Intelligent condition monitoring method for wind turbine gearbox based on virtual index
Patent Information
- Application Number
- CN202410201305.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-23
- Publication Date
- 2026-11-03
- Estimated Expiration
- 2044-02-23
AI Technical Summary
Existing wind turbine gearbox condition monitoring methods are based on a single physical indicator, which cannot fully characterize the operating status, leading to frequent false alarms and failing to meet the requirements for real-time monitoring and accurate alarms.
An intelligent state monitoring model based on convolutional neural networks, long short-term memory neural networks, and autoencoders is constructed. It integrates the temporal and spatial characteristics of SCADA data to generate virtual indicators to indicate the health status of the gearbox, and sets early warning thresholds based on low-probability event assumptions.
It enables intelligent monitoring of wind turbine gearbox status and real-time feedback of health status, reducing false alarms and improving the accuracy and fault tolerance of operational status assessment.
Smart Images
Figure CN118094374B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of intelligent manufacturing, and in particular, it is a method for intelligent condition monitoring of wind turbine gearboxes based on virtual indicators. Background Technology
[0002] With the continuous increase in energy consumption and the urgent need for new clean energy, the wind power industry has developed rapidly. However, due to the long-term operation of wind turbines in extreme environments such as deserts, Gobi, and oceans, various components of wind turbines are prone to failure. As the core component of the wind turbine's transmission system, the gearbox is responsible for converting wind energy into mechanical energy; failure in this system can lead to prolonged downtime for maintenance and significant economic losses. Timely detection of abnormal operation in the wind turbine gearbox and the implementation of corresponding preventative and control measures can effectively reduce operating and maintenance costs, ensuring continuous power generation and economic benefits for the wind farm. Most existing wind turbine gearbox condition monitoring methods are based on establishing condition monitoring models using single physical indicators. However, a single physical indicator cannot fully characterize the gearbox's operating state, leading to frequent false alarms in practical applications and failing to meet the requirements for real-time monitoring and accurate alarms in wind farms. Summary of the Invention
[0003] To address at least one technical problem in the existing technology, this invention provides a method for intelligent condition monitoring of wind turbine gearboxes based on virtual indicators. By constructing an intelligent condition monitoring model based on convolutional neural networks, long short-term memory neural networks, and autoencoders, the method can fully mine the temporal and spatial feature information contained in SCADA data and integrate it into virtual indicators to achieve condition monitoring and fault early warning of wind turbine gearboxes. To achieve the above technical objectives, the technical solution adopted in this invention is as follows:
[0004] This invention provides a method for intelligent condition monitoring of wind turbine gearboxes based on virtual indicators, comprising the following steps:
[0005] Step S10: Obtain SCADA data of the wind turbine;
[0006] Step S20: Select parameters related to the gearbox operating status as feature parameters from the acquired SACDA data, and normalize the feature parameters.
[0007] Step S30: Construct and train an intelligent state monitoring model based on convolutional neural network, long short-term memory neural network and autoencoder; input the normalized feature parameters into the intelligent state monitoring model and fuse them to obtain virtual parameters;
[0008] Step S40: Construct virtual indicators based on the virtual parameters. The virtual indicators are used to indicate the health status of the wind turbine gearbox.
[0009] Step S50: Analyze the probability density distribution of virtual indicators during the operation of the wind turbine gearbox, set an early warning threshold based on the assumption of low probability events, and determine that the gearbox is abnormal when the virtual indicators continuously exceed the early warning threshold.
[0010] Even better, after acquiring the SCADA data of the wind turbine, the process also includes data cleaning of the acquired SCADA data.
[0011] Furthermore, the characteristic parameters include: average oil temperature of the gearbox, average temperature of the high-speed bearing of the gearbox, average temperature of the low-speed bearing of the gearbox, average inlet pressure of the gearbox filter, and average outlet pressure of the gearbox filter.
[0012] Furthermore, the normalization of the feature parameters is performed using formula (1):
[0013]
[0014] Where S is the characteristic parameter, S norm S represents the normalized feature parameters, and min(S) and max(S) are the minimum and maximum values of the feature parameters, respectively.
[0015] Furthermore, the intelligent state monitoring model based on convolutional neural networks, long short-term memory neural networks, and autoencoders includes: an input layer + an encoder + a fully connected layer + a decoder + an output layer;
[0016] The input layer includes a convolutional layer; the input layer is used to input feature parameters.
[0017] The encoder structure includes: one convolutional layer + one pooling layer + two long short-term memory neural network layers;
[0018] The fully connected layer is used to connect the encoder and decoder; the output of the fully connected layer is the virtual parameter.
[0019] The decoder structure includes: two long short-term memory neural network layers + one pooling layer + one convolutional layer;
[0020] The output layer consists of a convolutional layer; the output layer is used to output the reconstructed feature parameters.
[0021] Furthermore, the number of neurons in the convolutional layer is 32; the number of neurons in the long short-term memory neural network layer is 32; the kernel size of the convolutional layer is 1; and the kernel size of the pooling layer is 1.
[0022] Furthermore, the step of constructing a virtual index based on the virtual parameters includes: constructing a virtual index using the root mean square of the virtual parameters, and setting a sliding window to eliminate random spikes when constructing the virtual index;
[0023] The virtual indicator is shown in formula (2):
[0024]
[0025] Among them, VI(t) d ) represents a virtual indicator; v represents a virtual parameter; t d Indicates the current time; L represents the length of the sliding window.
[0026] Furthermore, the analysis of the probability density distribution of virtual indicators during the operation of the wind turbine gearbox, and the setting of the early warning threshold based on the assumption of low probability events, specifically includes: calculating the probability density distribution of the virtual indicators, and using the probability density function value when the confidence probability is 99.95% as the early warning threshold.
[0027] The beneficial effects of the technical solution provided by the embodiments of the present invention are as follows:
[0028] 1) Intelligent monitoring of the gearbox status of wind turbine units has been achieved.
[0029] 2) The virtual indicators used for health monitoring are accurate and provide real-time feedback on the operational health status of the wind turbine gearbox.
[0030] 3) The warning threshold has a reasonable fault tolerance rate, avoiding frequent false alarms and realizing accurate assessment of the operating health status of the wind turbine gearbox. Attached Figure Description
[0031] Figure 1 This is a flowchart of the status monitoring method in an embodiment of the present invention.
[0032] Figure 2 This is a schematic diagram of the CNN-LSTM-AE model in an embodiment of the present invention.
[0033] Figure 3 This is a virtual indicator monitoring result diagram of the health monitoring of the wind turbine gearbox in an embodiment of the present invention. Detailed Implementation
[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0035] In the description of the embodiments of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. In addition, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.
[0036] In the description of the embodiments of the present invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can also refer to the internal connection of two components; and they can refer to a wireless connection or a wired connection. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.
[0037] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0038] This invention proposes an intelligent condition monitoring method for wind turbine gearboxes based on virtual indicators, such as... Figure 1 As shown, it includes the following steps:
[0039] Step S10: Obtain SCADA data of the wind turbine;
[0040] SCADA is a data acquisition and monitoring system. SCADA data is mainly used to comprehensively understand the overall operating status and performance of wind turbine units, and is suitable for daily monitoring and maintenance.
[0041] Optionally, before proceeding to the next step, the acquired SCADA data may be cleaned to improve data quality; the principles of data cleaning are as follows:
[0042] Average wind speed ≤ 3m / s; or,
[0043] Average wind speed ≥ 25 m / s; or,
[0044] Average spindle speed ≤10rpm;
[0045] Step S20: Select parameters related to the gearbox operating status as feature parameters from the acquired SACDA data, and normalize the feature parameters.
[0046] In this embodiment, the characteristic parameters include: average oil temperature of the gearbox, average temperature of the high-speed bearing of the gearbox, average temperature of the low-speed bearing of the gearbox, average inlet pressure of the gearbox filter, and average outlet pressure of the gearbox filter.
[0047] The normalization of the feature parameters is performed using formula (1):
[0048]
[0049] Where S is the characteristic parameter, S norm S represents the normalized feature parameters, and min(S) and max(S) represent the minimum and maximum values of the feature parameters, respectively.
[0050] Use a portion of the normalized feature parameters from the past two months as the training set and the other portion as the test set.
[0051] Step S30: Construct and train an intelligent state monitoring model based on convolutional neural network, long short-term memory neural network and autoencoder; input the normalized feature parameters into the intelligent state monitoring model and fuse them to obtain virtual parameters;
[0052] The intelligent state monitoring model based on convolutional neural networks, long short-term memory neural networks, and autoencoders is abbreviated as the CNN-LSTM-AE model, such as... Figure 2 As shown, it includes: input layer + encoder + fully connected layer + decoder + output layer;
[0053] The input layer includes a convolutional layer; the input layer is used to input feature parameters, namely the five feature parameters mentioned above.
[0054] The encoder structure includes: one convolutional layer + one pooling layer + two long short-term memory neural network layers;
[0055] The fully connected layer is used to connect the encoder and decoder; the output of the fully connected layer is the virtual parameter.
[0056] The decoder structure includes: two long short-term memory neural network layers + one pooling layer + one convolutional layer;
[0057] The output layer consists of one convolutional layer; the output layer is used to output the restored feature parameters; that is, the five feature parameters restored corresponding to the five input feature parameters.
[0058] The number of neurons in the convolutional layer is 32; the number of neurons in the long short-term memory neural network layer is 32; the kernel size of the convolutional layer is 1; and the kernel size of the pooling layer is 1.
[0059] The learning rate is set to 0.01; the number of iterations is set to 100; the five input feature parameters and the five restored feature parameters are compared. When the difference is less than the preset range or the maximum number of iterations is reached, the intelligent state monitoring model is trained.
[0060] Step S40: Construct virtual indicators based on the virtual parameters. The virtual indicators are used to indicate the health status of the wind turbine gearbox.
[0061] The step of constructing a virtual index based on the virtual parameters includes: constructing a virtual index using the root mean square of the virtual parameters, and setting a sliding window to eliminate random spikes when constructing the virtual index;
[0062] The virtual indicator is shown in formula (2):
[0063]
[0064] Among them, VI(t) d ) represents a virtual indicator; v represents a virtual parameter; t d Indicates the current time; L represents the length of the sliding window;
[0065] Step S50: Analyze the probability density distribution of virtual indicators during the operation of the wind turbine gearbox, set an early warning threshold based on the assumption of low probability events, and determine that the gearbox is abnormal when the virtual indicators exceed the early warning threshold continuously.
[0066] In this embodiment, the step of analyzing the probability density distribution of virtual indicators during the operation of the wind turbine gearbox and setting an early warning threshold based on the assumption of low probability events specifically includes: calculating the probability density distribution of the virtual indicators and using the probability density function value when the confidence probability is 99.95% as the early warning threshold.
[0067] like Figure 3 As shown, in the following specific embodiment, the method proposed in this application is verified using SCADA data of wind turbines in a cooperative wind farm; the rated power of the wind turbines in the wind farm is 2MW, and the SCADA data sampling interval is 5min; Unit 001 is selected as the verification unit, and the time span for the operation of the verification unit is from July 6, 2020 to December 31, 2020;
[0068] Based on the probability distribution chart of virtual indicators for health monitoring of the gearbox of wind turbine unit 001, the warning threshold determined at a confidence level of 99.95% is 2.79. Figure 3As can be seen, from September 16, 2020, the virtual indicators fluctuated around the warning threshold, indicating an early fault. As the wind turbine continued to operate, the upward trend of the virtual indicators became more obvious, and all of them exceeded the warning threshold, indicating that the fault was gradually becoming more serious. At this point, it was determined that the gearbox was abnormal.
[0069] Finally, it should be noted that the above specific embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for intelligent condition monitoring of wind turbine gearboxes based on virtual indicators, characterized in that, Includes the following steps: Step S10: Obtain SCADA data of the wind turbine; Step S20: Select parameters related to the gearbox operating status as feature parameters from the acquired SACDA data, and normalize the feature parameters. Step S30: Construct and train an intelligent state monitoring model based on convolutional neural network, long short-term memory neural network and autoencoder; input the normalized feature parameters into the intelligent state monitoring model and fuse them to obtain virtual parameters; Step S40: Construct virtual indicators based on the virtual parameters. The virtual indicators are used to indicate the health status of the wind turbine gearbox. Step S50: Analyze the probability density distribution of virtual indicators during the operation of the wind turbine gearbox, set an early warning threshold based on the assumption of low probability events, and determine that the gearbox is abnormal when the virtual indicators exceed the early warning threshold continuously. The intelligent state monitoring model based on convolutional neural networks, long short-term memory neural networks and autoencoders includes: an input layer + encoder + fully connected layer + decoder + output layer; The input layer includes a convolutional layer; the input layer is used to input feature parameters. The encoder structure includes: one convolutional layer + one pooling layer + two long short-term memory neural network layers; The fully connected layer is used to connect the encoder and decoder; the output of the fully connected layer is the virtual parameter. The decoder structure includes: two long short-term memory neural network layers + one pooling layer + one convolutional layer; The output layer consists of one convolutional layer; the output layer is used to output the reconstructed feature parameters. The number of neurons in the convolutional layer is 32; the number of neurons in the long short-term memory neural network layer is 32; the kernel size of the convolutional layer is 1; the kernel size of the pooling layer is 1. The step of constructing a virtual index based on the virtual parameters includes: constructing a virtual index using the root mean square of the virtual parameters, and setting a sliding window to eliminate random spikes when constructing the virtual index; The virtual indicator is shown in formula (2): (2) in, VI(t d ) Indicates virtual indicators; Indicates virtual parameters; Indicates the current time; L This indicates the length of the sliding window.
2. The intelligent condition monitoring method for wind turbine gearboxes based on virtual indicators as described in claim 1, characterized in that, After acquiring the SCADA data of the wind turbine, the process also includes data cleaning of the acquired SCADA data.
3. The intelligent condition monitoring method for wind turbine gearboxes based on virtual indicators as described in claim 1, characterized in that, The characteristic parameters include: average oil temperature of the gearbox, average temperature of the high-speed bearing of the gearbox, average temperature of the low-speed bearing of the gearbox, average inlet pressure of the gearbox filter, and average outlet pressure of the gearbox filter.
4. The intelligent condition monitoring method for wind turbine gearboxes based on virtual indicators as described in claim 1, characterized in that, The normalization of the feature parameters is performed using formula (1): (1) in, For characteristic parameters, These are the normalized feature parameters. and These are the minimum and maximum values among the characteristic parameters, respectively.
5. The intelligent condition monitoring method for wind turbine gearboxes based on virtual indicators as described in claim 1, characterized in that, The analysis of the probability density distribution of virtual indicators during the operation of wind turbine gearboxes, and the setting of early warning thresholds based on the assumption of low probability events, specifically includes: calculating the probability density distribution of virtual indicators, and using the probability density function value with a confidence probability of 99.95% as the early warning threshold.