Auto-Tuning of Real-Time Dynamic Gas Turbine Models

Abstract

Real-time gas turbine engine models are integral part of techniques such as model-based control and diagnostics. The use of model based techniques to diagnose and adaptively manage degradation of engine components is crucial for operational effectiveness of gas turbines. Since the gas turbine model represents “nominal” engine, it must be adapted or tuned to the performance of the real engine as it deviates from nominal baseline. Implementation of a method for auto-tuning of dynamic gas turbine engine models was considered in this paper, and a non-linear physics based component level model was facilitated as an on-line gas turbine model. Real-time nonlinear dynamic model of an industrial twin-shaft gas turbine with tracking filter was deployed onto the dedicated hardware platform and integrated with the engine control system. In presented application the identified gas turbine health parameters were obtained by the performance estimation tool and included in the observer design. The designed observer detects changes in the engine health parameters and generates model tuners. The model tuning process based on Kalman filtering technique was applied to secure robust execution of real-time dynamic models. Proposed auto-tuning methodology provides a tool for model adaptation, capable of addressing abrupt and gradual degradation of engine performance and at the same time offers a means for model compensation of performance deviation caused by engine-to-engine variation. Although most of performance tracking and diagnostic methods are developed for gas turbine operating at steady state, current trend demonstrates increasing interest in diagnostics during transient operation. Devised method estimates dynamic behaviour of gas turbine health parameters enabling in that way performance tracking under transient conditions. Examples of model adaptation during gas turbine engine transient operation are given in the paper.