Abstract
The hydrocarbon industry is constantly challenged to improve subsurface description and operating practices to maximize returns on capital employed. Well locations are typically identified by continuously improving the geologic and reservoir engineering modeling. As a consequence, and on an on-going basis, refinements in drilling and stimulation practices are introduced based on a better understanding of rock and fluid interaction.
However, over the near term it is difficult to validate the impact of changing processes. It takes typically several months before stabilized flow conditions are established such that well performance can be determined with confidence from decline curve analysis. Moreover, it takes several years before robust and reliable estimated ultimate recovery (EUR) predictions can be made. Hence, the findings are obtained often too late to influence ongoing unconventional field development, especially in manufacturing pad development that is characterized by high levels of drilling and completion activities.
In this work we propose a benchmarking system that enhances the forecasting of long-term well performance, based on metrics of short-term well behavior. The metrics are routinely recorded from the first day of production such as peak month production or 3, 6, 12 and 24 month cumulative production. The process is refined continuously as and when new information becomes available and the added value of the information is quantified. A new key performance indicator (KPI) condenses the range of predicted outcomes into a single number via a simple numerical function that allocates increasing cost for increasing misclassification. The range is 0% = no predictive capability of the system to 100% = perfect predictive capability of the system. This approach enables a transparent and unbiased benchmarking of the system’s prediction capability for each metric.
In adding to the analytical toolkit, the key objective of this benchmarking method is to support decision making on an ongoing development, well before the entire program has been executed. Possible applications include:
Early confirmation of successful well placement.
Early indication of the impact on well performance as a result of changes to drilling and stimulation procedures.
A ‘conditional probabilistic’ outlook of long-term well behavior to better define well/field economic scenarios and to guide reserve bookings.
This process has been developed using public data from the data rich fields Barnett in Texas, Fayetteville in Arkansas and Woodford in Oklahoma. This process is also viable for plays with scarce data and is able to be refined with increasing data availability.
