| In recent years,wind power energy as a clean energy has great potential for development in the world.With the hidden danger of an energy crisis all over the world,wind energy has also been valued,and the installed capacity of global wind turbines has also increased year by year.Wind turbines will fail during operation,and when shutdown will reduce power generation and cause economic losses,the operation and maintenance evaluation of wind turbines has also received significant attention.The transmission system is an important part of wind turbines,and many faults are concentrated in the transmission system,so the modeling and evaluation of the transmission system has received widespread attention in recent years.Supervisory Control and Data Acquisition(SCADA)systems are widely installed in wind turbines,and the data collected and transmitted by SCADA systems is also widely used in various wind turbine operation and maintenance inspections.The working conditions of the transmission system are under complex working conditions,and the sensor parameters installed on it are collected and monitored in real time with rich information related to working conditions.SCADA data can reflect the rich healthy operation related data of the transmission system,and has a large number of timing characteristics and working condition characteristics information.According to the timing characteristics of SCADA data,a conditional convolutional self-coding-Gaussian hybrid network wind turbine transmission system health assessment model is proposed,and according to the SCADA data working conditions,a deep convolutional self-coding multi-door expert network wind turbine transmission system health assessment model is proposed,which solves the problem that SCADA timing characteristics and working condition characteristics are difficult to effectively use,and provides a new scheme for transmission system health assessment.Multiple sets of experiments were used to demonstrate the effectiveness and reliability of the method.The data used are the EDP dataset and the leaf icing public dataset.The main work of the thesis is summarized as follows:(1)Analyze and summarize the composition,fault type and operation principle of the wind turbine transmission system.SCADA data is selected as the basis of health assessment data,SCADA data is analyzed and processed,and working condition characteristics are obtained,which is the basis for the construction of subsequent health assessment model.(2)Aiming at the time series characteristics of SCADA data,the time series information of SCADA data is obtained by one-hot encoding of SCADA data timestamp,and a transmission system health assessment model based on fusion timing special certificate is designed.The CCAE-GMM model is proposed to learn the temporal characteristics and sample distribution through the front-end feature learning network.The back-end evaluates the network,outputs health indicators through a Gaussian mixture model,and finally divides the health location.This method is verified on the EDP dataset,which proves the effectiveness and reliability of the proposed method(3)Aiming at the relevant characteristics of SCADA data,the SCADA data is labeled with working condition characteristics to obtain working condition characteristics by dividing working conditions,and a transmission system health assessment model based on the special certificate of integrated working conditions is designed.The DCAE-MMo E model is proposed,and the network structure uses MMo E to adjust the learning ability of dual tasks,and promotes the learning of health assessment tasks through working condition characteristics.This method is verified on the leaf icing public dataset,which proves the effectiveness and reliability of the method. |
