Team Drillpy: winners of the 2023 Energy AI Hackathon

I'm thrilled to announce that my team, Drillpy, won the 2023 Energy AI Hackathon! The competition focused on predicting failures in electrical submersible pumps (ESPs) used in oil and gas production. We analyzed time-series data from sensors installed on the pumps, in the wells, and on the surface. This data was essential for understanding the pumps' performance and identifying potential issues.

To tackle the challenge, we created a set of features by calculating various statistical properties from the data, which helped us highlight key aspects of the signals, like their variability and how they deviated from a normal distribution. We also addressed the issue of imbalanced data since there were few instances of pump failures in the dataset. We used a random forest classifier, a machine learning model that combines multiple decision trees, to predict whether a pump would fail.

Our solution was evaluated based on the model's accuracy, the clarity of our presentation, and the organization of our code. We achieved the highest accuracy, and our presentation and well-structured code impressed the judges, securing us first place. This experience was not only a great chance to apply our skills in data analysis and machine learning but also to leverage our knowledge in production engineering to develop a meaningful feature set that effectively distinguished between failing and non-failing pumps.

For more information, please visit: 3rd Annual Energy AI Hackathon