59082 - Prediction Enhancement of Machine Learning Using Time Series Modeling in Gas Turbines 

Turbine temperature profile distortion can serve as a precursor of anomalies/faults in combustion system and/or cooling system. This distortion is quantified using blade-path temperature spread. In the current study, given spatio-temporal observation from blade path temperature sensors of a heavy-duty gas turbine, we consider real-time prediction of the temperature for each sensor based on its time history and the corresponding fuel flow. The only extraneous predictor included is the combustion turbine fuel flow, while measurements of other potential predictors related to combustion process itself are unavailable. Long-memory behavior and heterogeneous variance are observed from the residuals of the generalized additive model (GAM). Autoregressive fractionally integrated moving average (ARFIMA) and generalized autoregressive conditional heteroskedasticity (GARCH) models are employed to fit the residuals process of GAM, which significantly improve the prediction.

Rolling one-step-ahead forecast is studied for each of the sixteen univariate blade path temperature sensors. Their conditional variances are also estimated. Difference between the real-time forecast and observation can be considered as a proxy for anomalies. Multiple test cases are simulated mathematically with manually generated perturbation to evaluate the specificity and sensitivity of the prediction. Abrupt changes in the temperature are considered in the study with various jump size. We also consider slowly increasing trend in the blade path temperature with different slopes. Our prediction is sensitive given reasonable signal-to-noise ratio. It also has a much lower false positive rate than the generalized additive model prediction from the combustion turbine fuel flow.