Logo image
Back
Distributed implementation of model-based methods on large-scale systems
Dissertation   Open access

Distributed implementation of model-based methods on large-scale systems

Leila Samandari Masooleh
Doctor of Philosophy (Ph.D.), Drexel University
Mar 2022
DOI:
https://doi.org/10.17918/00001015
pdf
Samandari-Masooleh_Leila_20225.25 MBDownloadView
PDF Open Access Open Access (License Unspecified)

Abstract

Large scale systems--Data processing Large scale systems--Mathematical models
The considerable advances in information and communication technologies in the past two decades have made possible the immediate transfer of information across the globe. These advances, together with the global competition for higher-quality chemical products at lower costs, require process automation tools that can quickly and effectively compensate for sudden global changes. To address this challenge, improved automation systems are needed to ensure the safe and optimal operation of the processes. As chemical processes are typically highly nonlinear and have increasingly more integrated subsystems with strong material and energy interactions, the automation systems should account for the interactions to be robust and effective and should be computationally efficient. Motivated by these needs, this dissertation focuses on developing methods that address the challenges associated with the implementation of model-based methods like model predictive safety (MPS) and model predictive control, on large-scale, highly integrated chemical processes. First, an efficient decomposition approach based on community detection is introduced to optimally divide large-scale process systems into smaller subsystems. The proposed algorithm detects all potential community structures in a weighted network by solving a multi-objective optimization problem using the whale optimization algorithm. Second, the problem of large-scale robust state-estimate prediction is studied to predict the present and future values of state estimate at every time instant in large-scale processes. An efficient algorithm that systematically identifies subsystems is developed for the distributed implementation of the state predictor. This algorithm allows for the use of computers with parallel processes to further improve its computational efficiency. Third, the problem of state prediction in multi-unit processes with multi-rate and delayed measurements is studied. Fourth, to solve the min-max optimization problems of MPS, a nested particle swarm optimization (PSO) algorithm is developed. Computational efficiency and quick convergence are attractive features of the PSO algorithm. The applicability and effectiveness of the methods and algorithms are illustrated using chemical process examples.

Metrics

41 File views/ downloads
41 Record Views

Details

Logo image