The Error in Estimating Reserves Using Decline Curves

Abstract

SPE Members Abstract A common method of estimating remaining reserves is decline curve analysis. There is a tendency to believe that estimates of remaining reserves will improve (become closer to truth) as more production data becomes available. This is not necessarily true even though our estimate of the ultimate recovery will generally improve with increased production history. The use of curve fitting techniques will not necessarily increase the accuracy of estimating remaining reserves and may, in some cases, lead to unreasonable answers. Reserve estimates will also vary among reserve estimators using the same data. This variance can be significant as demonstrated by the results of a controlled experiment presented in this paper. Introduction Reserve estimates are used for both internal decision making and public reporting. In 1979 with the introduction of Reserve Recognition Accounting greater public attention was given to reserve estimates. Unfortunately, the general public does not understand the uncertainty involved in estimating reserves and generally believes that figures reported as "Proved Reserves" in annual reports are certain numbers. Now, under a scenario of decreasing prices, the reserve estimators numbers are again under attack by the public. Decline curve analysis is a common method of estimating remaining reserves. Decline curve analysis consists of plotting production rate versus time and extrapolating the established trend into the future using any one of a number of techniques. Usually the logarithm of production rate is plotted versus time. The decline performance of a well is usually described as exponential, hyperbolic, or harmonic. Exponential and harmonic decline are special cases of hyperbolic decline where the decline exponent "b" is equal to 0 and 1 respectively. Exponential decline plots as a straight line on semi-log paper (log rate versus time) and is often used to forecast future production and reserves because it is easy to apply. The sensitivity of reserve estimates to the value of the decline exponent "b" is discussed by Thompson and Wright. How accurate is decline curve analysis? Can better estimates of remaining reserves be made as more production history is recorded? This is analogous to asking if our estimates will get closer to "truth" as more production data is available. Should we expect close agreement among different evaluators having access to the same data? Finally, can our ability to estimate reserves using decline curve analysis be improved by using curve fitting techniques? Possible answers to the above questions and the results presented in this paper are based on a controlled experiment in which a large group of engineering students individually analyzed the decline data from four actual wells for each year of the fields 12 year life. At the time of the study the field was essentially depleted thus permitting a comparison of estimated values to actual results. Based on the results of this study, it can be stated that:The variance in reserve estimates obtained from several engineers interpreting the same data can be large. P. 93^