Towards an integrated wide approach for upstream field recovery
File(s)
Author(s)
Ramjanee, Shakeel
Type
Thesis
Abstract
Integrated asset modelling is the modelling of an entire production facility comprising of both subsurface and surface elements. Historically, asset modelling has entailed discrete reservoir, well and facility models accompanied by silo discipline ownership. Adopting an integrated wide approach provides a holistic overview capturing the complex interactions between subcomponents thereby enabling the assessment of system constraints and identification of production optimisation opportunities.
The objective of this research study is to develop a viable, representative alternative to the industry state of the art integrated asset modelling tool which can be deployed for short term surveillance and medium to long term field optimisation.
A series of integrated asset frameworks were constructed using an industry integrated asset modelling tool; this served as a basis for the development of proxy models via traditional approaches such as the response surface methodology and regression to complex techniques such as Artificial Neural Networks, RS-HDMR and ALAMO. Functional relationships defining production and pressure responses were postulated for a gas field case and an oil field supplemented by water injection drive case. Global sensitivity analyses of selected input parameters were investigated to validate the postulated functional relationships and smoothing splines were deployed to alleviate instabilities in the pressure prediction outputs and improve the surrogate response.
The optimisation of the surrogates via a constrained nonlinear optimisation framework in addition to gradient free optimisation techniques of neural networks yielded favourable results. Whilst the surrogates provided suitably accurate predictions, deviations were noted at late field life conditions suggesting suitability in early phases of the field development or production life cycle.
The developed surrogate models are proven to predict production and pressure profiles within specified production windows at a fraction of the simulation time and computational efforts of the industry counterpart and can be extended to further field development types in future workflows.
The objective of this research study is to develop a viable, representative alternative to the industry state of the art integrated asset modelling tool which can be deployed for short term surveillance and medium to long term field optimisation.
A series of integrated asset frameworks were constructed using an industry integrated asset modelling tool; this served as a basis for the development of proxy models via traditional approaches such as the response surface methodology and regression to complex techniques such as Artificial Neural Networks, RS-HDMR and ALAMO. Functional relationships defining production and pressure responses were postulated for a gas field case and an oil field supplemented by water injection drive case. Global sensitivity analyses of selected input parameters were investigated to validate the postulated functional relationships and smoothing splines were deployed to alleviate instabilities in the pressure prediction outputs and improve the surrogate response.
The optimisation of the surrogates via a constrained nonlinear optimisation framework in addition to gradient free optimisation techniques of neural networks yielded favourable results. Whilst the surrogates provided suitably accurate predictions, deviations were noted at late field life conditions suggesting suitability in early phases of the field development or production life cycle.
The developed surrogate models are proven to predict production and pressure profiles within specified production windows at a fraction of the simulation time and computational efforts of the industry counterpart and can be extended to further field development types in future workflows.
Version
Open Access
Date Issued
2023-07
Date Awarded
2024-03
Copyright Statement
Creative Commons Attribution NonCommercial Licence
License URL
Advisor
Shah, Nilay
Muggeridge, Ann
Chachuat, Benoit
Publisher Department
Chemical Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)