MVEM Based on Back-propagation Neural Network and Model Predictive Control of Air System

The control of air systems has consistently been a crucial aspect of managing internal combustion engines. Conventional control strategies often rely on PID (Proportional-Integral-Derivative) controllers and MAP feedforward calibrations, which struggle to accommodate the varying time scales, pronounced interdependencies, and nonlinear dynamics inherent to engine components. This article introduces an innovative method that combines PID control with Model Predictive Control (MPC), aiming to address the complexity of PID parameter tuning, control delay, limited adaptability, and MPC's dependence on model accuracy. Through simulations, this research has acquired steady-state characteristic data of the engine across a spectrum of operating conditions. A hybrid modeling technique, leveraging both physical mechanisms and empirical data, was applied. The air system was decomposed into modules and modeled using a Backpropagation Neural Network (BPNN) to create an analog Mean Value Engine Model (MVEM). Following this, the control inputs for both MPC and PID were dynamically allocated based on the rate of change in operating conditions. Ultimately, the efficacy of the proposed control strategy was substantiated through rigorous simulation testing.