ESTIMATION OF COEFFICIENT OF ISOTHERMAL OIL COMPRESSIBILITY AT RESERVOIR PRESSURE GREATER THAN BUBBLE POINT PRESSURE USING GROUP METHOD OF DATA HANDLING

The procedures of determining the oil compressibility through PVT analyses are usually costly and time consuming. Therefore, various empirical correlations were designed in order to accurately estimate the oil compressibility. This project focuses on designing a model correlation which could accurately predict the oil compressibility coefficient above bubble point pressure. The new model correlation using the Polynomial GMDH method will be compared in terms of its accuracy and reliability with actual PVT data and also current oil compressibility correlation using different types of analysis such as trend analysis, group error analysis, statistical error analysis and graphical error analysis. A total number of 195 data sets were collected from the Mediterranean Basin, Africa, Persian Gulf and the North Sea and after data filtration 183 data were used and divided into 3 sections of training, validation and testing data sets in a ratio of 2:1:1. Using the Polynomial GMDH technique, 3 input parameters were found to be affecting the outputs which are Reservoir Pressure, Solution GOR and Bubble Point Pressure. The new model correlations were then compared with the other correlations and it surpasses all of the other correlations in terms accuracy by having the lowest Root Mean Square Error, Average Percent Relative Errors, Average Absolute Percent Relative Error and Standard Deviations. Furthermore, it is worth to note that the Absolute Percent Relative Error was the main statistical analysis criteria in this study and the Polynomial GMDH model obtained the lowest value. Overall, the Polynomial GMDH model is a robust model and can be affectively applied within its trained data ranges