Use cumulative production as observations

[wouterjdb]

Currently we see that the cumulative production is always off. Could it help steering the model in the right direction by condition the model to cumulative "observations" as well?

[olwijn]

Cumulative rates (Eclipse keywords WOPT, WWPT and WGPT) were added as possible observation types. A first test using only WOPT data at the final time with a 10% relative error standard deviation suggests that this might help to move the ensemble in the right direction, perhaps as a first step that could be followed by further fine tuning. The following figure is an example showing WOPR for 1 production well after 4 iterations with equal weights and an initial ensemble size of 240 (some members failed during the process):

Next test: add WWP and WGPT at the final time, and use smaller errors.