Investments in the LNG Value Chain : A Multistage Stochastic Optimization Model focusing on Floating Liquefaction Units

In this thesis, we have developed a strategic optimization model of investments in infrastructure in the LNG value chain. The focus is on floating LNG production units: when they are a viable solution and what value they add to the LNG value chain. First a deterministic model is presented with focus...

Full description

Bibliographic Details
Main Authors: Røstad, Lars Dybsjord, Erichsen, Jeanette Christine
Format: Others
Language:English
Published: Norges teknisk-naturvitenskapelige universitet, Institutt for industriell økonomi og teknologiledelse 2012
Online Access:http://urn.kb.se/resolve?urn=urn:nbn:no:ntnu:diva-20980
Description
Summary:In this thesis, we have developed a strategic optimization model of investments in infrastructure in the LNG value chain. The focus is on floating LNG production units: when they are a viable solution and what value they add to the LNG value chain. First a deterministic model is presented with focus on describing the value chain, before it is expanded to a multistage stochastic model with uncertain field sizes and gas prices. The objective is to maximize expected discounted profits through optimal investments in infrastructure. A dataset based on a set of potential fields on the Norwegian continental shelf, with shipping of LNG to three markets in the Atlantic basin, is used to solve the model. The results illustrate when FLNG units can add value to the value chain. They are used as a supplement to onshore processing plants; for example expanding peak capacity or to react to the resolution of uncertain parameters. The floating liquefaction option is especially attractive for fields located far from shore. We also find that the main reason for using FLNG units is their lower liquefaction costs, not the ability to move between fields. The stochastic version of the model results in solutions very similar to the solutions of the deterministic model, even though it is significantly harder to solve. Dantzig-Wolfe decomposition is implemented to reduce run times, but does not converge.