This paper presents the first experiments undertaken with a stochastic programming model of gas contract portfolio management. This model is intended to be the core of a decision support system for helping gas marketers of American or Canadian firms, facing the deregulated North-American market, in their decisions to commit their reserves to different market segments. The data set used for these simulations is adapted from a previous paper in which the approach was based on an adaptation of the Markowitz model. The simulation results presented here permit a comparison between the open-loop and the closed-loop decision making contexts. A closed-loop solution produces a policy adapted to the random changes of the state of the markets. This possibility to adapt is an important feature of the stochastic control model. Various perturbations of the base case run are considered for a first evaluation of the sensitivity of the model to variations in some key parameters. A discussion of the further developments needed for the construction of a decision support system based on this methodology is also included. Finally the implementation of the model in the GAMS modeling framework is fully described.
Published February 1992 , 27 pages