| Abstract | In this study, well log data were collected from two study locations in the Wichian-buri (WB) sub-basin, Thailand and the Cuu Long (CL) basin, Vietnam. The objective of this study is to develop a GA code in Python to analyze the fracture porosity of fractured igneous rock reservoirs based on well log data. The GA-based calculation results were compared and validated with fracture porosity calculated by conventional method. First, the fracture porosity was calculated by the method proposed by Elkewidy and Tiab (1998). The results of this conventional approaches were further used to train the machine-learning model using genetic algorithm (GA). Two sets of analysis were conducted for two study sites, and namely, group A and group B, respectively. The best performed models of each site were selected based on least total prediction error, cost and execution time. In the WB sub basin, twelve models were investigated with CAL, GR, RHOB, NPHI, LLM and DT as input parameters. The best GA model consist of 600-training dataset, 200-population when the analysis depth was from 1,545m to 1,698m. The conventionally-calculated and GA analysis fracture porosities matched quite well in the range from 0 to 0.04. For CL basin, the input well log data differ a bit from the dataset used for the WB basin consisting of CAL, GR, RHOB, NPHI, LLS and LLD as inputs. Two GA models showed best performance among the fifteen models, they consist of 120 and 1080-training data with 100 population. The conventionally-calculated fracture porosity for both models was found in range from 0 to 0.03, which matched relatively well with that predicted by the two models mentioned. |