A wind turbine fault analysis system based on SCADA data modeling

Abstract

The invention relates to the technical field of fault analysis, and discloses a wind turbine fault analysis system based on SCADA data modeling. The wind turbine fault analysis system based on SCADA data modeling comprises a central processing unit (CPU), which is used to transmit and store all data in the wind turbine fault analysis system. After performing fault analysis on the wind turbine, the invention uses an intelligent adjustment system to combine data from a qualitative analysis database, a quantitative analysis database, and a networked large database for intelligent calculation to obtain an intelligent adjustment value of abnormal data, and generates an instruction for adjusting relevant data through a single-chip microcomputer. In this way, pre-adjustment of the wind turbine can be achieved, and the occurrence of fault conditions caused by untimely data adjustment can be avoided. At the same time, no manual adjustment is required. After the adjustment, an adjustment log is generated as a database support, thereby forming intelligent adjustment, reducing the failure rate of the wind turbine and improving the service life of the wind turbine.

Description

Technical Field

[0001] The invention belongs to the technical field of fault analysis, and in particular relates to a wind turbine fault analysis system based on SCADA data modeling. Background Art

[0002] The wind power source consists of a wind turbine generator set, a tower supporting the generator set, a battery charge controller, an inverter, an unloader, a grid controller, a battery pack, etc. The wind turbine generator set includes a wind wheel and a generator; the wind wheel contains blades, a hub, reinforcements, etc.; it has the functions of generating electricity by rotating the blades under wind force and rotating the generator head.

[0003] According to the patent number "CN202110602643.3", a "wind turbine fault analysis system based on SCADA data modeling is proposed, including a gateway device, a data processing unit and a model building unit; the data processing unit and the model building unit are integrated and developed based on the SCADA system platform, and the data processing unit obtains the data information of the research object and generates a fault database for extraction by the model building unit; the gateway device includes a memory, a central processing unit and a computer program; the data processing unit includes a standardized preprocessing module and a fault data automatic labeling module; the model building unit includes an information extraction module, a fault warning module and a visualization module, and are connected in sequence. Through the present invention, the automatic mining and comprehensive utilization of related information, the autonomous learning and modeling of the complex temporal and spatial correlations inherent in SCADA data, and the automatic detection and alarm of faults are realized."

[0004] According to the patent number "CN202011009681.X", a "motor operation fault analysis system is proposed, which includes a temperature sensor, a voltage sensor, a vibration sensor, a first conditioning circuit, a second conditioning circuit, a third conditioning circuit, an alarm device, a centralized control module, a display device, a storage device, a communication device and a power supply device, wherein a temperature sensor is used to test the casing temperature of the motor to be tested, and the temperature signal of the motor rotor can be obtained by combining the corresponding signal curve of the rotor temperature of the motor to be tested and the casing temperature of the motor to be tested. A vibration sensor is used to test the vibration signal of the motor to be tested, and a voltage sensor is used to test the output voltage of the motor to be tested, and then the motor fault is effectively diagnosed based on the motor rotor temperature, casing vibration and working voltage."

[0005] The above solution solves the problem of fault analysis during the operation of wind turbines. However, after the fault analysis, the analysis results are generally sent to the operation and maintenance personnel for corresponding operation and maintenance management. This may lead to wind turbine failures due to untimely data processing, thereby increasing maintenance costs and reducing the service life of the wind turbine. Therefore, in view of the current situation, it is necessary to improve it. Summary of the Invention

[0006] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a wind turbine fault analysis system based on SCADA data modeling, which effectively solves the problem that the above solution solves the fault analysis during the operation of the wind turbine. However, after the fault analysis, the analysis results are generally sent to the operation and maintenance personnel for corresponding operation and maintenance management, which may lead to wind turbine failures caused by untimely data processing, thereby increasing maintenance costs and reducing the service life of the wind turbine.

[0007] To achieve the above object, the present invention provides the following technical solutions: a wind turbine fault analysis system based on SCADA data modeling, comprising a central processing unit, the central processing unit being used to transmit and store all data in the wind turbine fault analysis system;

[0008] Data monitoring system: The data monitoring system is used to monitor all data of the wind turbine in real time. The data monitoring system is based on the SCADA data system.

[0009] Data screening system: The data screening system is used to pre-process all data within a fixed time interval monitored by the above-mentioned data monitoring system. The pre-processing steps specifically include interval calibration, resampling and general pre-processing;

[0010] Data processing system: The data processing system is used to perform feature sampling and extraction on the data processed by the data screening system. The feature sampling and extraction data specifically includes power deviation feature extraction, cabin internal and external environment deviation feature extraction, pitch deviation feature extraction, and time series feature extraction;

[0011] Model analysis system: The model analysis system is used to perform fault prediction model analysis on the data monitored by the data monitoring system and obtain fault prediction results;

[0012] Intelligent adjustment system: The intelligent adjustment system is used to analyze the fault prediction results obtained by the fault prediction model constructed in the above-mentioned model analysis system, and combine the database data to perform intelligent data adjustment control and send it to the corresponding single-chip computer end;

[0013] Result feedback system: After the intelligent adjustment system sends the intelligent data adjustment control instruction, the result feedback system is used to perform a fault prediction model on the data that has been adjusted to have fault signs, compare it with the data before adjustment, obtain the comparison result, and generate an adjustment log for storage.

[0014] Preferably, all data of the wind turbine monitored by the data monitoring system specifically include: wind speed, generator speed, power, wind angle, average wind direction angle, yaw position, yaw speed, ambient temperature, cabin temperature, ng51, 2, 3 temperatures respectively and charger DC current, blades 1, 2, 3 speed, angle and temperature respectively, acceleration in X and Y directions. The data monitoring system specifically uses wind speed sensor, power sensor, speed sensor, wind direction sensor, yaw sensor, speed sensor, temperature sensor, current sensor and acceleration sensor when performing data monitoring.

[0015] Preferably, the interval calibration is specifically to perform normal data calibration and fault data calibration on all data within a fixed time interval, wherein the uncalibrated data is eliminated; the resampling is specifically to perform centralized summing and averaging of all data within a fixed time interval per minute, thereby obtaining resampled sample data; the general preprocessing in the data screening system is specifically to normalize the data after interval calibration and resampling, and perform feature correlation data analysis on the normalized data, and finally aggregate, expand and reduce the dimension of the data.

[0016] Preferably, the algorithm used for the normalization processing is specifically one or a combination of the maxminmap algorithm, the z-score algorithm or the quantile normalization algorithm; the method used for the data analysis of the feature correlation is specifically one or a combination of the Pearson correlation coefficient, the Spearman correlation rank, the chi-square test and related feature engineering operations.

[0017] Preferably, the data processing system specifically adopts one or a combination of filtering feature sampling extraction methods, encapsulation feature sampling extraction methods or mosaic feature sampling extraction methods when performing feature sampling extraction.

[0018] Preferably, the fault prediction model analysis system is a comprehensive feature model that combines qualitative analysis and quantitative analysis. When constructing the fault prediction model, the above-mentioned data after feature extraction is used as training data, and the data after qualitative analysis is used as test data to perform data testing on the established model, and after the test, the model performance evaluation result is compared with the qualitative analysis to obtain the model performance evaluation result. When there is no abnormality in the evaluation result, the model construction is completed. The construction of the fault prediction model in the model analysis system is specifically based on one or a combination of traditional classification models, logistic regression models, support vector models, K-nearest neighbor models, random forest models or integrated classification models.

[0019] Preferably, the model analysis system specifically adopts one or a combination of polynomial regression method, difference judgment method or decision tree learning method when performing predictive analysis on data.

[0020] Preferably, the intelligent adjustment system is specifically based on qualitative analysis and quantitative analysis databases and networked large databases when performing intelligent data adjustment and control, and the algorithm adopted by the intelligent adjustment system is specifically one or a combination of integrated learning algorithms, simulated annealing, genetic algorithms or neural networks.

[0021] Preferably, the result feedback system generates an adjustment time log simultaneously with the adjustment log, and the adjustment time log specifically adopts a clock model.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. After performing fault analysis on a wind turbine, the present invention uses an intelligent adjustment system to combine data from a qualitative analysis database with data from a quantitative analysis database and a networked large database for intelligent calculation to obtain an intelligent adjustment value for abnormal data, and generates an instruction to adjust relevant data through a single-chip microcomputer, thereby achieving pre-adjustment of the wind turbine, avoiding the occurrence of faults caused by untimely data adjustment, and at the same time eliminating the need for manual adjustment. After the adjustment, an adjustment log is generated as a database support, thereby forming intelligent adjustment, reducing the failure rate of the wind turbine, and increasing the service life of the wind turbine;

[0023] 2. When analyzing wind turbine faults, the present invention uses a data screening system to remove redundant data from SCADA data, thereby reducing the amount of data processed by the model and improving the model's working efficiency. At the same time, resampling is used to collect new data samples, thereby avoiding inaccurate model recognition caused by missing sample data, thereby ensuring the efficiency and accuracy of data processing by the fault analysis system.

[0024] 3. The present invention can reduce the management work of operation and maintenance personnel through intelligent adjustment, thereby improving operation and maintenance efficiency and providing convenience for centralized management. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0026] In the attached figure:

[0027] Figure 1 This is a block diagram of a wind turbine fault analysis system based on SCADA data modeling in the present invention;

[0028] Figure 2 This is a flow chart of the wind turbine fault analysis system of the present invention;

[0029] Figure 3 A diagram showing the steps for constructing a fault prediction model of the present invention;

[0030] Figure 4 This is a step diagram for feature sampling and extraction of the present invention. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0032] like Figure 1 As shown, the present invention provides a technical solution: a wind turbine fault analysis system based on SCADA data modeling, including a central processing unit, the central processing unit is used to transmit and store all data in the wind turbine fault analysis system;

[0033] Data monitoring system: The data monitoring system is used to monitor all data of wind turbines in real time. The data monitoring system operates based on the SCADA data system.

[0034] Data screening system: The data screening system is used to pre-process all data within a fixed time interval monitored by the above-mentioned data monitoring system. The pre-processing steps include interval calibration, resampling and general pre-processing;

[0035] Data processing system: The data processing system is used to perform feature sampling and extraction on the data processed by the data screening system. The feature sampling and extraction data specifically includes power deviation feature extraction, cabin internal and external environment deviation feature extraction, pitch deviation feature extraction, and time series feature extraction;

[0036] Model analysis system: The model analysis system is used to perform fault prediction model analysis on the data monitored by the data monitoring system and obtain fault prediction results;

[0037] Intelligent adjustment system: The intelligent adjustment system is used to analyze the fault prediction results obtained by the fault prediction model constructed in the above model analysis system, and combine the database data to perform intelligent data adjustment control and send it to the corresponding single-chip computer end;

[0038] Result feedback system: After the intelligent adjustment system sends the intelligent data adjustment control instruction, the result feedback system is used to perform a fault prediction model on the data that has been adjusted to have fault signs, compare it with the data before adjustment, obtain the comparison results, and generate an adjustment log for storage.

[0039] Among them, all the data of the wind turbines monitored by the data monitoring system include: wind speed, generator speed, power, wind angle, average wind direction angle, yaw position, yaw speed, ambient temperature, cabin temperature, ng51, 2, 3 temperatures and charger DC current, blade 1, 2, 3 speed, angle and temperature, X and Y direction acceleration. The data monitoring system uses wind speed sensor, power sensor, speed sensor, wind direction sensor, yaw sensor, speed sensor, temperature sensor, current sensor and acceleration sensor for data monitoring; interval calibration is to convert all the data within a fixed time interval into a fixed value. Normal data calibration and fault data calibration are performed on the data, and the uncalibrated data are eliminated; resampling is to centrally sum and average all data within a fixed time interval per minute to obtain resampled sample data. The general preprocessing in the data screening system is to normalize the data after interval calibration and resampling, and perform feature correlation data analysis on the normalized data, and finally aggregate, expand and reduce the dimension of the data; the algorithm used for normalization is the maxminmap algorithm, z-score algorithm or quantile algorithm. One or a combination of normalization algorithms; the method used for data analysis of feature correlation is specifically one or a combination of Pearson correlation coefficient, Spearman correlation rank, chi-square test and related feature engineering operations; the data processing system specifically adopts one or a combination of filtering feature sampling extraction method, encapsulation feature sampling extraction method or mosaic feature sampling extraction method when performing feature sampling extraction; the fault prediction model analysis system is a comprehensive feature model that combines qualitative analysis and quantitative analysis. When constructing the fault prediction model, the above-mentioned feature-extracted data is used as training data, and the data after qualitative analysis is used as test data to test the established model, and after the test, it is compared with the qualitative analysis to obtain the model performance evaluation result. When there is no abnormality in the evaluation result, the model construction is completed. The construction of the fault prediction model in the model analysis system is specifically based on one or a combination of traditional classification models, logistic regression models, support vector models, K-nearest neighbor models, random forest models or integrated classification models; the model analysis system specifically adopts one or a combination of polynomial regression method, difference judgment method or decision tree learning method when performing predictive analysis on data; the intelligent adjustment system is specifically based on qualitative analysis and quantitative analysis databases and networked large databases when performing intelligent data adjustment and control, and the algorithm adopted by the intelligent adjustment system is specifically one or a combination of integrated learning algorithms, simulated annealing, genetic algorithms or neural networks; the result feedback system generates an adjustment time log at the same time as generating an adjustment log, and the adjustment time log specifically adopts a clock model.

[0040] Example 1

[0041] like Figure 2 As shown in the figure, the use process of the wind turbine fault analysis system is as follows:

[0042] ①: The data monitoring system monitors the wind speed, generator speed, power, wind angle, average wind direction angle, yaw position, yaw speed, ambient temperature, cabin temperature, ng51, 2, 3 temperatures, charger DC current, blade 1, 2, 3 speed, angle and temperature, and X and Y acceleration data of the wind turbine in real time, and sends the real-time monitoring data to the data screening system at a fixed period of 5-10 minutes;

[0043] ②: The data screening system calibrates the normal data and fault data of the real-time monitoring data, and centrally sums and averages all the calibration data every minute. Finally, it normalizes the interval-calibrated and resampled data, analyzes the feature correlation of the normalized data, and finally aggregates, expands, and reduces the dimensionality of the data.

[0044] ③: The data processing system extracts power deviation features, cabin internal and external environment deviation features, pitch deviation features, and time series features from the data processed by the above data screening system, and analyzes the extracted data through the fault prediction model in the model analysis system to obtain fault prediction results;

[0045] ④: The intelligent adjustment system performs intelligent data adjustment and control based on the above fault prediction results and in combination with the database data, and sends it to the corresponding single-chip microcomputer end. The single-chip microcomputer adjusts the relevant data of the wind turbine. After the adjustment, the data monitoring system is used again to transmit, filter, process and analyze the data model to obtain the results, and compare them with the data before adjustment to obtain the comparison results, and generate an adjustment log for storage.

[0046] By using the above method, when performing fault analysis on a wind turbine, redundant data in the SCADA data is eliminated by a data screening system, thereby reducing the amount of data processed by the model and improving the working efficiency of the model. At the same time, resampling is used to collect new samples of data, thereby avoiding the occurrence of inaccurate model recognition caused by missing sample data, thereby ensuring the efficiency and accuracy of data processing by the fault analysis system. After the fault analysis of the wind turbine is performed, the intelligent adjustment system combines the data of the qualitative analysis and quantitative analysis databases and the networked large database to perform intelligent calculations to obtain intelligent adjustment values ​​for abnormal data, and generates instructions for adjusting relevant data through a single-chip microcomputer, thereby achieving pre-adjustment of the wind turbine, avoiding the occurrence of faults caused by untimely data adjustment, and at the same time eliminating the need for manual adjustment. After the adjustment, an adjustment log will be generated as database support, thereby forming intelligent adjustment, reducing the failure rate of the wind turbine, and increasing the service life of the wind turbine. Through intelligent adjustment, the management work of operation and maintenance personnel can be reduced, thereby improving operation and maintenance efficiency and providing convenience for centralized management.

[0047] Example 2

[0048] like Figure 3 As shown in Figure 2, the steps for building a fault prediction model are as follows:

[0049] S1: Data processing: After interval calibration, resampling and general preprocessing, the SCADA data is used as training data for power deviation feature extraction, cabin internal and external environment deviation feature extraction, pitch deviation feature extraction and time series feature extraction;

[0050] S2: Data training: After sampling and extracting the feature data in step S1 above, model training is performed, and the model is adjusted according to the actual situation and the algorithm is learned by the machine to obtain the model;

[0051] S3: Model verification: After performing qualitative and quantitative analysis on 20-30 groups of SCADA data, perform model verification on them using the model from step S2 above. Compare the obtained results with the qualitative and quantitative analysis results. If the difference is within 1-2 groups, the evaluation is completed. If the difference is within 2 groups or more, continue to adjust the model until it is completed.

[0052] Example 3

[0053] like Figure 4 As shown, the steps of feature sampling and extraction are as follows:

[0054] ①: Perform sample training on the data;

[0055] ②: Get feature attributes and category attributes;

[0056] ③: Calculate the mutual information between feature attributes and category attributes, and arrange the features according to the size of the mutual information to obtain feature subsets and complete feature extraction.

[0057] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0058] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A wind turbine fault analysis system based on SCADA data modeling, characterized by: It includes a central processing unit, which is used to transmit and store all data in the wind turbine fault analysis system; Data monitoring system: The data monitoring system is used to monitor all data of the wind turbine in real time. The data monitoring system is based on the SCADA data system. Data screening system: The data screening system is used to pre-process all data within a fixed time interval monitored by the above-mentioned data monitoring system. The pre-processing steps specifically include interval calibration, resampling and general pre-processing; Data processing system: The data processing system is used to perform feature sampling and extraction on the data processed by the above-mentioned data screening system. The feature sampling and extraction data specifically includes power deviation feature extraction, cabin internal and external environment deviation feature extraction, pitch deviation feature extraction, and time series feature extraction; Model analysis system: The model analysis system is used to perform fault prediction model analysis on the data monitored by the data monitoring system and obtain fault prediction results; The model analysis system is a comprehensive feature model that combines qualitative analysis and quantitative analysis. When constructing a fault prediction model, the data extracted from the above features is used as training data, and the data after qualitative analysis is used as test data to perform data testing on the established model. After the test, the model performance evaluation result is compared with the qualitative analysis to obtain a model performance evaluation result. When there is no abnormality in the evaluation result, the model construction is completed. The construction of the fault prediction model in the model analysis system is specifically based on one or a combination of traditional classification models, logistic regression models, support vector models, K-nearest neighbor models, random forest models or integrated classification models; Intelligent adjustment system: The intelligent adjustment system is used to analyze the fault prediction results obtained by the fault prediction model constructed in the above-mentioned model analysis system, and combine the database data to perform intelligent data adjustment control and send it to the corresponding single-chip computer end; Result feedback system: After the intelligent adjustment system sends the intelligent data adjustment control instruction, the result feedback system is used to perform a fault prediction model on the data that has been adjusted to have fault signs, compare it with the data before adjustment, obtain the comparison result, and generate an adjustment log for storage.

2. A wind turbine fault analysis system based on SCADA data modeling according to claim 1, characterized in that: All data of the wind turbine generator set monitored by the data monitoring system specifically include: wind speed, generator speed, power, wind angle, average wind direction angle, yaw position, yaw speed, ambient temperature, cabin temperature, blade temperature, charger DC current, blade rotation speed, angle and temperature and blade acceleration in the X and Y directions. The data monitoring system specifically uses wind speed sensors, power sensors, speed sensors, wind direction sensors, yaw sensors, speed sensors, temperature sensors, current sensors and acceleration sensors when performing data monitoring.

3. The wind turbine fault analysis system based on SCADA data modeling according to claim 1, characterized in that: The interval calibration is specifically to perform normal data calibration and fault data calibration on all data within a fixed time interval, wherein the uncalibrated data is eliminated; the resampling is specifically to perform centralized summing and averaging of all data within a fixed time interval per minute, thereby obtaining resampled sample data; the general preprocessing in the data screening system is specifically to normalize the data after interval calibration and resampling, and perform feature correlation data analysis on the normalized data, and finally aggregate, expand and reduce the dimension of the data.

4. The wind turbine fault analysis system based on SCADA data modeling according to claim 3, characterized in that: The algorithm used in the normalization processing is specifically one or a combination of the maxminmap algorithm, the z-score algorithm or the quantilenormalization algorithm; the method used in the data analysis of the feature correlation is specifically one or a combination of the Pearson correlation coefficient, the Spearman correlation rank, the chi-square test and related feature engineering operations.

5. The wind turbine fault analysis system based on SCADA data modeling according to claim 1, characterized in that: The data processing system specifically adopts one or a combination of filtering feature sampling extraction methods, encapsulation feature sampling extraction methods or mosaic feature sampling extraction methods when performing feature sampling extraction.

6. The wind turbine fault analysis system based on SCADA data modeling according to claim 1, characterized in that: The model analysis system specifically adopts one or a combination of polynomial regression method, difference judgment method or decision tree learning method when performing predictive analysis on data.

7. The wind turbine fault analysis system based on SCADA data modeling according to claim 1, characterized in that: The intelligent adjustment system is specifically based on qualitative analysis and quantitative analysis databases and networked large databases when performing intelligent data adjustment and control, and the algorithm used by the intelligent adjustment system is specifically one or a combination of integrated learning algorithms, simulated annealing, genetic algorithms or neural networks.

8. The wind turbine fault analysis system based on SCADA data modeling according to claim 1, characterized in that: The result feedback system generates an adjustment time log when generating an adjustment log, and the adjustment time log specifically adopts a clock model.

Images