核心概念
The core message of this article is to develop an iterative partition-based distributed moving horizon estimation (DMHE) scheme for linear systems, where the local estimators penalize both system disturbances and measurement noise to provide accurate and stable state estimates.
摘要
The article presents two DMHE designs for linear systems:
- DMHE-1 for the unconstrained case:
- The global objective function of centralized MHE is partitioned into local objective functions for the subsystem estimators.
- The local objective functions incorporate penalties on both subsystem disturbances and measurement noise from the interacting subsystems.
- The local estimators are executed iteratively within each sampling period to converge to the centralized MHE estimates.
- Sufficient conditions are derived for the convergence of the subsystem state estimates and the stability of the estimation error dynamics.
- DMHE-2 for the constrained case:
- Hard constraints on subsystem states and disturbances are incorporated into the local objective functions.
- The local estimators are executed iteratively to solve the constrained optimization problems.
- Stability of the entire distributed estimation scheme is proven for the constrained case.
The proposed DMHE schemes leverage the advantages of distributed architectures, such as improved fault tolerance, computational efficiency, and organizational flexibility, while providing accurate state estimates convergent to the centralized MHE. A benchmark chemical process example is used to illustrate the effectiveness of the proposed methods.
統計資料
The article does not contain any explicit numerical data or metrics. It focuses on the theoretical development and analysis of the proposed distributed estimation algorithms.
引述
"The major advantage of these iterative distributed state estimation approaches lies in that they have the potential to provide estimates convergent to the centralized counterpart; this is favorable when more accurate estimates or faster convergence are needed."
"In this work, the entire system is partitioned into subsystems that interact with each other. An individual objective function incorporating penalties on both subsystem disturbances and measurement noise is formulated, and local MHE-based estimators are developed to provide estimates of the subsystem states in a collaborative manner."