Screening is the step that makes the rest tractable. A reservoir model may have dozens of uncertain parameters, and sampling them all thoroughly requires more simulation runs than any schedule allows. Screening designs identify which parameters actually move the responses that matter, typically a small subset, and allow the detailed work to concentrate on those. Skipping screening and applying an optimisation algorithm to the full parameter set wastes most of the computational effort on parameters that do not matter.

Proxies are useful and hazardous in equal measure. A response surface fitted to a set of simulation runs can evaluate thousands of parameter combinations instantly, which makes optimisation and uncertainty quantification practical. But a proxy is only valid where it was trained, it degrades badly in regions of the parameter space the training runs did not cover, and a proxy fitted to too few runs with too many parameters will fit noise. Validating a proxy against runs held out from its training is not optional.

Ensemble methods have become the standard approach for large problems and carry their own failure mode. Ensemble collapse, where repeated updating drives the ensemble members together until they no longer represent uncertainty, produces an ensemble that matches the history and understates the forecast range severely. Localisation, inflation and appropriate ensemble size are the countermeasures, and applying an ensemble method without them produces confident and wrong forecasts.

Finally, the automation does not remove the engineering judgement. The parameters chosen, the ranges assigned, the observations weighted and the prior distributions specified all encode assumptions, and the algorithm optimises within them without questioning them. An assisted workflow applied to a badly parameterised problem produces a well-optimised bad answer faster than manual matching would. The methods amplify judgement rather than replacing it.

By the end of this training, participants will be able to:

• Design screening experiments to identify influential parameters from a large uncertain parameter set
• Select and apply experimental designs appropriate to the number of parameters and available runs
• Construct response surface and proxy models and validate them against held-out simulation runs
• Select optimisation algorithms appropriate to the misfit surface and evaluate their convergence behaviour
• Formulate the history matching problem in Bayesian terms with priors and likelihood
• Apply ensemble Kalman filter and ensemble smoother methods to reservoir matching
• Diagnose and mitigate ensemble collapse using localisation, inflation and ensemble sizing
• Generate posterior parameter distributions and ensemble-based production forecasts
• Assess whether an assisted workflow has produced a physically defensible result
• Integrate assisted matching and uncertainty workflows into reservoir management practice

The training builds an assisted workflow end to end on a reservoir case, with participants designing screening experiments, running them, building and validating proxies, applying optimisation and ensemble methods and generating forecast ranges. The effect of poor choices is demonstrated directly: proxies validated and unvalidated, ensembles with and without localisation, optimisations with and without screening. Results from assisted and manual matching of the same field are compared. Published field applications are examined for the workflow design and the outcome achieved.

Organisations sending participants to this training will:

• Produce forecast ranges from history matching rather than single deterministic forecasts
• Reduce the engineering time consumed by manual history matching
• Improve the credibility of uncertainty ranges presented to decision makers
• Use computational resources more efficiently through screening and proxy methods
• Reduce the risk of confident forecasts from collapsed or poorly designed ensembles
• Build internal capability in workflows increasingly expected in reservoir studies

Participants will:

• Design and execute assisted history matching workflows
• Screen parameters efficiently before committing computational effort
• Build and validate proxy models properly
• Apply ensemble methods and recognise their failure modes
• Produce and defend ensemble-based forecast ranges
• Build a quantitative capability in growing demand across reservoir engineering

• Reservoir simulation engineers
• Reservoir engineers responsible for forecasting and uncertainty quantification
• Technical staff working on uncertainty and decision support
• Data and computational specialists supporting reservoir workflows
• Reserves and evaluation engineers using probabilistic simulation output
• Technical staff reviewing assisted matching studies
• Engineers with manual history matching experience moving to assisted methods

Upon successful completion of this training course, delegates will be awarded an official Certificate of Completion issued by the Petroleum Engineers Association (PEA), an ISO 9001:2015 certified training organization. The certificate carries 10 Credits and formally records the total learning hours completed.

Each certificate is signed by the Course Facilitator and the CEO of the Petroleum Engineers Association, and serves as verifiable proof of professional training that delegates can present to employers and professional bodies worldwide.