Wind turbine condition monitoring method and system

Patent metadata

Field	Value
Publication number	US-10288043-B2
Application number	US-201615359976-A
Country	US
Kind code	B2
Filing date	Nov 23, 2016
Priority date	Nov 18, 2014
Publication date	May 14, 2019
Grant date	May 14, 2019

Abstract

The present application includes wind turbine condition monitoring method and system. The method includes: acquiring historical SCADA data, and wind turbine reports corresponding to the historical SCADA data; training an overall model for overall diagnosing the wind turbine, and training different individual models for analyzing different components of the wind turbine based on the historical SCADA data and the corresponding wind turbine report, by establishing relationship between the historical SCADA data and the wind turbine report; acquiring real time SCADA data, inputting the real time SCADA data to the trained overall model, obtaining the health condition of the wind turbine from the trained overall model, and performing individual diagnosing step if the trained overall model determines wind turbine as defective status; inputting the real time SCADA data to the trained individual model corresponding to the defective component, and obtaining the fault details of the defective component from the trained individual model corresponding to the defective component.

First claim

The invention claimed is:

1. A wind turbine condition monitoring method, comprising the following steps:
historical data acquiring step: acquiring historical SCADA data, and wind turbine reports corresponding to the historical SCADA data, wherein the historical SCADA data covers operation data of the wind turbine, and wherein the wind turbine reports covers: health condition of wind turbine diagnosed as normal or defective status, and defective component with corresponding fault details if health condition of wind turbine is diagnosed as defective status;
model training step: training an overall model for overall diagnosing the wind turbine, and training different individual models for analyzing different components of the wind turbine based on the historical SCADA data and the corresponding turbine report, by establishing relationship between the historical SCADA data and the wind turbine report;
overall diagnosing step: acquiring real time SCADA data, inputting the real time SCADA data to the trained overall model, obtaining the health condition of the wind turbine from the trained overall model, and performing individual diagnosing step if the trained overall model determines wind turbine as defective status;
individual diagnosing step: inputting the real time SCADA data to the trained individual model corresponding to the defective component, and obtaining the fault details of the defective component from the trained individual model corresponding to the defective component;
wherein the model training step comprises:
selecting overall data mining algorithm;
training an overall model for overall diagnosing the wind turbine with the overall data mining algorithm to establish relation between input and output of the overall model, input of the overall model is the historical SCADA data, output of the overall trained model is health condition of the wind turbine which includes normal and defective status, and defective component if health condition of the wind turbine is diagnosed as defective status;
selecting individual data mining algorithms for different components of the wind turbine;
training different individual models for analyzing different components of the wind turbine with the corresponding individual data mining algorithms to establish relation between input and output of the individual model, input of each individual model is the historical SCADA data, output of each individual model if the individual component corresponding to the individual model is defective component.

2. The method according to claim 1 , wherein the historical data acquiring step comprises:
acquiring historical SCADA data, and wind turbine reports corresponding to the historical SCADA data, wherein the historical SCADA data covers operation data of the wind turbine, and wherein the wind turbine reports covers: health condition of wind turbine diagnosed as normal or defective status, and defective component with corresponding fault details if health condition of wind turbine is diagnosed as defective status;
verifying the wind turbine report to identify which data section of the historical SCADA data is normal status and which data section of the historical SCADA data is defective status.

3. The method according to claim 1 , wherein the model training step comprises:
verifying the effectiveness of the overall model and individual models using the historical SCADA data.

4. The method according to claim 1 , wherein the overall diagnosing step comprises:
acquiring real time SCADA data;
inputting the real time SCADA data to the trained overall model;
running the trained overall model to implement the overall data mining algorithm;
obtaining the health condition of the wind turbine from the trained overall model;
performing individual diagnosing step if the trained overall model determines wind turbine as defective status.

5. The method according to claim 1 , wherein the individual diagnosing step comprises:
selecting the trained individual model corresponding to the defective component as the trained defective model;
inputting the real time SCADA data to the trained defective model;
running the trained defective model to implement the corresponding individual data mining algorithm;
obtaining the fault details of the defective component from the trained defective model.

6. A wind turbine condition monitoring system, comprising the following modules:
historical data acquiring model, used for acquiring historical SCADA data, and wind turbine reports corresponding to the historical SCADA data, wherein the historical SCADA data covers operation data of the wind turbine, and wherein the wind turbine reports covers: health condition of wind turbine diagnosed as normal or defective status, and defective component with corresponding fault details if health condition of wind turbine is diagnosed as defective status;
model training module, used for training an overall model for overall diagnosing the wind turbine, and training different individual models for analyzing different components of the wind turbine based on the historical SCADA data and the corresponding wind turbine report, by establishing relationship between the historical SCADA data and the wind turbine report;
overall diagnosing module, used for acquiring real time SCADA data, inputting the real time SCADA data to the trained overall model, obtaining the health condition of the wind turbine from the trained overall model, and performing individual diagnosing module if the trained overall model determines wind turbine as defective status;
individual diagnosing module, used for inputting the real time SCADA data to the trained individual model corresponding to the defective component, and obtaining the fault details of the defective component from the trained individual model corresponding to the defective component;
wherein the model training module comprises:
module used for selecting overall data mining algorithm;
module used for training an overall model for overall diagnosing the wind turbine with the overall data mining algorithm to establish relation between input and output of the overall model, input of the overall model is the historical SCADA data, output of the overall trained model is health condition of the wind turbine which includes normal and defective status, and defective component if health condition of the wind turbine is diagnosed as defective status;
module used for selecting individual data mining algorithms for different components of the wind turbine;
module used for training different individual models for analyzing different components of the wind turbine with the corresponding individual data mining algorithms to establish relation between input and output of the individual model, input of each individual model is the historical SCADA data, output of each individual model is the fault details for the individual component corresponding to the individual model if the individual component corresponding to the individual model is defective component.

7. The system according to claim 6 , wherein the historical data acquiring module comprises:
module used for acquiring historical SCADA data, and wind turbine reports corresponding to the historical SCADA data, wherein the historical SCADA data covers operation data of the wind turbine, and wherein the wind turbine reports covers: health condition of wind turbine diagnosed as normal or defective status, and defective component with corresponding fault details if health condition of wind turbine is diagnosed as defective status;
module used for verifying the wind turbine report to identify which data section of the historical SCADA data is normal status and which data section of the historical SCADA data is defective status.

Assignees

• Abb Schweiz Ag

Inventors

• Yu Rongrong
• Chen Niya
• Chen Yao

Classifications

• G06N99/005

  Physics · mapped topic

• F05B2260/83

  Testing, e.g. methods, components or tools therefor · CPC title

• F05B2260/84

  Modelling or simulation · CPC title

• F03D17/00 Primary

  Monitoring or testing of wind motors, e.g. diagnostics (testing during commissioning of wind motors F03D13/30) · CPC title

• F05B2260/80

  Diagnostics · CPC title