In the current market of drilling programs that emphasise cost effectiveness, making optimum decisions in a dynamic real-time setting is very important. We have observed a disproportional reliance on intuitive decision making in upstream oil & gas operations. In this paper, we propose a process that enables data driven decision making. We then implement this process for a specific situation.

In homogeneous unconventional shale drilling, the number of factors that effect drilling performance are greatly limited. Well drilling times are relatively shorter and so many more wells are drilled in a short period of time. This allows for an opportunity to learn quickly from past wells. We developed a decision-making dashboard that uses data from past wells and helps predict the outcome of a change before it is made. We used multiple linear regression (MLR) as the statistical tool to derive correlations between parameters and drill rate (ROP). Historically, MLR has been used as a tool to optimize the conventional rotary drilling process. However, these analyses are based on limited field data and can yield inaccurate results. The introduction of new drilling technologies such as polycrystalline diamond compact (PDC) bits, rotary steerable systems (RSS), and horizontal well placement have render past models obsolete when applied to modern unconventional homogenous shale wells.

Real-time data were collected from the lateral sections of eight horizontal Marcellus Shale wells drilled from a single pad located in Greene County, Pennsylvania. Twenty-seven variables were identified and evaluated. A MLR of these variables was calculated for each of the eight wells. Weighted averages of the resulting coefficients were then used to derive a predictive ROP model. A decision-making dashboard was created using the coefficients generated through MLR on field data specific to the rig, well-plan, and area being drilled.

In conclusion, a novel method of making data-driven decisions in real-time was developed. This dashboard is especially useful because it formulates a method of extracting actionable information from historical data. Care must be taken to ensure data used for analysis are of good quality and specifically related to real-time applications. These considerations include rig limitations, formation types, well profiles, mud properties. Updating the dashboard with more data as it is gathered enables continued learning from historical well-data in a systematic, data-driven method that has not existed in the past.

You can access this article if you purchase or spend a download.