Water breakthrough indicator

How we helped one of our partners automate water breakthrough indications

Value proposition

  • Increased certainty in detecting indicators of water breakthrough using data-driven models
  • Reduced routine work by condition-based alerts of water breakthrough indications
  • Real time overview of indicators and risk analysis of water breakthrough for each well


Water loading in wells can be caused by water coning, or water breakthrough, leading to reduced production of hydrocarbons. In the worst case it can cause a shut-in of a well. Indication of water coning can be found in the wells’ signals, but it demands a trained eye to observe it in the raw data. Such monitoring is time consuming, and becomes increasingly difficult to track when the number of wells is high.

Early detection of increased water flow from the well enables faster response to avoid water loading and thus avoid a costly problem.


Our partner wanted to better utilize their live available sensor data from the field, to help identify indications of water breakthrough. Together we have developed a solution that continuously performs several analyses on key features to predict a water breakthrough, based on their historical and live data. All raw data is fed through Squashy, our data compressor, helping us mine out high quality data. The data is then analyzed by a combination of machine learning models and covariation trends that continuously run on each well.

The model is meant to assist production/reservoir engineers in their search of early water breakthroughs. A notification is sent to the user once the model perceives there is a high possibility of a coming water breakthrough so the engineers have time to investigate and take countermeasures. All wells are displayed in a ranked overview, where the highest ranked wells are those that the algorithm currently perceives with the highest risk for a breakthrough. Deep dive pages can also be accessed to visualize the critical features per well, to quickly assess why a well gets its current ranking from the model and if there is need for further actions.


With our solution the production or reservoir engineers get a fact-based overview to support them in making decisions about intervening in a well’s production to take precautionary measures against water breakthrough.

Having an alarm go off when wells reach a risk threshold has given the engineers more certainty about their wells and it’s no longer necessary to check for water breakthrough as often as before. The real-time monitoring of all the wells also acts as a time saver, as the engineers can quickly identify the wells that have the highest likelihood of water breakthrough.