Wind power generation equipment monitoring device, wind power generation equipment monitoring method, and program

Abstract

The present invention provides a wind power generation equipment monitoring device that can ensure the accuracy of anomaly detection even when it is not possible to continuously acquire operating data from the equipment. [Solution] A wind power generation equipment monitoring device according to one embodiment includes: a feature extraction unit that extracts features from an acoustic signal indicating the detection result of sound generated by at least one piece of equipment installed in the power generation facility; a storage unit that stores machine learning models pre-classified according to the operating state of the equipment using sound; an equipment state estimation unit that identifies the operating state of the equipment based on the acoustic signal and estimates whether or not there is an abnormality in the equipment using a machine learning model corresponding to the identified operating state; and a notification unit that outputs an alert signal indicating an abnormality in the equipment.

Description

Technical Field

[0001] Embodiments of the present invention relate to a wind power equipment monitoring device, a wind power equipment monitoring method, and a program.

Background Art

[0002] In power generation facilities such as wind farms and hydroelectric power plants where access is difficult, when an abnormality occurs in the equipment, there is a cost for workers to access, so the cost associated with maintenance increases. Therefore, simple and accurate detection of equipment abnormalities is required. As a means of ensuring the accuracy of abnormality detection, a method of constantly acquiring the operation data of the equipment installed in the power generation facility and detecting abnormalities can be considered. However, when the equipment in the power generation facility is a product of another company or owned by another operator, it is not always possible to constantly acquire the operation data of the equipment.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Embodiments of the present invention provide a wind power equipment monitoring device, a wind power equipment monitoring method, and a program capable of ensuring the accuracy of abnormality detection of the equipment even when the operation data of the equipment cannot be constantly acquired.

Means for Solving the Problems

[0005] A wind power generation equipment monitoring device according to one embodiment includes: a feature extraction unit that extracts features from an acoustic signal indicating the detection result of sound generated by at least one piece of equipment installed in the power generation facility; a storage unit that stores machine learning models pre-classified according to the operating state of the equipment using sound; an equipment state estimation unit that identifies the operating state of the equipment based on the acoustic signal and estimates whether or not there is an abnormality in the equipment using a machine learning model corresponding to the identified operating state; and a notification unit that outputs an alert signal indicating an abnormality in the equipment. [Effects of the Invention]

[0006] According to this embodiment, it is possible to ensure the accuracy of abnormality detection in equipment even if operating data of the equipment cannot be acquired at all times. [Brief explanation of the drawing]

[0007] [Figure 1] This is a block diagram showing the configuration of a wind power generation equipment monitoring device according to the first embodiment. [Figure 2] (A) is a schematic diagram showing an example of a fixed-bottom offshore wind turbine, and (B) is a schematic diagram showing an example of a floating offshore wind turbine. [Figure 3] This is a schematic diagram illustrating the interior of a wind turbine nacelle. [Figure 4] (A) is a figure showing the waveform of the acoustic signal output from the acoustic sensor, (B) is a figure showing the result of processing the acoustic signal with a Fast Fourier Transform, and (C) is a figure showing an example of dividing the frequency band of the acoustic signal into 1 / 3 octave bands. [Figure 5] This is a data distribution diagram showing an example of wind turbine operating conditions. [Figure 6] This is a correlation map diagram showing an example of the correlation between the rotational speed of the main engine and acoustic frequency. [Figure 7] This diagram shows an example of the operation classification of the main engine and auxiliary engine. [Figure 8] A flowchart illustrating the method for monitoring wind turbines is shown. [Figure 9] This figure shows an example of the relationship between the rotational speed component of a rotating machine and its acoustic frequency. [Figure 10] This figure shows an example of the waveform of the acoustic signal from an acoustic sensor that detected wind noise from the blade. [Figure 11] This is a block diagram showing the configuration of the yaw control device according to the fourth embodiment. [Figure 12] This is a block diagram showing an example configuration of a mobile device. [Figure 13] This is a block diagram illustrating a wind power generation equipment monitoring method according to the fifth embodiment. [Figure 14] This is a block diagram illustrating a wind power generation equipment monitoring method according to the sixth embodiment. [Figure 15] This is a schematic diagram of the main engine and auxiliary engine according to the sixth embodiment. [Figure 16] This is a block diagram illustrating a wind power generation equipment monitoring method according to the seventh embodiment. [Modes for carrying out the invention]

[0008] Embodiments of the present invention will be described below with reference to the drawings. The embodiments described below are not intended to limit the present invention.

[0009] (First Embodiment) Figure 1 is a block diagram showing the configuration of a wind power generation equipment monitoring device according to the first embodiment. The wind power generation equipment monitoring device 10 according to this embodiment is a device that monitors the main engine (blades, gearbox, generator, bearings, etc.) and auxiliary equipment installed in a wind turbine for wind power generation. However, the equipment monitored by the wind power generation equipment monitoring device 10 is not limited to the main engine and auxiliary equipment, and may also include, for example, rotating equipment such as turbines that rotate using water power installed in hydroelectric power generation, or equipment that needs to be operated unattended.

[0010] Figure 2(A) is a schematic diagram showing an example of a fixed-bottom offshore wind turbine. Figure 2(B) is a schematic diagram showing an example of a floating offshore wind turbine. Figure 3 is a schematic diagram showing the interior of the nacelle of the wind turbine 20 shown in Figures 2(A) and 2(B). Note that the wind turbine 20 is not limited to offshore installations, but may also be a ground-based wind turbine.

[0011] The wind turbine 20 shown in FIGS. 2(A) and 2(B) includes a plurality of blades 21, a hub 22, a nacelle 23, and a tower 24. Each blade 21 is radially arranged so as to be connected to a rotor shaft 25 (see FIG. 2) at the hub 22. The pitch angle of each blade 21 is adjusted with respect to the wind inflow direction so that the kinetic energy of the wind can be efficiently converted into rotational energy. A drive mechanism (not shown) such as a motor, a brake, and an emergency power source for adjusting this pitch angle is provided inside the hub 22. The tower 24 is built vertically on the upper surface of a foundation 26 that is built on the seabed and exposed from the sea surface.

[0012] The nacelle 23 is provided at the top of the tower 24 via a yaw control device 411 that automatically causes the rotor shaft 25 to follow the wind direction.

[0013] The nacelle 23 houses the main machine 30 and some auxiliary machines 40. The main machine 30 is a device directly involved in wind power generation. The main machine 30 includes, for example, a rotor shaft 25, a main bearing 311, a speed increaser 312, a generator 313, and a PCS 314 (Power Conditioning System). The rotor shaft 25 rotates together with the blade 21. The main bearing 311 supports the rotor shaft 25. The speed increaser 312 is connected to the end of the rotor shaft 25 and increases the rotational speed of the rotor shaft 25. The generator 313 converts the rotational energy of the rotor shaft 25 into electrical energy. The PCS 314 modulates and transforms the output of the generator 313 and outputs power adjusted to the system frequency and system voltage.

[0014] On the one hand, the auxiliary machine 40 is a device that assists the main machine 30. The auxiliary machine 40 includes, for example, a yaw control device 411, a pitch motor 412, and a cooling device 413. The yaw control device 411 has a yaw motor for changing the yaw angle of the wind turbine 20 and a yaw brake for fixing the yaw angle of the wind turbine 20. The pitch motor 412 is provided in the hub 22 and changes the pitch angle of the blade 21. The cooling device 413 has a cooling fan for cooling the main machine 30 and a cooling water pump motor for circulating cooling water. Note that the main machine 30 and the auxiliary machine 40 are not limited to the above-described devices and may include other devices. For example, the auxiliary machine 40 includes an oil circulation pump.

[0015] An acoustic sensor 50 is installed in each of the main machine 30 and the auxiliary machine 40. The acoustic sensor 50 is composed of, for example, a directional microphone. Each acoustic sensor 50 outputs an acoustic signal indicating the measurement result of the sound generated by the main machine 30 or the auxiliary machine 40. The acoustic signal of each acoustic sensor 50 is acquired by the wind power generation equipment monitoring device 10.

[0016] In this embodiment, the wind power generation equipment monitoring device 10 is housed in the nacelle 23. However, the wind power generation equipment monitoring device 10 may be installed at a location away from the wind turbine 20. The wind power generation equipment monitoring device 10 monitors the states of the main machine 30 and the auxiliary machine 40 based on the above acoustic data signal.

[0017] The wind power generation equipment monitoring device 10 has hardware resources such as a CPU, ROM, RAM, HDD, and GPU, and can be configured as a computer in which information processing by software is realized using the hardware resources by the CPU executing various programs. Furthermore, the wind power generation equipment monitoring method using the wind power generation equipment monitoring device 10 can be realized by causing a computer to execute various programs.

[0018] Now, returning to Figure 1, the configuration of the wind power generation equipment monitoring device 10 according to this embodiment will be described. The wind power generation equipment monitoring device 10 according to this embodiment includes a feature extraction unit 11, a model creation unit 12, a storage unit 13, an equipment status estimation unit 14, and a notification unit 15.

[0019] The feature extraction unit 11 extracts features from the acoustic signal of the acoustic sensor 50. Here, an example of the feature extraction process will be explained with reference to Figures 4(A) to 4(C).

[0020] Figure 4(A) shows the waveform of the acoustic signal output from the acoustic sensor 50. The feature extraction unit 11 performs a Fast Fourier Transform (FFT) on the acoustic signal shown in Figure 4(A). As a result, the frequency components of the acoustic signal are extracted as features, as shown in Figure 4(B).

[0021] Furthermore, in this embodiment, the feature extraction unit 11 uses a noise analysis method to reduce the number of data points while maintaining the characteristics of the acoustic signal. As a result, the frequency band of the acoustic signal shown in Figure 4(B) is divided into 1 / 1 octave bands and 1 / 3 octave bands, as shown in Figure 4(C).

[0022] The model creation unit 12 creates a machine learning model, which is an equipment state estimation model 120, for each operating state of the main unit 30 and the auxiliary unit 40. Here, the method of creating the equipment state estimation model 120 by the model creation unit 12 will be explained with reference to Figures 5 to 7.

[0023] Figure 5 is a data distribution diagram showing an example of the operating conditions of the wind turbine 20. In Figure 5, the vertical axis represents the rotational speed of the main engine 30 (e.g., the rotor shaft 25), and the horizontal axis represents the frequency of the sound generated by the main engine 30.

[0024] When creating the equipment state estimation model 120, the model creation unit 12 collects rotational speed data from the operating data D1 for a certain period of time. The operating data D1 may be provided, for example, by the manufacturer of the wind turbine 20, or it may be collected by temporarily installing a simple rotation detector. Note that the operating data D1 may include not only the rotational speed of the rotor shaft 25 but also the power output of the generator 313. Therefore, the vertical axis may represent power output instead of rotational speed.

[0025] Furthermore, the model creation unit 12 obtains frequency data corresponding to the rotation speed data by collecting the results of feature extraction processing of the acoustic signals from the acoustic sensor 50 from the feature extraction unit 11 for a certain period of time.

[0026] The model creation unit 12 calculates the regression equation for the approximation line L using, for example, the least squares method, for the data distribution shown in Figure 5. At this time, the model creation unit 12 sets the slope of the created regression equation as the correlation coefficient. Note that the regression equation for the approximation line L is not limited to linear approximation, but may also be polynomial approximation. Furthermore, although the bandwidth of the frequency data shown in Figure 5 corresponds to the bandwidth of the normal operation data D1, it is desirable to select a frequency bandwidth that does not change even under abnormal conditions when calculating the regression equation for the approximation line L.

[0027] Figure 6 is a correlation map diagram showing an example of the correlation between the rotational speed of the main engine 30 and the acoustic frequency. In Figure 6, the vertical axis corresponds to the horizontal axis in Figure 5 and shows the frequency band of the sound generated by the main engine 30. The vertical axis corresponds to the horizontal axis in Figure 6 and shows the rotational speed of the main engine 30. In Figure 6, the correlation coefficient obtained from the regression equation of the approximation line L is mapped onto a two-dimensional coordinate plane consisting of the rotational speed of the main engine 30 and the acoustic frequency.

[0028] The model creation unit 12 uses the correlation map shown in Figure 6 to remove data with a correlation coefficient lower than a predetermined value. Subsequently, the model creation unit 12 uses supervised machine learning methods such as support vector machines and random forests to classify and label the operating states of the main unit 30 and auxiliary unit 40, as shown in Figure 7.

[0029] Figure 7 shows an example of the operation classification of the main engine 30 and auxiliary engine 40. In Figure 7, multidimensional data is visualized as a two-dimensional distribution using the t-SNE method. In Figure 7, the main engine 30 is classified into three types: "load operation," "no-load operation," and "off-board operation."

[0030] Furthermore, the auxiliary equipment 40 is classified into two types: operating and not operating. Note that the correlation map shown in Figure 6 is not created for the auxiliary equipment 40. Whether the auxiliary equipment 40 is operating or not is classified according to the acoustic frequencies extracted by the feature extraction unit 11. The rotational speed of the auxiliary equipment 40 is higher than that of the main unit 30. Therefore, the acoustic frequency of the auxiliary equipment 40 (for example, power supply frequency of 60 Hz or higher) is higher than the acoustic frequency of the main unit 30 (for example, 0.25 Hz). In this way, the acoustic frequency bands of the auxiliary equipment 40 and the main unit 30 do not overlap, so the operating states of the two types of equipment can be distinguished.

[0031] In Figure 7, the separation boundary for each operating classification corresponds to the equipment state estimation model 120. If the data falls inside the separation boundary of an operating classification, that data belongs to that operating classification.

[0032] The model creation unit 12 does not necessarily have to be included as a component of the wind power generation equipment monitoring device 10. In other words, the model creation unit 12 may be provided in a separate device that can be connected to the wind power generation equipment monitoring device 10.

[0033] The memory unit 13 stores the equipment state estimation model 120 created by the model creation unit 12. In addition to the equipment state estimation model 120, the memory unit 13 also stores various other data, such as programs for operations performed by the wind power generation equipment monitoring device 10.

[0034] The equipment state estimation unit 14 identifies the operating state of the main unit 30 and auxiliary unit 40 based on the features extracted by the feature extraction unit 11. The equipment state estimation unit 14 also uses the equipment state estimation model 120, which is classified according to the identified operating state, to perform a One-class SVM ( The presence or absence of abnormalities in the main unit 30 or auxiliary unit 40 is estimated using methods such as support vector graphics (SVM) or autoencoders. For example, the equipment state estimation unit 14 calculates the separation boundary surface shown in the equipment state estimation model 120 as the data variance σ, and estimates features that deviate by 1.2σ from the normal distribution of features as abnormal values. If the above feature is, for example, the acoustic level of the equipment, the equipment state estimation unit 14 may estimate that the equipment is abnormal when the acoustic level exceeds a threshold.

[0035] If the equipment status estimation unit 14 estimates that the main unit 30 and auxiliary unit 40 are malfunctioning, the notification unit 15 outputs an alert signal indicating this. This alert signal is received, for example, by equipment connected to the wind power generation equipment monitoring device 10 via a network. Upon receiving the alert signal, this equipment performs a predetermined warning action. This predetermined warning action includes, for example, outputting a warning image indicating that a malfunction has occurred in the main unit 30 and auxiliary unit 40, and outputting a warning sound.

[0036] Next, we will explain how to monitor the wind turbine 20 using the wind power generation equipment monitoring device 10.

[0037] Figure 8 shows a flowchart of the wind turbine 20 monitoring method. In this flowchart, first, the model creation unit 12 of the wind power generation equipment monitoring device 10 creates the equipment state estimation model 120 shown in Figure 7 (step S101). The method for creating the equipment state estimation model 120 has been described above, so the explanation is omitted here. The model creation unit 12 stores the created equipment state estimation model 120 in the storage unit 13.

[0038] Next, the acoustic sensor 50 detects the sound generated by the main unit 30 and the auxiliary unit 40 (step S102). The acoustic sensor 50 outputs an acoustic signal indicating the detection result to the wind power generation equipment monitoring device 10.

[0039] Next, the feature extraction unit 11 of the wind power generation equipment monitoring device 10 performs feature extraction processing (step S103). In this feature extraction processing, as described above, the feature extraction unit 11 performs frequency analysis processing on the acoustic signal (see Figure 4(A)) using the Fast Fourier Transform (see Figure 4(B)). Subsequently, the feature extraction unit 11 divides the frequency band of the acoustic signal into 1 / 3 octave bands (see Figure 4(C)).

[0040] Next, the equipment state estimation unit 14 performs an abnormality estimation process (step S104). In this abnormality estimation process, as described above, the equipment state estimation unit 14 estimates whether there is an abnormality in the main unit 30 and the auxiliary unit 40 based on whether the operating state identified based on the features extracted by the feature extraction unit 11 falls within or outside the boundary of the operating states classified by the equipment state estimation model 120 stored in the memory unit 13.

[0041] If a feature variable deviates from the equipment state estimation model 120, the equipment state estimation unit 14 estimates that the main unit 30 or auxiliary unit 40 corresponding to that feature variable is abnormal (step S105: YES), and in this case, the notification unit 15 performs an abnormality notification process (step S106). In this abnormality notification process, as described above, the notification unit 15 notifies the user of the abnormality of the main unit 30 or auxiliary unit 40 by displaying an image, outputting sound, etc.

[0042] According to the embodiment described above, the model creation unit 12 pre-creates equipment state estimation models 120 classified according to the normal operating state of the equipment (main unit 30 and auxiliary unit 40) installed on the wind turbine 20, based on acoustic and operating data of the equipment. Then, the equipment state estimation unit 14 identifies the operating state of the equipment based on the characteristic quantities of the acoustic signal from the acoustic sensor 50, and estimates whether or not there is an abnormality in the equipment using the equipment state estimation model 120 corresponding to the identified operating state.

[0043] Therefore, even if it is not possible to continuously acquire operating data from the equipment, it is possible to ensure the accuracy of anomaly detection in that equipment.

[0044] (Second Embodiment) In the second embodiment, the main machine 30 or auxiliary machine 40 is a rotating device. Also, the process by which the feature extraction unit 11 extracts feature quantities from the acoustic signal acquired from the acoustic sensor 50 differs from that of the first embodiment.

[0045] In this embodiment, the feature extraction unit 11 performs frequency analysis of the acoustic signal using FFT processing. Subsequently, the feature extraction unit 11 extracts fluctuating components that depend on the rotational speed of the rotating equipment using methods such as waterfall analysis, and identifies the frequency of the rotational speed component. Here, the method for identifying the acoustic frequency from the rotational speed component will be explained with reference to Figure 9.

[0046] Figure 9 shows an example of the relationship between the rotational speed component of a rotating machine and the acoustic frequency. In Figure 9, the horizontal axis represents time, and the vertical axis represents the acoustic frequency. In this embodiment, the feature extraction unit 11 performs frequency analysis on the acoustic signal within a predetermined time window (e.g., 3 seconds). Subsequently, the feature extraction unit 11 performs frequency analysis while shifting the results of the frequency analysis within the time window by predetermined time intervals (e.g., 0.1 seconds). As a result, as shown in Figure 9, when the rotational speed of the rotating machine increases, the level of a certain acoustic frequency increases with time. The feature extraction unit 11 extracts this fluctuating component.

[0047] The feature extraction unit 11 detects peak values ​​of acoustic frequencies in each time window. Next, the feature extraction unit 11 classifies the detected peak values ​​into those that have fluctuated significantly more than a preset reference value and those that have not changed significantly from the reference value. Subsequently, for the former peak values, the feature extraction unit 11 identifies the acoustic frequencies corresponding to the rotation speed component as features.

[0048] Subsequently, the equipment state estimation unit 14 identifies the operating state of the rotating equipment based on the acoustic frequency, and uses the equipment state estimation model 120 classified by the identified operating state to determine whether or not there is an abnormality in the rotating equipment.

[0049] According to this embodiment described above, similar to the first embodiment, the equipment state estimation model 120 is created in advance by the model creation unit 12, so the equipment state estimation unit 14 can estimate whether or not there is an abnormality in the rotating equipment without having to constantly acquire operating data of the rotating equipment.

[0050] (Third embodiment) In the third embodiment, the wind power generation equipment monitoring device 10 monitors the main unit 30 and auxiliary unit 40 based on the wind noise of the blades 21. Therefore, the acoustic sensor 50 is installed in a location where it can detect the wind noise of the blades 21.

[0051] Figure 10 shows an example of the waveform of the acoustic signal from the acoustic sensor 50 that detects wind noise from the blade 21. For example, when the rotation speed of the rotor shaft 25 is set to 0.25 Hz, the blade 21 rotates once every 4 seconds. Therefore, as shown in Figure 10, the acoustic sensor 50 detects wind noise as a peak value of the acoustic signal every 4 seconds.

[0052] As shown in Figures 2(a) and 2(b), when the wind turbine 20 has three blades 21, the acoustic sensor 50 detects wind noise three times in four seconds. Therefore, when the rotor shaft 25 is rotating at its rated speed, the interval between wind noises detected by the acoustic sensor 50 is 0.75 Hz.

[0053] As the wind speed increases, the wind noise from the blades 21 increases, and the output power of the generator 313 increases. Therefore, there is a correlation between the acoustic level of the wind noise detected by the acoustic sensor 50 and the output power. In the equipment state estimation model 120 created by the model creation unit 12, the operating state of the main unit 30 is classified as either under load or stopped, according to the value of the output power corresponding to the acoustic level of the wind noise.

[0054] In this embodiment, the feature extraction unit 11 extracts the level of the acoustic signal as a feature.

[0055] Furthermore, the equipment state estimation unit 14 identifies the operating state of the main unit 30 based on the output power corresponding to the sound level extracted by the feature extraction unit 11. Subsequently, the equipment state estimation unit 14 estimates whether or not there is an abnormality using the equipment state estimation model 120 corresponding to the identified operating state. At this time, as in the first embodiment, if the identified operating state deviates from the equipment state estimation model 120, the equipment state estimation unit 14 estimates that the main unit 30 is abnormal.

[0056] According to this embodiment described above, similar to the first embodiment, the equipment state estimation model 120 is created in advance by the model creation unit 12, so the equipment state estimation unit 14 can estimate whether or not there is an abnormality in the rotating equipment without having to constantly acquire operating data of the rotating equipment.

[0057] (Fourth Embodiment) Figure 11 is a block diagram showing the configuration of a yaw control device 411 according to the fourth embodiment. Multiple yaw motors 421 are arranged in this yaw control device 411. The sound generated by each yaw motor 421 is detected by a mobile device 60. Now, with reference to Figure 12, the configuration of the mobile device 60 will be described.

[0058] Figure 12 is a block diagram showing an example configuration of the mobile device 60. The mobile device 60 shown in Figure 12 is, for example, a multi-rotor drone and includes a positioning unit 61, a distance measuring unit 62, a camera 63, a microphone 64, a drive unit 65, a processing unit 66, a flight controller 67, and a communication unit 68. Note that the mobile device 60 is not limited to unmanned aerial vehicles such as drones, but may also be, for example, a mobile robot.

[0059] The positioning unit 61 receives position data transmitted from, for example, GPS (Global Positioning System) or GNSS (Satellite positioning, navigation and timing system). In indoor environments, the position is obtained by SLAM (Simultaneous Localization and Mapping), which estimates the position from Lidar or camera information. The positioning unit 61 also periodically transmits the received position data to the flight controller 67.

[0060] The distance measuring unit 62 measures, for example, the distance to each yaw motor 421 being inspected. The distance measuring unit 62 is composed of, for example, a 3D LiDAR (Light Detection And Ranging). The 3D LiDAR measures the time from when a laser beam is emitted from a light source until the laser beam reflected by an object is received by a light receiving unit, and calculates the distance to each yaw motor 421 based on the measured time. The distance measuring unit 62 periodically transmits data showing the calculation result to the processing unit 66. The distance measuring unit 62 is used so that the mobile device 60 can avoid collisions with each yaw motor 421 and maintain a constant distance.

[0061] Camera 63 can photograph each yaw motor 421. Camera 63 periodically transmits image data showing the captured image to the processing unit 66.

[0062] Microphone 64 is an acoustic sensor that detects the sound generated by each yaw motor 421. Microphone 64 periodically transmits acoustic signals to the processing unit 66. There may be one microphone or multiple microphones in an array type.

[0063] The drive unit 65 is composed of several components necessary for driving the mobile device 60. These components include, for example, a storage battery, a motor that rotates using power supplied from the storage battery, an ESC (Electric Speed ​​Controller) that controls the motor's rotational speed, and a propeller connected to the motor.

[0064] The processing unit 66 processes various data acquired from the distance measuring unit 62, the camera 63, and the microphone 64. The processing unit 66 is composed of, for example, a computer that processes data according to a predetermined program.

[0065] The flight controller 67 determines the current position of the mobile device 60 based on position data from the positioning unit 61. The flight controller 67 also controls the drive unit 65 to make the mobile device 60 fly based on instructions from the processing unit 66.

[0066] The communication unit 68 communicates with the wind power generation equipment monitoring device 10. The communication unit 68 transmits, for example, acoustic signals and position data to the feature extraction unit 11.

[0067] As described above, the mobile device 60 detects the sound of each yaw motor 421 using the microphone 64 while flying. When the mobile device 60 moves to a location near an operating yaw motor 421, the level of the sound signal increases at that location. In this embodiment, map data showing the arrangement of the main engine 30 and auxiliary engine 40 is stored in the storage unit 13. Therefore, the equipment state estimation unit 14 uses the map data to check the location data of the location received from the mobile device 60 and identifies the operating yaw motor 421.

[0068] Subsequently, the equipment state estimation unit 14 estimates whether or not there is an abnormality in the identified yaw motor 421 using the equipment state estimation model 120.

[0069] According to this embodiment described above, similar to the first embodiment, the equipment state estimation model 120 is created in advance by the model creation unit 12, so the equipment state estimation unit 14 can estimate whether or not there is an abnormality in the rotating equipment without having to constantly acquire operating data of the rotating equipment.

[0070] Furthermore, in this embodiment, by detecting sound while the microphone 64 is moving, it is possible to identify the device that is in operation from among multiple devices of the same type.

[0071] In this embodiment, a mobile device 60 is used to move one microphone 64 while monitoring for abnormalities in multiple yaw motors 421. However, the monitored devices are not limited to yaw motors 421, but may include multiple devices of the same type included in the main unit 30 and auxiliary unit 40.

[0072] (Fifth embodiment) Figure 13 is a block diagram illustrating a wind power generation equipment monitoring method according to the fifth embodiment. In this embodiment, multiple acoustic sensors 50 fixed at positions far apart from each other are used to detect the sounds of the main unit 30 and auxiliary unit 40.

[0073] When the positions of multiple acoustic sensors 50 are fixed, a time difference (phase difference) occurs between the multiple acoustic sensors 50 when sound generated from a main unit 30 or auxiliary unit 40 installed at a certain point is detected by the multiple acoustic sensors 50. Therefore, the equipment state estimation unit 14 identifies the distance to the sound source, i.e., the equipment that detected the sound, by multiplying this time difference by the speed of sound. At this time, regarding the level difference of the acoustic signals between the multiple acoustic sensors 50, the closer the acoustic sensor 50 is to the sound source, the larger the signal level. Therefore, the equipment state estimation unit 14 also uses the level difference of the acoustic signals to identify the equipment of the sound source.

[0074] Subsequently, the equipment status estimation unit 14 estimates whether or not there is an abnormality in the identified equipment using the equipment status estimation model 120.

[0075] According to this embodiment described above, similar to the first embodiment, the equipment state estimation model 120 is created in advance by the model creation unit 12, so the equipment state estimation unit 14 can estimate whether or not there is an abnormality in the main unit 30 and the auxiliary unit 40 without having to constantly acquire operating data of the main unit 30 and the auxiliary unit 40.

[0076] Furthermore, in this embodiment, it is possible to identify the device currently in operation from among multiple devices based on the time difference of sound detected by multiple acoustic sensors 50 fixed in different locations.

[0077] (Sixth Embodiment) Figure 14 is a block diagram illustrating a wind power generation equipment monitoring method according to the sixth embodiment. The wind power generation equipment monitoring device 10 according to this embodiment uses a light receiving sensor 51 to classify the operating status of the main unit 30 and auxiliary unit 40.

[0078] Figure 15 is a schematic diagram of the main unit 30 and auxiliary unit 40 according to the sixth embodiment. In this embodiment, the main unit 30 and auxiliary unit 40 are rotating machines. A reflective member 52 is attached to a part of the surface of the rotating part of this rotating machine. The reflective member 52 reflects ambient light. The reflected light is received by a light receiving sensor 51. At this time, the light receiving sensor 51 receives the reflected light once when the rotating member rotates once.

[0079] The light receiving sensor 51 outputs a photodetection signal, obtained by photoelectrically converting the detected amount of reflected light, to the feature extraction unit 11. In the feature extraction process, the feature extraction unit 11 performs frequency analysis on the photodetection signal using FFT or the like, and extracts the frequency as a feature.

[0080] The equipment state estimation unit 14 identifies the operating state of the rotating equipment based on the characteristic quantities of the light detection signal. For example, when the rotating equipment is under load, reflected light is detected at intervals of 0.25 Hz. Subsequently, the equipment state estimation unit 14 determines whether or not there is an abnormality based on the equipment state estimation model 120 corresponding to the identified operating state.

[0081] According to this embodiment described above, similar to the first embodiment, the equipment state estimation model 120 is created in advance by the model creation unit 12, so the equipment state estimation unit 14 can estimate whether or not there is an abnormality in the rotating equipment without having to constantly acquire operating data of the rotating equipment.

[0082] Furthermore, in this embodiment, the equipment state estimation unit 14 uses not only the acoustic signal from the acoustic sensor 50 but also the optical detection signal from the light receiving sensor 51 to determine the operating state of the rotating equipment. Therefore, the operating state of the rotating equipment can be determined more accurately, thereby improving the accuracy of abnormality detection.

[0083] (Seventh Embodiment) Figure 16 is a block diagram illustrating a wind power generation equipment monitoring method according to the seventh embodiment. In this embodiment, the wind power generation equipment monitoring device 10 uses a thermographic camera 53 in addition to an acoustic sensor 50 to classify the operating status of the main unit 30 and auxiliary unit 40.

[0084] The thermographic camera 53 measures the temperature of the main unit 30 and the auxiliary unit 40. The measured temperature is sent to the equipment state estimation unit 14. When the main unit 30 and the auxiliary unit 40 are operating, the temperature measured by the thermographic camera 53 rises. Also, immediately after the main unit 30 and the auxiliary unit 40 stop, the temperature falls. Therefore, in this embodiment, the equipment state estimation unit 14 determines the operating state based on the change in the temperature measured by the thermographic camera 53 over time.

[0085] For example, if the temperature change per minute is +5°C / min or more, the device being measured is determined to be in operation. Similarly, if the measured temperature is 40°C or higher and the temperature change per minute is -5°C / min or more, the device is determined to be in operation. On the other hand, if the measured temperature is lower than 40°C and the temperature change per minute is less than -5°C / min, the device is determined to be stopped.

[0086] Subsequently, the equipment status estimation unit 14 determines whether or not there is an abnormality based on the equipment status estimation model 120 corresponding to the identified operating state.

[0087] According to this embodiment described above, similar to the first embodiment, the equipment state estimation model 120 is created in advance by the model creation unit 12, so the equipment state estimation unit 14 can estimate whether or not there is an abnormality in the rotating equipment without having to constantly acquire operating data of the rotating equipment.

[0088] Furthermore, in this embodiment, the equipment state estimation unit 14 identifies the operating state of the rotating equipment not only using the acoustic signal from the acoustic sensor 50 but also using the temperature change measured by the thermographic camera 53. Therefore, the operating state of the rotating equipment can be identified more accurately, thereby improving the accuracy of abnormality detection.

[0089] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents. [Explanation of symbols]

[0090] 10: Wind power generation equipment monitoring device 11: Feature extraction unit 12: Model Creation Department 13: Storage part 14: Equipment status estimation unit 15:Notification Department 20: Windmill 30: Main engine 40: Auxiliary equipment 50: Acoustic sensor 51: Light receiving sensor 52: Reflective material 53: Thermal imaging camera 60: Mobile equipment 120: Equipment state estimation model

Claims

1. A feature extraction unit that extracts feature quantities from an acoustic signal showing the detection result of sound generated by at least one device installed in the power generation facility, A storage unit that stores machine learning models pre-classified according to the operating state of the equipment using the aforementioned sound, A device state estimation unit identifies the operating state of the device based on the aforementioned acoustic signal and estimates whether or not there is an abnormality in the device using a machine learning model corresponding to the identified operating state. A notification unit that outputs an alert signal indicating an abnormality in the aforementioned equipment, A wind power generation equipment monitoring device equipped with the following features.

2. The system further comprises a model creation unit for creating the aforementioned machine learning model, The wind power generation equipment monitoring device according to claim 1, wherein the model creation unit collects operating data of the equipment and the characteristic quantities of the acoustic signal for a certain period of time, and creates the machine learning model by classifying the operating state based on the correlation between the collected operating data and the characteristic quantities using machine learning.

3. The aforementioned device is a rotating device, The wind power generation equipment monitoring device according to claim 1 or 2, wherein the feature extraction unit further performs frequency analysis on the acoustic signal while shifting the results of frequency analysis within a predetermined time window by a predetermined time to extract fluctuating components that depend on the rotational speed of the rotating equipment, and identifies the acoustic frequencies corresponding to the fluctuating components.

4. The aforementioned acoustic signal indicates the detection result of wind noise from the blades installed on the wind turbine. The wind power generation equipment monitoring device according to claim 1 or 2, wherein the equipment status estimation unit determines the operating state based on the interval of the wind noise indicated in the acoustic signal.

5. The wind power generation equipment monitoring device according to claim 1 or 2, wherein the equipment state estimation unit estimates that the equipment is abnormal when the feature quantity deviates from the separation boundary surface shown in the machine learning model.

6. Multiple devices are arranged in the power generation facility. An acoustic sensor that detects the sound from the aforementioned multiple devices is installed on the mobile device. Map data showing the arrangement of the multiple devices is stored in the storage unit. The wind power generation equipment monitoring device according to claim 1 or 2, wherein the equipment status estimation unit identifies the equipment currently in operation from the plurality of equipment using the location data from which the sound was detected and the map data.

7. Multiple devices are arranged in the power generation facility. Multiple acoustic sensors, which detect the sounds of the aforementioned multiple devices, are fixed in locations that are far apart from each other. The wind power generation equipment monitoring device according to claim 1 or 2, wherein the equipment status estimation unit identifies the sound source equipment from the plurality of equipment based on the time difference in which the sound was detected among the plurality of acoustic sensors and the speed of sound.

8. The aforementioned device is a rotating device to which a reflective member is attached, The reflected light from the reflective member is received by the light receiving sensor. The wind power generation equipment monitoring device according to claim 1 or 2, wherein the equipment status estimation unit determines the operating state of the rotating equipment based on the interval at which the light receiving sensor receives the reflected light.

9. The temperature of the aforementioned device is measured by a thermographic camera. The wind power generation equipment monitoring device according to claim 1 or 2, wherein the equipment status estimation unit identifies the operating status of the equipment based on the temperature change measured by the thermographic camera.

10. Using the sound generated by at least one device installed in the power generation facility, a machine learning model is created that is pre-classified according to the operating state of the device. Feature quantities are extracted from the acoustic signal showing the detection results of the aforementioned sound, Based on the aforementioned acoustic signal, the operating state of the equipment is identified. A machine learning model corresponding to the identified operating conditions is used to estimate whether or not there is an abnormality in the equipment. The device outputs an alert signal indicating an abnormality. Method for monitoring wind power generation equipment.

11. A process to create a machine learning model that is pre-classified according to the operating state of at least one device installed in a power generation facility, using the sound generated by the device. A process for extracting feature quantities from the acoustic signal showing the detection results of the aforementioned acoustics, A process to identify the operating state of the equipment based on the aforementioned acoustic signal, A process to estimate whether or not there is an abnormality in the equipment using a machine learning model corresponding to the identified operating state, A process that outputs an alert signal indicating an abnormality in the aforementioned equipment, A program that causes a computer to execute something.

Images