oa Maximum Likelihood & Multiple Imputation of Incomplete Static and Dynamic Reservoir Data

The problem of incomplete data is a crucial issue that should be handled efficiently for accurate evaluation and prediction in a reservoir characterization. Maximum likelihood method has been adopted to handle the incomplete data in well logs and core measurements in addition to production and injection rates of ten wells in a heterogeneous reservoir in an Iraqi oil field. Expectation Maximization algorithm (EM) is the concept of ML imputation and it starts with some initial values for the mean and the covariance matrix and iterates through imputing missing values (imputation step) and re-estimating the mean and the covariance matrix from the complete data set (estimation step). The iteration process ends when the maximum relative difference in all of the estimated means, variances between two iterations is less than or equal to a given value. Furthermore, multi-regression linear models have been set for the log porosity as a response and function of all other factors including Vsh, Gamma ray, formation density, and resistivity. Then, the core porosity has been estimated for all intervals as a function of log porosity in order to estimate the corrected permeability. These linear models have performed good quality of test indicators especially p-value, adjusted-R, and F-ANOVA.