A method and related device for alarming faults in wind turbine generators

Abstract

This application discloses a method and related apparatus for alarming faults in wind turbine generators. For wind turbine blades, a video monitoring terminal, after receiving a detection request to acquire a target image of the target blade at a preset time, sends the detection request to a server. The server performs fault detection on the target blade according to the detection request and sends the detection result to the video monitoring terminal. The video monitoring terminal displays the detection result and obtains a re-inspection result based on the detection result. If the re-inspection result indicates a fault in the target blade, a fault alarm is issued. Therefore, by correcting the less accurate detection result based on the more accurate re-inspection result, the number of false alarms can be reduced, and the accuracy of fault alarms for the target blade can be improved.

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, and in particular to a method and related device for alarming faults in wind turbine generators. Background Technology

[0002] New energy sources such as solar and wind power are gradually gaining attention, and the number of wind turbines is also increasing. Wind turbines are electrical devices that convert wind energy into mechanical work, which drives the rotor to rotate and ultimately output alternating current.

[0003] Due to the unique installation location of wind turbines, components such as blades may experience problems such as icing, breakage, cracking, fissures, and lightning strikes, resulting in significant power generation losses and maintenance costs.

[0004] In related technologies, image recognition is used to detect faults in wind turbine blades. However, due to the uncertain installation locations of different wind turbines in a wind turbine generator set, the collected images are diverse, and the blade conditions are varied, often resulting in false alarms. Summary of the Invention

[0005] To address the aforementioned issues, this application provides a wind turbine fault alarm method and related apparatus to reduce the number of false alarms and improve the accuracy of fault alarms.

[0006] Based on this, the embodiments of this application disclose the following technical solutions:

[0007] In a first aspect, embodiments of this application provide a wind turbine fault alarm method, applied to a video monitoring terminal, the method comprising:

[0008] Obtain a detection request, the detection request being used to request the acquisition of a target image at a preset time, the target image including the target blades of a wind turbine;

[0009] The receiving server performs fault detection on the target blade based on the detection request;

[0010] Display the detection results;

[0011] Obtain a retest result based on the test result, wherein the accuracy of the retest result is higher than the accuracy of the test result;

[0012] If the re-inspection result indicates a fault in the target blade, a fault alarm will be issued.

[0013] Optionally, the method further includes:

[0014] If the re-inspection result is inconsistent with the detection result, the re-inspection result is sent to the server so that the server can update the fault detection algorithm based on the re-inspection result.

[0015] Secondly, this application provides a wind turbine fault alarm method, applied to a server, the method comprising:

[0016] A target image is acquired according to a detection request, the target image including the target blades of a wind turbine;

[0017] The target image is subjected to feature processing to obtain the attribute features of the target image;

[0018] The state of the target blade is determined based on the attribute characteristics;

[0019] If the target blade is in a pre-fault state, the detection result of the target blade is determined based on the target image and the set of suppressed images; wherein, the set of suppressed images includes images of false alarms of the target blade, and the images of false alarms are determined based on the re-inspection result and the detection result;

[0020] If the detection result indicates a fault in the target blade, the detection result is sent to the video detection terminal so that the video monitoring terminal can issue a fault alarm.

[0021] Optionally, determining the detection result of the target leaf based on the target image and the set of suppressed images includes:

[0022] The target image and the suppressed image set are input into a similarity detection model to determine the similarity between the attribute features of the target image and the attribute features of other images, wherein the other images are the images in the suppressed image set that falsely alarm the target leaf.

[0023] If the attribute features of the target image and the attribute features of other images meet the similarity condition, the detection result is that the target blade is not faulty; if the attribute features of the target image and the attribute features of other images do not meet the similarity condition, the detection result is that the target blade is faulty.

[0024] Optionally, the method further includes:

[0025] If the detection result indicates that the target blade is faulty, and the re-inspection result indicates that the target blade is not faulty, the target image is added to the suppression image set as a false alarm image.

[0026] Optionally, the step of performing feature processing on the target image to obtain the attribute features of the target image includes:

[0027] The target image is input into the diagnostic model to obtain the attribute features of the target leaf;

[0028] The method further includes:

[0029] If the detection result indicates that the target blade is not faulty, and the re-inspection result indicates that the target blade is faulty, the diagnostic model is updated based on the target image and the re-inspection result.

[0030] Optionally, the diagnostic model extracts the attribute characteristics of the target image in the same way as the similarity detection model.

[0031] Optionally, the detection request to acquire the target image includes:

[0032] A call command is sent to the camera acquisition device according to the detection request, and the call command includes the target position information of the target blade;

[0033] Receive target video, including the target blade, acquired by the camera acquisition device based on the target location information;

[0034] Target images for the target blades are extracted from the target video.

[0035] Optionally, the method further includes:

[0036] Obtain the device status of the camera acquisition device;

[0037] If the device is in normal condition, the step of receiving the target video, including the target blade, acquired by the camera acquisition device based on the target location information is executed.

[0038] Thirdly, this application provides a wind turbine fault alarm system, characterized in that the system includes a terminal and a server:

[0039] The terminal device is used to execute the method described in any one of the first aspects;

[0040] The server is used to perform the method described in any of the second aspects.

[0041] Optionally, the system may also include a camera acquisition device;

[0042] The camera acquisition device is used to acquire target video including the target blade according to the target position information of the target blade included in the call instruction after receiving the call instruction.

[0043] Fourthly, this application provides a wind turbine fault alarm device for use in a video monitoring terminal. The device includes: an acquisition unit, a receiving unit, a display unit, and an alarm unit.

[0044] The acquisition unit is used to acquire a detection request, the detection request being used to request the acquisition of a target image at a preset time, the target image including the target blade of a wind turbine.

[0045] The receiving unit is used to receive the detection result of the server performing fault detection on the target blade according to the detection request;

[0046] The display unit is used to display the detection results;

[0047] The acquisition unit is further configured to acquire a retest result for the detection result, wherein the accuracy of the retest result is higher than the accuracy of the detection result;

[0048] The alarm unit is used to issue a fault alarm if the re-inspection result indicates that the target blade is faulty.

[0049] Fifthly, this application provides a wind turbine fault alarm device applied to a server, the device comprising: an acquisition unit, an algorithm unit, and a transmission unit;

[0050] The acquisition unit is used to acquire a target image according to a detection request, wherein the target image includes the target blades of a wind turbine.

[0051] The algorithm unit is used to perform feature processing on the target image to obtain the attribute features of the target image; determine the state of the target blade based on the attribute features; if the target blade is in a pre-fault state, determine the detection result of the target blade based on the target image and the set of suppressed images; wherein, the set of suppressed images includes images of false alarms of the target blade, and the images of false alarms are determined based on the re-inspection result and the detection result;

[0052] The sending unit is used to send the detection result to the video detection terminal if the detection result indicates that the target blade is faulty, so that the video monitoring terminal can issue a fault alarm.

[0053] Sixth aspect: This application provides a computer device, the device including a processor and a memory:

[0054] The memory is used to store program code and transmit the program code to the processor;

[0055] The processor is configured to execute the method described in either the first aspect or the second aspect according to the instructions in the program code.

[0056] In a seventh aspect, this application provides a computer-readable storage medium for storing a computer program for performing the method described in either the first or second aspect.

[0057] Eighthly, embodiments of this application provide a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method described in either the first or second aspect.

[0058] Compared with the prior art, the advantages of the above-mentioned technical solution of this application are as follows:

[0059] For wind turbine blades, after receiving a detection request to acquire target images of the target blade at a preset time, the video monitoring terminal sends the request to the server. The server performs fault detection on the target blade based on the detection request and sends the detection results to the video monitoring terminal. The video monitoring terminal displays the detection results and obtains the re-inspection results. If the re-inspection results indicate a fault in the target blade, a fault alarm is issued. Therefore, by correcting the less accurate detection results based on the more accurate re-inspection results, the number of false alarms can be reduced, and the accuracy of alarms for target blade faults can be improved. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 A schematic diagram of a wind turbine alarm system provided in this application;

[0062] Figure 2 A signaling interaction diagram of a wind turbine fault alarm system provided in this application embodiment;

[0063] Figure 3 A flowchart of a fault detection algorithm provided in an embodiment of this application;

[0064] Figure 4 A schematic diagram of a diagnostic algorithm provided in an embodiment of this application;

[0065] Figure 5A schematic diagram of a fault detection algorithm provided in an embodiment of this application;

[0066] Figure 6 This application provides a schematic diagram illustrating an application scenario for a wind turbine alarm method.

[0067] Figure 7 A schematic diagram of a wind turbine alarm device provided in an embodiment of this application;

[0068] Figure 8 A schematic diagram of a wind turbine alarm device provided in an embodiment of this application;

[0069] Figure 9 This is a structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0070] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0071] For wind turbine blade fault detection, image recognition is often used to avoid significant power generation losses and maintenance costs. However, due to the uncertain installation locations of different wind turbines within a wind turbine generator set, the collected images are diverse, and the blade states within these images vary. If the image recognition algorithm cannot be updated in a timely manner, false alarms can occur, and even continuous false alarms can result. For example, for a blade image that triggers a false alarm, if the blade has been patched after repair, causing a significant difference between the image and the standard image, it might be judged as a faulty blade when it is actually not faulty. This misjudgment leads to false alarms, or even continuous false alarms. In field applications, the real fault is buried among numerous false alarm data, hindering timely and rapid repairs.

[0072] Based on this, this application provides a wind turbine fault alarm method. The server, according to a detection request, performs feature processing on the target blade in the acquired target image to obtain the attribute features of the target image, thereby determining the state of the target blade. If the target blade is in a pre-fault state, the detection result of the target blade is determined based on the target image and a set of suppressed images. If the target blade is faulty, the detection result is sent to a video detection terminal so that the video monitoring terminal can issue a fault alarm. The video monitoring terminal obtains a re-inspection result based on the detection result. This re-inspection result has high accuracy, and the set of suppressed images determined based on this re-inspection result also has high accuracy. This improves the accuracy of determining the fault status of the target blade based on the set of suppressed images and the target image, reduces the number of false alarms and consecutive false alarms, and improves the timeliness of repairing truly faulty blades.

[0073] The wind turbine fault alarm method provided in this application embodiment is applied to, for example, Figure 1 The wind turbine alarm system shown includes a video monitoring terminal and a server. The video monitoring terminal can be a smartphone, tablet, laptop, PDA, personal computer, etc., but is not limited to these. The video monitoring terminal and the server can be connected directly or indirectly through wired or wireless communication, and this application does not impose any restrictions.

[0074] As one possible implementation, the functions implemented by the server can also be achieved through a video monitoring terminal to save network requests and ensure that fault alarms continue to function normally when the server malfunctions. However, when the computational load required by the fault detection method is large, it may consume a lot of resources, which may not be able to be handled by the video monitoring terminal alone. Therefore, a simplified version of the fault detection method can be built into the video monitoring terminal as a fallback solution when the server fails to identify the fault or malfunctions.

[0075] For ease of explanation, the following detailed description uses the interaction scheme between the video monitoring terminal and the server. See [link / reference] Figure 2 The figure is a signaling interaction diagram of a wind turbine fault alarm system provided in an embodiment of this application.

[0076] S201: The video monitoring terminal receives a detection request.

[0077] In practical applications, owners or on-site maintenance personnel can use the video monitoring terminal to create an automated inspection plan. This plan corresponds to a detection request and is used to request the acquisition of target images, including the target blades of the wind turbine, at a preset time. The video monitoring terminal reads and displays the automated inspection plan.

[0078] The preset time is a pre-set time, the wind turbine is a wind turbine in a wind power plant, and the target blade is the blade of the aforementioned wind turbine used for fault detection. This application does not specifically limit the preset time, the wind turbine, and the target blade; those skilled in the art can set them according to actual needs. For example, images of every blade of all wind turbines in the wind power plant can be collected at 8:00 AM every day.

[0079] S202: The video monitoring terminal sends the detection request to the server.

[0080] S203: The server obtains the target image based on the detection request.

[0081] The video monitoring terminal sends the automatic inspection plan to the server, which then writes the automatic inspection plan into the database so that it can collect target images according to the detection requests corresponding to the automatic inspection plan.

[0082] As one possible implementation, the target image can be acquired by calling the camera acquisition device, as detailed in S2031-S2033.

[0083] S2031: The server sends a call command to the camera acquisition device based on the detection request.

[0084] The server reads the automatic inspection plan, generates a call instruction based on the corresponding detection request, and sends it to the camera acquisition device so that the camera acquisition device can acquire target video including the target blade according to the target blade position information included in the call instruction.

[0085] The camera acquisition device can be a device within the video monitoring terminal system or a device outside the video monitoring terminal system. It is used to acquire target images including the target blades based on the position information of the target blades included in the call command after receiving the call command.

[0086] S2032: The camera acquisition device acquires target video, including the target blades, based on the target location information.

[0087] It should be noted that in a wind power plant, there can be one or more camera acquisition devices. When there is only one camera acquisition device, target video can be acquired based on the target location information. When there are multiple camera acquisition devices, different camera acquisition devices are responsible for acquiring video of the blades at different locations in the wind power plant. Thus, based on the target location information, the camera acquisition device responsible for acquiring the target blade can be determined, and target video including the target blade can be acquired.

[0088] S2033: The server receives the target video sent by the camera acquisition device.

[0089] As one possible implementation, before executing S2033, the device status of the camera acquisition device can be obtained, and the reliability of the target video acquired by the camera acquisition device can be determined based on the device status. If the device status of the camera acquisition device is normal, then S2033 is executed.

[0090] S2034: The server extracts target images for the target blades based on the target video.

[0091] The server can use a frame-slicing algorithm to extract the target video and obtain the target image for the target leaf.

[0092] As one possible implementation, the camera acquisition device can extract a target image of the target blade based on the target video and send the target image to the server.

[0093] S204: The server performs feature processing on the target image to obtain the attribute features of the target image.

[0094] Attribute features are characteristics used to describe image attributes, such as image feature values.

[0095] S205: The server determines the state of the target blade based on its attribute characteristics.

[0096] In related technologies, after deploying the fault detection algorithms described in S204 and S205, wind turbine fault alarm systems typically collect image data periodically and then upgrade the algorithm based on the collected image data. This means the fault detection algorithm cannot be updated in real time, and algorithm developers cannot obtain data on the algorithm's operational status. Once a wind turbine fault alarm system issues a false alarm, and the blade state remains unchanged, false alarms will typically continue, severely impacting the system's effectiveness. Even manually clearing the algorithm's state does not resolve the false alarm problem.

[0097] Based on this, after determining the state of the target blade, the embodiments of this application no longer directly send it to the video monitoring terminal for display. Instead, they perform fault detection again based on the state. In the second fault detection, the image of the false alarm is used as the basis for whether the target blade is actually faulty. That is, if the target image is similar to the image of the false alarm, it means that the target blade is not actually faulty, but a false alarm, thereby reducing the number of false alarms and the number of consecutive false alarms.

[0098] It should be noted that since there is a second fault detection in this application embodiment, when the first fault detection result, that is, when the target blade is in a fault state, the target blade may not be a real fault, and can be called a "pre-fault state". Similarly, when the target blade is in a normal state, the target blade may not be a real normal state, and can be called a "pre-normal state". This will be explained in detail below.

[0099] S206: If the target blade is in a pre-fault state, the server determines the detection result of the target blade based on the target image and the set of suppressed images.

[0100] The image set for suppression includes images of false alarms about the target blade. These false alarm images are determined based on the re-inspection results and the detection results. The accuracy of the re-inspection results is higher than that of the detection results. If the re-inspection results differ from the detection results, it indicates that fault alarms based solely on the detection results will result in false alarms. Therefore, corrections can be made based on the re-inspection results.

[0101] Specifically, when performing fault detection on an image, if the re-inspection result differs from the detection result, it indicates that the image is a false alarm. This image can be added to the suppression image set, and the images in the suppression image set can be used as the basis for correction so that subsequent target blades are in a pre-fault state. In order to further determine whether the target blade is a false alarm, the detection result of the target blade can be determined based on the target image and the suppression image set, thereby improving the accuracy of the detection result.

[0102] As one possible implementation, the fault detection algorithm can also be updated based on the re-inspection results. It should be noted that the fault detection algorithm proposed in this application includes the first fault detection algorithm as described in S204-S205, and the second fault detection algorithm as described in S206. The fault detection algorithm will be described in detail later, and will not be repeated here.

[0103] S207: If the detection result indicates a fault in the target blade, the server sends the detection result to the video monitoring terminal.

[0104] S208: The video monitoring terminal displays the detection results.

[0105] As one possible implementation, the video detection terminal can issue a fault alarm based on the detection results, or when the server determines that the detection result is a fault in the target blade, it can not only send the detection result to the video monitoring terminal, but also send the fault alarm indication information to the video monitoring terminal so that the video monitoring terminal can issue a fault alarm based on the fault alarm indication information. This application does not impose any specific restrictions on this.

[0106] S209: The video monitoring terminal acquires the re-inspection results of the detection results.

[0107] Among them, the accuracy of the retest results is higher than that of the test results.

[0108] In one possible implementation, the re-inspection result can be the result of the video monitoring terminal re-inspecting the detection result. For example, the terminal device has a unit health diagnosis algorithm (such as support vector machine algorithm, wavelet neural network algorithm, etc.) for re-inspecting the detection result. The signal reflecting the state of the unit blades collected by the sensor (such as the icing signal collected by the icing sensor, etc.) is input into the unit health diagnosis algorithm to obtain the re-inspection result for the detection result.

[0109] In one example, severe blade icing is often accompanied by wind power mismatch faults, leading to unit shutdown. Unit health diagnostic algorithms based on unit status information, such as cold wave algorithms and icing sensors collecting icing signals (not standard), also reflect the icing status of the unit blades. Theoretically, unit status information and unit health diagnostic algorithms can be combined to replace the manual re-inspection process. Assuming a cold wave occurs and the unit's wind power mismatch status is acquired for timing-based judgment, this serves as a backup judgment basis for the manual re-inspection of the icing identification algorithm in the current blade video monitoring system.

[0110] In another example, turbine blade fracture is often accompanied by blade cracks. The vibration signal generated by a cracked blade differs from that generated by a blade without cracks. Vibration sensors can be used to collect vibration signals from the turbine nacelle and main shaft, which can then be input into a turbine health diagnostic algorithm (such as a diagnostic algorithm) to obtain a re-inspection result based on the initial detection.

[0111] In another example, if the blades are struck by lightning or cracked, abnormalities will occur that are different from usual. Sound signals can be collected by sound sensors and input into the unit health diagnosis algorithm (such as lightning strike, crack, and fissure algorithm) to obtain a re-inspection result based on the detection results.

[0112] In another example, a machine learning model or a neural network model can be used. Furthermore, for different business needs, multiple sub-models can be included, each sub-model corresponding to the identification of a function (icing, fracture, lightning strike, cracking, fissure, etc.). In this case, the terminal can set and display different options for different functions, and the owner or maintenance personnel can select different functions. The terminal calls different models for identification and re-inspection.

[0113] In another possible implementation, the re-inspection results may come from the owner or on-site maintenance personnel, etc.

[0114] In related technologies, data collection requires manual screening of data, observation of operational status, and labeling. That is, after collecting images of wind turbine blades, manual labeling is necessary to retrain the model and optimize the fault detection algorithm, resulting in a significant workload and a long optimization time. Using this application, the process of obtaining re-inspection results is equivalent to manually labeling the images, thus reducing the workload for optimizing the fault detection algorithm.

[0115] In one example, field maintenance personnel do not need to review all test results. They can review only the test results for the target blade fault to determine whether they are false alarms. This not only reduces the number of false alarms but also reduces the workload of field maintenance personnel.

[0116] In another example, for cases of continuous false alarms, the number of times each type of fault alarm blade image appears can be counted. When the number of occurrences exceeds a preset threshold, the blade image of that type of fault alarm is more likely to be a continuous false alarm. Therefore, such blade images can be reviewed. Through targeted review, the workload of on-site maintenance personnel can be further reduced.

[0117] S210: If the re-inspection result shows that the target blade is faulty, issue a fault alarm.

[0118] As described above, this application provides a wind turbine fault alarm system, including a video monitoring terminal and a server. The video monitoring terminal acquires a detection request for a target image including a target blade and sends it to the server. The server performs feature processing on the target image based on the detection request to obtain the attribute features of the target image. Based on the attribute features, it determines the state of the target blade. If the target blade is in a pre-fault state, it determines the detection result of the target blade based on the target image and a set of suppressed images, and sends the detection result to the video monitoring terminal. The video monitoring terminal acquires a re-inspection result based on the detection result. The accuracy of the re-inspection result is higher than that of the detection result, making the fault alarm notification based on the re-inspection result more accurate. Specifically, the images that trigger false alarms can be identified based on the re-inspection result and the detection result, thereby constructing a set of suppressed images. This improves the accuracy of determining the fault condition of the target blade based on the set of suppressed images and the target image, reduces the number of false alarms and the number of consecutive false alarms, and improves the timeliness of repairing genuinely faulty blades.

[0119] The following example, using the detection of whether blades are icing, illustrates the aforementioned fault detection algorithm. (See also...) Figure 3 The figure is a flowchart of a fault detection algorithm provided in an embodiment of this application.

[0120] S301: Start.

[0121] S302: Acquire the target image.

[0122] Understandably, the target image can be an image of any blade included in any wind turbine in a wind power plant.

[0123] S303: Execute the blade diagnostic algorithm.

[0124] It can be understood that this blade diagnostic algorithm is used to implement the function of the first fault detection algorithm as described in S204-S205. Taking the diagnostic model as an example, the target image is input into the diagnostic model to obtain the attribute features of the target blade.

[0125] See Figure 4 The figure is a schematic diagram of a diagnostic algorithm provided in an embodiment of this application. The target image is preprocessed to obtain a distance BGR (blue, green, red) three-color map, which is then input into the diagnostic model to obtain the binary classification one-hot confidence score of each pixel, thereby determining the state of the target leaf.

[0126] The state is a binary classification result for the fault detection type, which can be divided into pre-fault state and non-fault state. For example, if the fault detection for the target blade is icing detection, the detection result can be that the target blade is iced or not iced. The pre-fault state is that the target blade is iced, and the non-fault state is that it is not iced.

[0127] S304: Whether a pre-fault state exists.

[0128] The state of the target blade, i.e., icing or non-icing, is determined based on its properties. Icing is considered a pre-failure state.

[0129] S305: Execute the similarity detection algorithm.

[0130] If the target blade is in a pre-fault state, a similarity detection algorithm is executed. It is understood that this similarity detection algorithm is used to implement the function of the second fault detection algorithm as described in S206.

[0131] Taking a similarity detection model as an example, the target image and the set of suppressed images are input into the similarity detection model. The similarity between the attribute features of the target image and the attribute features of the other images is determined. If the attribute features of the target image and the attribute features of the other images meet the similarity condition, the detection result is that the target blade is not faulty; if the attribute features of the target image and the attribute features of the other images do not meet the similarity condition, the detection result is that the target blade is faulty. Here, the other images are those from the set of suppressed images that falsely alarmed the target blade.

[0132] The embodiments of this application do not specifically limit the similarity conditions. For example, if the similarity between the attribute features of the target image and the attribute features of other images is greater than 90%, then the attribute features of the target image and the attribute features of other images satisfy the similarity conditions.

[0133] As one possible implementation, the diagnostic model extracts the attribute characteristics of the target image in the same way as the similarity detection model. That is, the similarity detection model can use the same feature extraction steps as the diagnostic model. For example... Figure 5 As shown, after data preprocessing, the target image is determined to be in a pre-fault state (icing) by the diagnostic model. The similarity detection model takes the last layer of the diagnostic model as the image feature value, resulting in 1280 data points. Through the similarity detection method, the target image is found to be similar to the image of the leaf with slight snow cover in the suppressed image set (the image of the false alarm). That is, the detection result of the target leaf is that it is not faulty and can be processed without alarm.

[0134] It should be noted that similarity detection methods include, but are not limited to, Minkowski distance, Euclidean distance, Manhattan distance, Chebyshev distance, Mahalanobis distance, cosine similarity, Pearson correlation coefficient, Hamming distance, Jaccard similarity coefficient, edit distance, DTW distance, and KL distance. This application does not make any specific limitations on these methods.

[0135] S306: Are they similar?

[0136] By performing two fault detections, if the target image is in a pre-fault state but is similar to an image in the suppressed image set that triggers a false alarm, no alarm is triggered. This effectively reduces false alarms for special repetitive images and allows maintenance personnel to decide not to trigger an alarm for certain images, thereby significantly improving the accuracy and effectiveness of the early warning algorithm.

[0137] There are two situations where the retest results differ from the original test results, which will be explained below.

[0138] The first scenario: the test result indicates that the target blade is faulty, but the retest result indicates that the target blade is not faulty.

[0139] At this time, if a fault alarm is triggered based solely on the detection results, false alarms or even consecutive false alarms may occur. Therefore, in this embodiment, a fault alarm is not triggered based on the re-inspection results, and the target image is added to the suppression image set as the image for false alarms, so that such false alarms can be avoided during the subsequent second fault detection.

[0140] The second scenario: the initial test result shows that the target blade is not faulty, but the retest result shows that the target blade is faulty.

[0141] This indicates that the results obtained through the diagnostic model are not accurate enough. In this case, the diagnostic model can be updated based on the target image and the re-inspection results to avoid this type of missed alarm. As one possible implementation, the images of missed alarms can be added to the fault optimization set.

[0142] Therefore, by using the set of suppressed images and the set of fault optimizations that are regularly fed back from the field, algorithm engineers can effectively upgrade the fault detection algorithm, which not only improves the algorithm's performance but also reduces the workload in the algorithm development and optimization process.

[0143] To make the technical solutions provided in the embodiments of this application clearer, the following is in conjunction with... Figure 6 The wind turbine fault alarm method provided in this application is illustrated with an example.

[0144] Step 1: The front end (video detection terminal) of the blade video monitoring system (wind turbine fault alarm system) reads the automatic inspection plan and displays it on the front end of the blade video monitoring system.

[0145] Step 2: The owner or on-site maintenance personnel develop an automated inspection plan.

[0146] Step 3: The written automatic inspection plan will be written to the front-end database of the server.

[0147] Step 4: The algorithm service reads the automatic inspection plan from the front-end database.

[0148] Step 5: The algorithm service generates a call command and sends it to the camera control service so that the target image can be acquired through the camera acquisition device.

[0149] Step 6: The camera control service, following the call command, acquires video of the target blade based on its location, executes a frame extraction algorithm, and sends the results (such as the current operating status) and device status (such as whether the device is normal) back to the algorithm service. At the same time, it transmits the extracted image data (target image) to the host computer's image data folder (the area where the server stores the target image).

[0150] Step 7: The algorithm service calls the fault detection algorithm to perform fault detection using image data and a fault suppression database (a set of suppressed images).

[0151] Step 8: The fault detection algorithm sends the detection results back to the algorithm service.

[0152] Step 9: The algorithm service writes the automatic inspection results (including detection results) into the front-end database.

[0153] Step 10: The front-end of the blade video monitoring system updates and displays the automatic inspection results in the front-end database.

[0154] Step 11: On-site maintenance personnel discover and review the test results.

[0155] On-site maintenance personnel review the test results indicating a fault in the target blade (alarm data review), and even conduct regular spot checks on the test results indicating that the target blade is not faulty (normal data regular spot checks).

[0156] Step 12: Automatic inspection results (including re-inspection results) are updated to the front-end database.

[0157] Step 13: The algorithm service detects the manual review results (re-inspection results) of the front-end database and processes the results.

[0158] Step 14: The algorithm service adds false alarm data to the alarm suppression library and false negative data to the optimization database.

[0159] Step 15: Algorithm developers regularly recycle the fault suppression database and fault optimization database, optimize the fault detection algorithm, and upgrade the field fault detection after the fault detection algorithm has been optimized to a certain extent.

[0160] Therefore, after fault diagnosis, the blade video monitoring system incorporates two elements: an alarm suppression library and manual alarm screening. It also includes manual sampling and a fault optimization database. Regularly collecting data from the alarm suppression and fault optimization databases from the field allows for optimization of the blade fault detection algorithm, resulting in increasingly better overall performance of the blade video monitoring system. This effectively avoids the adverse effects of false alarms caused by the fault detection algorithm detecting images it hasn't learned from; it highlights the blade video monitoring system's ability to identify known fault alarms; and it reduces the workload of algorithm development or optimization engineers in screening field data.

[0161] In addition to the wind turbine fault alarm method provided in this application, a wind turbine fault alarm device is also provided, which is applied to a video monitoring terminal, such as... Figure 7 As shown, it includes: an acquisition unit 701, a receiving unit 702, a display unit 703, and an alarm unit 704;

[0162] The acquisition unit 701 is used to acquire a detection request, the detection request being used to acquire a target image at a preset time, the target image including the target blade of a wind turbine.

[0163] The receiving unit 702 is used to receive the detection result of the server performing fault detection on the target blade according to the detection request;

[0164] The display unit 703 is used to display the detection results;

[0165] The acquisition unit 701 is further configured to acquire a retest result for the detection result, wherein the accuracy of the retest result is higher than the accuracy of the detection result;

[0166] The alarm unit 704 is used to issue a fault alarm prompt if the re-inspection result is that the target blade is faulty.

[0167] As one possible implementation, the device further includes a sending unit, configured to send the re-inspection result to the server if the re-inspection result is inconsistent with the detection result, so that the server updates the fault detection algorithm based on the re-inspection result.

[0168] As described above, for wind turbine blades, after receiving a detection request to acquire target images of the target blade at a preset time, the video monitoring terminal sends the request to the server. The server performs fault detection on the target blade based on the detection request and sends the detection results to the video monitoring terminal. The video monitoring terminal displays the detection results and obtains a re-inspection result based on the detection results. If the re-inspection result indicates a fault in the target blade, a fault alarm is issued. Therefore, by correcting the less accurate detection results based on the more accurate re-inspection results, the number of false alarms can be reduced, and the accuracy of alarms for target blade faults can be improved.

[0169] In addition to the wind turbine fault alarm method provided in this application, a wind turbine fault alarm device is also provided, which is applied to a server, such as... Figure 8 As shown, it includes: an acquisition unit 801, an algorithm unit 802, and a sending unit 803;

[0170] The acquisition unit 801 is used to acquire a target image according to a detection request, wherein the target image includes the target blade of a wind turbine.

[0171] The algorithm unit 802 is used to perform feature processing on the target image to obtain the attribute features of the target image; determine the state of the target blade according to the attribute features; if the target blade is in a pre-fault state, determine the detection result of the target blade according to the target image and the set of suppressed images; wherein, the set of suppressed images includes images of false alarms of the target blade, and the images of false alarms are determined according to the re-inspection result and the detection result;

[0172] The sending unit 803 is used to send the detection result to the video detection terminal if the detection result indicates that the target blade is faulty, so that the video monitoring terminal can issue a fault alarm.

[0173] As one possible implementation, the algorithm unit 802 is used for:

[0174] The target image and the suppressed image set are input into a similarity detection model to determine the similarity between the attribute features of the target image and the attribute features of other images, wherein the other images are the images in the suppressed image set that falsely alarm the target leaf.

[0175] If the attribute features of the target image and the attribute features of other images meet the similarity condition, the detection result is that the target blade is not faulty; if the attribute features of the target image and the attribute features of other images do not meet the similarity condition, the detection result is that the target blade is faulty.

[0176] As one possible implementation, the device further includes an update unit, configured to add the target image as a false alarm image to the suppression image set if the detection result indicates that the target blade is faulty and the re-inspection result indicates that the target blade is not faulty.

[0177] As one possible implementation, the algorithm unit 802 is used to input the target image into the diagnostic model to obtain the attribute features of the target leaf; the update unit is used to update the diagnostic model according to the target image and the re-inspection result if the detection result is that the target leaf is not faulty and the re-inspection result is that the target leaf is faulty.

[0178] As one possible implementation, the diagnostic model extracts the attribute characteristics of the target image in the same way as the similarity detection model.

[0179] As one possible implementation, the acquisition unit 801 is used for:

[0180] A call command is sent to the camera acquisition device according to the detection request, and the call command includes the target position information of the target blade;

[0181] Receive target video, including the target blade, acquired by the camera acquisition device based on the target location information;

[0182] Target images for the target blades are extracted from the target video.

[0183] As one possible implementation, the device further includes a checking unit for:

[0184] Obtain the device status of the camera acquisition device;

[0185] If the device is in normal condition, the step of receiving the target video, including the target blade, acquired by the camera acquisition device based on the target location information is executed.

[0186] As described above, the server performs feature processing on the acquired target image and target blade based on the detection request, obtaining the attribute features of the target image and thus determining the state of the target blade. If the target blade is in a pre-fault state, the detection result of the target blade is determined based on the target image and the set of suppressed images. If the target blade is faulty, the detection result is sent to the video detection terminal so that the video monitoring terminal can issue a fault alarm. The video monitoring terminal obtains a re-inspection result based on the detection result. This re-inspection result has high accuracy, and the set of suppressed images determined based on this re-inspection result also has high accuracy. This improves the accuracy of determining the fault status of the target blade based on the set of suppressed images and the target image, reduces the number of false alarms and the number of consecutive false alarms, and improves the timeliness of repairing truly faulty blades.

[0187] This application also provides a computer device, see [link to relevant documentation] Figure 9 The figure illustrates a structural diagram of a computer device provided in an embodiment of this application, such as... Figure 9 As shown, the device includes a processor 910 and a memory 920:

[0188] The memory 910 is used to store program code and transmit the program code to the processor;

[0189] The processor 920 is used to execute any of the wind turbine fault alarm methods provided in the above embodiments according to the instructions in the program code.

[0190] This application provides a computer-readable storage medium for storing a computer program that executes any of the wind turbine fault alarm methods provided in the above embodiments.

[0191] This application also provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the wind turbine fault alarm method provided in the various optional implementations of the above aspects.

[0192] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

[0193] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0194] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0195] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0196] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for alarming faults in a wind turbine generator, characterized in that, The method, applied to a video monitoring terminal, includes: Obtain a detection request, the detection request being used to request the acquisition of a target image at a preset time, the target image including the target blades of a wind turbine; The receiving server receives the detection result of fault detection on the target blade according to the detection request; the detection result is determined using a similarity detection model based on the target image and a set of suppressed images; the set of suppressed images includes images of false alarms from the target blade. Display the test results; Obtain a retest result based on the test result, wherein the accuracy of the retest result is higher than the accuracy of the test result; If the re-inspection result indicates a fault in the target blade, a fault alarm is issued; the re-inspection result is also used to jointly determine the set of suppressed images with the detection result.

2. The method according to claim 1, characterized in that, The method further includes: If the re-inspection result is inconsistent with the detection result, the re-inspection result is sent to the server so that the server can update the fault detection algorithm based on the re-inspection result.

3. A method for alarming faults in a wind turbine generator, characterized in that, Applied to a server, the method includes: A target image is acquired according to a detection request, the target image including the target blades of a wind turbine; The target image is subjected to feature processing to obtain the attribute features of the target image; The state of the target blade is determined based on the attribute characteristics; If the target blade is in a pre-fault state, a similarity detection model is used to determine the detection result of the target blade based on the target image and the set of suppressed images; a re-inspection result is obtained based on the detection result, and the accuracy of the re-inspection result is higher than that of the detection result; wherein, the set of suppressed images includes images of false alarms of the target blade, and the images of false alarms are determined based on the re-inspection result and the detection result; If the detection result indicates a fault in the target blade, the detection result is sent to the video detection terminal so that the video monitoring terminal can issue a fault alarm.

4. The method according to claim 3, characterized in that, The step of determining the detection result of the target leaf based on the target image and the set of suppressed images includes: The target image and the suppressed image set are input into a similarity detection model to determine the similarity between the attribute features of the target image and the attribute features of other images, wherein the other images are the images in the suppressed image set that falsely alarm the target leaf. If the attribute features of the target image and the attribute features of other images meet the similarity condition, the detection result is that the target blade is not faulty; If the attribute features of the target image and the attribute features of the other images do not meet the similarity condition, the detection result is that the target blade is faulty.

5. The method according to claim 4, characterized in that, The method further includes: If the detection result indicates that the target blade is faulty, and the re-inspection result indicates that the target blade is not faulty, the target image is added to the suppression image set as a false alarm image.

6. The method according to claim 5, characterized in that, The step of performing feature processing on the target image to obtain the attribute features of the target image includes: The target image is input into the diagnostic model to obtain the attribute features of the target leaf; The method further includes: If the detection result indicates that the target blade is not faulty, and the re-inspection result indicates that the target blade is faulty, the diagnostic model is updated based on the target image and the re-inspection result.

7. The method according to claim 6, characterized in that, The diagnostic model extracts the attribute characteristics of the target image in the same way as the similarity detection model.

8. The method according to claim 3, characterized in that, The detection request acquires the target image, including: A call command is sent to the camera acquisition device according to the detection request, and the call command includes the target position information of the target blade; Receive target video, including the target blade, acquired by the camera acquisition device based on the target location information; Target images for the target blades are extracted from the target video.

9. The method according to claim 8, characterized in that, The method further includes: Obtain the device status of the camera acquisition device; If the device is in normal condition, the step of receiving the target video, including the target blade, acquired by the camera acquisition device based on the target location information is executed.

10. A wind turbine generator fault alarm system, characterized in that, The system includes a terminal and a server: The terminal device is used to perform the method according to claim 1 or 2; The server is used to perform the method according to any one of claims 3-9.

11. The system according to claim 10, characterized in that, The system also includes a camera acquisition device; The camera acquisition device is used to acquire target video including the target blade according to the target position information of the target blade included in the call instruction after receiving the call instruction.

12. A wind turbine generator fault alarm device, characterized in that, The device, used in video monitoring terminals, includes: an acquisition unit, a receiving unit, a display unit, and an alarm unit; The acquisition unit is used to acquire a detection request, the detection request being used to request the acquisition of a target image at a preset time, the target image including the target blade of a wind turbine. The receiving unit is configured to receive the detection result of the server performing fault detection on the target blade according to the detection request; the detection result is determined using a similarity detection model based on the target image and a set of suppressed images; the set of suppressed images includes images of false alarms from the target blade; The display unit is used to display the detection results; The acquisition unit is further configured to acquire a retest result for the detection result, wherein the accuracy of the retest result is higher than the accuracy of the detection result; The alarm unit is used to issue a fault alarm if the re-inspection result indicates that the target blade is faulty; the re-inspection result is also used to jointly determine the set of suppressed images with the detection result.

13. A wind turbine generator fault alarm device, characterized in that, Applied to a server, the device includes: an acquisition unit, an algorithm unit, and a sending unit; The acquisition unit is used to acquire a target image according to a detection request, wherein the target image includes the target blades of a wind turbine. The acquisition unit is further configured to acquire a retest result for the detection result, wherein the accuracy of the retest result is higher than the accuracy of the detection result; The algorithm unit is used to perform feature processing on the target image to obtain the attribute features of the target image; determine the state of the target blade based on the attribute features; if the target blade is in a pre-fault state, determine the detection result of the target blade using a similarity detection model based on the target image and a set of suppressed images; wherein, the set of suppressed images includes images of false alarms of the target blade, and the images of false alarms are determined based on the re-inspection result and the detection result; The sending unit is used to send the detection result to the video detection terminal if the detection result indicates that the target blade is faulty, so that the video monitoring terminal can issue a fault alarm.

14. A computer device, characterized in that, The device includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the method of claim 1 or 2 according to the instructions in the program code, or to execute the method of any one of claims 3-9.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method of claim 1 or 2, or the method of any one of claims 3-9.

16. A computer program product, characterized in that, Includes a computer program or instructions; when the computer program or instructions are executed by a processor, the method of claim 1 or 2 is performed, or the method of any one of claims 3-9 is performed.

Images