Disturbance rejection control of air–fuel ratio with transport-delay in engines

In this paper, an air–fuel ratio (AFR) controller is proposed for gasoline engines, to address the control challenges in the engine air flow rate estimation uncertainty and the exhaust gas transport delay. A physics-based AFR model is developed, in which two online correction factors are introduced for model accuracy improvement. The discrepancy of the model from the real plant is lumped as an extended state denoted as “total disturbance”. An extended state predictor observer (ESPO) is then deployed to estimate the total disturbance and predict the AFR. By canceling the total disturbance in real-time and using the predicted AFR as the output to be controlled, the plant is enforced to behave as a first-order linear system, which is easy to control. Since the estimation capability of ESPO is limited by the transport delay, a recursive least square (RLS) estimator is designed to reduce the model estimation error and total disturbance through adjusting model parameters. This is achieved by using the historical data of the predicted AFR and the corresponding fuel amount. Experimental validation is conducted in a gasoline engine test bench. Results show that the ESPO reduces the settling time of AFR by 41% and 35% respectively, compared to the results with the conventional extended state observer (ESO) and input-delay ESO in step tests of the target AFR. The deviation of AFR from the target is reduced by 35%, 27%, and 32% respectively by using the RLS estimator in step tests of throttle position, exhaust valve closing timing, and intake valve closing timing.