CN117744859A - Wind turbine generator fault early warning method based on self-adaptive double control strategy - Google Patents
Wind turbine generator fault early warning method based on self-adaptive double control strategy Download PDFInfo
- Publication number
- CN117744859A CN117744859A CN202311662364.1A CN202311662364A CN117744859A CN 117744859 A CN117744859 A CN 117744859A CN 202311662364 A CN202311662364 A CN 202311662364A CN 117744859 A CN117744859 A CN 117744859A
- Authority
- CN
- China
- Prior art keywords
- data
- early warning
- training
- wind turbine
- control strategy
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000011217 control strategy Methods 0.000 title claims abstract description 52
- 238000012549 training Methods 0.000 claims abstract description 62
- 238000012544 monitoring process Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000003044 adaptive effect Effects 0.000 claims description 20
- 238000004422 calculation algorithm Methods 0.000 claims description 20
- 230000009977 dual effect Effects 0.000 claims description 16
- 238000012795 verification Methods 0.000 claims description 15
- 238000012360 testing method Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 238000009826 distribution Methods 0.000 claims description 6
- 238000007689 inspection Methods 0.000 claims description 5
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 description 8
- 238000011161 development Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
Landscapes
- Wind Motors (AREA)
Abstract
The invention relates to the technical field of fault early warning, in particular to a wind turbine generator fault early warning method based on a self-adaptive double control strategy, which comprises the following steps: s1, acquiring historical operation data of a wind turbine generator by utilizing a data acquisition and monitoring control system, and preprocessing the historical operation data; s2, importing the data set into a training model, setting the input and output of the training model, performing model training, and storing the training model as a class gradient lifting reference model after training; s3, searching an optimal super-parameter variable of the class gradient lifting reference model by utilizing random search; s4, calculating residual errors between the predicted value and the actual value, and then constructing a self-adaptive double-control strategy to dynamically judge the variation trend of the residual errors; s5, judging the fault point through the self-adaptive double control strategy. The method can be used for real-time early warning and alarming of the offshore wind turbine group and the land wind turbine group, and has the advantages of robustness, universality, accuracy and high efficiency in fault early warning.
Description
Technical Field
The invention relates to the technical field of fault early warning, in particular to a wind turbine generator fault early warning method based on a self-adaptive double control strategy.
Background
Wind energy is one of the renewable energy sources with the largest development and commercial development prospects at present, and is also the renewable energy source with the fastest development speed. However, the working environment of the wind turbine generator is very severe, is usually located in remote suburban areas and severe coastal areas, and runs around the clock under alternating load, so that faults occur frequently, and huge maintenance cost and economic loss are caused.
The most important part in the wind turbine generator is a generator, so that aiming at intelligent operation and maintenance of the wind turbine generator, data acquisition and monitoring control system (Supervisory Control And Data Acquisition, SCADA) data are utilized for processing and analysis, most algorithm early warning is to perform early warning on related parts through statistics indexes or setting hard threshold indexes, the setting of the hard threshold needs experience setting, and the false alarm rate is high. Therefore, SCADA is analyzed through a big data algorithm, data characteristics are extracted, and the accuracy of the algorithm is increased.
The traditional random forest, GBDT and XGBoost algorithms require a large number of parameters to be adjusted, and the training time and cost are high. The required algorithm is continuous data of a wind turbine generator system to be processed in batches, the data size is large, the dimension is high, and the universality of the traditional algorithm is not high. Therefore, the early warning algorithm with strong adaptability, high stability and quick data processing is an indispensable component in the current wind farm health monitoring.
Disclosure of Invention
Based on the above, it is necessary to provide a wind turbine generator fault early warning method based on an adaptive double control strategy for the above technical problems.
The invention provides a wind turbine generator fault early warning method based on a self-adaptive double control strategy, which comprises the following steps:
s1, acquiring historical operation data of a wind turbine generator by utilizing a data acquisition and monitoring control system, preprocessing the historical operation data, and finally combining the historical operation data to form a new data set;
s2, importing the data set into a training model, setting the input and output of the training model, performing model training, and storing the training model as a class gradient lifting reference model after training;
s3, searching an optimal super-parameter variable of the class gradient lifting reference model by utilizing random search;
s4, calculating residual errors between the predicted value and the actual value, and then constructing a self-adaptive double-control strategy to dynamically judge the variation trend of the residual errors, so as to prevent early warning false alarm and early warning;
s5, preloading the class gradient lifting reference model and the predicted data in the process of executing new data prediction, judging fault points through a self-adaptive double control strategy, and outputting early warning information.
Further, preprocessing the historical operating data, and finally combining to form a new data set comprises the following steps:
s11, eliminating special point data in the historical operation data;
s12, extracting data features of historical operation data by using a Pearson correlation coefficient algorithm, selecting a data column with a correlation value greater than 0.5, and qualitatively deleting parameters of the data column;
s13, combining the finally reserved data columns to form a new data set.
Furthermore, the special point data comprise negative values, zero values, infinite values, null values and messy codes, and the qualitative deletion parameters comprise data of the blades 2 and 3 in the wind turbine generator.
Further, setting the input and output of the training model, performing model training, and storing the training model as a class gradient lifting reference model after training is completed, wherein the method comprises the following steps:
s21, converting the data format of the data set into the data set format, and carrying out data standardization processing on the data in the data set by utilizing a maximum and minimum normalization algorithm;
s22, the data set is represented by 7:2:1 is divided into a training set, a testing set and a verification set;
s23, setting a training set as an input end of a training model, setting the predicted temperature of a bearing at a driving end of a generator as an output end of the training model, selecting default basic super parameters, performing model training, and finally storing the model training as a class gradient lifting reference model after finishing the training.
Further, searching the optimal super-parameter variable of the class gradient lifting reference model by using random search comprises the following steps:
s31, selecting key parameters of the class gradient lifting reference model for setting and adjusting, selecting default values for other parameters, and checking fitting or underfilling before parameter adjustment;
s32, automatically searching optimal parameters by adopting a grid searching method, arranging and combining variables, traversing each combination, verifying various super parameter selections on a training set and a verification set in a cross verification mode, and selecting the super parameter with the minimum average error;
s33, after the super-parameters are selected, combining the training set and the verification set, and then retraining the class gradient lifting reference model to obtain a final class gradient lifting reference model;
s34, testing the class gradient lifting reference model by using the test set.
Further, the key parameters include the number of iterations, the minimum number of trees that the best model should have, and the tree depth.
Further, calculating residual errors between the predicted value and the actual value, reconstructing a self-adaptive double-control strategy, dynamically judging the variation trend of the residual errors, and preventing early warning false alarm and early warning, wherein the method comprises the following steps:
s41, importing the verification set into a class gradient lifting reference model to obtain a predicted value of the temperature of the bearing at the driving end of the generator, converting the predicted value into a dimension consistent with the temperature through an inverse normalization method, and performing residual calculation with an actual value of the temperature of the bearing at the driving end of the generator to obtain a residual curve;
s42, respectively solving a hard threshold and a soft threshold of the residual error, setting soft-hard double-layer threshold protection of the residual error value as a self-adaptive double-control strategy, and preventing early warning false alarm and early warning of the class gradient lifting reference model.
Further, the solving method of the hard threshold is to calculate the original data to obtain a standard deviation assuming that a group of residual data only contains random errors, then determine a judging section according to the preset probability, and if the errors exceed the judging section, judge that the data belongs to an abnormal value.
Further, the soft threshold solving method is to compare whether two sample data are from the same distribution, dynamically calculate the error distribution of each interval section, and the two sample data are represented by the actual temperature value and the predicted temperature value of the driving end of the generator.
Further, judging a fault point through a self-adaptive double control strategy, and outputting early warning information comprises the following steps:
s51, if the residual numerical value single point is larger than a soft threshold curve, observing the residual point position;
s52, if the residual numerical value single point is larger than the hard threshold curve, judging that the damage to the fault is large, and executing shutdown inspection of the wind turbine generator;
s53, if the residual numerical value single point is smaller than the soft threshold curve, judging that the time is normal;
s54, if the residual data single point is larger than the soft threshold curve and smaller than the hard threshold curve, judging that the fault is about to occur, and enabling the wind turbine generator to be always in an observed state in the alarm process.
The beneficial effects of the invention are as follows: by providing the self-adaptive double-control strategy wind turbine generator fault early warning method, redundancy of data can be prevented, and in the data cleaning process, data integration and dimension reduction are carried out on historical operation data by utilizing double-feature extraction; aiming at the problems of large data volume and high data dimension of SCADA monitoring acquisition, the selection of the basic parameter quantity is reduced by utilizing the characteristic of Catboost high efficiency, and the time cost problem caused by large data volume is well solved; aiming at alarm events, the method is divided into four parts of normal, early warning, time observation and shutdown inspection, and more reasonable reserves time for fault occurrence and controls economic cost generated by early shutdown; the method can be used for real-time early warning and alarming of the offshore wind turbine group and the land wind turbine group, and has the advantages of robustness, universality, accuracy and high efficiency in fault early warning.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention and do not constitute a limitation on the invention. In the drawings:
FIG. 1 is a flow chart of a wind turbine generator fault early warning method based on an adaptive dual control strategy according to an embodiment of the present invention;
FIG. 2 is a fault early warning flow chart of a wind turbine generator fault early warning method based on an adaptive dual control strategy according to an embodiment of the invention;
FIG. 3 is a flowchart of a temperature regression prediction algorithm of a wind turbine generator fault early warning method based on an adaptive double control strategy according to an embodiment of the invention;
FIG. 4 is a generator drive end bearing temperature prediction graph of a wind turbine generator fault early warning method based on an adaptive dual control strategy according to an embodiment of the invention;
FIG. 5 is one of the pre-warning effect graphs obtained by the adaptive dual control strategy of the wind turbine generator fault pre-warning method based on the adaptive dual control strategy according to the embodiment of the invention;
FIG. 6 is a second graph of early warning effects obtained by an adaptive dual control strategy of a wind turbine generator fault early warning method based on the adaptive dual control strategy according to an embodiment of the present invention;
FIG. 7 is a third graph of pre-warning effects obtained by an adaptive dual control strategy of a wind turbine generator fault pre-warning method based on the adaptive dual control strategy according to an embodiment of the present invention;
FIG. 8 is a fourth graph of pre-warning effects obtained by an adaptive dual control strategy of a wind turbine generator fault pre-warning method based on the adaptive dual control strategy according to an embodiment of the present invention;
FIG. 9 is a fifth graph of early warning effects obtained by an adaptive dual control strategy of a wind turbine generator fault early warning method based on the adaptive dual control strategy according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Referring to fig. 1-3, a wind turbine generator fault early warning method based on a self-adaptive double control strategy is provided, and the method comprises the following steps:
s1, acquiring historical operation data of the wind turbine generator by utilizing a data acquisition and monitoring control system, preprocessing the historical operation data, and finally combining the historical operation data to form a new data set.
A data acquisition and supervisory control system (Supervisory Control And Data Acquisition, SCADA) is a system for monitoring and controlling industrial processes in real time.
SCADA systems typically contain the following key components: remote Terminal Units (RTUs) or PLCs, sensors and actuators, communication networks, SCADA master stations, human-machine interfaces (HMI), database systems, alarm systems and remote access, SCADA systems are widely used in industry, energy, utilities and other fields, which provide the capability of real-time monitoring, fault diagnosis and remote control of industrial processes, contributing to improved production efficiency and system reliability.
In the description of the present invention, preprocessing the historical operating data and finally combining to form a new data set includes the steps of:
s11, eliminating special point data in the historical operation data.
Special point data includes negative values, zero values, infinite values, null values, and scrambling codes.
S12, extracting data features of the historical operation data by using a Pearson correlation coefficient algorithm, selecting a data column with a correlation value larger than 0.5, and qualitatively deleting parameters of the data column.
The pearson correlation coefficient algorithm is used for extracting the data characteristics, carrying out correlation analysis on the data, extracting the data columns with the correlation value larger than 0.5, carrying out qualitative deletion parameters on the extracted data columns, and being beneficial to reducing the data redundancy.
Among the deleted data are: data for blade 2 and blade 3, as the three blades are all identical with respect to the intrinsic function of the generator.
S13, combining the finally reserved data columns to form a new data set.
S2, importing the data set into a training model, setting the input and output of the training model, performing model training, and storing the training model as a class gradient lifting (Catboost) reference model after training.
In the description of the invention, setting the input and output of the training model, performing model training, and saving the training model as a class gradient lifting reference model after the training is finished, wherein the method comprises the following steps of:
s21, converting the data format of the data set into a data set (Dataset) format, and carrying out data standardization processing on the data in the data set by utilizing a maximum and minimum normalization algorithm.
Selecting a function MinMaxScaler ()' integrated by a maximum and minimum normalization algorithm in a sklearn library (the library is a machine learning algorithm library); where the maximum-minimum normalization algorithm scales the features between a given minimum and maximum value, the maximum absolute value of each feature may also be converted to a unit size. The method is to linearly transform the original data, and to normalize the data to the middle of 0,1, so that the regularity of abnormal value data points can be ensured to be displayed in a certain range.
S22, the data set is represented by 7:2: the ratio of 1 is divided into a training set, a test set and a verification set.
S23, setting a training set as an input end of a training model, setting the predicted temperature of a bearing at a driving end of a generator as an output end of the training model, selecting default basic super parameters, performing model training, and finally storing the model training as a class gradient lifting (Catboost) reference model after the training is finished.
CatBOOST is a machine learning framework based on gradient lifting algorithms, specifically designed to handle class features, which can achieve good performance in classification and regression problems, and which also performs quite well when dealing with large-scale datasets, as: the category characteristics can be automatically processed, independent thermal coding or other complicated preprocessing steps are not needed, and the method is more convenient in processing the actual data set; the method realizes an algorithm based on gradient lifting, can effectively learn nonlinear relations and complex modes, and has better performance, particularly on a large-scale data set; the missing values can be automatically processed without additional processing of a user, so that model training is easier and more robust; in addition, the Catboost adopts an effective regularization technology, so that overfitting is prevented, and the generalization capability of the model is improved.
And S3, searching an optimal super-parameter variable of the class gradient lifting reference model by utilizing random search.
In the description of the present invention, searching for the optimal hyper-parametric variables of the class gradient lifting reference model using random search comprises the steps of:
s31, selecting key parameters of the class gradient lifting reference model for setting and adjusting, selecting default values for the rest parameters, and checking fitting or underfilling before parameter adjustment.
Key parameters include the number of iterations (iterations), the minimum number of trees (best model min trees) that the best model should have, and the tree depth (depth).
S32, automatically searching optimal parameters by adopting a grid search method, arranging and combining variables, traversing each combination, verifying various super parameter selections on a training set and a verification set in a cross verification mode, and selecting the super parameter with the minimum average error.
And S33, after the super-parameters are selected, combining the training set and the verification set, and then retraining the class gradient lifting reference model to obtain a final class gradient lifting reference model.
S34, testing the class gradient lifting reference model by using the test set.
S4, calculating residual errors between the predicted value and the actual value, and then constructing a self-adaptive double-control strategy to dynamically judge the variation trend of the residual errors, so as to prevent early warning false alarm and early warning.
The cross-validation parameter cv is selected to be 3, the model evaluation standard is selected to be r2, and other default values are selected.
In the description of the invention, residual errors between predicted values and actual values are calculated, a self-adaptive double-control strategy is built again, the change trend of the residual errors is dynamically judged, and the early warning false alarm and early warning prevention comprises the following steps:
s41, importing the verification set into a class gradient lifting reference model to obtain a predicted value of the temperature of the bearing at the driving end of the generator, converting the predicted value into a dimension consistent with the temperature through an inverse normalization method, and performing residual calculation with an actual value of the temperature of the bearing at the driving end of the generator to obtain a residual curve.
S42, respectively solving a hard threshold (3 sigama hard threshold) and a soft threshold (K-S soft threshold) of the residual error, setting soft and hard double-layer threshold protection of the residual error value as a self-adaptive double-control strategy, and preventing early warning false alarm and early warning of the class gradient lifting reference model.
The solving method of the hard threshold value is to calculate the original data to obtain standard deviation assuming that a group of residual data only contains random errors, then determine a judging section according to the preset probability, and judge that the data belongs to an abnormal value if the error exceeds the judging section.
The method for solving the soft threshold value is to compare whether two sample data come from the same distribution, dynamically calculate the error distribution of each interval section, and the two sample data sub-tables represent the actual value and the predicted value of the temperature of the driving end of the generator.
The generation of early warning false alarm phenomenon of the model is prevented through soft and hard double-layer threshold protection, so that the designed double-control strategy greatly improves the accuracy of the alarm.
S5, preloading the class gradient lifting reference model and the predicted data in the process of executing new data prediction, judging fault points through a self-adaptive double control strategy, and outputting early warning information.
In the description of the invention, the fault point is judged through the self-adaptive double control strategy, and the output of the early warning information comprises the following steps:
and S51, if the residual numerical value single point is larger than the soft threshold curve, observing the residual point position.
And S52, if the residual numerical value single point is larger than the hard threshold curve, judging that the damage to the fault is large, and executing shutdown inspection of the wind turbine generator.
And S53, if the residual numerical value single point is smaller than the soft threshold curve, judging that the time is normal.
S54, if the residual data single point is larger than the soft threshold curve and smaller than the hard threshold curve, judging that the fault is about to occur, and enabling the wind turbine generator to be always in an observed state in the alarm process.
As shown in fig. 4, the temperature prediction effect diagram of the bearing at the generator driving end of the wind turbine generator is obtained by automatically searching the optimal parameters by using gridSearchCV, searching the optimal solution for predicting the data set, and using RMS and R2 as evaluation indexes, such as the prediction effect shown in fig. 4.
As shown in fig. 5-9, the early warning effect diagram obtained by using the self-adaptive double control strategy provided by the invention can clearly show that the invention can early warn in advance and prevent false alarm.
In summary, by means of the technical scheme, the method for early warning the faults of the wind turbine generator by providing the self-adaptive double-control strategy can prevent redundancy of data, and in the process of data cleaning, data integration and dimension reduction are performed on historical operation data by utilizing double-feature extraction; aiming at the problems of large data volume and high data dimension of SCADA monitoring acquisition, the selection of the basic parameter quantity is reduced by utilizing the characteristic of Catboost high efficiency, and the time cost problem caused by large data volume is well solved; aiming at alarm events, the method is divided into four parts of normal, early warning, time observation and shutdown inspection, and more reasonable reserves time for fault occurrence and controls economic cost generated by early shutdown; the method can be used for real-time early warning and alarming of the offshore wind turbine group and the land wind turbine group, and has the advantages of robustness, universality, accuracy and high efficiency in fault early warning.
It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited in order and may be performed in other orders, unless explicitly stated herein. Moreover, at least some of the steps in the flowcharts of the figures may include a plurality of sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, the order of their execution not necessarily being sequential, but may be performed in turn or alternately with other steps or at least a portion of the other steps or stages.
Claims (10)
1. The wind turbine generator fault early warning method based on the self-adaptive double control strategy is characterized by comprising the following steps of:
s1, acquiring historical operation data of a wind turbine generator by utilizing a data acquisition and monitoring control system, preprocessing the historical operation data, and finally combining the historical operation data to form a new data set;
s2, importing the data set into a training model, setting the input and output of the training model, performing model training, and storing the training model as a class gradient lifting reference model after training;
s3, searching for the optimal super-parameter variable of the class gradient lifting reference model by utilizing random search;
s4, calculating residual errors between the predicted value and the actual value, and constructing a self-adaptive double-control strategy to dynamically judge the variation trend of the residual errors so as to prevent early warning false alarm and early warning;
and S5, preloading the class gradient lifting reference model and the predicted data in the process of executing new data prediction, judging fault points through a self-adaptive double control strategy, and outputting early warning information.
2. The method for early warning of generator faults of a wind turbine generator based on an adaptive dual control strategy according to claim 1, wherein the preprocessing of the historical operating data and the final combination to form a new data set comprises the following steps:
s11, eliminating special point data in the historical operation data;
s12, extracting data features of the historical operation data by using a Pearson correlation coefficient algorithm, selecting a data column with a correlation value greater than 0.5, and qualitatively deleting parameters of the data column;
and S13, combining the finally reserved data columns to form a new data set.
3. The method for early warning of generator faults of a wind turbine generator based on a self-adaptive double control strategy according to claim 2, wherein the special point data comprise negative values, zero values, infinity values, null values and messy codes, and the qualitative deleting parameters comprise data of blades 2 and 3 in the wind turbine generator.
4. The method for early warning of generator faults of a wind turbine generator based on a self-adaptive double control strategy according to claim 1, wherein the steps of setting input and output of the training model, performing model training, and saving the training model as a class gradient lifting reference model after training are carried out include the following steps:
s21, converting the data format of the data set into a data set format, and carrying out data standardization processing on the data in the data set by utilizing a maximum and minimum normalization algorithm;
s22, mixing the data set with 7:2:1 is divided into a training set, a testing set and a verification set;
s23, setting the training set as an input end of the training model, setting the predicted temperature of the bearing at the driving end of the generator as an output end of the training model, selecting default basic super parameters, performing model training, and finally storing the model training as a class gradient lifting reference model after the training is finished.
5. The method for early warning of generator faults of a wind turbine generator based on an adaptive dual control strategy according to claim 1, wherein the searching for the optimal super-parameter variable of the class gradient lifting reference model by using random search comprises the following steps:
s31, selecting key parameters of the class gradient lifting reference model for setting and adjusting, selecting default values for other parameters, and checking fitting or under fitting before parameter adjustment;
s32, automatically searching optimal parameters by adopting a grid searching method, arranging and combining variables, traversing each combination, verifying various super-parameter selections on the training set and the verification set in a cross verification mode, and selecting the super-parameter with the minimum average error;
s33, after the super-parameters are selected, combining the training set and the verification set, and training the class gradient lifting reference model again to obtain a final class gradient lifting reference model;
s34, testing the class gradient lifting reference model by using the test set.
6. The method for early warning of generator faults of a wind turbine generator based on an adaptive dual control strategy according to claim 5, wherein the key parameters comprise the iteration number, the minimum number of trees that an optimal model should have and the tree depth.
7. The method for early warning of generator faults of a wind turbine generator based on a self-adaptive double control strategy according to claim 1, wherein the steps of calculating residual errors between predicted values and actual values, reconstructing the self-adaptive double control strategy, dynamically judging the variation trend of the residual errors, and preventing early warning false alarm and early warning comprise the following steps:
s41, importing the verification set into the class gradient lifting reference model to obtain a predicted value of the temperature of the bearing at the driving end of the generator, converting the predicted value into a dimension consistent with the temperature through an inverse normalization method, and performing residual calculation with an actual value of the temperature of the bearing at the driving end of the generator to obtain a residual curve;
s42, respectively solving a hard threshold and a soft threshold of the residual error, setting soft-hard double-layer threshold protection of the residual error value as a self-adaptive double-control strategy, and preventing early warning false alarm and early warning of the class gradient lifting reference model.
8. The method for early warning of generator faults of a wind turbine generator based on a self-adaptive double control strategy according to claim 7, wherein the solving method of the hard threshold is characterized in that a group of residual data only contains random errors, standard deviation is obtained by calculation processing of original data, then a judging interval is determined according to preset probability, and if the errors exceed the judging interval, the data is judged to belong to an abnormal value.
9. The method for early warning of generator faults of a wind turbine generator based on the self-adaptive double control strategy according to claim 8, wherein the solution method of the soft threshold is to compare whether two sample data are from the same distribution, dynamically calculate error distribution of each interval section, and the two sample data sub-tables represent actual values and predicted values of temperature of a driving end of the generator.
10. The method for early warning of generator faults of a wind turbine generator based on an adaptive double control strategy according to claim 7, wherein the judging of the fault point by the adaptive double control strategy and outputting early warning information comprise the following steps:
s51, if the residual numerical value single point is larger than a soft threshold curve, observing the residual point position;
s52, if the residual numerical value single point is larger than a hard threshold curve, judging that the damage to the fault is large, and executing shutdown inspection of the wind turbine generator;
s53, if the residual numerical value single point is smaller than a soft threshold curve, judging that the time is normal;
and S54, if the residual data single point is larger than the soft threshold curve and smaller than the hard threshold curve, judging that the fault is about to occur, and keeping the wind turbine generator in an observed state in the alarm process.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311662364.1A CN117744859A (en) | 2023-12-06 | 2023-12-06 | Wind turbine generator fault early warning method based on self-adaptive double control strategy |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311662364.1A CN117744859A (en) | 2023-12-06 | 2023-12-06 | Wind turbine generator fault early warning method based on self-adaptive double control strategy |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117744859A true CN117744859A (en) | 2024-03-22 |
Family
ID=90257140
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311662364.1A Pending CN117744859A (en) | 2023-12-06 | 2023-12-06 | Wind turbine generator fault early warning method based on self-adaptive double control strategy |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117744859A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117972547A (en) * | 2024-03-26 | 2024-05-03 | 华电电力科学研究院有限公司 | Fault early warning method, device and medium for key components of wind turbine generator |
-
2023
- 2023-12-06 CN CN202311662364.1A patent/CN117744859A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117972547A (en) * | 2024-03-26 | 2024-05-03 | 华电电力科学研究院有限公司 | Fault early warning method, device and medium for key components of wind turbine generator |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111596604B (en) | Intelligent fault diagnosis and self-healing control system and method for engineering equipment based on digital twinning | |
CN103323772B (en) | Based on the running status of wind generator analytical approach of neural network model | |
CN106779200A (en) | Based on the Wind turbines trend prediction method for carrying out similarity in the historical data | |
CN111720296B (en) | Steam feed pump equipment abnormal state early warning method based on self-encoder | |
CN112629585A (en) | Equipment on-line monitoring method and device based on multi-dimensional parameter estimation | |
CN117744859A (en) | Wind turbine generator fault early warning method based on self-adaptive double control strategy | |
CN117290800B (en) | Timing sequence anomaly detection method and system based on hypergraph attention network | |
CN110209144B (en) | Two-layer real-time monitoring and alarm tracing method based on dynamic and static cooperative difference analysis | |
US11120350B2 (en) | Multilevel pattern monitoring method for industry processes | |
CN113570138A (en) | Method and device for predicting residual service life of equipment of time convolution network | |
CN113436194B (en) | Abnormity detection method, device and equipment for wind turbine generator | |
CN115294671A (en) | Air compressor outlet pressure prediction method and prediction system | |
CN115238573A (en) | Hydroelectric generating set performance degradation trend prediction method and system considering working condition parameters | |
CN115717590A (en) | Intelligent abnormity detection method for compressor and related device | |
Wang et al. | An adaptive condition monitoring method of wind turbines based on multivariate state estimation technique and continual learning | |
CN117216640A (en) | Power time sequence data anomaly detection method | |
CN117523793A (en) | Power plant equipment fault early warning method and computer equipment | |
CN117034169A (en) | Power grid main transformer equipment abnormal state prediction method based on time sequence causality network | |
CN116363843A (en) | Laboratory equipment early warning system | |
Velasquez et al. | Machine Learning Approach for Predictive Maintenance in Hydroelectric Power Plants | |
CN115774847A (en) | Diesel engine performance evaluation and prediction method and system | |
Li et al. | Research on digital twin and collaborative cloud and edge computing applied in operations and maintenance in wind turbines of wind power farm | |
Peng et al. | A temporal convolutional network based method for fault diagnosis of DEH system | |
Zhu et al. | Research of system fault diagnosis method based on imbalanced data | |
CN117152355B (en) | Visual supervision system based on factory data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |