SENSOR VALIDATION AND RECOVERY FOR COMBUSTION TURBINES IN ELECTRICAL POWER GENERATION

摘要

Development of a comprehensive combustion turbine health management system will play a critical role in reducing the cost of electricity by improving reliability, availability, and maintainability. The real-time health management technologies under development use a combination of probabilistic and artificial intelligence-based tools to assess both thermodynamic and mechanical health of combustion turbines. These technologies include sensor validation, performance diagnostics and prognostics, vibration diagnostics, and critical component remaining useful life assessments. Sensor validation is an important front-end of the health management system that checks the integrity of sensed data before it is passed to the diagnostic and prognostic modules. The sensor validation software utilizes a combination or fusion of neural network model-based and generic signal-processing based approaches to ensure the highest possible sensor fault detection confidence with minimal false alarms. In the event that a gas path sensor fault is detected, neural network models are used to calculate proxy or "recovered" signal values that allow diagnostic and I component life assessments until the fault is corrected. The sensor validation and recovery module is demonstrated on a GE Frame 7FA application.