This work presents the practical application of the recently developed B-spline based deconvolution methodology to analyze variable-rate/variable pressure drop well performance data from gas wells. As deconvolution provides the corresponding constant rate pressure drawdown response for a well/reservoir system which is affected by variable flowrates, we intend to use deconvolution of production data to identify the reservoir model and perform an analysis for estimating reservoir properties and reservoir volume.
In this work we apply our B-spline based deconvolution methodology to production pressure and flowrate history data (which are typically available on a daily or monthly basis --- all field cases in this work (except Case 3) consider daily pro-duction data measurements). For this work, we apply our method using traditional gas well test data, as well as regularly measured gas well production data. We also demonstrate the appropriate handling of input data (particularly pressure test data and production data) to ensure stable/accurate deconvolution results.
The application cases in this work should be considered typical for a reservoir or production engineer, and we would expect similar performance/robustness of our methodology as it becomes a common analysis practice.
The following objectives are proposed for this work:
To apply and extend the recent B-spline based deconvolution methodology to analyze variable-rate/variable pressure drop gas well performance data.
To identify the critical issues which affect the success of deconvolution methodology when applied to production data.
To state specific recommendations for practice and/or future work
As orientation, we note that the conventional analysis of well test data involves the analysis of "high-frequency" pressure buildup data --- specifically, the derivative of the pressure drop function with respect to the logarithm of time --- using "superposition" or specialized time transforms. We can also perform a similar (albeit much simplified) approach for production data (i.e., boundary-dominated flow data). However, the purpose of using deconvolution is to "extract" the equivalent constant rate pressure drop (or pressure) function and to avoid the use of such "specialized" functions as described above. In short, the primary signature used to classify/establish the reservoir model is the constant-rate drawdown pressure behavior of a well/reservoir system --- and the goal of any deconvolution algorithm is to extract that signature with as little corruption as possible in the "extraction" process.
All of which leads us to our present work for the analysis/interpretation of gas well production data using deconvolution. In many ways the gas well performance problem is prototypical --- we generally have production rates and pressures available on a per well basis (due to regulations, data collection practices, or both). These data are often measured daily, but, unfortunately we are often faced with surface pressure measurements of unknown quality (gas flowrates are typically accurate, although some "manifold averaging" often occurs as well). Put simply, although the data are not ideal, the gas well performance scenario is "data rich" compared to most other cases (oil, volatile oil, gas condensate, commingled production, etc.).
Most deconvolution methods have been developed and applied to deconvolve "ideal" data (e.g., idealized pressure drawdown/buildup test sequences, monotonic or functional production rate decline sequences, etc.). Very few of these deconvolution methods perform well in practice due to the ill-conditioned nature of the deconvolution problem, which means that small changes in the input data (rate and pressure signals) cause large variations in the deconvolved, equivalent constant-rate pressures.