A q-learning-based multivariable active disturbance rejection control method for diesel engine air system

By using a Q-learning-based multivariable active disturbance rejection control method, the problems of complex cross-coupling and high-frequency interference in the diesel engine air system are solved, achieving higher control accuracy and anti-interference capability while reducing computational complexity.

CN117028040BActive Publication Date: 2026-06-23TIANJIN UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
TIANJIN UNIV
Filing Date
2023-06-28
Publication Date
2026-06-23

AI Technical Summary

Technical Problem

Existing diesel engine air system control strategies fail to effectively address the complex cross-coupling relationships and high-frequency interference effects between control loops, and have high computational complexity, making it difficult to achieve stable control in the presence of output signal noise.

Method used

A Q-learning-based multivariable active disturbance rejection control method is adopted. By establishing a TVA-VGT-EGR air system model, a three-input three-output controller and an extended state observer are designed. The Q-learning algorithm is used to adjust the observer bandwidth and optimize the control effect.

Benefits of technology

It improves control accuracy and anti-interference capability, achieves smooth dynamic response process and accurate tracking of key parameters, and reduces computational complexity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117028040B_ABST
    Figure CN117028040B_ABST
Patent Text Reader

Abstract

The application discloses a diesel engine air system multivariable active disturbance rejection control method based on Q-learning, and comprises the following steps: a TVA-VGT-EGR diesel engine air system control-oriented model is established to calculate turbocharging pressure p 22 , pre-vortex pressure p3 and EGR rate X EGE in real time; a simplified expression of the TVA-VGT-EGR diesel engine air system control-oriented model is written in a state space equation form, a three-input and three-output air system controller, a corresponding extended state observer and a control law are designed based on a multivariable active disturbance rejection control algorithm, and the control of p 22 , p3 and X EGR is realized; a state space S and an action set A are designed according to the actual situation of the air system controller, a corresponding state-action value function Q(s, a) = 0 is initialized, a discount factor gamma in the Q-learning algorithm is set as gamma belongs to (0, 1), a learning rate sequence is selected, and a state transition probability is selected; a reward function is designed, the parameters of the extended state observer are adjusted through Q learning, and the bandwidth of the extended state observer is adjusted. The application can improve the control precision and anti-interference ability of key parameters.
Need to check novelty before this filing date? Find Prior Art