Abstract
We applied Experimental Design methodologies to evaluate development options in a Siberian field with major uncertainties in the geological, reservoir and financial parameters. Our approach, used for the first time in Russia, demonstrates an inverted 7-spot injection pattern is expected to yield 11% more oil than a line drive – identified by conventional evaluation methods. Of more significance, for decision making, the analysis predicts that the expected net present value (NPV) of this pattern is 46% more than the pattern proposed previously.
The primary reservoirs consist of stacked fluvial, delta distributary channel and mouth bar sands. Although exploration tests confirmed oil rates exceeding 120 tonne/day, the location and orientation of the individual sand bodies could not be mapped from existing seismic. Recognising this was a risk to project success, we developed an Experimental Design technique for quantifying the uncertainties in oil rate predictions. The major stages were:
Identification of the 20 geological and reservoir parameters that influence the simulation production forecasts. Examples include the dimensions of the sand bodies, residual oil saturation, facies distributions, amongst others.
Determination which parameters have a major impact on oil rate predictions. In principle, this requires simulating all combinations of possible values for the 20 parameters. In practice, this is not realistic; hundreds of thousands of dynamic models would be needed. With the Plackett-Burman Experimental Design, we only needed 48 reservoir descriptions – each consistent with the available field data – to identify the 11 significant subsurface parameters.
Creation of a second order polynomial equation that expressed the net present value (NPV) as a function of the 15 most significant reservoir and financial parameters. This equation, intended to reproduce the response of the dynamic simulator and the financial model, gave a probabilistic distribution of NPVs for each development pattern. For the 6 patterns under consideration, we created the equation with a Composite Face Centred Experimental Design technique. This required a total of 378 simulation cases. Results from these cases were obtained in 3 days with the use of parallel processing technology that allowed submitting multiple models simultaneously.
Our experience shows that Experimental Design techniques integrated with advanced computer technology can cut evaluation times from months to weeks. More importantly, they give unbiased probabilities of production profiles and NPV so that downside risks can be quantified.
