An automated flowing bottom-hole pressure prediction for a vertical well having multiphase flow using computational intelligence techniques

The need for accurate estimation of flowing bottom-hole pressure (FBHP) is owning an immense importance in petroleum engineering applications such as continuous field production optimization, cost per barrel of oil reduction, assess reservoir performance and to quantify workover remedial operations. Induction of pressure down-hole gauges in oil wells to measure FBHP is a common practice especially in wells artificially lifted with downhole electrical submersible pumps. However, intervening a producing well is a quite exhaustive and expensive task which is involved with production risk, and disruptions. For these reasons, various empirical correlations and mechanistic models were developed to estimate FBHP. Majority of these models were formed under laboratory scale conditions and are, therefore, imprecise when scaled-up to field conditions. Because of the complexity associated with the numerical modeling and physical implementation these models are also computationally very expensive to run. In this study, an empirical model based on computational intelligence (CI) technique is developed to quantify FBHP in a vertical well with multiphase flow. The proposed model is based on only surface production data which includes; oil flow rate, gas flow rate, water flow rate, oil API gravity, perforation depth, surface temperature, bottom-hole temperature, and tubing diameter. The data used to develop empirical model covered a wide range of values and are collected from published sources and several wells from different locations. The proposed model is then tested against new field data and results were compared statistically with the estimations of some commonly used empirical correlations and mechanistic models in oil industry. The comparison results show that the proposed empirical model significantly outperforms all other existing models and delivers the predictions with high accuracy. A small average absolute percentage error of less than 2% was found with new proposed empirical correlation, while comparing the existing published correlations on the same data gives more than 15% error. The novelty of proposed empirical model is that it is very simple and only require surface production data while previous models requires exhaustive computationally expensive calculations. The new model is accurate enough and can serve as a handy tool for the production engineers to forecast the FBHP in wells with high level of certainty.