Adiga, Rishi

Stochastic Mixed Integer Programming for Geothermal Well Decision Optimization

Rishi Adiga, Andy Philpott, and John O’Sullivan

Department of Engineering Science, University of Auckland

Drilling geothermal wells has a very high capital cost, and as such, it is imperative to maximize value from wells by selecting them optimally. An important technology used when making well placement and scheduling decisions is computer simulation of production. This is usually done manually, with experts creating reservoir models, and simulating and comparing different production scenarios. These reservoir models have the problem of non-uniqueness; different models can fit available data equally well but give differing predictions.

This paper uses Stochastic Mixed Integer Programming models to address this problem. An economic model was created to calculate Net Present Values for a set of candidate wells at different start times, and the interactions between them, using AUTOUGH2 simulation results of an example geothermal system. Binary decision variables were used to select the combination of wells and start times that would maximize total NPV. Stochastic, multistage optimization models were formulated to account for uncertainty in the numerical simulation data, reflecting a realistic geothermal project, and to make hedging decisions based on this uncertainty.

This presentation is eligible for the ORSNZ Young Practitioners Prize.