Active Disturbance Rejection Control of Diesel Engine Air System Based on Online Optimization of Observation Bandwidth and Control Gain

In this paper, a parameter self-learning multivariable active disturbance rejection control (ADRC) algorithm is proposed for the Diesel engine air system equipped with a variable exhaust gas turbocharger (VGT), exhaust gas recirculation (EGR), and intake throttle valve (TVA). Firstly, based on the principle of simplified Diesel engine operation, a three-in-three-out linear variable parameter air system prediction model is established, with the model deviation equivalent to the total disturbance. A multivariable extended state observer is employed for active observation to enhance the model's adaptability. Secondly, to address the issues of strong noise and variable signal-to-noise ratio of the air signal, the gain matrix of the extended state observer is dynamically adjusted to achieve a real-time trade-off between observation speed and disturbance fluctuation amplitude. Finally, in order to improve tracking accuracy while compensating for the total disturbance, extremum seeking is utilized to optimize the control input gain online and enhance the algorithm's adaptability to nonlinearities. The algorithm is verified on a Diesel engine stand, and compared with a PID controller with global parameter optimization, it achieves the same response speed while reducing the overshoot of intake manifold pressure and EGR rate tracking by 24.9% and 10.3%, respectively. The proposed algorithm's parameters can be automatically adjusted with the change of working conditions, avoiding the need for sectional calibration.