| CPC G01V 1/306 (2013.01) [E21B 49/005 (2013.01); E21B 49/0875 (2020.05); G06N 20/00 (2019.01)] | 19 Claims |
|
1. A method for characterizing a geological formation comprising:
a) training at least one artificial neural network (ANN) of a machine learning system using training data collected from well log data of a plurality of other formations, wherein, for each of the plurality of other formations, the training data includes resistivity log data, gamma ray log data, formation porosity log data and volume fractions of total clay;
b) obtaining well log data of the formation derived from a plurality of different-type logging measurements of the formation, wherein the well log data of the formation comprises resistivity log data of the formation, formation total porosity log data of the formation, and gamma ray log data of the formation, wherein the well log data of the formation comprises a well depth in the formation at which the resistivity log data of the formation, the formation total porosity log data of the formation, and the gamma ray log data of the formation were measured;
c) providing the well log data of the formation as input to the at least one trained ANN of the machine learning system and determining or estimating a volume fraction of total clay in the formation as output from the at least one trained ANN, wherein the training data includes the resistivity log data measured, utilizing at least one tool, at different well depths of each of the plurality of other formations, the gamma ray log data measured at the different well depths of each of the plurality of other formations, the formation porosity log data measured at the different well depths of each of the plurality of other formations, and the volume fractions of total clay measured at the different well depths of each of the plurality of other formations, wherein the volume fraction of total clay in the formation determined or estimated as the output of the at least one trained ANN of the machine learning system corresponds to the same well depth in the formation, the method further comprising:
repeating the operations of b) and c) for multiple well depths to determine or estimate volume fractions of total clay in the formation at the multiple well depths;
using the volume fractions of total clay in the formation at the multiple well depths to determine amounts or concentrations of a predefined set of clay minerals in the formation, wherein the amounts or concentrations of the predefined set of clay minerals in the formation are determined using first and second computational models, wherein the first computational model relates total clay volume fraction of the formation and amounts or concentrations of the predefined set of clay minerals in the formation to an apparent cation exchange capacity (CEC) value, and wherein the second computational model relates induction log data of the formation to a calculated CEC value;
using the first computational model to generate multiple apparent CEC profiles of the formation from the volume fractions of total clay in the formation at the multiple well depths, wherein the multiple apparent CEC profiles assume varying amounts or concentrations of the predefined set of clay minerals, and wherein each apparent CEC profile includes apparent CEC values for the multiple well depths in the formation:
using the second computational model and induction log data of formation to generate a calculated CEC profile that includes calculated CEC values for the same multiple well depths in the formation;
comparing the multiple apparent CEC profiles to the calculated CEC profile to identify a particular apparent CEC profile of the multiple apparent CEC profiles that best matches the calculated CEC profile; and
using the assumed amounts or concentrations of the predefined set of clay minerals for the particular apparent CEC profile to determine the amounts or concentrations of the predefined set of clay minerals in the formation.
|