Development And Research Of Data Stream-based Wind Turbine Fault Real-time Monitoring System

In recent years,with the deepening of the concept of environmental protection,clean energy represented by wind power generation has occupied an increasing proportion in human energy consumption.In order to improve the efficiency of wind power generation,wind turbines are often installed in remote suburbs far from the urban area,which causes the problems of slow fault detection and difficult maintenance of wind turbines.The stable operation of wind turbines is of great significance to improve the economic and social benefits of power generation enterprises.At present,wind turbine fault-related research is mostly based on offline data through SAS,Matlab and other analysis software,which can not meet the needs of real-time monitoring of industrial field faults.On the basis of reading a large number of literatures about wind turbine fault diagnosis and data stream processing technology,this thesis presents a set of solutions for real-time monitoring system of wind turbine fault.A data stream processing platform based on Flink and Kafka is built in this thesis.On the basis of this platform,the corresponding operators are developed to realize the real-time monitoring of the operation status of wind turbines combined with the characteristics of vibration signals and machine learning algorithm.Firstly,according to the characteristics of the original vibration signal of wind turbine,this thesis realizes a fault warning method based on the characteristics of vibration signal waveform.In order to measure the similarity of waveform features,the similarity-based distance is introduced as a new distance measurement method,and combined with machine learning classification algorithm to realize the effective recognition of the normal operation state and fault state of wind turbines.In order to further analyze the type and degree of wind turbine faults,the traditional empirical mode decomposition method is improved according to the characteristics of data stream.and an online empirical mode decomposition method(SWEMD)based on sliding window is implemented.On this basis,the frequency characteristics and energy characteristics of the decomposed vibration signal are extracted by Hilbert transform.Then the fault diagnosis of wind turbines is realized by combining the random forest algorithm and the frequency characteristics of vibration signals,and the health assessment of wind turbines is realized by regression analysis of the energy characteristics of vibration signals.Finally,the thesis introduces the process of platform construction and deployment,and tests the function and performance of the whole system to verify the effectiveness of the system.This has great reference value for real-time monitoring of faults of wind turbines in practical engineering applications.