Machine learning-assisted tight gas forecasting
Challenge
- Forecasting tight gas production is challenging due to complex reservoir behavior, changing operating conditions, and pressure dynamics.
- Traditional decline curve analysis (DCA) assumes constant decline trends, limiting forecast accuracy under real-world reservoir conditions.
- Manual forecasting workflows are time-consuming and make it difficult to respond quickly to changing production conditions.
- Reservoir engineers need a more adaptive and data-driven approach to improve forecast reliability and support timely production planning and operational decisions.
Web application containing various tabs, including a pressure analysis module that provides essential engineering diagnostics such as the Horner plot, Bourdet derivative, linear-equivalent drawdown, and p/z plot. These analyses are automatically computed in the backend and processed to support in-depth reservoir analysis.
Solution
- Delivers an automated, end-to-end workflow for short-term gas production forecasting by combining physics-based analysis with machine learning.
- Provides an intuitive web application where users can upload production, reservoir, and operational data, perform pressure diagnostics, train forecasting models, and generate predictions for both existing and new wells.
- Integrates advanced engineering analysis, including Horner analysis, Bourdet derivatives, p/z material balance, linear-equivalent drawdown, and choke-control dynamics, to enhance model accuracy.
- Applies the most appropriate machine learning model, including Gradient Boosting, Random Forest, and N-HiTS (deep learning model for time series forecasting), based on well characteristics and forecasting requirements.
- Enables efficient field-scale forecasting with map-based well selection, automated pressure buildup identification, operational schedule inputs, and exportable prediction results.
Results
- Reduces reliance on manual engineering workflows while overcoming the limitations of traditional decline curve analysis (DCA).
- Accelerates the generation of accurate and reliable production forecasts.
- Enables reservoir engineers to generate rapid, repeatable forecasts grounded in both engineering physics and historical production trends.
- Improves forecasting accuracy for both existing and new wells, supporting more reliable planning for choke management, drilling schedules, and production targets.
- Supports faster decision making, greater forecast confidence, and improved field development planning.
The prediction tab enables production forecasting for both existing and new wells, providing reliable prediction capabilities through an intuitive interface.
