An optimization algorithm for neural predictive control of air-fuel ratio in SI engines

This work presents an optimization algorithm to solve quadratic sub problem in neural predictive control of AFR in SI engines. The solution of quadratic programming is computationally efficient and works in conjunction with offline trained NNARX model for AFR identification. Use of offline trained model and its linearization can invite some model mismatch in presence of engine uncertainties. This mismatch is taken care of by incorporating a PID feedback correction scheme. It has been shown that neural predictive control with online linearization using PID feedback correction scheme gives satisfactory results.