Improving Reservoir Production Forecast Accuracy — Application of Lessons Learned From Conventional Reservoirs

Abstract

Abstract

Nandurdikar and Wallace (2011) reported that the production forecasts used to sanction projects are significantly optimistic. They note that the industry produces about 75% of the forecast production. In projects that have subsurface “issues” such as overestimated hydrocarbons in place or underestimated compartmentalization, actual recovery was about 55% of the forecast. In a time of low prices, improving capital efficiency via accurate production forecasting is critical.

Four factors were evaluated to determine their potential contributions to forecast optimism. As many projects utilize reservoir models as part of their forecasting workflows an analysis of modeling components such as the semivariogram parameters, model grid definition (size, number of cells), and upscaling show that only the model grid definition is a major contributor to forecast optimism and may contribute 20-30% of the observed forecast optimism. Sparse data is a major contributor to forecast optimism, particularly in the early stages of appraisal when well locations may be “biased” to be in higher quality reservoir volumes. Analysis of multiple projects shows that sparse data may contribute 25-50% of the observed optimism. Human bias specifically associated with a real or perceived “need” to have a project move forward is a major contributor to forecast optimism. Statistical analysis along with experiments with “actors” in technical team and management roles show that human bias may contribute 25-50% of the forecast optimism. Well optimization workflows may contribute 10-15% of the observed optimism particularly for heterogeneous reservoirs.

Improving reservoir forecasts involves evaluation of the forecast sensitivity to model parameters; particularly grid size as models with many small cells yield less optimistic forecasts. It makes almost no sense to use coarse models simply to decrease run times if the results will be optimistic. Lowering the impact of sparse data must go beyond assessing the uncertainty of what is known to include assessing the impact of what is not known by more consistent use of analog reservoirs. Well location optimization induced optimism is reduced by incorporating multiple uncertainties in the workflows. Human bias can be reduced by using truly independent peer review processes and larger catalogs of analogs. The lessons learned from conventional reservoirs are largely applicable to unconventional reservoirs.

AAPG Datapages/Search and Discovery Article #90260 © 2016 AAPG/SEG International Conference & Exhibition, Cancun, Mexico, September 6-9, 2016