Granted Patent B2

US 12,590,570 · App. 18/260,075 · Granted Mar 31, 2026

Inventors: Yong Zhao (Beijing, CN); Xinle Li (Beijing, CN); Xinyuan Niu (Beijing, CN)

Assignee: BEIJING GOLDWIND SCIENCE & CREATION WINDPOWER EQUIPMENT CO., LTD.

F03D17/024F03D17/028G01M99/005G06F18/10F05B2260/80

Abstract

Claims (61)

1 . A blade fault diagnosis method, comprising:

acquiring a blade rotation audio collected by an audio collection device during operation of a wind turbine generator system;

preprocessing the blade rotation audio based on a wind noise filtering algorithm to obtain a blade rotation audio filtered out of wind noise;

dividing the blade rotation audio filtered out of wind noise to obtain audio segments corresponding to blades of the wind turbine generator system respectively;

diagnosing, based on the audio segments, whether the blades each corresponding to one of the audio segments are faulty,

transmitting blade fault diagnosis result to a server in a central control room; and

causing the server to display the blade fault diagnosis result through an interface program, and to transmit an alarm prompt when there is faulty blade,

wherein diagnosing, based on the audio segments, whether the blades each corresponding to one of the audio segments are faulty comprises:

processing each of the audio segments by a Fourier transform to obtain a second feature value of each of the audio segments, wherein the second feature value is configured to characterize a frequency domain feature of each of the audio segments;

inputting the second feature value of each of the audio segments to a blade fault diagnosis model to obtain a first fault diagnosis result of each of the blades, wherein the blade fault diagnosis model is a pre-trained model configured to identify whether a blade corresponding to the audio segment is faulty according to the second feature value of the audio segment;

counting durations of the audio segments corresponding to the blades respectively;

obtaining a second fault diagnosis result by judging whether there is a faulty blade according to whether a difference between durations of every two audio segments exceeds a preset threshold value; and

judging, by combining the first fault diagnosis result of each of the blades and the second fault diagnosis result, whether there is a faulty blade.

2 . The blade fault diagnosis method according to claim 1 , wherein dividing the blade rotation audio filtered out of wind noise to obtain the audio segments corresponding to the blades respectively comprises:

processing, by a short-time Fourier transform, the blade rotation audio filtered out of wind noise to obtain a first feature value, wherein the first feature value is configured to characterize a frequency domain feature of the blade rotation audio filtered out of wind noise;

inputting the first feature value to a blade recognition model to obtain dividing time points of the blade rotation audio filtered out of wind noise, wherein the blade recognition model is a pre-trained model configured to identify switching time points of rotation sounds of different blades according to the first feature value;

dividing the blade rotation audio filtered out of wind noise according to the dividing time points to obtain the audio segments corresponding to the blades respectively.

3 . The blade fault diagnosis method according to claim 1 , wherein before diagnosing, based on the audio segments, whether the blades each corresponding to one of the audio segments are faulty, the method further comprises:

acquiring an environmental parameter of the wind turbine generator system, wherein the environmental parameter is configured to indicate a season and/or weather;

determining, among a plurality of candidate fault diagnosis models configured to identify different fault types respectively, a model configured to identify a fault type corresponding to the environmental parameter, so as to determine a fault diagnosis model for use.

4 . The blade fault diagnosis method according to claim 1 , wherein acquiring the blade rotation audio collected by the audio collection device during the operation of the wind turbine generator system comprises acquiring blade rotation audios collected at a plurality of locations;

preprocessing the blade rotation audio based on the wind noise filtering algorithm to obtain the blade rotation audio filtered out of wind noise comprises processing each of the blade rotation audios respectively based on the wind noise filtering algorithm to obtain a plurality of blade rotation audios filtered out of wind noise;

after the plurality of blade rotation audios filtered out of wind noise are obtained, the method further comprises calculating a wind noise parameter of each of the plurality of blade rotation audios filtered out of wind noise respectively by a wind noise recognition model, wherein the wind noise parameter is configured to represent a wind noise level in the audio, and the wind noise recognition model is a pre-trained model configured to evaluate the wind noise parameter of the audio;

dividing the blade rotation audio filtered out of wind noise to obtain the audio segments corresponding to the blades respectively comprises selecting, according to the wind noise parameter and among the plurality of blade rotation audios filtered out of wind noise, a blade rotation audio filtered out of wind noise with a minimum wind noise, and dividing the blade rotation audio filtered out of wind noise with the minimum wind noise to obtain the audio segments corresponding to the blades respectively.

5 . The blade fault diagnosis method according to claim 4 , wherein calculating the wind noise parameter of each of the plurality of blade rotation audios filtered out of wind noise respectively by the wind noise recognition model comprises:

processing each of the plurality of blade rotation audios filtered out of wind noise by a Fourier transform to obtain a third feature value of each of the plurality of blade rotation audios filtered out of wind noise, and the third feature value is configured to characterize a frequency domain feature of a corresponding blade rotation audio filtered out of wind noise;

inputting the third feature value of each of the plurality of blade rotation audios filtered out of wind noise to the wind noise recognition model to obtain the wind noise parameter of each of the plurality of blade rotation audios filtered out of wind noise.

6 . A blade fault diagnosis apparatus, comprising:

an acquisition module configured to acquire a blade rotation audio collected by an audio collection device during an operation of a wind turbine generator system;

a preprocessing module configured to preprocess the blade rotation audio based on a wind noise filtering algorithm to obtain a blade rotation audio filtered out of wind noise;

a dividing module configured to divide the blade rotation audio filtered out of wind noise to obtain audio segments corresponding to blades of the wind turbine generator system respectively;

a diagnosis module configured to diagnose, based on the audio segments, whether the blades each corresponding to one of the audio segments are faulty,

an alarm module configured to transmitting blade fault diagnosis result to a server in a central control room, and causing the server to display the blade fault diagnosis result through an interface program, and to transmit an alarm prompt when there is faulty blade,

wherein the diagnosis module is further configured to:

process each of the audio segments by a Fourier transform to obtain a second feature value of each of the audio segments, wherein the second feature value is configured to characterize a frequency domain feature of each of the audio segments:

input the second feature value of each of the audio segments to a blade fault diagnosis model to obtain a first fault diagnosis result of each of the blades, wherein the blade fault diagnosis model is a pre-trained model configured to identify whether a blade corresponding to the audio segment is faulty according to the second feature value of the audio segment:

count durations of the audio segments corresponding to the blades respectively;

obtain a second fault diagnosis result by judging whether there is a faulty blade according to whether a difference between durations of every two audio segments exceeds a reset-threshold value; and

judge, by combining the first fault diagnosis result of each of the blades and the second fault diagnosis result, whether there is a faulty blade.

7 . A blade fault diagnosis system, comprising:

an audio collection device comprising a plurality of audio sensors, wherein at least one of the audio sensors is disposed at a prevailing wind direction position or a lee wind direction position of a tower barrel of a wind turbine generator system;

a processor disposed inside of the tower barrel and connected to the audio sensors, wherein the processor is configured to: receive a blade rotation audio collected by the audio sensors; preprocess the blade rotation audio based on a wind noise filtering algorithm to obtain a blade rotation audio filtered out of wind noise; divide the blade rotation audio filtered out of wind noise to obtain audio segments corresponding to blades of the wind turbine generator system respectively; diagnose, based on the audio segments, whether the blades each corresponding to one of the audio segments are faulty; transmit blade fault diagnosis result to a server in a central control room; and cause the server to display the blade fault diagnosis result through an interface program, and to transmit an alarm prompt when there is faulty blade,

wherein the processor is further configured to:

process each of the audio segments by a Fourier transform to obtain a second feature value of each of the audio segments, wherein the second feature value is configured to characterize a frequency domain feature of each of the audio segments:

input the second feature value of each of the audio segments to a blade fault diagnosis model to obtain a first fault diagnosis result of each of the blades wherein the blade fault diagnosis model is a pre-trained model configure to identify whether a blade corresponding to the audio segment is faulty according to the second feature value of the audio segment;

count durations of the audio segments corresponding to the blades respectively;

obtain a second fault diagnosis result by judging whether there is a faulty blade according to whether a difference between durations of every two audio segments exceeds a preset threshold value; and

judge, by combining the first fault diagnosis result of each of the blades and the second fault diagnosis result, whether there is a faulty blade.

8 . A storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement a blade fault diagnosis method, comprising:

acquiring a blade rotation audio collected by an audio collection device during operation of a wind turbine generator system:

preprocessing the blade rotation audio based on a wind noise filtering algorithm to obtain a blade rotation audio filtered out of wind noise:

dividing the blade rotation audio filtered out of wind noise to obtain audio segments corresponding to blades of the wind turbine generator system respectively;

diagnosing, based on the audio segments, whether the blades each corresponding to one of the audio segments are faulty,

transmitting blade fault diagnosis result to a server in a central control room; and

causing the server to display the blade fault diagnosis result through an interface program, and to transmit an alarm prompt when there is faulty blade,

wherein diagnosing, based on the audio segments, whether the blades each corresponding to one of the audio segments are faulty comprises:

processing each of the audio segments by a Fourier transform to obtain a second feature value of each of the audio segments, wherein the second feature value is configured to characterize a frequency domain feature of each of the audio segments;

inputting the second feature value of each of the audio segments to a blade fault diagnosis model to obtain a first fault diagnosis result of each of the blades wherein the blade fault diagnosis model is a pre-trained model configured to identify whether a blade corresponding to the audio segment is faulty according to the second feature value of the audio segment;

counting durations of the audio segments corresponding to the blades respectively;

obtaining a second fault diagnosis result by judging whether there is a faulty blade according to whether a difference between duration of every two audio segments exceeds a preset threshold value; and

judging, by combining the first fault diagnosis result of each of the blades and the second fault diagnosis result, whether there is a faulty blade.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (18)