Novel Approach for Porosity Quantification in Carbonates Using Borehole Images

Abstract Porosity quantification in carbonates rocks represents one of the main challenges in assessing potential carbonate reservoirs. Conventional well logs have low sampling rate and do not allowing to get an accurate porosity distribution. Currently, there are different methods for quantifying secondary and primary porosities by integrating different measurements and logs. Borehole image can solve the carbonate heterogeneity challenge when it comes to quantifying accurately the complex pore system distribution. Single source of information – borehole image (BHI) is used across the full wellbore to calculate the contribution of matrix and heterogeneities to the total porosity and subdivide it into primary and secondary porosities. BHI is transformed into porosity image using only static image with normalized pixel values. Primary porosity refers to the porosity related to the matrix, while secondary porosity is related to all the heterogeneity regions extracted through the semantic segmentation module. For each row in the static image, pixels are classified as "heterogeneity" if they belong to a heterogeneity region, otherwise they are classified as "matrix". The primary and secondary porosity in the row are calculated as the sum of pixel values in the static image that belong respectively to matrix and heterogeneity, normalized by the total number of pixels in the row. The technique has been used in this study to quantify the secondary porosity. Since there can be large porosity variation between consecutive rows, the results are presented in terms of a moving average. In this application it was used a window span of 1ft, although this can be changed depending on the length scale of heterogeneities in the borehole images. The outputs of this technique (image total porosity, primary porosity and secondary porosity curves) can be used in an empirical equation to improve permeability calculations, core calibration is required. The technique can give sufficient information to characterize the carbonate reservoirs in cases where the porosity system is dominated by one secondary porosity type, e.g. vuggy porosity or fracture porosity with low clay content and no borehole artifacts and performed with careful selection for the parameters and cut-offs. Applying machine learning reduced these limitations as depending on using database of many wells characterizing the carbonate reservoirs it could be overcome following limitations in comparison with exciting techniques: The primary/secondary porosity distribution greatly cutoff method selection.Separation is possible to vuggy porosity and fracture porosity.Possibility to separate the isolated vugs from the connected vugs.