Research On Health Index Observation And Fault Prediction Of Wind Turbine Based On Laplacian Eigenmaps

In recent years,with the rapid development of the field of wind power generation technology,the installed capacity of single wind turbine is increasing,and the scale of wind farm construction is also expanding.The problems of high failure rate and high operation and maintenance cost of wind turbine gradually appear.Therefore,monitoring the real-time operating status of wind turbines and accurately determining the location and cause of faults has important application value for ensuring safe and efficient operation of wind turbines and reducing operation and maintenance costs.A large number of historical data generated during the operation of the wind turbine are recorded in the SCADA system.Based on the operating data recorded by the SCADA system,a health monitoring and fault prediction model for the wind turbine is established.The specific research contents include:1)Brief description of SCADA system for wind turbines and statistical analysis of fault conditions.Briefly describe the basic structure of wind turbines and the basic working principles of wind power generation,then introduce the state parameters recorded by the SCADA system,and perform statistical analysis on common fault types and fault causes through the maintenance log of wind turbine.2)Data processing of health monitoring model for wind turbines.The wind speed-power scatter plot of the wind turbine is analyzed to classify the abnormal data points,and the cleaning of the abnormal data points is realized based on the quartile algorithm.As the output power of wind turbine is the most intuitive index reflecting its health status,partial least squares algorithm is used to preliminarily screen the characteristic parameters that affect the output power.Finally,based on the mutual information rules,the feature parameters with obvious correlation are redundantly removed to prepare for the establishment of health monitoring model.3)A health monitoring model for wind turbines based on Laplacian Eigenmaps.By conducting a dimensionality reduction study on the Swiss volume high-dimensional data set,it is demonstrated that Laplacian Eigenmaps is more suitable for analyzing the health status change trend of wind turbines than other manifold learning algorithms.Then introduced the health monitoring models of LE-EVD and PCA-DEV respectively.Comparative studies show that the LE-EVD model can detect the degradation of the wind turbine earlier and more accurately.Finally,the advantages of the LE-EVD model over the PCA-DEV model are analyzed.4)Fault prediction model of wind turbine based on operating conditions..The soft fuzzy C-means clustering algorithm is used to cluster the SCADA data of the wind turbine,and the number of clusters is determined based on the clustering validity index to realize the interval division of the operating conditions of the wind turbine.Then take the gearbox as an example to build a LE-SOM fault prediction model for each component of the wind turbine.And then input the extracted feature parameters into the Laplacian Eigenmaps.The self-organizing map algorithm is used to divide the fault types and determine the clustering centers of each cluster.By monitoring the Euclidean distance between the clustering centers of online state data sets and the clustering centers of various fault types in real time,the possible failure types can be predicted.