Introduction I have always been captivated by calculations performed at depth in wells. The numerous correlations and curves that were built in the golden years of production engineering are just fascinating. Thank you Mr. Brown. Thank you Mr. Beggs. From all the various algorithms, I particularly liked one calculating the pressure losses in the tubing as the hydrocarbon fluids ascend to the surface, also called tubing performance, or vertical lift performance (VLP).
Continuing with my transition from network folders to a Rmarkdown blog system, I found a presentation I gave on few best practices for the construction of well models. After three years of building wells and network models, I came up with this list of recommended practices. It is unique in the sense that it took lessons learned by tackling the continuous well optimization process using Data Science.
The previous article Scripts for Well Modeling and Batch Automation is part of the execution of well modeling which is just a component of the whole process.
Motivation I am moving my collection of slides and previous data science work in production engineering to a static type website using Hugo and Rmarkdown. I came to the realization that is too much work to perform the calculations in one environment (R or Python), and then having to copy-paste code and plots. I think it will be more efficient to have R and Rmarkdown do that for me instead. I think that reproducibility is one of the most important aspects of data science that petroleum engineers have to strive for work in machine learning and artificial intelligence.