About this Training Course
The oil and gas industry is increasingly adopting artificial intelligence (AI) and machine learning to improve reservoir characterisation, production optimisation, predictive maintenance, and field development planning. While these technologies have demonstrated significant value in extracting insights from large and complex datasets, conventional machine learning models often produce point predictions without indicating the level of confidence or uncertainty associated with the results. As petroleum engineering decisions involve substantial technical and financial risks, understanding prediction uncertainty has become essential for supporting reliable and risk-informed decision-making.
This 5-day comprehensive training introduces modern uncertainty quantification techniques specifically designed for petroleum engineering applications. Participants will explore the sources of uncertainty in subsurface and production data before progressing through practical methods such as Monte Carlo simulation, conformal prediction, uncertainty-aware regression, probabilistic classification, and time-series forecasting. Participants will be able to demonstrates how these approaches enhance model reliability, improve prediction confidence, and provide greater transparency in engineering analyses.
Delivered through a combination of concise theory and hands-on exercises, the trainer will be using realistic petroleum engineering case studies. Participants will develop practical machine learning workflows using widely adopted Python libraries such as scikit-learn, MAPIE, Statsmodels, Darts, and skforecast, with emphasis on model calibration, interpretability, interval quality, reproducibility, and effective communication of results. By the end of this training programme, participants will be equipped to build robust, uncertainty-aware predictive models that support better reservoir evaluation, production forecasting, operational optimisation, and data-driven decision-making across the upstream oil and gas value chain.
1. What is uncertainty quantification in petroleum engineering?
Uncertainty quantification (UQ) helps engineers handle unknown values in reservoir and production data. Rather than relying on one result, engineers can review several possible outcomes. As a result, teams can make better choices about drilling, reserves, and field plans. In addition, UQ helps teams see risks before making major decisions.
2. Why should petroleum engineers use machine learning with confidence ranges?
Many machine learning tools provide only one result. However, engineers also need to know how much they can trust that result. Confidence ranges show possible risks around a prediction. Therefore, teams can compare options more easily. They can also make better choices for field work and business goals.
3. Who should attend this course?
This course suits engineers, earth science teams, data analysts, and AI teams in the oil and gas sector. In addition, supervisors and managers can use these skills to support planning and improve field results.
4. How does Monte Carlo simulation improve hydrocarbon reserve estimates?
Monte Carlo simulation tests many possible combinations of reservoir data. For example, these inputs include area, thickness, porosity, saturation, and recovery factor. As a result, engineers can see a range of reserve results instead of one estimate. They can then use P10, P50, and P90 values to better understand risk. Therefore, teams can plan projects with greater confidence.
5. What is conformal prediction and why is it useful?
Conformal prediction creates reliable prediction ranges for machine learning models. Engineers can use it for property estimates, pressure forecasts, rock type prediction, and production planning. In turn, teams gain more trust in model results. Moreover, they can see how certain or uncertain a result may be. Therefore, it becomes easier to make sound decisions.
6. What will participants learn during this training course?
Participants learn practical ways to measure risk in petroleum engineering data. The course covers Monte Carlo simulation, conformal prediction, forecasting methods, regression models, and class-based models. In addition, participants work on real oil and gas examples. For example, they study PVT analysis, rate transient analysis (RTA), gas rate prediction, rock type prediction, and production forecasts. As a result, they gain skills that they can apply at work right away.
7. Which software and Python tools are used in the course?
Participants use popular Python tools such as scikit-learn, MAPIE, Statsmodels, Darts, and skforecast. Throughout the course, they build models and measure risk. They also improve production forecasts through hands-on exercises. In addition, they work with real field data, so they can apply what they learn to real projects.