Remote Monitoring and Diagnostic for Combined-Cycle Power Plants

Abstract

Reliability and availability are both critical for competitive operation of high efficiency combined-cycle power plants in the liberalized electricity markets. Monitoring and diagnostic of operational data is successfully used to prevent unexpected plant shut downs and to schedule maintenance activities. A complete set of tools is presented, which is used to monitor combined-cycle power plant operation. The daily incoming data is analyzed with a physics-based performance model, with a neural network-based novelty detection tool and with an experience-based failure detection algorithm, which looks for fingerprints of known and possible component faults. A knowledge management system is used to support the assessment of the findings and the recommended actions are communicated to the power plant operators in the form of early warnings. The application of the different methods in parallel shall ensure that most of the emerging problems are detected before they result in damage that leads to a forced plant shut down. Most of the cases are simple measurement errors, although component deterioration and failures can also be detected. Experience with the ALSTOM early warning system for the GT24/GT26 power plants is shown to demonstrate the successful application of this approach.