Uncertainty and risk analysis is an inseparable part of any decision making process in the field development planning. This study sheds light on the available approaches to capture the range of uncertainties but digs deep into the misuses of the probabilistic approach that renders the method difficult and time consuming to implement with little added value for risk mitigation and proper decision making.
Probabilistic modeling using dynamic simulation models has been adopted in recent decades to address the variations in forecasted production profiles and to capture the uncertainties. However, there are misuses in the approach that pose questions on the outcome and its meaningfulness. Lack of enough spread in the forecast, history-matched models with physically incorrect parameter ranges/ combinations and models suggesting contradicting development scenarios are among examples. These in turn make the probabilistic forecasting output inconclusive and considering the high computational cost and time required to perform the exercise makes it unattractive to management. In this paper four case studies including mature and green fields have been described and a number of main issues and pitfalls of using probabilistic dynamic modeling in those cases are analyzed. General workflows are then presented for green and brown fields based on experimental design, proxy modeling, optimization and prediction candidates selection that provides solution for proper selection and implementation of the probabilistic dynamic modeling.
It is argued that probabilistic modeling can help better capture the uncertainties and reduce the risk in field development planning provided that a fit-for-purpose approach is taken with correct understanding of the data requirement according to the reservoir complexity, the physical processes being modeled and assumptions used in the methodologies and simulation engines. This is in contrast to the attempts to capture the ranges of recoverables based on deterministic high and low cases that is often inefficient as the optimistic high-case of ‘hole-in-one’, may suggest an ideal but not plausible scenario whereas the pessimistic low-case of ‘train-wreck’ may be economically unattractive. The exercise then leaves the companies with the best technical estimate model to make the final call and the numbers from other models are only used for reserve booking purposes.
The published papers in the literature include discussions on deterministic vs. probabilistic approaches and selection of base case models, the detailed algorithms and also case studies done using the published methods available in the commercial softwares. This paper however discusses the misuses of the probabilistic dynamic modelling approach and tries to inform the audience of the pitfalls of not understanding the reservoir and/or the tools used in implementing the methods and in this sense it is novel.