Identification And Early Warning Of Health Status Of Large Direct-Drive Wind Turbine Based On SCADA Data Relation

The identification and early warning of large wind turbine operating state is of great significance for timely maintenance,reduction of downtime and operation and maintenance cost.The application of wind turbine SCADA data to the research of identification and early warning of large wind turbine operating state can avoid the inconvenience of installation and increase of cost caused by monitoring device.It has developed into the main research direction of large wind turbine operating state identification and early warning.In this paper,the existing wind turbine operating state identification and early warning based on SCADA data are mainly focused on physical modeling and intelligent modeling of abnormal(including fault)state data training.Based on the parameter relation of SCADA data and the principle of small probability,this paper presents a method of operation state recognition and early warning to avoid the dilemma caused by the shortage of abnormal state training data.The main contents of this thesis are as follows:(1)System discusses the structure and principle of a 2MW large direct drive wind turbine and SCADA system.Firstly,the structure and principle of a 2MW large-scale direct-drive wind turbine are studied.The energy,information transmission and interaction mechanism and energy transfer mathematical model of the operation process are analyzed.Then the wind farm and its large direct-drive wind turbine SCADA system structure are discussed.,hardware and software,analyze the relationship between data format and data parameters of SCADA system,and propose the principle of operation status recognition and early warning of large direct-drive wind turbines based on SCADA data relationship.(2)Propose the health status indicators of wind turbines based on SCADA data relationship.Firstly,the sliding window model is combined with the Bin method to process the data.The polynomial fitting modeling method based on the SCADA data relationship is proposed.Then,based on the Euclidean distance of the data relationship curve,the health index of the dimensionless wind turbine is presented.The calculation method,health indicators have good stability and sensitivity;finally,the discussion of data window width,window increment,data fitting modeling,etc.all have an impact on health status indicators.(3)A method for identifying and warning the operation status of large wind turbines based on the principle of small probability is proposed.Firstly,the principle of small probability in probability and statistics is introduced,and the kernel density estimation method is used to establish the random distribution model of health state indicators.Then,the SCADA data of two 2MW direct-drive wind turbines of the same model in the same wind farm are analyzed and discussed in order to realize the operation of wind turbines.Abnormal state identification and early warning.Finally,according to the hierarchical structure of the input/output parameter relationship of the operating state of the wind turbine and its components,the operating state input/output parameter relationship of the components in the abnormal wind turbine is traced from top to bottom,and the relevant component information of the abnormal state is obtained.