SAS has developed a data-driven framework and method to optimize heavy oil production from SAGD operations. A data-driven method, it is believed can provide significant value to operators by speeding up the process of analyzing well data to make better operational decisions. We show how our approach and analysis on actual well data, vetted by production and reservoir engineers, is used to determine optimal values for control variables such as pump speed, steam in the short and long tubing, header pressure, and casing gas that could significantly increase oil production and reduce water production leading to a doubling of NPV and reduction in the payback on the investment in SAGD operations.
Our approach to analyzing SAGD data is believed to be one of the few, if not the only study, using this type of approach. Specifically, preliminary results for well-pair one show that using optimally chosen values for control variables results in a decrease in water production by 33% (15.4 m3/h to 10.2 m3/h) while maintaining a similar oil rate when compared to actual average value from this well-pair, and a 6% reduction in water cut. Results for well-pair two were more impressive. Specifically, optimal control values resulted in an increase in oil rate from an actual average value of 5.5 m3/h to 7.9 m3/h which represents a 43% increase, water production was decreased from 13.4 m3/h to 7.4 m3/h (45% reduction), and water cut was reduced from 70.7% to 48.4% (32% reduction). Our results may also indicate that producers are likely to be “over-steaming”.
While this study used 5 control variables – the modeling approach can accommodate any number and type of variables. Our solution is also able to capture the dynamical changes in the operating conditions by continuously learning and adapting to these changing conditions using a non-parametric predictive modeling approach together with an optimization approach to maximize oil or minimize water cut subject to the constraints imposed by the control variables. We also show that performing what-if analysis production engineers are able to determine if the optimal values for the control variables are global or local optima’s. This also enables engineers to set or change the limits or thresholds for all control variables, constraints ranges, and optimization variables in the what-if analysis.
From the use of optimal values for the control variables, a notional financial calculation was conducted. For well-pair one, NPV increased from $98M to $106M, and the payback on the investment is reduced from 16 years to 9 years. For well-pair two, NPV increased from $149M to $319M, and the payback on the investment on the well is cut by more than half from 11 to 4 years.
Our findings are further evidence that a data-driven approach can optimize heavy oil production operations in a way that would have a positive impact on the bottom-line for producers as well as have a positive impact on the environment from reduced water use.