Estimating Fluid Properties in North Sea Using a Causal AI Agent

Summary CausX AI is an epistemic, causal artificial intelligence agent capable of learning from both data and expert hypotheses. It performs well with sparse and incomplete data sets rather than requiring the very large data sets typically needed by traditional machine learning models. It creates a multi-view hypothesis set subject to situation-specific guardrails building upon initial training data and ground truths. Using CausX AI, Senslytics created an AI application for fluid property estimation, ResVoirX, which uses gas composition views from mud-gas logs along with related log and drilling data to identify the reservoir fingerprints and uses that for training purposes. The trained AI agent uses mud gas logs from standard or advanced mud gas traps as its data source and applies both geological age and density clustering to generate higher reliability gas/oil ratio (GOR) and reservoir fluid density predictions. This novel approach has been used to model reservoir fluid in the Gulf of America most often within 15% error margin. In this paper, we discuss the results of our first application of this approach in the North Sea which has shown similar excellent results. This method offers a major advantage by de-risking decision-making and boosting operational efficiency.