Title : Decision-making technique for water shutoff using random forests and its application in high water cut reservoirs
Most of the oilfields are now experiencing intermediate to late stages of their oil production by waterflooding method. The water injection effects in creating channels between wells are serious and the oil recovery efficiency are decreasing. The usage of water plugging profile control is required to manage excessive water production from an oil reservoir. This research examines the well selection for profile control using the fuzzy evaluation method (FEM) and its enhancement utilizing the random forest (RF) classification model. Using a machine learning RF technique, candidate wells for profile control operations were performed. The data source consists of 18 injection wells, with 70% of the well data used for training and the remaining 30% for model testing. The new factor weights are established when the model has been fitted, and choices are made. Thus, 4 out of 18 wells are selected for profile control using a new factor weight created by RF, while 7 out of 18 wells are selected using FEM. The profile control is then carried out using a polymer gel plugging method that was studied by laboratory experiments. The selected wells are profiled using both techniques in a numerical simulation model, and the effects of gel system plugging on daily oil output, water cut, and cumulative oil production are then contrasted. According to the study, compared to 7 wells that were selected using the FEM model, the reservoir performed better when 4 wells were selected using a novel weighting system created by RF. The decision-making abilities of the profile control wells were improved by the new weighting model.
Audience Take Away:
- The well selection for water shutoff process by fuzzy mathematic method
- Factor weight generated by Random Forests and it is usage for well selection for water shutoff
- Chemical agent Polymer based starch gel introduction for water shutoff numerical simulation and experiments