Early Warning of Stator Winding Overheating Faults in Water-Cooled Turbogenerator Based on Dynamic Network Markers

The defects of the stator winding water cooling system of turbo-generator will cause the overheating of stator winding. In this paper, a method of early warning of overheating faults in stator windings based on DCS data of the turbogenerator is proposed by combining grey correlation analysis method with dynamic network marker. Firstly, the water temperature measurement points at each outlet of the stator winding water cooling system are mapped into a complex dynamic network, and the key nodes in the network are also analyzed by using grey relation methods. Then, by analyzing the dynamic changes in the standard deviation and correlation of key nodes in the dynamic network, the dynamic network markers during the phase change process are extracted as warning signals for overheating defects in the generator stator winding. Finally, the method was verified by using the historical DCS data prior to the failure of a faulty turbogenerator, the results show that the proposed method can provide timely and accurate warning of overheating defects in stator windings, which can also overcome the shortcomings of traditional machine learning algorithms, such as misdiagnosis and missed diagnosis, caused by the difficulty in obtaining training samples and unreliable models.