Abstract
The order of drilling a collection of wells matters. One possible sequence gives you an NPV and a total duration and another slightly different sequence of the same wells can yield an altogether different outcome, even simply swapping two wells in order. The nonlinearity of optimizing the sequence means that above a very small number of wells it is impossible for a human to intuitively identity the optimal sequence—the one sequence that yields the "best" possible outcome. As a guide, sequencing 13 wells has a surprisingly large solution space of 6.2 billion possibilities.
We must harness computing power to understand the optimal sequence of wells to drill and complete. Facilities have to be timed perfectly to be in place when the wells are ready for production. Zipper frac opportunities arise in a once-in-a-billion chance when two adjacent wells allow it to be done in a practical way. The special math of optimization is used within a computer model to calculate the optimal sequence, and the associated KPIs such as profitability and total program duration. In effect, it is as if the computer model tries every single one of the billions of possible sequences, and places the leaders on an ever-shifting leaderboard as it goes along.
In our talk we will elaborate on three past projects where we have applied analytical engines to discover optimal well sequences with lessons on how to reproduce this work in your own drilling operations.
Conventional approaches to optimization, such as those found in downstream applications like refining do not always work adequately for well sequencing due to the highly non-linear nature of the constraints, such as spatial interference between bit positions of adjacent wells. We will discuss a heuristic (non-traditional) approach that was required in one particular case of well sequence optimization.
