Porosity estimation at depths below the borehole bottom from resistivity logs and electromagnetic resistivity

摘要

A neural network approach to porosity prediction below the well bottom from either the electrical resistivity logs or electromagnetic resistivity profiles is developed using data from Soultz-sous-Forets area (France). It is shown that the neural network approach enables the porosity to be predicted at depths below the bottom of the borehole from resistivity well-logging data and by resistivity profiles determined by inversion of electromagnetic sounding data collected in the vicinity of the well. The results indicate that the forecasts based on the electrical logging data are more accurate (average relative errors being equal to 3%-5%) when the target depths do not exceed the double length of the well log used for calibration. In the absence of the resistivity logs at target depths or when the target to well depth ratio is large enough, the forecasts based on the electromagnetic resistivity data are more preferable.