A breakthrough methodology has been implemented in the San Francisco field, Colombia, to yield an additional 12% in remaining reserves by optimizing production and injection based on key principles of the ‘learning theory’.
This paper presents a new massive optimization methodology applied in the San Francisco field. Such techniques are especially well suited to mature fields, as these have substantial infrastructure in place, which can be used in ways different to the usual.
Mature fields represent a challenge in terms of investment and allocation of resources. Still, there remain opportunities to improve production, since past strategic choices have created some heterogeneity in field pressure and saturation. These can be drastically revisited and production parameters rearranged. Since many production avenues have been explored in the past, a learning process can be applied: rearranged parameters can be implemented with very low risk. Still, there are billions of ways to operate mature fields: injection and production rates can be varied, conversions made.
Massive optimization of a field means selecting and playing with thousands of production scenarios and identifying the best. This is possible only with a reliable and fast simulator. The "statistical learning theory" defines under which conditions such a simulator can be devised. The Production Simulator is designed in such a way that it complies with the conditions of a reliable long-term forecast.
This new approach has been successfully implemented in the San Francisco field in order to re-organize the water-flooding regime, in particular through the conversion of producers into injectors. The entire process took only three months.
Our methodology has led to an expected additional production of 1.2 mmbo, a +12% increase in remaining reserves, with a total development cost of 4.5 US$/bbl. Current production data, one year after implementation, is in line with the forecast.