Abnormal sensing method and system for yaw system of wind turbine generator
By extracting the temporal and spatial characteristics of the operating parameters of the yaw system and fusing them into multi-dimensional spatiotemporal and spatial characteristics, using the CNN model of efficient channel attention for abnormal state perception, the problems of low early warning accuracy and neglected time and space correlation in the existing technology are solved, and higher abnormal state perception accuracy and operating stability of wind turbines are achieved.
Patent Information
- Application Number
- CN202411594554.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-09
- Publication Date
- 2025-03-18
AI Technical Summary
The existing yaw system fault monitoring methods have problems such as low early warning accuracy, low fault monitoring efficiency, low fan operation efficiency and stability, and ignoring the time and space correlation in the operation of yaw system.
By extracting the temporal and spatial characteristics of the operating parameters of the yaw system, the time and spatial correlations were extracted using Bi-LSTM and ResNet models, fused into multi-dimensional spatiotemporal and spatial features, and using the CNN model of efficient channel attention for abnormal state perception.
It improves the accuracy of abnormal state perception of yaw system, enhances local temporal feature expression and spatial feature representation, can predict potential abnormalities more accurately, and improves the operating stability and power generation efficiency of wind turbines.
Smart Images
Figure CN119646695A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wind power operation and maintenance, and in particular to a method and system for sensing abnormalities in a yaw system of a wind turbine generator set. Background Art
[0002] In the operation and maintenance of wind turbines, the yaw system is an important component of the wind turbine. Its main function is to adjust the wind rotor to align with the wind direction according to the wind direction to ensure that the wind turbine can efficiently and stably obtain wind energy, thereby improving power generation efficiency. However, abnormalities and faults in the yaw system, especially potential minor faults, may lead to serious consequences if they are not detected and handled in time at an early stage. These potential abnormalities are usually not obvious in the early stages of operation, but with the long-term and high-intensity operation of the unit, these minor faults may gradually worsen, eventually causing the yaw system to shut down or be damaged, affecting the operating life and power generation efficiency of the entire unit.
[0003] At present, traditional yaw system fault monitoring methods mostly use threshold comparison methods, that is, based on the thresholds set for the operating parameters of wind turbines, monitor and warn whether the yaw system has abnormalities. For example, by setting a reasonable range for parameters such as wind speed, wind direction angle, and yaw angle, when the actual monitored parameters exceed the range, an alarm will be triggered. However, due to the complex operating environment of wind turbines, the abnormal state of the yaw system is not easy to accurately warn with a fixed threshold. Therefore, the threshold comparison method is easily interfered by noise data and abnormal data, and the warning accuracy is low. In addition, the threshold method has limited perception of potential abnormal states, especially for signs of gradually emerging faults, which often cannot provide timely warnings. On the other hand, the application of machine learning and deep learning has been gradually introduced into the fault monitoring of wind turbines, but most of the current algorithms still focus on single-dimensional data analysis, ignoring the temporal and spatial correlations in the operation of the yaw system when extracting features, and these correlations are crucial for fault detection. The state of the yaw system is usually affected by spatial factors such as meteorological conditions and geographical location, as well as temporal factors. Ignoring these correlations will lead to insufficient representation of fault features and affect the accuracy of prediction. Summary of the invention
[0004] In view of the above-mentioned problems, the present invention is proposed.
[0005] Therefore, the technical problem solved by the present invention is that the existing yaw system fault monitoring method has the problems of low early warning accuracy, low fault monitoring efficiency, low operating efficiency and stability of the fan, and ignoring the time and space correlation in the operation of the yaw system.
[0006] To solve the above technical problems, the present invention provides the following technical solutions: a method for sensing abnormality of a yaw system of a wind turbine, comprising extracting time characteristics of operating parameters of the yaw system; extracting spatial characteristics of operating parameters of the yaw system; and sensing abnormal state of the yaw system.
[0007] As a preferred solution of the method for sensing abnormality of the yaw system of a wind turbine according to the present invention, wherein: the extracting time characteristics of the yaw system operating parameters comprises collecting a sequence of yaw system operating parameters;
[0008] The collected parameter sequence includes wind speed, wind direction angle, and yaw angle. The parameter sequence collected within the time range t is recorded as:
[0009] S=[s1,s2,s3,…,s t ]
[0010] Among them, s i =[a1,a2,a3], (i=1,2,3,…t), a1,a2,a3 represent wind speed, wind direction angle and yaw angle respectively.
[0011] As a preferred solution of the abnormality perception method of the yaw system of a wind turbine according to the present invention, wherein: the extraction of the time characteristics of the yaw system operation parameters includes then inputting the parameter sequence S into the Bi-LSTM, and outputting the extracted time characteristics P = [p1, p2, p3, ... p t ], there are two LSTMs in Bi-LSTM, one processes the sequence from front to back and the other processes the sequence from back to front, and Bi-LSTM uses both forward and backward hidden states at each time step.
[0012] As a preferred solution of the abnormality perception method of the yaw system of a wind turbine according to the present invention, the extraction of the spatial characteristics of the yaw system operation parameters includes reshaping the collected parameter sequence S into a three-dimensional matrix U, which is expressed as:
[0013]
[0014] Wherein, i, j are constants, and i*j=t.
[0015] As a preferred solution of the abnormality perception method of the yaw system of a wind turbine according to the present invention, the extraction of the spatial characteristics of the yaw system operating parameters includes normalizing all elements in the three-dimensional matrix U to the RGB color space [0, 255] and outputting an RGB image I, which is expressed as:
[0016]
[0017] Among them, Min(U) represents the minimum function of matrix U, and Max(U) represents the maximum function of matrix U;
[0018] Input the RGB image I into the 50-layer ResNet to extract the spatial features in the image and output the feature Q = [q1, q2, q3, ...q t], ResNet captures spatial features in images.
[0019] As a preferred solution of the method for sensing abnormality of the yaw system of a wind turbine according to the present invention, the sensing of abnormal state of the yaw system includes fusing the time feature P and the space feature Q extracted from the yaw system operation parameter sequence, and outputting a multi-dimensional space-time feature F, which is expressed as:
[0020] F=[[p1,p2,…,p t ],[q1,q2,…,q t ]]
[0021] Among them, P represents the time feature, Q represents the spatial feature;
[0022] The CNN with added channel attention is trained using the multi-dimensional spatiotemporal features F, and the yaw system abnormality perception model based on the improved CNN is output.
[0023] As a preferred solution of the method for sensing abnormality of the yaw system of a wind turbine described in the present invention, the sensing of the abnormal state of the yaw system includes collecting the operating parameters of the yaw system at the current moment, extracting the time features and the spatial features and fusing them into the multi-dimensional spatiotemporal features of the operating parameters of the yaw system at the current moment, and using the multi-dimensional spatiotemporal features as the input of the yaw system abnormality perception model based on the improved CNN to sense whether there is a potential abnormal problem in the yaw system.
[0024] Another object of the present invention is to provide a wind turbine yaw system abnormality perception system, which can solve the problem of ignoring the time and space correlation in the operation of the yaw system in the current yaw system fault monitoring method by extracting the time and space characteristics of the yaw system operation parameters.
[0025] As a preferred solution of the wind turbine yaw system abnormality perception system described in the present invention, it includes: a yaw system operating parameter time feature extraction module, a yaw system operating parameter spatial feature extraction module, and a yaw system abnormal state perception module; the yaw system operating parameter time feature extraction module is used to extract the time correlation in the yaw system operating parameters through a bidirectional long short-term memory network; the yaw system operating parameter spatial feature extraction module is used to extract the spatial features of the yaw system using a residual network; the yaw system abnormal state perception module is used to fuse the time and space features to form multi-dimensional spatiotemporal features, which are input into a convolutional neural network with efficient channel attention added to identify the abnormal state of the yaw system.
[0026] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a method for detecting abnormalities in a yaw system of a wind turbine.
[0027] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for sensing abnormalities in a yaw system of a wind turbine.
[0028] Beneficial effects of the present invention: The abnormality perception method of the yaw system of a wind turbine provided by the present invention uses Bi-LSTM to extract the time features in the operating parameters of the yaw system, which can effectively obtain the past and future contextual time correlations in the parameter sequence and enhance the expression of local time features. The spatial features in the operating parameters of the yaw system are extracted using a deep ResNet, which can effectively reflect the regional correlations in the yaw system parameters in the extracted spatial features. The time features extracted from the yaw system parameters are fused with the spatial features to obtain multi-dimensional spatiotemporal features. The multi-dimensional spatiotemporal features can characterize the operating state of the yaw system from two dimensions, time and space, and have richer feature representations than single time features or single space features. The abnormal state of the yaw system is perceived using a CNN model with efficient channel attention added. The efficient channel attention can highlight some important features in the multi-dimensional spatiotemporal features, help capture the spatiotemporal correlations in the multi-dimensional spatiotemporal features, and give full play to the advantages of the multi-dimensional spatiotemporal features, thereby improving the accuracy of the perception of the abnormal state of the wind turbine yaw system. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0030] Figure 1 A flow chart of a method for extracting spatial features of yaw system operating parameters based on ResNet for a method for sensing abnormalities in a yaw system of a wind turbine provided in the first embodiment of the present invention.
[0031] Figure 2 A flow chart of a method for sensing anomalies in a yaw system of a wind turbine provided in the first embodiment of the present invention, which is based on an improved CNN with efficient channel attention.
[0032] Figure 3 A flow chart of a method for sensing anomalies in a yaw system of a wind turbine based on multi-dimensional spatiotemporal characteristics, which is provided as a first embodiment of the present invention.
[0033] Figure 4 An overall flow chart of a wind turbine yaw system abnormality sensing system provided in accordance with the third embodiment of the present invention. DETAILED DESCRIPTION
[0034] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.
[0035] Example 1, reference Figure 1-Figure 3 , as an embodiment of the present invention, provides a method for sensing abnormality of a yaw system of a wind turbine, comprising:
[0036] S1: Extract the time characteristics of the yaw system operating parameters.
[0037] Furthermore, the present invention proposes a method for sensing abnormalities in the yaw system of a wind turbine. A Bi-LSTM-based yaw system operating parameter time feature extraction method is used to extract the time correlation in a continuous yaw system operating parameter sequence. Specifically, a bidirectional LSTM is used to obtain the past and future context time correlations in the parameter sequence, and the extracted time features are enhanced. The extracted time features are used to fuse with the subsequently extracted spatial features into spatiotemporal features.
[0038] The present invention uses Bi-LSTM to extract the time features in the yaw system parameter sequence, which can effectively obtain the context time correlation in the parameter sequence and enhance the expression of local time features.
[0039] It should be noted that the present invention proposes a Bi-LSTM-based method for extracting temporal features of yaw system operating parameters. The yaw system operating parameter (wind speed, wind direction angle, yaw angle) sequence is a set of temporally continuous data, and therefore has a strong contextual temporal correlation. The above-mentioned contextual temporal correlation can be used as an important unique feature for distinguishing the abnormal state of the yaw system. This method uses Bi-LSTM to effectively extract the contextual temporal correlation in the yaw system operating parameter sequence, providing a stronger temporal feature expression for distinguishing yaw system anomalies, thereby improving the accuracy of yaw system anomaly perception.
[0040] Specifically, the yaw system operation parameter sequence is first collected from the Supervisory Control and Data Acquisition (SCADA) system. The collected parameter sequences are wind speed, wind direction angle, and yaw angle. The parameter sequence within the collection time range t is recorded as S = [s1, s2, s3, ..., s t ]. Where s i=[a1,a2,a3], (i=1,2,3,…t), a1,a2,a3 correspond to wind speed, wind direction angle and yaw angle respectively.
[0041] Then the parameter sequence S is input into Bi-LSTM to obtain the extracted time feature P = [p1, p2, p3, ... p t ], There are two LSTMs in Bi-LSTM, one processes the sequence from front to back and the other processes the sequence from back to front, so that the forward and backward contextual information can be utilized. Bi-LSTM has the same state equation as LSTM, but uses forward and backward hidden states at each time step.
[0042] S2: Extract the spatial characteristics of the yaw system operating parameters.
[0043] Furthermore, a ResNet-based yaw system operating parameter spatial feature extraction method is used to extract the spatial correlation in the yaw system operating parameter sequence. Specifically, the yaw system parameter sequence is converted into an image, and the spatial features in the image are extracted using a deep ResNet, which can reflect the regional correlation of the yaw system parameter sequence. The extracted spatial features and temporal features are fused into multi-dimensional spatiotemporal features as the input of the subsequent improved CNN model, providing richer feature representation for sensing yaw system anomalies.
[0044] Using deep ResNet to extract spatial features from the yaw system parameter sequence can effectively reflect the regional correlation of the yaw system parameter sequence in the extracted spatial features. The time features extracted from the yaw parameter sequence are fused with the spatial features to obtain multi-dimensional spatiotemporal features. The fused spatiotemporal features have richer feature representations than single time features or single spatial features.
[0045] Since each wind turbine is located in a different geographical location, that is, it is distributed in space, the parameters in the wind turbine yaw system parameter sequence have regional correlation, which can serve as an important spatial clue for distinguishing the abnormal state of the yaw system. The present invention proposes a spatial feature extraction method for yaw system operating parameters based on ResNet. The method can extract the regional correlation of the yaw system operating parameters of wind turbines at different locations, provide a stronger spatial feature expression for the abnormality judgment of the yaw system, and thus improve the accuracy of yaw system abnormality perception.
[0046] Specifically, the flowchart of the ResNet-based yaw system operating parameter space feature extraction method is as follows: Figure 1 First, reshape the parameter sequence S collected above into a three-dimensional matrix U. The formula of U is as follows:
[0047]
[0048] Wherein, i, j are constants, and i*j=t.
[0049] It should be noted that all elements in the three-dimensional matrix U are then normalized to the RGB color space [0,255] to obtain the RGB image I. The formula of I is as follows:
[0050]
[0051] Where Min(U) represents the minimum function of matrix U, and Max(U) represents the maximum function of matrix U.
[0052] Finally, the RGB image I obtained above is input into a 50-layer ResNet to extract the spatial features in the image and obtain the feature Q = [q1, q2, q3, …q t ], ResNet can solve the gradient disappearance, gradient explosion and degradation problems of neural networks and capture useful spatial features in images.
[0053] S3: Sense abnormal status of the yaw system.
[0054] Furthermore, in view of the problem that the spatiotemporal correlation in the yaw system operating parameter sequence is not considered when perceiving anomalies in the prior art, the present invention proposes a yaw system anomaly perception method based on an improved CNN with efficient channel attention. The temporal features and spatial features in the yaw system operating parameter sequence extracted above are fused to obtain multi-dimensional spatiotemporal features. The CNN with added channel attention is trained using the multi-dimensional spatiotemporal features to obtain a yaw system anomaly perception model based on the improved CNN. The temporal and spatial features in the current yaw system operating parameter data are extracted in real time and fused into multi-dimensional spatiotemporal features, which are used as the input of the trained yaw system anomaly perception model based on the improved CNN to perceive and warn of abnormal conditions of the yaw system.
[0055] The abnormal state of the yaw system is perceived using a CNN model improved based on efficient channel attention. Efficient channel attention can highlight some important features in the spatiotemporal characteristics, help capture the spatiotemporal correlation in the spatiotemporal characteristics, and give full play to the advantages of multi-dimensional spatiotemporal characteristics, thereby improving the accuracy of perception of the abnormal state of the wind turbine yaw system.
[0056] Specifically, the process of improving the CNN yaw system abnormality perception method based on efficient channel attention is as follows: Figure 2 First, the time feature P and space feature Q extracted from the yaw system operating parameter sequence are fused to obtain the multi-dimensional spatiotemporal feature F, which is represented as follows:
[0057] F=[[p1,p2,...,p t],[q1,q2,...,q t ]]
[0058] Then, the multi-dimensional spatiotemporal features F obtained above are used to train the CNN with added channel attention, and the yaw system abnormality perception model based on the improved CNN is obtained. The efficient channel attention can more effectively capture the interdependence between channels through the cross-channel interaction strategy, capture the spatiotemporal feature representation with the largest amount of information along the channel direction, and better extract the time correlation and spatial correlation in the multi-dimensional spatiotemporal features for perceiving the yaw system state.
[0059] It should be noted that the operating parameters of the yaw system at the current moment are finally collected, and the time features and spatial features are extracted and fused into the multi-dimensional spatiotemporal features of the operating parameters of the yaw system at the current moment. The multi-dimensional spatiotemporal features are used as the input of the above-mentioned trained yaw system abnormality perception model based on improved CNN to perceive whether there is potential abnormality in the yaw system, which helps to discover problems in the yaw system early and take countermeasures in advance.
[0060] Embodiment 2, an embodiment of the present invention, provides a method for sensing abnormalities in a yaw system of a wind turbine. To verify the beneficial effects of the present invention, scientific demonstration is performed through economic benefit calculation and simulation experiments.
[0061] The experimental data is obtained through monitoring equipment set up in the actual wind farm. The data collection range includes parameters such as wind speed, wind direction angle, yaw angle, etc., which are used as the operating parameters of the yaw system of the wind turbine. Specifically, the yaw system operating parameters in continuous time are obtained from the data acquisition and monitoring system (SCADA). The data collection frequency is once every minute, covering the operation status of 24 hours a day.
[0062] The collected yaw system operation data is input into the Bi-LSTM model, and the model is set to a bidirectional structure to fully capture the contextual correlation in the time series. The bidirectional unit of Bi-LSTM ensures that when processing each time point, it can refer to the previous and subsequent data features to enhance the representation ability of that time point. This process calculates and generates a feature vector with time correlation at each time step and stores it as the time feature P.
[0063] The collected time series data is reshaped into a two-dimensional image format, and parameters such as wind speed, wind direction angle, and yaw angle are mapped into a three-dimensional matrix, normalized, and converted into an RGB image. This image is used as the input of ResNet, and spatial features are extracted through a deep convolutional neural network to reflect the regional differences in the geographical location of wind turbines. The generated spatial feature Q provides important information on spatial distribution.
[0064] The temporal feature P and the spatial feature Q are fused to form a multi-dimensional spatiotemporal feature F. The multi-dimensional spatiotemporal feature F is input into the improved CNN model based on efficient channel attention for training. During the training process, the model focuses on the important information between channels and further strengthens the expression of abnormal features. After multiple rounds of training, the model can identify potential abnormalities of the yaw system under different states, ensuring timely warning when abnormalities occur to prevent system failures.
[0065] Table 1 Experimental data table
[0066]
[0067] It can be clearly seen from the tabular data that the yaw system abnormality perception method of the present invention can effectively identify abnormal conditions under different wind speed, wind direction and yaw angle combinations. Especially at the 30th and 60th minutes, when the wind speed or wind direction is lower than the average level, the abnormal perception result is displayed as "abnormal", which shows the sensitivity of the system in identifying potential faults. Compared with the traditional fault monitoring method based on a single threshold, the present invention captures time correlation through Bi-LSTM, extracts spatial features through ResNet, and improves the attention mechanism of the CNN model, making data processing more accurate and comprehensive.
[0068] Specifically, the changing trends of the time feature extraction values and the spatial feature extraction values reflect the abnormal state of the operating parameters. For example, at the 30th and 60th minutes, due to the wind speeds of 5.5m / s and 4.8m / s respectively, and the wind direction angle was small, the time feature extraction values and spatial feature extraction values detected by the system fluctuated, and the abnormal perception results were displayed as abnormal. These abnormality detections are different from traditional methods. Traditional threshold methods may ignore such subtle fluctuations and delay the detection time of potential faults.
[0069] The present invention utilizes the fusion of multi-dimensional spatiotemporal features, so that the yaw system not only takes into account single time or space information during operation, but also covers the correlation between operating parameters in time and space. Therefore, the innovation demonstrated by the present invention in the data lies in its joint analysis of time and space features, which enables the system to more accurately predict potential anomalies, and has obvious innovative effects and technical advantages. By using an efficient channel attention mechanism, the present invention highlights the key factors in spatiotemporal features in the perception model, further improving the accuracy of yaw system anomaly detection, and providing a higher level of protection for the continuous and stable operation of wind turbines.
[0070] Example 3, reference Figure 4 , is an embodiment of the present invention, and provides a wind turbine yaw system abnormality perception system, including a yaw system operating parameter time feature extraction module, a yaw system operating parameter space feature extraction module, and a yaw system abnormal state perception module.
[0071] The yaw system operating parameter time feature extraction module is used to extract the time correlation in the yaw system operating parameters through a bidirectional long short-term memory network. The yaw system operating parameter spatial feature extraction module is used to extract the spatial features of the yaw system using a residual network. The yaw system abnormal state perception module is used to fuse the time and space features to form multi-dimensional spatiotemporal features, which are input into a convolutional neural network with efficient channel attention added to identify the abnormal state of the yaw system.
[0072] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0073] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.
[0074] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.
[0075] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.
[0076] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A method for sensing abnormality of a yaw system of a wind turbine, characterized in that: include: Extract the time characteristics of the yaw system operating parameters; Extract spatial characteristics of yaw system operating parameters; Sense abnormal status of the yaw system.
2. The method for detecting abnormality of a yaw system of a wind turbine according to claim 1, characterized in that: The extracting of the time characteristics of the yaw system operating parameters includes collecting a sequence of the yaw system operating parameters; The collected parameter sequence includes wind speed, wind direction angle, and yaw angle. The parameter sequence collected within the time range t is recorded as: <h2 style=";text-align:left;direction:ltr">S = [s1,s2,s3,…,s<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> ] Among them, s i =[a1,a2,a3], (i=1,2,3,…t), a1,a2,a3 represent wind speed, wind direction angle and yaw angle respectively.
3. The method for detecting abnormality of a yaw system of a wind turbine according to claim 2, characterized in that: The method of extracting the time features of the yaw system operation parameters includes inputting the parameter sequence S into the Bi-LSTM, and outputting the extracted time features P = [p1, p2, p3, ... p t ], there are two LSTMs in Bi-LSTM, one processes the sequence from front to back and the other processes the sequence from back to front, and Bi-LSTM uses both forward and backward hidden states at each time step.
4. The method for detecting abnormality of a yaw system of a wind turbine according to claim 3, characterized in that: The extraction of the spatial features of the yaw system operating parameters includes reshaping the collected parameter sequence S into a three-dimensional matrix U, which is expressed as: Wherein, i, j are constants, and i*j=t.
5. The method for detecting abnormality of a yaw system of a wind turbine according to claim 4, characterized in that: The method of extracting the spatial features of the yaw system operating parameters includes then normalizing all elements in the three-dimensional matrix U to the RGB color space [0, 255], and outputting an RGB image I, which is expressed as: Among them, Min(U) represents the minimum function of matrix U, and Max(U) represents the maximum function of matrix U; Input the RGB image I into a 50-layer ResNet to extract the spatial features in the image and output features Q = [q1, q2, q3, …q t ], ResNet captures spatial features in images.
6. The method for detecting abnormality of a yaw system of a wind turbine according to claim 5, characterized in that: The sensing of the abnormal state of the yaw system includes fusing the time feature P and the space feature Q extracted from the yaw system operation parameter sequence, and outputting a multi-dimensional time-space feature F, which is expressed as: F=[[p1,p2,…,p t ],[q1,q2,…,q t ]] Among them, P represents the time feature, Q represents the spatial feature; The CNN with added channel attention is trained using the multi-dimensional spatiotemporal features F, and the yaw system abnormality perception model based on the improved CNN is output.
7. The method for detecting abnormality of a yaw system of a wind turbine according to claim 6, characterized in that: The method of sensing the abnormal state of the yaw system includes collecting the operating parameters of the yaw system at the current moment, extracting the time features and the space features and fusing them into the multi-dimensional spatiotemporal features of the operating parameters of the yaw system at the current moment, and using the multi-dimensional spatiotemporal features as the input of the yaw system abnormality perception model based on the improved CNN to sense whether there is a potential abnormal problem in the yaw system.
8. A system using the method for sensing abnormality of a yaw system of a wind turbine as claimed in any one of claims 1 to 7, characterized in that: It includes a yaw system operating parameter time feature extraction module, a yaw system operating parameter space feature extraction module, and a yaw system abnormal state perception module; The yaw system operating parameter time feature extraction module is used to extract the time correlation in the yaw system operating parameters through a bidirectional long short-term memory network; The yaw system operation parameter spatial feature extraction module is used to extract the spatial features of the yaw system using a residual network; The yaw system abnormal state perception module is used to fuse time and space features to form multi-dimensional time and space features, which are input into a convolutional neural network with efficient channel attention added to identify the abnormal state of the yaw system.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for detecting abnormalities in a yaw system of a wind turbine set according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for detecting abnormalities in a yaw system of a wind turbine set according to any one of claims 1 to 7 are implemented.