Production rate is the single most important parameter that indicates performance of wells to subsurface/reservoir engineers so that they can efficiently monitor as well as optimize the behavior of the system (well and reservoir). Ironically, production rate in individual wells is not monitored usually in real-time, and wherever they are, flow transmitters are employed which carry their own inherent errors when multiphase flow occurs. Therefore, as an industry practice, daily conditions and well tests data is utilized to determine flow rates using developed correlations like Gilbert's. Nevertheless, the effectiveness of such correlations has been reduced due to lack of frequent production well tests and operational technical and cost issues.

For this reason, a model that can predict fluid rates, gives great advantage to the production engineers towards optimizing the well performance in real time. Accordingly, this paper focusses on adapting computational intelligence algorithms, to predict oil flow rate in artificial lift oil wells, which is robust, simple and can be applied universally. This study utilizes multiple computational intelligence (CI) techniques, which have not been utilized previously for this topic, namely; Artificial Neuro Fuzzy Inference Systems, Support Vector Machines alongwith Artificial Neural Network.

Separator production test data were acquired from a South Asian oil field operating on gas lift; followed by data cleaning and data type reduction to prepare the data for input to computational intelligence system. Consequently, only the readily available well head parameters were finalized to be used as the inputs. The output/target data was the separator measured oil rate. All the techniques are compared rigorously with each other and with currently used empirical models based on average absolute error and coefficient of determination. The analysis of the outcome of this work shows that the new model predicts oil flow rate with an accuracy of 98%, interestingly, example of such accuracy is not found in the current literature.

The results of this study show that the ANN outperforms all the current empirical correlations. This effort puts forth an industrial insight into the role of data-driven computational models for the production reconnaissance scheme, not only to validate the well tests but also as an effective tool to reduce qualms in production provisions.

You can access this article if you purchase or spend a download.