OLYMPUS field development optimization study-KFUPM

The well control and field development optimization on the Olympus reservoir model was carried out by initially selecting the five most representative ensemble out of the fifty different ensembles. The five representative ensembles were selected based on a clustering algorithm. The algorithm constructs a cocorrelation matrix, which is symmetric, by calculating the Euclidean distance between each ensemble. The L2 norm of the column (row) of co-correlation matrix was calculated and the five ensembles were selected based on the lowest values. Due to the huge computational cost associated with solving an optimization problem for a large sample space, this problem was solved using a trigonometric (Cosine) based model reduction approach to reduce the number of control variables. Due to the large simulation run time of the Olympus model, a proxy model based approach was adopted. The proxy model was developed employing ANN in which more than 400 data points were used for training and approximately 200 data points were selected for testing. Differential Evolution(DE) optimizer was used as the main optimization algorithm in this study.