توصيفگر ها :
نمودار رزونانس مغناطيسيهستهاي , نمودار MDT , كندي فشارشي , زون بندي هيدروليكي , فشارحفرهاي , شبكههاي عصبي
چكيده انگليسي :
Abstract
The nuclear magnetic resonance (NMR) log is a highly effective tool for extracting reservoir petrophysical parameters, including porosity, permeability, fluid saturation, pore size distribution, and fluid types. Given the close relationship between abnormal pore pressure variations, increased clay content, and reduced permeability, NMR data can assist in identifying and estimating pore pressure.
This study focuses on the Burgan sandstone reservoir in the Binaloud field. Initially, total porosity values were calculated by inverting transverse relaxation time series from NMR logs at depths ranging from 2179 to 2275 meters in Well No. 6 of Binaloud. The total porosity values were compared with density porosity in non-collapse depths, leading to a refinement of the number of bins and regularization coefficient in the NMR inversion process. This optimization ensured maximum alignment between NMR total porosity and density porosity at the studied depths. Subsequently, effective porosity was obtained by assuming a cutoff time of 33 milliseconds to differentiate free fluid from capillary-bound water. Using the effective porosity, permeability was calculated based on Timur-Coates and Schlumberger-Doll relationships.
Next, reservoir quality index (RQI) and flow zone index (FZI) were determined, and a cumulative normalized reservoir quality index plot was generated, leading to hydraulic zonation of the well and identification of seven hydraulic zones. To detect depths with abnormal pore pressure, compressional slowness data were analyzed. The analysis revealed a decreasing trend in compressional slowness up to a depth of 1600 meters, followed by an increasing trend, suggesting the presence of overpressure at greater depths. Consequently, Eaton's equation was used to estimate pore pressure across the entire well.
However, since NMR data had a higher resolution than compressional slowness data, a multilayer perceptron (MLP) neural network was trained using NMR data to estimate compressional slowness at a higher depth resolution. The neural network, trained with diverse NMR log data alongside caliper log and dipole sonic shear slowness data as the target, successfully estimated compressional slowness at 43,437 points along the well. The validation process showed a correlation coefficient (R²) of 0.84 between the neural network's predicted compressional slowness and the actual values from dipole sonic shear slowness logs.
Using Eaton's equation and MDT log-derived pore pressure values at 16 points, the exponent in Eaton’s equation, representing the ratio of normal slowness to existing slowness, was estimated as 0.1 through the least squares method. Applying this exponent and the available compressional slowness data, pore pressure was determined across the entire well. The obtained pore pressure profile showed a significant correlation with the cumulative normalized reservoir quality index. The estimated pore pressure curve indicated normal pressure conditions up to approximately 1400 meters, followed by a weak underpressure phenomenon with a maximum deviation of 100 psi up to 1750 meters. Beyond this depth, an overpressure regime emerged, reaching a maximum deviation of 400 psi from normal pressure up to the well’s total depth of 2300 meters. This overpressure phenomenon persisted non-uniformly toward the bottom of the well.
A comparison between shale volume, thorium-to-potassium ratio, and pore pressure variations demonstrated a significant correlation between these parameters and pore pressure. Furthermore, the presence of overpressure and underpressure zones, along with sharp thorium-to-potassium ratio variations, suggested fault-induced pore pressure changes. The existence of these faults was confirmed through caliper log analysis.
**Keywords:** Nuclear Magnetic Resonance (NMR) Log, MDT Log, Compressional Slowness, Hydraulic Zonation, Pore Pressure, Neural Networks
