Probabilistic Production Forecasting – RTA Meets Monte Carlo Simulation

Generating reliable production forecasts for multi-fractured horizontal wells drilled in low permeability reservoirs is a challenge currently facing our industry.

Traditional decline methods that have been established and proven for conventional reservoirs are largely unproven for unconventional rock. Rate transient analysis and more specifically, analytical modeling, add an element of physics that can increase confidence in forecasting for these more unpredictable reservoirs. However, analytical models are a non-unique solution, and a single forecast may not fully reflect the uncertainty of our model inputs. as illustrated below.

Pumping massive amounts of fracture fluid, into a formation creates an extremely complex reservoir that will likely never be fully understood. Rather than attempting to nail down each input parameter, admitting that there is uncertainty, and then quantifying it may be more pragmatic. In the history match above the fracture half-length (x_f) is estimated as 250ft, however a range of fracture half-length is possible for a given completion design. Similarly, varying degrees of uncertainty exist for most input parameters. In some cases this range of uncertainty can be quite large, and not accounting for it could lead to inaccurate forecasts.

A probabilistic analysis is available in IHS Harmony that accounts for both the uncertainty associated with modeling input parameters, as well as the inherent non-uniqueness of analytical modeling. The engine of the probabilistic analysis is a Monte Carlo simulator coupled with an analytical model. Probability distributions are specified for uncertain input parameters. Some parameters must be used in regression analysis in order to preserve the history match. The Monte Carlo simulation samples these distributions, generating hundreds of realizations. Each realization is matched to the production history, and a forecast is generated and stored. The output of the simulation is a series of probabilistic forecasts, which are calculated from the stored forecasts.

The procedure described above is akin to an engineer history-matching the same data using hundreds of different reservoir descriptions, and generating hundreds of forecasts. The Monte Carlo simulation automates this process so nobody has to lose their mind! All of the history match parameters are within the ranges the engineer has specified, therefore if proper engineering judgment is applied, each reservoir description will be viable. Instead of trying to speculate if a given input like xf is 250ft, or 258.6ft, or 400ft, the simulation has generated a solution that acknowledges all of these fracture half-lengths are possible.

This probabilistic analysis is a technically sound means of generating true, defensible P90/P50/P10 forecasts for unconventional wells. This analysis can be applied to conventional reservoirs, the only difference being the input parameters of interest. Quantifying and accounting for the uncertainty in production forecasts is crucial for both reporting reserves, and making informed decisions on future field development.

Article created by Adam Chin, Principal Analyst/Researcher, Fekete, Energy Technical.

This article was published by S&P Global Commodity Insights and not by S&P Global Ratings, which is a separately managed division of S&P Global.