Application of Neural Networks in Developing an Empirical Oil Recovery Factor Equation for Water Drive Niger Delta Reservoirs

Abstract The importance of estimation of reservoir recovery factors to a high level of accuracy cannot be overstated in field appraisal or development. Estimation of recovery factor depends on a combination of static and dynamic parameters which in themselves come with a lot of uncertainties and in most cases there is insufficient or poor quality data to enable effective and accurate estimation of a recovery factor. Empirically derived recovery factor equations have been developed for cases were there is limited data to be able to get an estimate of recovery factor. Empirical Oil Recovery Regression Equations were initially developed by the American Petroleum Institute (API) on sandstone reservoirs in the United States and adapted by Arps, popularly called Arps equation. Since then numerous other equations have been developed but there has been lack of a robust equation developed specifically for the Niger Delta environment, although Arps equation has been tested on Niger Delta Reservoirs and work done shows it can be used in the environment although with some caution. This paper shows a new approach to developing an Empirical Oil Recovery Factor Equation for water drive reservoirs in the Niger Delta using reservoir characterization and Artificial Neural Networks (ANN). The ANN is a mathematical model inspired by biological neural networks. In this approach reservoir recovery factors from mature water drive reservoirs in the Niger Delta Region were characterized based on facies type and properties into groups to draw inferences on similar reservoir types tied to different recovery factor ranges. The data from selected group set was randomly divided into three parts, of which 60% was used for training, 20% for validation, and 20% for testing. The final ANN model for particular groups was tested for robustness against the original empirical correlation developed by API.