Error-Triggered Adaptive Sparse Identification for Predictive Control and Its Application to Multiple Operating...

Area of Science:

• Process Control
• System Identification
• Adaptive Control

Background:

• Digital transformation in process manufacturing relies on model-based predictive control.
• Traditional methods struggle with changing and unknown operating conditions, leading to low accuracy.
• Adapting to dynamic environments is crucial for robust process control.

Purpose of the Study:

• To develop an adaptive predictive control method for dynamic process environments.
• To address challenges posed by changing and unknown operating conditions.
• To improve control accuracy during operating condition transitions.

Main Methods:

• Proposed an error-triggered adaptive sparse identification for predictive control (ETASI4PC).
• Utilized sparse identification to establish an initial model.
• Implemented a prediction error-triggered mechanism for real-time monitoring of condition changes.
• Introduced an elastic feedback correction strategy for improved transition accuracy.

Main Results:

• The ETASI4PC method demonstrated rapid adaptation to frequent operating condition changes.
• Achieved precise control across multiple operating conditions, including unknown ones.
• Significantly improved control accuracy during the transition periods.
• Outperformed state-of-the-art methods in numerical and CSTR simulations.

Conclusions:

• The ETASI4PC method offers robust and accurate predictive control in dynamic process environments.
• Effective adaptation to both known and unknown operating conditions is achieved.
• The method enhances real-time control capabilities for modern process manufacturing.