Condition Monitoring of Turbine Generator Using Stator Winding Temperature

Condition Monitoring (CM) is critical for safety related complex components in thermal power plants. Stator winding temperature is one of the key monitoring indicators of turbine generator. A robust condition monitoring method that integrates AAKR-based empirical estimation and SPRT-based detection is proposed to monitor stator winding temperature in this paper. Auto Associative Kernel Regression (AAKR) is a modeling method to construct normal behavior of turbine generator stator winding temperature. Sequential Probability Ratio Test (SPRT) detects the residual between AAKR estimation and measured temperature, determining whether the temperature is behaving as expected. A data set is applied to validate that the condition monitoring method can detect incipient failure accurately and simply.