The application of the PNN algorithm to evaluate the petrophysical properties of the heterogeneous E-sand reservoirs, Lower Goru Formation, Pakistan

Abstract

The Lower Goru Formation (LGF) has produced from E-sand in Kadanwari gas field, composed of sand and shale. The estimation of reservoir properties, including porosity, mineralogy, and thickness, which could help to identify new possible drilling sites, is hindered by the heterogeneous nature of LGF, and intermix shale, and sand. Integration of elastic and petrophysical properties worked as an effective tool for evaluating reservoir properties, especially for handling the heterogeneities of sand reservoirs associated with pores and their fluids. The probabilistic neural networking (PNN) technique has been proven the most effective and is a widely applied algorithm for predicting petrophysical parameters away from control points. It can create nonlinear correlations between input and estimated properties by efficiently handling the intermix shale. For this study, PNN prediction demonstrated a reliable statistical approximation. The trends of the actual and predicted properties such as effective porosities (PHIE ​= ​∅e) and water saturation (Sw) for E-sand level exhibited an excellent correlation of 0.98 and 0.99 with an error rate of 0.006 and 0.03 respectively, which is also evident through the regression lines. The promising zones of the reservoir are further assessed by the multivariate seismic attributes, i.e., amplitudes, sweetness, and RGBA blending of appropriate frequency bands, which illuminated the potential channelized sands. The methodology applied in this study eventually minimized the uncertainty associated with petrophysical properties such as ∅e and Sw of the producing sands in LGF. The findings suggest that when dealing with heterogeneous reservoirs, a similar approach could be employed to locate feasible prospects.