Fuel Pressure Control Algorithm Based on Neural Network-Accelerated Extended State Linear Quadratic Tracker

Accurate fuel pressure control is critical for optimizing diesel engine efficiency and emissions. To address the limitations of Linear Quadratic Tracker(LQT) in handling dynamic disturbances within high-pressure common rail systems, this paper proposes a Physics-Informed Neural Network (PINN)-accelerated Extended State Observer (ESO)-LQT algorithm (PINN-LQT-ESO). The key innovations lie in three aspects: an integrated ESO-LQT framework that dynamically estimates and compensates for lumped disturbances through real-time feedback; an incremental fuel pressure model enabling adaptive linearization under varying operating conditions; and a PINN architecture designed to solve the Riccati equation 15.3% faster than conventional Runge-Kutta methods. Experimental validations demonstrate that the proposed method reduces dynamic response time by 30%, decreases tracking error by 2.05%, and maintains steady-state pressure fluctuations below 0.5 MPa compared to conventional LQT. These advancements significantly enhance both control precision and computational efficiency, providing a robust solution for real-time fuel pressure regulation.