Study On Dynamic Modeling Methods Oriented Control For A Gas Fuel Engine

The engine management system based on the digital controller exhibits many inherently control problems, which mainly include the sensor signal sampling and processing, fluctuation noises, sensor/antialiasing filter response times and the bandwidth limit for transient control. The model-based control method can solve a large number of control problems mentioned above. At the same time, in order to satisfy the application requirement of the advanced control theories to engine control system charactered by multivariable, nonlinearity and time-varying, three different control oriented engine models are proposed for a special engine with coal-bed gas fuel. The main contents of studying are shown as follows:1) Aiming at high-frequency sampling signals related to air fuel ratio (AFR) control, the characteristics of sensor were studied in both time and frequency domains. The flow fluctuation models, which pick up the high order harmonics of pumping fluctuations, were established based on the air and gas flow signals characteristics. Several sampling algorithms of event average based sampling (EABSn) were tested by simulating the air and gas flow fluctuation models respectively, then the antialiasing sampling strategies of flow signals were selected.2) The reasons of AFR excursions from stoichiometry during the transient operation of throttle step changes were analyzed as viewed from the transmission and dynamic response delays of air and gas in intake system and engine as well as exhaust system. The AFR Hammerstein models were built after compensating for the time delays. The static nonlinear models from air and gas flow to the AFR were fitted by the steady state experimental data respectively; the dynamic linear models were identified using subspace methods by the dynamic state experimental data, and the order of the subspace dynamic models were determined by three different error criterion functions.3) The difference equation model structures of the intake manifold pressure and AFR were derived from the physical state equation of intake system. The adaptive models based on forgetting factor (FF) algorithm and Kalman filter (KF) algorithm, which could predict the averaging intake manifold pressure and suppress the pumping noises, were gained by choosing a proper update step (forgetting factor or covariance matrix); an adaptive model based on FF algorithm for precisely predicting the AFR was built by choosing a relative small forgetting factor, and the real-time parameters estimations for adaptive control were estimated at the same times.4) The feedforward model structure of two inputs and two outputs for intake system flows control was established based on a diagonal form matrix fraction description (MFD). The submodels orders of air and gas flows were selected using asymptotic criterion (ASYC) respectively, and the model parameters were estimated by the asymptotic identification method. Using two kinds of conventional model validation methods in time domain (residual analysis and cross validation) and two kinds of control oriented model validation methods in frequency domain (model error model and quantifying upper error bound), the established models were evaluated, the results show that the accuracy of the asymptotic flow model is high in both time domain and frequency domain, and this method is suitable for model predictive control.