Adding new wells and new production in existing fields under EOR is particularly important in matured fields that are characterized by a long history of field activity. Different drilling programs, variety of field treatments, well conversions, and new injectors add many layers of complexity and uncertainty on top of the existing effects of geological, completion, and production factors.
Surveillance and prediction of responses caused by injected fluids, in fields with dozens of patterns and hundreds of wells, calls for computer based systems that estimate responses based on numerical and statistical solutions. This is especially important when geological understanding is very weak (no core, no log data).
This paper shows how results from EOR surveillance programs can be integrated with the geological data. Furthermore, this paper shows how to build predictive models for the production estimates based on the injection responses and geology. These models support two to three times more accurate selection of wells with high oil production during EOR than historically implemented selections. Included in the paper are practical tips on how to select the best model and derive solutions with decision trees that are equivalent to sets of English based rules. Solutions from decision trees are compared with solutions from logistic regression and neural networks. This comparison deals with the statistical accuracy of model predictions, interpretation ability, and assist in applying these models to support field decisions.
Finally, the new results presented here have come from the two versions of the improved decision tree model. In the first version a better version of the model was built. This model was characterized by a lift curve above four (400% improvement in the first decile). Such a result represents a significant improvement in comparison to the model from the 2001 report and presented during CIPC 2002 conference. In the second version new core and completion data set were added when building the model.
The project's main goal was to develop a predictive model for production rates during waterfloods. In predictive modeling, regression is traditionally applied to predict continuos target variables1. Models that predict binary response variables use logistic regression2. A binary target variable is characterized by two events. They can be of numerical nature 0 and 1 where zero represents non-event and one represents an event. Alternatively, a character string with two outcomes (e.g. No and Yes) is often applied.
In the case of a continuos target variable, we may predict the fluid rates of oil, water, gas, or the total fluids. The binary 0/1- target variable can represent a low or high production output respectively. A production cut-off can be based on economical or engineering criteria applied to the actual rates or volumes in a specific time period.
This paper presents the history of a model development for predicting waterfood responses in the Pekisko B. field3. These predictions were done in two forms. The first form predicted the actual normalized volume of production.