Neural Network-Based Thrust Predictive Control for Turbofan Engines

Model Predictive Control (MPC) algorithm offers distinct advantages in dealing with constrained MultiInput, Multi-Output systems. Direct thrust control can effectively improve engine performance by estimating and predicting the unmeasurable thrust of the engine. In particular, to overcome the barrier of fast calculation, this article presents a neural network-based thrust predictive control (NNTPC) approach for a turbofan engine. Additionally, we exploit the existing M-RTCN model and further design an improved predictive model capable of forecasting multiple time steps online. The effectiveness of the overall strategy is validated through comprehensive simulation experiments. Notably, the results reveal that the multi-step predictive model reduces computational time for solving the optimization problem as compared to conventional models that update states at each individual step. NNTPC exhibits robust control performance in the presence of disturbances and outperforms PID control in terms of energy efficiency, achieving fuel savings exceeding 10%.