Practical Application of Pareto-Based Multi-Objective Optimization and Proxy Modeling for Steam Alternating Solvent Process Design
- Luis Coimbra (University of Alberta) | Zhiwei Ma (University of Alberta) | Juliana Y. Leung (University of Alberta)
- Document ID
- Society of Petroleum Engineers
- SPE Western Regional Meeting, 23-26 April, San Jose, California, USA
- Publication Date
- Document Type
- Conference Paper
- 2019. Society of Petroleum Engineers
- numerical simulation, machine learning, thermal methods, solvent injection, optimization
- 4 in the last 30 days
- 104 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 5.00|
|SPE Non-Member Price:||USD 28.00|
Steam alternating solvent (SAS) process is a thermal heavy oil recovery technique, where steam and solvent (e.g., propane) are injected alternatively through the same well configuration as in the steam-assisted gravity drainage (SAGD) process. The SAS process is deemed to be more energy-efficient and environment-friendly with less greenhouse gas emission and water usage. However, proper design of the SAS process is challenging as multiple conflicting objectives need to be optimized simultaneously. Conventional optimization methods that aggregate multiples objectives into a single weighted objective are not appropriate. In this work, a novel workflow is developed to identify a set of Pareto-optimal operational parameters for the SAS process.
First, a synthetic base model is constructed based on data gathered from the cold lake reservoir. Sensitivity analysis is carried out to determine the main decision variables [e.g. solvent (propane) concentration and duration of solvent injection in each cycle] and to formulate the objective function (e.g., recovery factor and cumulative propane injection). Next, a set of initial SAS models encompassing a wide range of decision variables are generated and subjected to flow simulation, and the corresponding objective functions are computed. Third, a response surface (proxy) model is calibrated to approximate the non-linear relationship between the multiple objective functions and the decision variables. Finally, a non-dominated sorting genetic algorithm II (NSGA-II) is applied as a multi-objective optimizer to obtain a set of optimal decision parameters.
The predictions from the base model are corroborated by several previous SAS simulation studies in the literature, where comparable production trends and patterns are observed. It is observed that both the solvent compositions and duration of solvent injection in each cycle would have significant impacts on the objective functions. The proposed hybrid optimization workflow can facilitate the identification of a set of Pareto-optimum solutions with considerable savings in computational costs.
|File Size||1 MB||Number of Pages||13|
Souraki, Y., Torsater, O., Jahanbani Ghahfarokhi, A., & Ashrafi, M. (2013). Application of Solvent Alternating SAGD Process to Improve SAGD Performance in Athabasca Bitumen Reservoir. In SPE Western Regional & AAPG Pacific Section Meeting 2013 Joint Technical Conference. Society of Petroleum Engineers, Monterey, USA.