Research On Large Wind Turbine Condition Monitoring And Intelligent Fault Diagnosis System

In recent years, the wind power has been developing rapidly and swiftly in the whole world and the total installed capacity has set new records year after year. With the development of the global wind power, a series of problems has been produced to require to be solved. As the wind turbines stay long-term working in the hostile environments, in addition to randomly changeable temperature and uncertain sand wind, wind speed, loading and other factors, components failure frequently occurs. With wind turbines scattering generally, it is hard to note the faults duly and efficiently. Furthermore, when faults occur, it normally takes longtime to maintain and repair the same, which seriously affects the economy of wind power. Therefore, it has important and long-term significance to carry out the research on condition monitoring and fault diagnosis for wind turbines in order to improve the reliability safety and economy of wind power operations.From the perspective of the actual demand of wind turbine condition monitoring, this article focuses on independently designing and developing the simulation experiment rig of wind turbine drive train faults in accordance with the demand of research and development of condition monitoring system for 3MW wind turbines to aim at gearbox fault types and characteristics. Based on experimental hardware conditions and innovated fault diagnosis methods, in all the characteristics of non-stationary, non-linear and complex modulation component of failure vibration signals of wind turbine drive train, this article proposes new fault diagnosis methods of combine which eliminates the band disorder and employs better resolution with envelope spectrum to improve node reconstruction wavelet packet algorithm.To solve the fault-identifying problem caused by unsteady fault signals of wind turbines and nonlinear mappings between faults and symptoms, proposes a new method of fault-diagnosing, combining improved wavelet packet band energy spectrum and the PNN (probabilistic neural network). Through using experimental data to verify two new fault diagnosis methods, the feasibility and accuracy thereof are confirmed and helpful support of tools and techniques are provided to thoroughly study the condition monitoring and fault diagnosis of wind turbines in the future.This paper determines the sensor type and the location of monitoring points and completes the whole on-site testing system. High-speed data acquisition analyzer involved in the design and development owns fault analysis and diagnosis functions and remote data transmission for synchronization, and real-time acquisition of vibration, high speed and other signals in connection of Multi-channel.Fully considering the actual needs on the scene, the self-designed system interface not only offers real-time fault alarms at the parts of wind turbines as well as specific faults and provides proposed guidance to the field staff, but also has the function to display the curves of monitoring a single or multiple wind turbines, which is useful for the on-site management staff to make contrast to determine faults.