Particle filter combined with data reconciliation for nonlinear state estimation with unknown initial conditions in...

Area of Science:

• Process Systems Engineering
• Control Engineering
• Computational Science

Background:

• State estimation is critical for dynamic process control and optimization.
• Particle filters (PFs) are effective for nonlinear dynamic systems but typically require known initial conditions.
• Industrial processes often lack known initial states for nonlinear dynamical systems.

Purpose of the Study:

• To develop a novel methodology for state estimation in nonlinear dynamic systems with unknown initial conditions.
• To integrate particle filters with data reconciliation for improved initial state estimation.
• To enhance the accuracy of state estimation in industrial processes.

Main Methods:

• A novel methodology combining Particle Filter (PF) with data reconciliation is proposed.
• A measurement test criterion and sequentially increasing data information are used for data reconciliation.
• Iterative refinement of initial state values through interaction between PF and data reconciliation.

Main Results:

• The proposed method effectively estimates states in nonlinear dynamic systems even with unknown initial conditions.
• Reliable initial state values are derived using measurement data and reconciliation techniques.
• The interactive approach between PF and data reconciliation leads to accurate state estimation.

Conclusions:

• The combined PF and data reconciliation approach provides accurate state estimation for nonlinear dynamic systems with unknown initial conditions.
• This methodology addresses a key limitation of conventional PFs in real-world industrial applications.
• The technique demonstrates effectiveness in improving process control and optimization through accurate state estimation.