Efficient Design of Warm-VAPEX using a Neural Network Coupled with a Multi-objective Optimization Algorithm for an Oilsands Reservoir in Alberta, Canada
Abstract
This study develops a hybrid workflow that couples a neural network with a multi-objective optimization algorithm for finding a set of Pareto-optimal operating conditions of the Warm vapor extraction process (Warm-VAPEX) at an oil sands reservoir. The proposed workflow is tested for a reservoir model representing the characteristics of Athabasca oildsands, Alberta, Western Canada. A sensitivity analysis is conducted to analyze the effects of heterogeneities on bitumen production. We formulate a set of decision variables (e.g., well spacing, bottom-hole pressure, solvent type) and a set of objective functions (i.e., RF, solvent efficiency, and greenhouse gas emissions). The multi-objective optimization algorithm is applied to evaluate the performance of each decision-variable set and search for the Pareto-optimal front in the objective space. For a cost-effective evaluation, a neural network-based proxy is designed to compute the objective functions at an affordable computational cost. The final Pareto-optimal front is derived by repeating the hybrid workflow in both decision and objective spaces. This workflow contributes to the efficient design of a solvent-based heavy oil recovery scheme, revealing the balance between accuracy and cost.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.H34A..01K