The Development of an Optimal Artificial Neural Network Model for Estimating Initial, Irreducible Water Saturation – Australian Reservoirs

Abstract

Initial, irreducuble water saturation, Swir is an important parameter that needs to be determined accurately when attempting to characterize hydrocarbon reservoirs. Swir is also one of the key parameters in relative permeability relationships. Furthermore, an unrepresentative value of Swir may lead to invalid residual oil saturation estimates when the latter is correlated with the former. Swi may have a dependence on several other parameters, including: absolute rock permeability, porosity, pore size distribution and capillary pressure. The above parameters are directly influenced by geological deposition and subsequent changes, such as diagenesis effects (for example clay-filled pores). It is a common practice to measure Swir utilizing representative core plugs by measuring capillary pressure with a centrifuge, at speeds equivalent to the maximum representative (reservoir) capillary pressure. However, a semi-empirical model that could estimate Swir to a good degree of accuracy would be of significant value. Over the last few years, artificial neural networks have found their application in petroleum engineering. In some cases such models have outperformed models employing conventional statistical and regression analysis. In this study, an Artificial Neural Network (ANN) model has been developed for the prediction of Swi (specifically irreducible saturation, Swir) using data from a number of onshore and offshore Australian hydrocarbon basins. The paper outlines a methodology for developing ANN models and the results obtained indicate that the ANN model developed is successful in predicating values of Swir over the range of data used for calibration. This neural network based model is believed to be unique for Australian reservoirs Introduction The accurate estimation of initial water saturation for a hydrocarbon reservoir is an essential practice in the oil and gas industry. The exact determination of Swi leads to a precise evaluation of initial hydrocarbon in place, which in turn provides valuable insight into further field development plans. Moreover, and as an important aspect in multiphase flow problems, the calculation of valid Swir values is an important requirement for obtaining accurate relative permeability relationships and for determining residual oil saturation. Swir is measured as part of primary drainage capillary pressure experiments on core plugs obtained from the reservoir under consideration. In a centrifuge test, a plug is saturated with formation or synthetic brine and positioned in the centrifuge cup filled with oil. The centrifuge is then spun at different speeds, pushing the water phase outward. The volume of the expelled water is measured and the centrifuge speed is increased until it reaches a pre-determined limit, which represents the maximum capillary pressure. Having that in mind, a careful calculation of the maximum reservoir capillary pressure and consequently the maximum capillary pressure under laboratory conditions are of importance (1). As an alternative to the centrifuge test described above, a core-flood test may be conducted. In such tests, the plug is initially saturated with brine. The oil phase is then introduced, resulting in a reduction in water saturation. The test continues until the plug reaches a stabilized condition where no more water is expelled. Care must be taken when interpreting the results from both tests described above. This is attributed to the fact that both tests employ different mechanisms and consider different influencing forces, including gravitational forces for centrifuge tests and viscous forces in core flooding tests(1).

Permeability Prediction of Tight Sandstone Reservoirs Using Improved BP Neural Network

“…However, it has defects such as easy falling into local minimum value and slow convergence rate. In order to adjust these two defects occurring in iterative process, Additional Momentum Method and Self-adaptive Learning Rate Method are adopted [14][15][16][17][18].…”

A New Approach for Estimating Free Point in Fishing of Stuck Pipe using Artificial Neural Network

“…Since the drilling parameters that are involved in stuck pipe are numerous, it is essential to find the variables that are closely related to stuck pipe. In ANN studies, the following criteria have been mentioned [12].…”

Artificial neural networks workflow and its application in the petroleum industry

“…On the other hand, selecting few parameters might not be enough to model the problem under investigation. There are many approaches available to select the input parameters [38,39]. These include the following:…”

“…Neural networks have the disadvantage of being less transparent compared with other conventional models [39,46]. To make the ANNs more transparent, it is important to understand the relevance and relative importance of model inputs.…”