Thanks to exploring data science, Tecpetrol has developed a Recommendations System to improve drilling efficiency at its operations in Fortín de Piedra and El Tordillo.

Big Data as a driver of innovation

Over recent decades, the amount of information available worldwide has grown exponentially. Although we may not necessarily realize this, we are in fact surrounded by data—and constantly producing it. Each time we send an email, make a comment on Facebook, listen to a song on Spotify, or connect to the GPS when we take the car out, we are creating information. It’s estimated that a single person generates 1.7 MB of information every second.

To measure how this process is speeding up, the International Data Corporation (IDC) quantifies and forecasts the amount of data produced annually. For example, in 2013, there were 4 trillion gigabytes of data generated worldwide; just seven years later, the digital universe had ballooned tenfold, reaching 44 trillion Gb of data.

This is what is known as Big Data, the new engine powering the world economy, and an inspiring technological challenge that is transforming the way we live and work.

The relevance of Big Data is not about how much data is available, but how it can be analyzed and combined to help with taking smart decisions. In short, how data can be turned into useful knowledge through Data Science, a novel area of expertise developing in a range of different industries such as commerce, entertainment, finance, and more recently the energy sector.

Data science at Tecpetrol

Over the last few years, Tecpetrol has been taking a Research & Development approach to its investigation of data science in order to identify opportunities to add value and create knowledge to support decision-making.

"To start with, in 2018 we carried out an assessment of different areas of the company, interviewing various people in order to analyze opportunities for applying data science methodologies or techniques," explains Pablo Blasco, Industrial Management Manager. The interviews led to a list of initiatives for the first pilot projects, which were prioritized thanks to feasibility and impact evaluations.

Currently, Tecpetrol has completed twelve of these pilot projects and has another three in the pipeline. These projects are about solving problems encountered in drilling, completion, reservoirs, production, and maintenance, among others. In addition, the company is getting involved in different forums and conferences in the industry as a way of pooling new knowledge on innovation.

One of the most innovative projects to come out of the company’s Data Science initiative was presented as a best practice at the Argentine Petroleum Institute’s recent Digital Revolution Conference. The Recommendations System was developed in conjunction with Tecpetrol’s Drilling and Work Over Department, involving the use of algorithms to improve drilling efficiency and performance.

Ezequiel Urdampilleta, head of process analysis at the Industrial Management Department explains that "When you design a good plan, you draw up a roadmap, essentially a static model analyzing the drilling parameters at neighboring wells in order to establish the best parameters for the current well proposed. This means that for each formation, the best operating parameters are sought: weight on bit (WOB), revolutions per minute (RPM), flows, etc."  However, as wells and formations don’t always respond as expected, it’s necessary to undertake corrective action in real-time to improve conditions. Alternatives are tested on the basis of the experience of the operators, drilling engineers, and the company man. In general, the parameters achieved tend to improve the outcome, but there’s no way of establishing whether these are in fact optimal or if, on the contrary, there could be another kind of combination able to produce better results.

To shorten learning cycles and achieve more efficient corrections at the operations, the Recommendations System provides an automated solution calculated on the basis of the high volume of data available from past successful or high-performing drilling operations. It then applies data science algorithms using machine learning to enable decision-making during the course of drilling aimed at optimizing the rate of penetration (ROP).

The information used to generate the dataset is drawn from historical drilling variables at Fortín de Piedra and the Mina del Carmen formation in the El Tordillo field. "These data have to have a specific shape and features to meet the algorithm’s input requirements," points out Pablo, adding that these earlier projects provided key information for consolidating data governance and management as well as advanced analytics processes at Tecpetrol.

Pablo Anicich, who is a big data specialist and external consultant from global tech consultancy Practia, accompanied the process. He goes into detail to explain how they processed data from both efficient and inefficient drilling configurations at some thirty wells in Mina del Carmen were processed. "The number of samples involved in the configurations comes to about 60,000 events every 2 minutes, that is to say, that we extracted 720,000 observations from the formation," he details.

At the state-of-the-art control room located at the company’s brand new corporate facilities in Neuquen, the drilling engineer can communicate with the company man to identify drilling efficiency points and suggest changes in real-time, such as bit weight. Additionally, drill sensors provide continuous information on variables that are fed into the algorithm so that it can update itself and continue to learn. “Just as having a drilling control room is quite unusual in the industry, so is using these tools to optimize operations,” adds Pablo.

In fact, the longer-term objective is to take this algorithm model to other fields and for the Recommendations System to be scaled to other operations in order to continue improving drilling efficiency, which is crucial for well construction and development.