ML-based Analysis of Fracture Porosity using Genetic Algorithm (GA) for a Shallow Fractured Igneous HC Reservoir

In this research, a machine learning (ML)-based analysis was conducted, using genetic algorithm (GA) to analyse the fracture porosity of a shallow fractured igneous reservoir in the Wichian-buri (WB) sub-basin, Thailand. The GA code was developed in Python, in which the multiple linear regression (MLR) and root mean square error (RMSE) were employed as objective & fitness functions, in addition the genetic operators were set up according to the tournament selection, arithmetic crossover and random resetting mutation. Conventionally calculated fracture porosity values by well log analysis were used to train the GA models, whose performance was evaluated based on the least total prediction error, cost and execution time. Twelve GA models were run for three reservoir intervals, using different well log data (i.e. CAL, GR, RHOB, NPHI, LLM and DT) as the input parameters. It was found that the GA model consisting of 600-training data with 200 population showed the best performance for the target shallow fractured igneous reservoir interval, located from 1,545m to 1,698m deep. The fracture porosity was determined from well log and the GA analyses was found to be from 0 to 0.04.