CN117390573B - Wind turbine generator operation abnormality early warning method based on time sequence prediction - Google Patents
Wind turbine generator operation abnormality early warning method based on time sequence prediction Download PDFInfo
- Publication number
- CN117390573B CN117390573B CN202311697304.3A CN202311697304A CN117390573B CN 117390573 B CN117390573 B CN 117390573B CN 202311697304 A CN202311697304 A CN 202311697304A CN 117390573 B CN117390573 B CN 117390573B
- Authority
- CN
- China
- Prior art keywords
- state
- abnormal
- obtaining
- abnormal state
- neighborhood
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000005856 abnormality Effects 0.000 title claims abstract description 19
- 230000002159 abnormal effect Effects 0.000 claims abstract description 214
- 238000012544 monitoring process Methods 0.000 claims description 9
- 238000012512 characterization method Methods 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000008929 regeneration Effects 0.000 description 1
- 238000011069 regeneration method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Wind Motors (AREA)
Abstract
The invention relates to the technical field of data processing, in particular to a wind turbine generator operation abnormality early warning method based on time sequence prediction; obtaining a neighborhood state difference value according to difference characteristics between adjacent state moments in the running state time sequence; and obtaining an overall state difference value, a suspected abnormal state index and a suspected abnormal state point according to the distribution characteristics of the neighborhood state difference value in the preset fitting window. Obtaining the deviation degree and the neighborhood abnormal frequency according to the data difference characteristics in a preset fitting window where the suspected abnormal state points are located; and obtaining the anomaly confidence and the final anomaly state point according to the deviation degree and the neighborhood anomaly frequency. According to the method, the self-adaptive weight and the fitting prediction sequence are obtained according to the final abnormal state point and the weighted moving average algorithm, the running state of the wind turbine generator is monitored, and the accuracy of data fitting and state early warning is improved.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a wind turbine generator operation abnormality early warning method based on time sequence prediction.
Background
The wind turbine generator system is a power generation equipment system formed by a plurality of wind generators, utilizes wind energy to generate cleaning capability, and has the characteristics of environmental protection, regeneration and low emission; because the abnormal condition of the wind turbine generator during operation can influence the safety and reliability of the wind turbine generator, equipment faults which are not detected in time or found in time delay can cause a larger range of loss, and therefore, the operation data of the wind turbine generator are monitored in time very critical.
And (3) curve fitting is usually carried out on time sequence monitoring data containing noise data by using a time sequence fitting processing method, trend change characteristics which accord with the running state of the wind turbine are generated, and early warning is carried out on the running state of the wind turbine according to the trend change. The time series fitting process typically uses an existing weighted moving average algorithm, the main idea being to estimate future values using data over a period of time, smooth the time series data and capture long term trends and periodic changes. However, the algorithm is difficult to distinguish abnormal data and noise data in the fitting process, so that the characteristics of the abnormal data are weakened; the abnormal operation characteristics of the wind turbine generator are difficult to highlight, and the operation state early warning accuracy of the wind turbine generator is influenced.
Disclosure of Invention
In order to solve the technical problem that abnormal operation characteristics of a wind turbine generator are difficult to highlight by fitting the time sequence monitoring data through a weighted moving average algorithm, and the early warning accuracy is low, the invention aims to provide a wind turbine generator operation abnormal early warning method based on time sequence prediction, and the adopted technical scheme is as follows:
acquiring an operation state time sequence of a monitoring wind turbine generator; obtaining a neighborhood state difference value of the state moment according to the difference characteristic between the state moment and the adjacent state moment in the running state time sequence;
obtaining an overall state difference value according to the distribution characteristics of the neighborhood state difference value in a preset fitting window of the running state time sequence; obtaining a suspected abnormal state index and a suspected abnormal state point at the state moment according to the difference characteristics of the neighborhood state difference value and the integral state difference value;
obtaining the deviation degree of the suspected abnormal state points according to the data difference characteristics in the preset fitting window where the suspected abnormal state points are located; obtaining neighborhood abnormal frequencies of the suspected abnormal state points according to the number characteristics of other suspected abnormal state points in a preset fitting window where the suspected abnormal state points are located; obtaining an abnormal confidence level and a final abnormal state point according to the deviation degree of the suspected abnormal state point and the neighborhood abnormal frequency;
obtaining self-adaptive weights and fitting prediction sequences according to the final abnormal state points and a weighted moving average algorithm; and monitoring the running state of the wind turbine according to the fitting prediction sequence.
Further, the step of obtaining the neighborhood state difference value of the state time according to the difference characteristic between the state time and the adjacent state time in the running state time sequence includes:
and calculating the absolute value of the data difference between the state time and the adjacent state time to obtain the neighborhood state difference value of the state time.
Further, the step of obtaining the overall state difference value according to the distribution characteristics of the neighborhood state difference value in the preset fitting window of the running state time sequence includes:
and calculating the average value of the neighborhood state difference values in a preset fitting window of the running state time sequence to obtain the whole state difference value of the preset fitting window.
Further, the step of obtaining the suspected abnormal state index and the suspected abnormal state point at the state time according to the difference characteristics of the neighborhood state difference value and the overall state difference value includes:
calculating and inversely correlating the difference value of the neighborhood state difference value and the integral state difference value at any state moment in a preset fitting window to obtain a difference characterization value; calculating the difference value between a preset constant and the difference characterization value to obtain the suspected abnormal state index at any state moment; and when the suspected abnormal state index exceeds a preset threshold, the arbitrary state time is a suspected abnormal state point.
Further, the step of obtaining the deviation degree of the suspected abnormal state point according to the data difference feature in the preset fitting window where the suspected abnormal state point is located includes:
calculating the numerical average value of all state moments in a preset fitting window where the suspected abnormal state points are located, and obtaining a window representation value of the suspected abnormal state points; and calculating and normalizing the absolute value of the difference between the value of the suspected abnormal state point and the window representation value to obtain the deviation degree of the suspected abnormal state point.
Further, the step of obtaining the neighborhood abnormal frequency of the suspected abnormal state point according to the number features of other suspected abnormal state points in the preset fitting window where the suspected abnormal state point is located includes:
calculating the number ratio of other suspected abnormal state points in the preset fitting window where the suspected abnormal state points are located to the preset fitting window, and carrying out negative correlation mapping to obtain the neighborhood abnormal frequency of the suspected abnormal state points.
Further, the step of obtaining the anomaly confidence and the final anomaly state point according to the deviation degree of the suspected anomaly state point and the neighborhood anomaly frequency comprises the following steps:
calculating a product of a preset first coefficient and the deviation degree to obtain a first confidence coefficient; calculating the product of a preset second coefficient and the neighborhood abnormal frequency to obtain a second confidence coefficient; calculating the sum of the first confidence coefficient and the second confidence coefficient to obtain an abnormal confidence coefficient of the suspected abnormal state point; and when the abnormality confidence exceeds a preset abnormality threshold, the suspected abnormal state point is a final abnormal state point.
Further, the step of obtaining adaptive weights and fitting a predicted sequence according to the final outlier and weighted moving average algorithm comprises:
obtaining different fitting weights in a preset fitting window of the state moment according to a weighted moving average algorithm, and calculating the product of the fitting weight of the final abnormal state point in the preset fitting window and a preset first constant to obtain the self-adaptive weight of the final abnormal state point; and obtaining a fitting value of the state moment according to a weighted moving average algorithm, the fitting weight and the self-adaptive weight, and obtaining a fitting prediction sequence according to the fitting value of the state moment.
The invention has the following beneficial effects:
in the embodiment of the invention, the neighborhood state difference value of the state moment is obtained to represent the difference characteristic and the whole state difference value between the adjacent state moments, so that whether the state moment is a suspected abnormal state index and a suspected abnormal state point is determined through the neighborhood state difference value, and the accuracy of final data fitting is improved. The deviation degree of the suspected abnormal state points and the neighborhood abnormal frequency can be obtained to further determine whether the abnormal state points are abnormal data, determine the final abnormal state points, and improve the accuracy of final data fitting. The self-adaptive weight of the final abnormal state point can be obtained to further scale the abnormal data, and the characteristics of the abnormal data after the data fitting are highlighted. Finally, a fitting prediction sequence is obtained, so that the accuracy of data fitting is improved, and abnormal data characteristics are highlighted; and the running state of the wind turbine is monitored according to the fitting prediction sequence, so that the accuracy of early warning of the running state of the wind turbine is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a method for early warning of abnormal operation of a wind turbine generator based on time sequence prediction according to an embodiment of the present invention.
Detailed Description
In order to further explain the technical means and effects adopted by the invention to achieve the preset aim, the following is a detailed description of specific implementation, structure, characteristics and effects of the wind turbine generator running abnormality early warning method based on time sequence prediction according to the invention, which is provided by combining the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The specific scheme of the wind turbine generator system abnormal operation early warning method based on time sequence prediction provided by the invention is specifically described below with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of a method for early warning of abnormal operation of a wind turbine generator based on time sequence prediction according to an embodiment of the present invention is shown, and the method includes the following steps:
step S1, acquiring a time sequence for monitoring the running state of a wind turbine generator; and obtaining a neighborhood state difference value of the state moment according to the difference characteristic between the state moment and the adjacent state moment in the running state time sequence.
In the embodiment of the invention, the implementation scene is to perform abnormal early warning on the running state of the wind turbine generator; firstly, acquiring an operation state time sequence of a monitoring wind turbine, acquiring the instantaneous power of the wind turbine and constructing the operation state time sequence, wherein the operation state time sequence can reflect the change characteristics of the instantaneous power of the wind turbine, and an implementer can determine the acquisition frequency of the instantaneous power of the wind turbine according to implementation scenes. After the running state time sequence is obtained, the abnormal running state condition of the wind turbine generator can be pre-warned, the prior art mainly gives weight to the obtained data through a weighted moving average algorithm, the obtained data are summed to obtain a fitting value at the future moment, and the pre-warning is carried out according to the difference degree between the fitting value and the normal running data range. The weighted moving average algorithm is to give different weights according to the distances between the historical data in the sequence and the fitting time, and the weights given are smaller as the distances are far. Abnormal data and noise data exist in the running state time sequence acquired in the actual application process, and the purpose of smoothing the noise data is achieved by giving weight in the weighted summation process; however, the noise data is smoothed, and the characteristics of the abnormal data are weakened, so that the characteristics of the abnormal data are difficult to highlight in the fitting result, the abnormal condition in the period is difficult to highlight by the fitting value, the fitting value is inaccurate, and the early warning accuracy of the wind turbine generator is affected; therefore, the running state time sequence of the wind turbine generator needs to be improved through the process of a weighted moving average algorithm, and the accuracy of a fitting result is improved.
Firstly, in order to improve the characteristics of abnormal data in the fitting process through a weighted moving average algorithm, the positions of the abnormal data in the running state time sequence need to be preliminarily determined; under the condition that the running state of the wind turbine generator is normal, the variation amplitude of the instantaneous power at the adjacent moment is small, and the variation amplitude is large for noise data and abnormal data of running abnormality, so that the neighborhood state difference value of the state moment can be obtained based on the characteristic by the difference characteristic between the state moment and the adjacent state moment in the running state time sequence.
Preferably, in an embodiment of the present invention, obtaining the neighborhood state difference value includes: and calculating the absolute value of the data difference between the state time and the adjacent state time to obtain the neighborhood state difference value of the state time. The state time refers to any one of the data time representing the running state of the wind turbine generator set in the running state time sequence; in the embodiment of the invention, the neighborhood state difference value is calculated by the state time and the adjacent last state time, and when the neighborhood state difference value is larger, the larger the data difference between the state time and the adjacent state time is, the more abnormal data or noise data is likely to appear.
S2, obtaining an overall state difference value according to the distribution characteristics of the neighborhood state difference value in a preset fitting window of the running state time sequence; and obtaining a suspected abnormal state index and a suspected abnormal state point at the state moment according to the difference characteristics of the neighborhood state difference value and the overall state difference value.
Even if the data of the wind turbine generator at adjacent state moments in the normal operation process also have small changes, in order to analyze whether the neighborhood state difference value at the state moment is larger than the normal data changes, the overall state difference value needs to be obtained according to the distribution characteristics of the neighborhood state difference value in a preset fitting window of the operation state time sequence, and the method specifically comprises the following steps: calculating the average value of neighborhood state difference values in a preset fitting window of the running state time sequence to obtain the overall state difference value of the preset fitting window; the overall state difference value reflects overall data change characteristics between adjacent state moments within a period of time in running of the wind turbine. It should be noted that, the preset fitting window is a window for performing data fitting in a weighted moving average algorithm, and the fitting value is obtained by performing weighted summation on the data of the state moments in the window.
Further, when the neighborhood state difference value at a certain state moment is larger than the whole state difference value of the preset fitting window, the state moment is more likely to be abnormal data or noise data, so that the suspected abnormal state index and the suspected abnormal state point at the state moment can be obtained according to the difference characteristics of the neighborhood state difference value and the whole state difference value.
Preferably, in one embodiment of the present invention, acquiring the suspected abnormal state index and the suspected abnormal state point includes: calculating and inversely correlating the difference value of the neighborhood state difference value and the whole state difference value at any state moment in the preset fitting window for any state moment in the preset fitting window to obtain a difference characterization value; the larger the neighborhood state difference value at any state moment is, the smaller the difference characterization value is, and the more likely the neighborhood state difference value at any state moment is abnormal data or noise data. Calculating the difference value between a preset constant and a difference characterization value to obtain a suspected abnormal state index at any state moment; when the suspected abnormal state index exceeds a preset threshold, the arbitrary state time is a suspected abnormal state point; in the embodiment of the present invention, the preset constant is 1, the preset threshold is 0.75, and the practitioner can set the preset constant according to the implementation scenario, and the larger the suspected abnormal state index is, the more likely the abnormal data or noise data point is at any state moment. The formula for obtaining the suspected abnormal state index comprises the following steps:
in the method, in the process of the invention,a suspected abnormal state index indicating any state time in a preset fitting window, < >>Neighborhood state difference value representing the arbitrary state moment, < ->Representing the number of state moments in the preset fit window,/for>Representing the +.f. in the preset fit window>Neighborhood state difference value of individual state moments, +.>Representing the overall state difference value; />Represents an exponential function based on natural constants, < ->Representing a difference characterization value; />Representing a preset constant.
Step S3, obtaining the deviation degree of the suspected abnormal state points according to the data difference characteristics in the preset fitting window where the suspected abnormal state points are located; obtaining neighborhood abnormal frequencies of the suspected abnormal state points according to the number characteristics of other suspected abnormal state points in a preset fitting window where the suspected abnormal state points are located; and obtaining the abnormal confidence and the final abnormal state point according to the deviation degree of the suspected abnormal state point and the neighborhood abnormal frequency.
In step S2, a suspected abnormal state point in a preset fitting window at the state moment to be fitted is obtained, and because the suspected abnormal state point may be abnormal data or noise data, in order to improve the data fitting and early warning accuracy, further feature fractal is required for the suspected abnormal state point, and whether the suspected abnormal state point is abnormal data is distinguished.
Firstly, when abnormal data appear, the data difference between the abnormal data and adjacent state moments is large, even the data range of normal operation is exceeded, noise data tend to randomly fluctuate with small amplitude, and the fluctuation degree is kept in the normal fluctuation range; therefore, the deviation degree of the suspected abnormal state point can be obtained according to the data difference characteristic in the preset fitting window where the suspected abnormal state point is located, and the method specifically comprises the following steps: calculating the numerical average value of all state moments in a preset fitting window where the suspected abnormal state points are located, and obtaining a window representation value of the suspected abnormal state points; calculating and normalizing the absolute value of the difference between the value of the suspected abnormal state point and the window representation value to obtain the deviation degree of the suspected abnormal state point; the greater the degree of deviation, the more likely the suspected abnormal state point is abnormal data.
Further, for noise data, the occurrence frequency is higher, and for abnormal suspected situations, the occurrence frequency is usually in a random form, so that the neighborhood abnormal frequency of the suspected abnormal state point can be obtained according to the number characteristics of other suspected abnormal state points in a preset fitting window where the suspected abnormal state point is located, specifically including: calculating the ratio of other suspected abnormal state points in a preset fitting window where the suspected abnormal state points are located to the number of the preset fitting window, and carrying out negative correlation mapping to obtain the neighborhood abnormal frequency of the suspected abnormal state points, wherein when the number of the other suspected abnormal state points in the preset fitting window is larger, the occurrence frequency is larger and more likely to be noise data, and conversely, when the number of the other suspected abnormal state points is smaller, the occurrence frequency is smaller and more likely to be abnormal data, and the neighborhood abnormal frequency is larger.
After the deviation degree and the neighborhood abnormal frequency of the suspected abnormal state point are obtained, the abnormal confidence degree and the final abnormal state point can be obtained according to the deviation degree and the neighborhood abnormal frequency of the suspected abnormal state point; the method specifically comprises the following steps: calculating a product of a preset first coefficient and a deviation degree to obtain a first confidence coefficient; calculating the product of a preset second coefficient and the neighborhood abnormal frequency to obtain a second confidence coefficient; calculating the sum of the first confidence coefficient and the second confidence coefficient to obtain an abnormal confidence coefficient of the suspected abnormal state point; in the embodiment of the invention, the preset first coefficient and the preset second coefficient are 0.5, and an implementer can determine according to implementation scenes by himself, and when the abnormality confidence is larger, the suspected abnormal state point is more likely to be abnormal data. When the abnormality confidence exceeds a preset abnormality threshold, the suspected abnormality state point is a final abnormality state point, and in the embodiment of the invention, the preset abnormality threshold is 0.6, and an implementer can determine according to implementation scenes. The formula for obtaining the anomaly confidence comprises the following steps:
in the method, in the process of the invention,abnormality confidence indicating a suspected abnormal status point, +.>A value representing a point of suspected abnormal state, +.>Representing window representation values>Indicating the degree of deviation>Representing a normalization function->Representing the number ratio of other suspected abnormal state points in a preset fitting window where the suspected abnormal state points are located to the preset fitting window, and (I)>Represents an exponential function based on natural constants, < ->Representing a preset first coefficient,/->Representing a preset second coefficient,/->Representing a first confidence level->Representing a second confidence level.
S4, obtaining self-adaptive weights and fitting prediction sequences according to the final abnormal state points and a weighted moving average algorithm; and monitoring the running state of the wind turbine according to the fitting prediction sequence.
The method comprises the steps of determining a final abnormal state point in a preset fitting window of a state moment to be fitted, adaptively adjusting the weight of the final abnormal state point, highlighting abnormal characteristics represented by a fitting value, and improving early warning accuracy, so that an adaptive weight and a fitting prediction sequence are obtained according to the final abnormal state point and a weighted moving average algorithm, and specifically comprising the following steps: and obtaining fitting weights of different state moments in a preset fitting window of the state moments to be fitted according to a weighted moving average algorithm, wherein the weights of the state moments which are closer to the state moment to be fitted are larger. Calculating the product of the fitting weight of the final abnormal state point in the preset fitting window and a preset first constant to obtain the self-adaptive weight of the final abnormal state point; in the embodiment of the invention, when the value of the final abnormal state point is larger than the corresponding window representation value, a first constant is preset to be 1.5; when the value of the final abnormal state point is smaller than the corresponding window representation value, the preset first constant is 0.5. The self-adaptive weight changes the fitting weight of the final abnormal state point, and the numerical value of the final abnormal state point is further scaled, so that the abnormal characteristics of abnormal data are highlighted in the fitting process. And obtaining a fitting value of the state moment to be fitted according to a weighted moving average algorithm, the fitting weight and the self-adaptive weight, wherein the normal state point uses the self-fitting weight to participate in fitting, and finally the abnormal state point uses the self-adaptive weight to participate in fitting. The fitting prediction sequence is obtained according to the fitting value of the state moment, and it should be noted that the weighted moving average algorithm belongs to the prior art, and specific calculation steps are not repeated. The fitting value can characterize the characteristics of the abnormal data in a preset fitting window at the moment of the state to be fitted, so that the abnormal characteristics in the fitting value are highlighted,
further, after the fitting prediction sequence is obtained, the running state of the wind turbine generator can be monitored according to the fitting prediction sequence, and because the weighted moving average algorithm can fit the numerical value at the next moment in the future according to the historical data, if the fitting value at the next state moment in the fitting prediction sequence exceeds the preset early warning range, early warning of the running state is carried out, and the early warning accuracy is improved. The implementer can set the preset early warning range or other early warning conditions according to the implementation scene by himself according to the fitting prediction sequence, and the implementation method is not limited.
In summary, the embodiment of the invention provides a wind turbine generator system operation abnormality early warning method based on time sequence prediction; obtaining a neighborhood state difference value according to difference characteristics between adjacent state moments in the running state time sequence; and obtaining an overall state difference value, a suspected abnormal state index and a suspected abnormal state point according to the distribution characteristics of the neighborhood state difference value in the preset fitting window. Obtaining the deviation degree and the neighborhood abnormal frequency according to the data difference characteristics in a preset fitting window where the suspected abnormal state points are located; and obtaining the anomaly confidence and the final anomaly state point according to the deviation degree and the neighborhood anomaly frequency. According to the method, the self-adaptive weight and the fitting prediction sequence are obtained according to the final abnormal state point and the weighted moving average algorithm, the running state of the wind turbine generator is monitored, and the accuracy of data fitting and state early warning is improved.
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
Claims (7)
1. A wind turbine generator system operation abnormality early warning method based on time sequence prediction is characterized by comprising the following steps:
acquiring an operation state time sequence of a monitoring wind turbine generator; obtaining a neighborhood state difference value of the state moment according to the difference characteristic between the state moment and the adjacent state moment in the running state time sequence;
obtaining an overall state difference value according to the distribution characteristics of the neighborhood state difference value in a preset fitting window of the running state time sequence; obtaining a suspected abnormal state index and a suspected abnormal state point at the state moment according to the difference characteristics of the neighborhood state difference value and the integral state difference value;
obtaining the deviation degree of the suspected abnormal state points according to the data difference characteristics in the preset fitting window where the suspected abnormal state points are located; obtaining neighborhood abnormal frequencies of the suspected abnormal state points according to the number characteristics of other suspected abnormal state points in a preset fitting window where the suspected abnormal state points are located; obtaining an abnormal confidence level and a final abnormal state point according to the deviation degree of the suspected abnormal state point and the neighborhood abnormal frequency;
obtaining self-adaptive weights and fitting prediction sequences according to the final abnormal state points and a weighted moving average algorithm; monitoring the running state of the wind turbine according to the fitting prediction sequence;
the step of obtaining the neighborhood state difference value of the state moment according to the difference characteristic between the state moment and the adjacent state moment in the running state time sequence comprises the following steps:
and calculating the absolute value of the data difference between the state time and the adjacent state time to obtain the neighborhood state difference value of the state time.
2. The method for early warning of abnormal operation of a wind turbine generator system based on time sequence prediction according to claim 1, wherein the step of obtaining the overall state difference value according to the distribution characteristics of the neighborhood state difference value in a preset fitting window of an operation state time sequence comprises:
and calculating the average value of the neighborhood state difference values in a preset fitting window of the running state time sequence to obtain the whole state difference value of the preset fitting window.
3. The method for early warning of abnormal operation of a wind turbine generator set based on time sequence prediction according to claim 1, wherein the step of obtaining the suspected abnormal state index and the suspected abnormal state point at the state time according to the difference characteristics of the neighborhood state difference value and the overall state difference value comprises the following steps:
calculating and inversely correlating the difference value of the neighborhood state difference value and the integral state difference value at any state moment in a preset fitting window to obtain a difference characterization value; calculating the difference value between a preset constant and the difference characterization value to obtain the suspected abnormal state index at any state moment; and when the suspected abnormal state index exceeds a preset threshold, the arbitrary state time is a suspected abnormal state point.
4. The method for early warning of abnormal operation of a wind turbine generator system based on time sequence prediction according to claim 1, wherein the step of obtaining the deviation degree of the suspected abnormal state point according to the data difference characteristic in the preset fitting window where the suspected abnormal state point is located comprises the following steps:
calculating the numerical average value of all state moments in a preset fitting window where the suspected abnormal state points are located, and obtaining a window representation value of the suspected abnormal state points; and calculating and normalizing the absolute value of the difference between the value of the suspected abnormal state point and the window representation value to obtain the deviation degree of the suspected abnormal state point.
5. The method for early warning of abnormal operation of a wind turbine generator system based on time sequence prediction according to claim 1, wherein the step of obtaining the neighborhood abnormal frequency of the suspected abnormal state point according to the number features of other suspected abnormal state points in a preset fitting window where the suspected abnormal state point is located comprises the steps of:
calculating the number ratio of other suspected abnormal state points in the preset fitting window where the suspected abnormal state points are located to the preset fitting window, and carrying out negative correlation mapping to obtain the neighborhood abnormal frequency of the suspected abnormal state points.
6. The method for early warning of abnormal operation of a wind turbine generator system based on time sequence prediction according to claim 1, wherein the step of obtaining the abnormal confidence and final abnormal state point according to the deviation degree of the suspected abnormal state point and the neighborhood abnormal frequency comprises the steps of:
calculating a product of a preset first coefficient and the deviation degree to obtain a first confidence coefficient; calculating the product of a preset second coefficient and the neighborhood abnormal frequency to obtain a second confidence coefficient; calculating the sum of the first confidence coefficient and the second confidence coefficient to obtain an abnormal confidence coefficient of the suspected abnormal state point; and when the abnormality confidence exceeds a preset abnormality threshold, the suspected abnormal state point is a final abnormal state point.
7. The method for early warning of abnormal operation of a wind turbine generator based on time sequence prediction according to claim 1, wherein the step of obtaining adaptive weights and fitting a predicted sequence according to the final abnormal state point and a weighted moving average algorithm comprises:
obtaining different fitting weights in a preset fitting window of the state moment according to a weighted moving average algorithm, and calculating the product of the fitting weight of the final abnormal state point in the preset fitting window and a preset first constant to obtain the self-adaptive weight of the final abnormal state point; and obtaining a fitting value of the state moment according to a weighted moving average algorithm, the fitting weight and the self-adaptive weight, and obtaining a fitting prediction sequence according to the fitting value of the state moment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311697304.3A CN117390573B (en) | 2023-12-12 | 2023-12-12 | Wind turbine generator operation abnormality early warning method based on time sequence prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311697304.3A CN117390573B (en) | 2023-12-12 | 2023-12-12 | Wind turbine generator operation abnormality early warning method based on time sequence prediction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117390573A CN117390573A (en) | 2024-01-12 |
| CN117390573B true CN117390573B (en) | 2024-02-23 |
Family
ID=89441407
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311697304.3A Active CN117390573B (en) | 2023-12-12 | 2023-12-12 | Wind turbine generator operation abnormality early warning method based on time sequence prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117390573B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117760593B (en) * | 2024-02-22 | 2024-05-10 | 张家港江海粮油港务有限公司 | Abnormal monitoring method for storage temperature of grain crops |
| CN117786584B (en) * | 2024-02-27 | 2024-04-30 | 西安中创博远网络科技有限公司 | Big data analysis-based method and system for monitoring and early warning of water source pollution in animal husbandry |
| CN117851955B (en) * | 2024-03-07 | 2024-05-24 | 辽宁亿金电子有限公司 | Flexible-direct-current capacitor state detection method and system |
| CN117874685B (en) * | 2024-03-11 | 2024-05-24 | 天津市塘沽第一阀门有限公司 | Pressure data abnormality early warning method for direct-current variable-frequency electric half ball valve |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6405486B1 (en) * | 2018-04-04 | 2018-10-17 | 株式会社日立パワーソリューションズ | Wind power generation facility abnormality sign diagnosis system, management device, and abnormality sign diagnosis method |
| CN109118384A (en) * | 2018-07-16 | 2019-01-01 | 湖南优利泰克自动化系统有限公司 | A kind of Wind turbines healthy early warning method |
| CN113410911A (en) * | 2021-08-02 | 2021-09-17 | 广东电网有限责任公司中山供电局 | Working state monitoring method and system of power factor compensation device |
| CN116610482A (en) * | 2023-07-18 | 2023-08-18 | 山东理工大学 | Intelligent monitoring method for operation state of electrical equipment |
| CN116717434A (en) * | 2023-05-30 | 2023-09-08 | 武汉盈风能源科技有限公司 | Wind power generation system, early warning method of wind turbine generator system and readable storage medium |
| CN116771611A (en) * | 2023-06-07 | 2023-09-19 | 华能大理风力发电有限公司祥云分公司 | Wind turbine generator power monitoring method and device |
-
2023
- 2023-12-12 CN CN202311697304.3A patent/CN117390573B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6405486B1 (en) * | 2018-04-04 | 2018-10-17 | 株式会社日立パワーソリューションズ | Wind power generation facility abnormality sign diagnosis system, management device, and abnormality sign diagnosis method |
| CN109118384A (en) * | 2018-07-16 | 2019-01-01 | 湖南优利泰克自动化系统有限公司 | A kind of Wind turbines healthy early warning method |
| CN113410911A (en) * | 2021-08-02 | 2021-09-17 | 广东电网有限责任公司中山供电局 | Working state monitoring method and system of power factor compensation device |
| CN116717434A (en) * | 2023-05-30 | 2023-09-08 | 武汉盈风能源科技有限公司 | Wind power generation system, early warning method of wind turbine generator system and readable storage medium |
| CN116771611A (en) * | 2023-06-07 | 2023-09-19 | 华能大理风力发电有限公司祥云分公司 | Wind turbine generator power monitoring method and device |
| CN116610482A (en) * | 2023-07-18 | 2023-08-18 | 山东理工大学 | Intelligent monitoring method for operation state of electrical equipment |
Non-Patent Citations (1)
| Title |
|---|
| 基于SCADA数据的风电机组传动系统故障状态监测;李霸军等;《西北水电》;20180228(第1期);116-119页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117390573A (en) | 2024-01-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117390573B (en) | Wind turbine generator operation abnormality early warning method based on time sequence prediction | |
| CN116702081B (en) | Intelligent inspection method for power distribution equipment based on artificial intelligence | |
| CN108376298B (en) | Early warning and diagnosing method for temperature faults of engine of wind turbine generator | |
| CN110378042B (en) | Wind turbine generator gearbox oil temperature anomaly detection method and system based on SCADA data | |
| CN112629585A (en) | Equipment on-line monitoring method and device based on multi-dimensional parameter estimation | |
| CN108802525A (en) | Equipment fault intelligent Forecasting based on small sample | |
| CN117411189B (en) | Monitoring data enhancement method of micro-grid coordination controller | |
| CN105825289A (en) | Prediction method for wind power time sequence | |
| CA2300546A1 (en) | Method for assessing the reliability of technical systems | |
| JP2009044857A (en) | Apparatus and method for diagnosing stability of electric power system | |
| CN117235557B (en) | Electrical equipment fault rapid diagnosis method based on big data analysis | |
| CN113821938B (en) | Short-term prediction method and device for metering error state of mutual inductor | |
| CN117421610B (en) | Data anomaly analysis method for electric energy meter running state early warning | |
| CN114841580A (en) | Generator fault detection method based on hybrid attention mechanism | |
| CN114062993A (en) | CVT error state prediction method based on time convolution network | |
| CN111931354A (en) | Transformer top layer oil temperature prediction method based on gray-autoregressive differential moving average model | |
| CN114046869B (en) | Broadband oscillation information online monitoring method and system based on daily disturbance response of power system | |
| CN109635430B (en) | Power grid transmission line transient signal monitoring method and system | |
| CN114740303A (en) | Fault monitoring system of wireless passive high-voltage switch cabinet | |
| CN112801533B (en) | Power system operation reliability assessment method considering uncertainty of decision dependence | |
| CN116894151B (en) | Intelligent power grid abnormal data cleaning method based on artificial intelligence | |
| CN117365869A (en) | Self-adaptive early warning strategy design method for wind turbine blade tower sweeping faults | |
| CN108562828A (en) | The method for improving electrical network low voltage ride-through capability based on Wavelet Detection | |
| CN112290887A (en) | BP neural network arc detection model improved based on L-M algorithm | |
| CN117172601A (en) | Non-invasive load monitoring method based on residual total convolution neural network |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |