Condition Monitoring of Electrical Machines via Virtual Measurements and Machine Learning

To increase the stability and the reliability of the electrical system, reliable condition monitoring and fault diagnosis of electrical machine is necessary since electrical machine is an indispensable part of power grid. Due to the special structure and working environment, there are several faults may occur while the electrical machine is working combined with some changes in the electrical and non-electrical parameters, which can be used to achieve the condition monitoring of electrical machine so as to predict the fault in advance and avoid the occurrence of faults. The main objectives in my research includes: 1. Through analyzing the structural characteristics and operating principles of electrical machine, determine the typical fault modes and characteristics of electrical machine, including inter-turn short or open circuit in stator winding, rotor eccentricities, broken rotor bar or cracked rotor end rings, bearing and gearbox failure and so on. 2. The Finite element model of DFIG would be established and the reduced-order finite element model of DFIG would be further obtained using the POD-type method. Through adding storage system, data interaction system and business system, the digital twin would be finally built to reproduce the function of the DFIG and predict the trend. 3. The fault characteristics of the stator and rotor of the generator will be extracted and converted into fault characteristic parameters to establish a complete knowledge base, using which the fault diagnosis algorithm based on machine learning will be come up and optimized using genetic algorithm.