ESP failure prediction: 19% net revenue improvement

Problem

ESP failures are expensive — lost production, intervention cost, and the staging of replacement equipment. Operators historically reacted to failures rather than predicting them. R2 was asked whether failure timing could be forecast far enough ahead to meaningfully reduce downtime.

Approach

R2 trained physics-informed predictive models on five years of historical SCADA, well-test, and intervention data. The pipeline ran in SubPUMP for hydraulic context plus MATLAB-based predictive models against live SCADA streams. The model surfaced rising-risk wells weeks before failure.

Result

Across the operator’s portfolio, the predictive-maintenance workflow delivered a 19% net revenue improvement — captured by intervention staging that reduced unplanned downtime and let intervention crews work optimal routes.