The Challenge:
Foaming events are an operational challenge felt across the entire oil and gas industry. They occur when separating out the oil/gas, and can increase the operating costs and reduce production efficiency. Usually a control room operator has little warning, perhaps a few minutes, before an event. Our client came to us looking to increase the delay and improve the ability to predict and prevent foaming from happening.

Our Approach:
We delivered a user interface allowing the control room to monitor in real time the likelihood of foaming within the next hour, along with contextual sensors data and features influencing the probability in real time.

The solution includes an automated retraining of the predictive modelling – along with performance monitoring – allowing the algorithm to update over time without manual intervention, taking into account any changes to operational or plant conditions.

The Results:
By using advanced AI techniques within a cloud environment, we were able to accurately predict the likelihood of a foaming event occurring within the next hour. The machine learning algorithm was deployed in the live environment (in real time), allowing the control room to be alerted immediately, and suggesting the factors that can be altered to prevent the foaming.