Prediction of compressor efficiency by means of Bayesian Hierarchical Models

The prediction of time evolution of gas turbine performance is an emerging requirement of modern prognostics and health management systems, aimed at improving system reliability and availability, while reducing life cycle costs. In this work, a data-driven Bayesian Hierarchical Model (BHM) is employed to perform a probabilistic prediction of gas turbine future behavior. The BHM approach is applied to field data, taken from the literature and representative of gas turbine degradation over time for a time frame of 7-9 years. The predicted variable is compressor efficiency collected from three power plants characterized by high degradation rate. The capabilities of the BHM prognostic method are assessed by considering two different forecasting approaches, i.e. single-step and multi-step forecast. For the considered field data, the prediction accuracy is very high for both approaches. In fact, the average values of the prediction errors are lower than 0.3% for single-step prediction and lower than 0.6% for multi- step prediction.