Fault detection and isolation system for boiler-turbine unit of a thermal power plant

Abstract

In this study, a sensor fault detection and isolation (FDI) system is presented for a boiler-turbine unit of a thermal power plant in Mexico. The FDI system is based on a Luenberger-like observer for residual generation. An adaptive threshold for residual evaluation is considered to avoid false alarms. The observer is based on a quasi-linear parameter variant (quasi-LPV) model of the boiler-turbine unit parameterized with real data of the plant in a wide range of operations, namely, at low, medium, and high loads. The quasi-LPV model adequately represents the dynamics of more critical variables including first stage turbine pressure, superheated steam pressure, drum pressure, and electric power. The performance of the FDI system is evaluated in a practical scenario by using real data from the thermoelectric plant. The main contribution of this study involves proposing a reliable fault diagnosis system to detect sensor faults in a wide operational range of the process based on the quasi-LPV framework.