(247z) Linking Production and Utilities – a Holistic View on Plants and Energy Supply | AIChE

(247z) Linking Production and Utilities – a Holistic View on Plants and Energy Supply


Pöllabauer, F. - Presenter, Graz University of Technology, NAWI Graz
Wallek, T., Graz University of Technology, NAWI Graz
Bachmann, G., OMV Refining & Marketing GmbH
In chemical engineering the use of commercial process simulation software and detailed models for each individual unit of a chemical plant is state of the art, especially during plant design and as a basis for tuning of operating points.

For cross-plant optimization tasks in a complex compound system of several plants, where each plant comprises numerous unit operations, this detailed approach is often limited because substantial interdependencies between the plants are not considered, in particular the connections among the production plants and from production plants to the energy supply chain.

An oil refinery is a typical example of a complex compound system in which several production plants are interconnected in order to produce numerous product streams such as fuels and petrochemicals from crude oil. Residue-fired power plants as the main utilities provide the production plants with steam and electrical power.

Due to the tight interconnection between production plants and utilities, a simultaneous treatment is obvious. In doing so, the usually huge number of streams, recycles, apparatuses and chemical compounds makes it difficult to combine detailed individual models of several plants while maintaining model stability at reasonable amounts of simulation time. The latter aspect suggests to reduce the modeling-depth in an appropriate way.

This paper presents a comprehensible methodology to reduce the modeling-depth of a compound system using a top-down approach. This is achieved by decomposing the real system into an aggregation of appropriate sub-systems. Each plant is regarded as a deterministic black box, in which product streams and steam consumption are described as linear functions of the feed. These model equations are obtained by fitting the linear functions to measured mass flow data of the real plants, using the fulfilment of the mass balance for each plant as constraint. The overall static model of the refinery is represented by mixers, splitters and the fitted functions.

This procedure results in a set of functions in which all product streams and steam streams only depend on 1.) the crude feed to the refinery and 2.) a set of several split fractions.

All links with the associated utilities, like steam network, electric supply network, hydrogen network and fuel network can simply be considered by summation of all imports and exports of each plant. The same procedure is carried out for the power plants, where steam production as well as power generation are described by linear or polynomial functions of fuel or steam flows, combined with several decision variables. The resulting set of functions describing the main characteristics of each process step in terms of product yield and energy consumption can be used to formulate one target criterion or a weighted sum of several target criteria that can be influenced by both production and utilities.

This approach is compared to more detailed models using commercial process simulation software. Different model concepts, representing different modeling depths, are investigated in terms of their ability to represent the interconnections between the energy demand of the production units and the utilities. In this context, the modeling depth is varied between rigorous thermodynamics and tray-to-tray simulations, shortcut methods, component splitters and the presented data-driven approach, which represents the most simplified model concept.

Proposals for an optimal choice of the modelling depth are given based on examples from industry, ensuring a well-balanced trade-off between complexity, stability and accuracy of the model.