Abstract
The global drive toward renewable energy is imposing challenging operating requirements on power turbines. Flexible load-leveling applications must accept more frequent and demanding start-stop cycles. Full transient analyses are too computationally expensive for real-time simulation across all operating regimes, so monitoring relies on sparse physical measurements. Alone, these sparse data lack the fidelity for real-time prediction of a complex thermal field. A new hybrid methodology is proposed, coupling data across a range of fidelities to bridge the limitations in the individual analyses. Combining several fidelity methods in parallel: low-order models, corrected by real-time physical measurements, are calibrated with high-fidelity simulations. The multifaceted hybrid approach enables the real-time speed of low-order analysis at high resolution. This paper series develops the critical enabling features of the hybrid method. Real-time definition of the simulation thermal boundary condition is fundamental to the methodology. A novel long-short-term-memory (LSTM) neural network is presented, enabling time series prediction of the turbine thermal profile. The remote sensing method utilizes standard plant measurements to reconstruct the turbine temperature history, allowing reliable thermal prediction whilst removing the need for direct monitoring of the casing temperature. Calculations within the training region demonstrated high accuracy, achieving a mean square error of 3.1 K and maximum instantaneous peak error of 7.6 K. Combining a reduced feature space and efficient two-step predictor-corrector structure, the LSTM method offers high-speed temperature calculation, successfully predicting one month of operating data in under a minute.