Analyzing shale wells using traditional decline curve methods is problematic because of the nature of reservoir properties and flow behavior in typical shale wells. New empirical methods were developed to model the special production decline of shales. These methods were formulated using different mathematical and statistical bases and result in different forecasts. Hence, engineers have a variety of methods that may give different estimates for ultimate recovery when analyzing shale wells. In this work, four recently developed decline curve methods, along with the traditional Arps method, were compared.
The four recent methods compared here were empirically formulated for shale wells and tight gas wells. They are: a) the Power Law Exponential Decline, b) the Stretched Exponential Decline, c) Duong’s Method, and d) the Logistic Growth Model. Each method has different tuning parameters and equation forms. In this work, the methods were programmed and automated by using nonlinear regression to match the production "history" of a well. In addition, they were compared in terms of "goodness of fit" to the history data and reliability of automation as well as production forecast and ultimate recovery estimation. These methods were compared with simulation models in addition to field data from Barnett Shale, Bakken Shale, and the Eagle Ford Shale.
Each of these methods may have application for different cases. It may be advisable to program each of these methods for optional usage in applications. But this current paper should allow engineers to understand better the characteristics of each method and to choose the method that best models their wells under various circumstances.