Fault diagnosis of pumped storage unit based on multi-physical parameter detection

Pumped storage units have become one of the most effective technical support means to ensure the safe operation of large-capacity renewable energy access power systems. Due to the complex structure, it is difficult to establish an overall model describing the characteristics of the motor/generator. Based on the vibration and temperature signal detection, a hierarchical method has been proposed. By designing a new data reconstruction strategy, the transient process in the time dimension is reconstructed into a stationary sequence and the vibration acceleration and temperature gradient of thrust bearing in pumped storage motor/generator extracted as the characteristic parameters during the start-up and breaking operation, and the fault diagnosis of non-stationary transient continuous processes has been realized based on BP neural network. Test results show that the proposed comprehensive judgment method that combines linear and nonlinear information of the electromechanical transition process can improve the accuracy and sensitivity of fault detection.