No abstract

This article describes a dynamic, control relevant, mechanistic model of the TEALARC liquified natural gas process. The model is to be used for both steady-state and dynamic controllability analysis. The model therefore needs to be computationally light, but still include enough complexity such as to study the impact of capacity constraints on the control structure. Structured assumptions have been used to obtain simplified representations of gas/liquid flows and thermodynamic properties. The steady-state operating points of the dynamic model have been adapted to a given steady-state process design model. The paper demonstrates that the model is well suited for operability analysis. Steady-state and dynamic characteristics are illustrated.

This article shows how controlled variables (CVs) of the regulatory control layer in a liquefied natural gas (LNG) plant can be chosen as linear combinations of measurements using self-optimizing control principles. By self-optimizing control, the CVs are chosen such that the set points of the CVs remain close to steadystate optimal despite disturbances, thus reducing the need for online reoptimization. Several methods for calculation of linear combinations within this framework are compared. Self-optimizing control design can also be used in the process design phase to place measurements by reducing a maximum candidate set of measurements to a best possible subset of measurements giving an acceptable loss. This article proposes a relatively simple method for successive selection (SS) of measurements and compares this approach to a more comprehensive branch-and-bound (BB) method for selection of measurements. The results indicate that, although the BB method gives lower average losses for very small subsets, the methods are comparable with respect to average losses for medium and large subsets, and the SS method outperforms the BB method in terms of computational load.

The paper addresses e¢ cient methods for parameter sensitivity analysis and ranking in large, nonlinear, mechanistic models requiring examination of many points in the parameter space. The paper shows how orthogonal decomposition and permutation of the sensitivity derivative is an intuitive and structured method for automatic ranking of the parameters within a candidate set. Provided the model error is Gaussian, and with the problem on a triangularized form, the additional variance associated with each parameter can easily be found. Ranking according to additional variance is therefore another option. The methods are tested on an industrially used simulator model.