Application of Numerical Simulation and Artificial Neural Network for Oil-Field Development, SI-A Field

Fariba Salehi¹, Ronak Azizi¹, Arnoosh Salehi², Amir Taheri³, and Vali A. Sajjadian⁴, (1) Karaj University, (2) Pars Oil & Gas Company, (3) Research & Development of NIOC, (4) Arvandan Oil & Gas Company  [email protected], [email protected]

SI-A field is a large offshore producing field located 100 km off the Iranian shore, close to the IranÐEmirate border. SI-A reservoir is situated in the Ilam formation, which is divided into several layers. The Ilam formation has been deposited in shallow marine conditions. This is a north-south elongated anticline of 46,000 feet x 23,000 feet with maximum vertical closure of 490 feet. The original oil in place is estimated to be around 2 MMMSTB. The sharp decline encountered in the field raised some concerns and prompted some reservoir studies on the field to possibly diagnose the problem and provide some remedies to stop further decline of the field. The pilot development performed is the basis for potential future field development, and more wells need to be drilled to ensure a good recovery in this low-permeability field. This study applies a methodology for optimizing well placement by numerical simulation and artificial neural network. Optimum location of an oil and gas well depends on many factors. Numerical simulation is the conventional and convenient way to evaluate these factors. Optimization techniques require an abundant number of function evaluations to find the optimum; thus, generally it is not possible to carry out a sufficient number of simulations. In field development studies, a large number of scenarios, which result in a time-consuming and expensive process must be considered. The objective of this paper is to structure the field-development schemes using an artificial neural network in conjunction with numerical reservoir simulation for the SI-A field. In this method, a few field development scenarios are studied using a numerical simulator. The results of these studies are used to train the ANN. The trained ANN is then used as a predictive tool for field-development purposes. Using NS-ANN, the number of numerical simulations is significantly reduced. The NS-ANN approach provides the flexibility of considering any location as a potential site in contrast to the conventional simulation approach when the well locations are restricted to the pre-defined block centers. The NS-ANN approach is faster and more efficient than its conventional counterpart. The results obtained from NS-ANN compare well with the results obtained from a reservoir simulator.