Current full-field models at Queensland Gas Company (QGC) typically contain thousands of coal-seam gas (CSG) wells, spanning ~4,000 sq.km. The goal of history matching individual wells in these models at a well level can present a significant challenge. Models adequately matched to a well-level is desirable as it helps to provide more confidence in the resultant forecast and well-level decisions to maximise recovery and manage Locked in Potential (LIP). Previous methods applied in QGC were to history match to a field level and more recently to apply well level property modifiers in a semi-automated workflow to match produced volumes at a well level (Chang, G. et al, 2018).

This new method is an extension to the previously mentioned well-level history matching workflow. A key disadvantage to the previous workflow was the ability to only focus on one match metric by changing one property at a time, leading to potentially many iterations of different properties to ensure an adequate match while being consistent with regional geological trends often being time consuming and resource intensive. The extended method can attempt to minimise multiple match metrics (pressure mismatch, rate mismatch, etc.) through the modification of multiple different reservoir grid properties in a single iteration, at a well-level, resulting in fewer iterations to converge on a satisfactory history-match. This workflow is run through python with many added levers to allow reservoir engineers to target different history matching goals.

The implementation requires a few key inputs; the properties and metrics to be changed and measured against, and a response function for how each desired metric change by a change in each property. From these an objective function is developed such that the function is at a minimum when the match metrics are at their desired value. With an objective function defined with property modifiers as inputs a minimisation algorithm can be applied to iterate through modifiers to find an optimal solution to step the simulation closer to a better history match.

This content is only available via PDF.
You can access this article if you purchase or spend a download.