A VALUATION MODEL FOR SUPPORTING THE DECISION TO ACQUIRE AN UNEXPLORED OILFIELD

• Main Author: Rian Pratikto, Fransiscus
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75458
• Institution: Institut Teknologi Bandung

<p align="justify">Acquisition of an oilfield is a repeat decision involving a substantial amount of funds covering a lengthy period with a high risk, especially for an unexplored one. In such a situation, investors usually use a valuation model to ensure a quality decision. A valuation model of an unexplored oilfield is typically more complex than the explored and developed ones. This complexity stems from the various types and the magnitude of uncertainties embedded in an unexplored oilfield. Previous unexplored oilfield valuation models use less realistic assumptions regarding these uncertainties to pursue mathematical tractability. This research aims to develop a more realistic valuation model while still preserving the solution optimality. The model is also operational in that it can be used by the parties involved in the oilfield acquisition decision, i.e., the government as the field's owner and oil companies as investors and operators. The valuation model considers uncertainty from exploration outcomes, reservoir conditions, and oil prices. The exploration outcome is modeled as a binomial process, while oil prices follow the two-factor mean-reverting process (the Schwartz-Smith model). The reservoir condition is represented using a set of probabilistic parameters, based on which production rates are estimated analytically using a dynamic compressible-liquid tank model. The joint probability distribution of reservoir parameters represented using a Gaussian copula is estimated using a machine learning approach combining the Gaussian mixture model and artificial neural networks. The valuation model also incorporates investors' flexibility to abandon the field when it becomes economically not viable. The objective function is to maximize the oilfield value based on the optimal field operation scenario. This scenario is represented in terms of the number of producing wells in each period and abandonment time. This valuation problem is a typical stochastic dynamic programming problem with a simulation-based reward function. The solution is obtained using the simulation-optimization approach by employing a random search method, i.e., the real-coded genetic algorithm. A hybrid variance reduction technique (combining the Latin hypercube sampling and antithetic variates) and a data-based approach to narrowing down the search space are used to increase the computational efficiency. The valuation model is implemented using a computer program written in R using the RStudio Server and run on the Google Cloud Platform using a compute engine consisting of 8 vCPU 32GB RAM. It takes 14 days to run the whole code. The model was parameterized using Brent crude oil's spot and future prices and proven reservoir data from the TORIS (Tertiary Oil Recovery Information System) database. The oil prices and reservoir condition submodels are shown to be valid. The model produces the distribution of the oilfield value represented using its mean and conditional value at risk (CVaR). The model has also been proven capable of discriminating oilfields with different risk levels. The novelty of this research lies in how the model simultaneously incorporates those three sources of uncertainty in the unexplored oilfield valuation and how the reservoir conditions are estimated and represented. The use of simulation optimization in an oilfield valuation model is also new. The weakness of this model is the lengthy computational time required to obtain the solution. Since analytical approaches to increasing computational efficiency have been incorporated into the model, further computation time reduction can only be achieved by using more sophisticated compute engines. For the model's users (the government and oil companies), the cost of the compute engines is relatively small compared to the benefit of the improved acquisition decision quality. This valuation model can be developed further for other related problems, such as the valuation of gas fields using proven reservoir data from the Gas Information Systems (GASIS), oilfield valuation under a partnership, and valuation of a portfolio of oilfields. The oilfield valuation under partnership can be developed by adding the share of participating interest as a decision variable. Meanwhile, in the portfolio model, this valuation model will be the submodel with the proportion of capital allocated in each oilfield as the additional decision variables.