Rapid and disruptive changes in the oil and gas industry necessitate a fresh look at all aspects of decision-making to achieve significantly higher capital and resource deployment efficiencies. Traditional tools such as reservoir simulation are ineffective for decision-making in unconventional assets, many operators have been using statistical type curves for years. However, statistical type curves are too simplistic and cannot capture the non-linear and complex effect of geology, completion, well spacing, and operational factors on well performance.

SpeedWise^® Unconventionals (SUN) is an AI-driven technology that optimally blends data-driven models with expert-based knowledge to optimize completion design, well spacing, workover activities, and subsurface targets for any given unconventional asset, applying robust predictive models to identify actionable future opportunities that maximizes the NPV of the field and minimizes potential risks.

How Does SUN Work?

SUN implements State-of-the-art Machine Learning practices to conduct unbiased data analytics with unique model explanation and uncertain quantification* capabilities. Within SUN, there are three different AI-based methodologies for robust and flexible production forecasting:

1. Automated Decline Curve Analysis (DCA): A data-driven workflow that can automatically perform multi-segment DCA for wells and is mainly used to estimate reserves. (figure 1)
2. ML-based Parametric Models: Assume a functional form for production profile (e.g., Arp equation) & build ML models to predict model parameters. It is mainly used for planning new acreages (e.g., well spacing, completion design optimization). (figure 2)
3. DL-based Time Series Models: Performs probabilistic multi-step forecasts for oil, gas, and water rates using a proprietary Encoder-Decoder deep-learning architecture. It is mainly used for actively managing producing wells, optimizing OPEX, and planning workovers. (figure 3)

SpeedWise^® Unconventionals Features

• SUN integrates the powerful capabilities of SpeedWise^® Decline Curve and more, such as vanilla machine learning and deep learning (Recurrent Neural Networks), for flexible probabilistic production forecasts.
• Automatic feature extraction (spacing calculations, geo-parameters extraction from logs and maps, etc.)
• Unmatched data processing capabilities (Data imputation, categorical encoding, decorrelator, dimensionality reduction, etc.).
• Integration of SpeedWise® ML platform that embeds robust, explainable, trained ML models for accurate predictive analysis.
• Economics sensitivity analysis with risk quantification.
• Interactive web-based dashboards.