missing data. If it is not high-frequency data, use the method based on the average value of the previous and next data to replace the missing data. S13. When the abnormal data is abnormal jump data, the abnormal jump data is processed by using a differential method, so that all abnormal jump data after the differential method fluctuate at a zero balance position.

3. The fan fault diagnosis method based on SCADA data as claimed in claim 2, It is characterized in that In step S13, if the abnormal jump data still deviates greatly from the zero balance position after being processed by the difference method, the abnormal jump data is directly eliminated, and the data is supplemented by the fitting interpolation method or the average value method of the previous and next data.

4. The fan fault diagnosis method based on SCADA data as claimed in claim 3, It is characterized in that In step S13, when the removed abnormal jump data is high-frequency data, a fitting interpolation method is used to complete it; if it is not high-frequency data, an average value of the previous and next data is used for replacement and completion.

5. The fan fault diagnosis method based on SCADA data as claimed in claim 4, It is characterized in that The deep neural network model pre-constructed in step S3 is obtained by training and optimization based on the simulation experiment data of the fan blades and the SCADA monitoring data during the actual operation of the fan blades. The deep neural network model is used to output the fault category of the fan blades.

6. The fan fault diagnosis method based on SCADA data as claimed in claim 5, It is characterized in that The wind turbine blade fault categories include blade surface crack fault, blade wear fault and blade deformation fault.

7. The fan fault diagnosis method based on SCADA data as claimed in claim 6, It is characterized in that The method also includes step S4, using an image acquisition unit to acquire in real time the surface image of the wind turbine blade in the operating state, and transmitting the acquired surface image of the wind turbine blade to a pre-built deep neural network model of blade appearance to judge the appearance defects and damages of the blade.

Images