Study of Pattern Search Optimization and Implementation of Hooke-Jeeves Direct Search for Production Optimization using Perforations
MetadataShow full item record
This thesis concerns the topic of petroleum production optimization using perforations.A plug-in by PhD student Einar Baumann was extended to include a second pattern search method: Hooke-Jeeves direct search. Originally only the compass search method was implemented. The thesis covers the topic of these two pattern search methods and how they are applied to optimize well perforation placement for a given objective function value. We present descriptions of how the existing plug-in was modified to make room for a second algorithm, and how the Hooke-Jeeves direct search was implemented. This thesis also presents the well perforation placement optimization problem. The plug-in only optimize for perforation placement, but can in the future be extended to optimize with other algorithms, or for other well completion types, e.g. ICDs. Two experiments are completed to analyze the behaviour of the two algorithms. The first experiment concerns the Hooke-Jeeves direct search, and how it reacts to changes in the input parameters. In the second experiment we compare the two pattern search algorithms when conducted with the same input parameters. The experiments show that two parameters, initial step length and segment resolution, are dependent on the simulation case.We also find that Hooke-Jeeves direct search is a more efficient pattern search method than compass search when it comes to total number of simulations.