Arps' decline model was developed for the boundary dominated flow (BDF) regime that is reached quickly in conventional reservoirs. Without modification, the Arps model does not work well for unconventional reservoirs with long-duration transient flow and variable bottom-hole pressure. This paper presents results of a study of different decline models that are used to forecast production in LRS reservoirs with long-duration transient flow.
The study used synthetic future production data generated using reservoir parameters typical of those observed in practice. In traditional DCA, a single curve is fitted to the entire production data without using the historical pressure data. However, the first step in reliable DCA is to identify the different flow regimes using appropriate diagnostic plots and then fitting different flow regimes with different models. This procedure allows us to forecast future production using appropriate decline models.
The results from the study show that log-log pressure-normalized vs. time diagnostic plots correctly identify flow regimes in low permeability LRS plays. Pressure normalization removes the confounding effects of multiphase flow and variable pressure and allows us to predict the time at which we need to switch from transient to boundary-dominated-flow (BDF) DCA models. Pressure normalization can be used on rate restricted wells which would otherwise be not possible using only rate data. Hybrid models, in which early transient linear flow is fitted by models like Duong, or SEPD, followed by the Arps' decline model during BDF, were significantly more accurate than single-flow-regime models. The results from the study also showed that there is a long transition period between early transient linear flow to BDF due to the onset of multiphase flow. Each flow regime, including the transition flow regime, was modeled by calculating appropriate model parameters from available rate-pressure-time data. Hybrid models forecasted the gas EUR accurately. Condensate production was also estimated by using a method proposed by Yu which led to accurate forecasts for lean gas condensate samples.
LRS gas condensate reservoirs have fundamentally different flow behavior than traditional black oil/dry gas reservoirs, and we cannot forecast rates using traditional decline curve analysis without modification. We can use diagnostic plots to select appropriate decline curve methodology based on the type of reservoir fluid.