Project title
Characterising the Fitness Landscapes of Seismic History Matching Problems
Project abstract
For the energy sector, assisted seismic history matching (ASHM) continues to be a challenging problem. The objective is to establish a subsurface reservoir model that is consistent with all available surveillance data. This enables more reliable predictions and forecasts to optimize fluid production and injection. The aim of this research is to understand ASHM at a more profound level. It is perceived as a multidimensional optimization problem in which the performance of the algorithm is related to the topology of the search space, or fitness landscape. We propose that understanding topological characteristics of the fitness landscape may lead to a deeper understanding of the ASHM problem, and inform the problem setup to achieve better subsurface models. In this work, various fitness landscape analysis (FLA) methods will be developed or adapted from computer science and applied to synthetic reservoir models of increasing complexity before application to real-field case study.