How do you update and refine your history matched model over time with new data and information?

One of the main objectives of HM is to reduce uncertainties in production forecasts, we need to consider this objective when new data is available.

The introduction is a good summary of the hm process. In terms of data will usually end up looking at everything again unless the updated data from all sources including production and pressure data falls inline with the existing model and assumptions. If not time to re-evaluate model assumptions.

To effectively update and refine a history-matched model, it is crucial to systematically identify and gather new data sources that can influence model performance. This includes a diverse range of inputs such as production data, well tests, and seismic surveys, each contributing unique insights. A thorough evaluation of the quality and relevance of this data is essential, as preprocessing may be required to ensure compatibility with the existing model framework. Ultimately, this meticulous approach enhances the model's predictive accuracy and reliability in dynamic environments.

Incorporating new data is going to depend on what type of data it is - static or dynamic and how the original model was created - did it apply a fully integrated method. If the original model was done in a fully integrated process then the new data will require less effort than if it didn’t. In particular was dynamic engineering data such as static pressures, transient pressure analyses, well completions/ work overs, fluid movement, changes in gas and water production included in the original static model build.

Data assimilation is a critical process for enhancing model accuracy by integrating new observations with prior estimates. Various techniques, including manual adjustments, assisted and automatic history matching, and ensemble methods, offer flexibility depending on the model's complexity and available data. Selecting the appropriate method is essential to align with specific objectives and computational resources. This strategic integration not only refines model parameters but also improves predictive capabilities in dynamic systems.

Model evaluation is a vital step in validating the effectiveness of an updated model against its predecessor and observed data. This process involves assessing the model's quality, accuracy, and inherent uncertainties through various criteria, including statistical measures, error analysis, and sensitivity analysis. Selecting appropriate evaluation metrics is crucial to ensure they align with specific goals and expectations. Ultimately, a thorough evaluation enhances confidence in the model's predictive capabilities and informs future adjustments.

Model evaluation needs to make sure you sort out reservoir performance which is affected by reservoir description versus production and pressure data affected by the condition of the wellbore over time.

Mathematically, history match is an ill posed problem. It means we don't need to get a perfect history match. The ranges of history match parameters should be as realistic as possible. The ranges of the parameters should be assigned using analogs or experts in that area. It is best to get a statistical distribution of the history match model, so you can provide a range of outcomes in your forecast.

Model improvement is essential for refining the updated model by addressing deficiencies identified during data assimilation and evaluation. This process involves modifying the model's structure, parameters, or assumptions to enhance fit and reduce errors, thereby increasing predictive confidence. Strategies for improvement may include adjusting feature sets, refining grid resolution, updating material properties, or integrating new insights. By systematically implementing these enhancements, the model can better reflect real-world conditions and improve its overall reliability.

Model predictions should take into consideration what is realistic and can be done in the field. If you can’t observe or modify well completions at the perf level in the field no use doing it during predictions.

Model application is the critical phase where the improved model is utilized for forecasting and informed decision-making. This process involves simulating various scenarios, testing strategies, and optimizing outcomes while ensuring consistency and transparency in application. Clear communication of results and associated uncertainties is essential for stakeholder understanding and trust. Additionally, ongoing monitoring of the model's performance and validity is necessary to facilitate timely updates and refinements, ensuring its continued relevance and accuracy in dynamic environments.

Model documentation is a crucial final step that involves systematically recording the updated model's details, including data sources, methodologies, results, uncertainties, assumptions, and limitations. Comprehensive and structured documentation ensures adherence to best practices in data management and quality assurance, facilitating transparency and reproducibility. Additionally, sharing and archiving the model alongside its documentation is essential for future reference and reuse, promoting collaborative efforts and continuous improvement in modeling practices. This thorough documentation process ultimately enhances the model's credibility and usability in various applications.

The reservoir study should start with an inventory of available data, analyses, and past studies and reports. This index and bibliography needs to be updated.

In my opinion from doing many studies over 45 years the base of this article is poorly organized, poorly written and pretty much useless.

This article is ok, probably could be improved, but also reflects the complexity and difficulties of history matching. It’s never the same. Also it takes time and interaction with g&g to figure out if the underlying geological description is the issue making this process challenging. Sometimes the model is too complex and others too simple.