Machine Learning & Data Science
Introduction
Our software and workflows are developed on a state-of-the-art platform, with machine learning embedded into the algorithms. We have adapted traditional reservoir engineering workflows through automation and analytics.
In addition, our platform allows for high interactivity, high-impact adaptive visualization, load-on-demand curated data, image and audio processing, 3D printing, device connectivity, and multi-language communication.
Our platform has built-in capability for a range of M/L and D/S capability including Anomaly Detection, Synthesis of Missing Data, Automated Structure Discovery, Advanced Statistics, Filtering and Aggregating Time Series, Fitting and interpolation, Time Series Process Modelling and Forecasting, and more.
Extrapolation & Outlier Detection
Our models use specially customized versions of Siegel’s repeated median, Theil-Sen , RANSAC and other specialist statistical algorithms for extremely robust outlier detection, interpolation and curve fitting. This allows for more accurate and fully automated “Flowing Material Balance”, “Decline Analysis” approaches. This results in a significant improvement to the business process, due to the reduction in processing time, and eliminating the need for subjective human interpretation.
Analysis & Forecasting
Our Machine Learning and data science techniques are the results of a combined efforts our petroleum engineering and data science team members. Our models and algorithms are bounded by the principles of fluid flow in porous media, and do not operate on as black box.
Smart Filtering
Predico and its team has begun the introduction of “Smart Filtering”. This is a physics-based approach based on the concepts of productivity index and material balance, which is combined with powerful statistical techniques, to automatically smooth, filter, and correlate pressure-rate data.