Solution Seeker granted funds to continue research of Transfer Learning methods for data-driven rate estimates

The Norwegian Research Council recently awarded Solution Seeker funds for an Innovation Project. The objective is to obtain better well flow rate estimates by using the machine learning concept of transfer learning.

Solution Seeker
Publish date
· 1 min read


typical error about half the time

The title of the project is «Transfer learning for oil and gas wells: unlocking the collective potential of production data from multiple oil fields». Using data-driven methods to solve the still existing challenge of live and accurate well flow rate estimates has many potential benefits. What is that still existing challenge do you ask? Analyses that we have done from multiple real fields in operation show a typical error of up to 20% on the rate estimates roughly half the time.

Automation, online learning, higher accuracy and additional insights through uncertainty and quality measurements are some of the benefits we are chasing with data-driven methods. But using straight out of the box, "vanilla" machine learning methods just wont do.

The Transfer Learning concept we will further develop in this Innovation Project directly adresses the typical machine learning challenge of "too little data". Stay tuned to our journal to learn more about these methods, or read more about the project here (only in Norwegian):