Validation and camparison of artificial neural network (ANN) and ΔLogR techniques in evaluating organic matter content of source rocks: Case study from Pabdeh Formation, Marun oilfield

Source rock intervals generally show a lower density، higher sonic transit time، higher porosity and higher resistivity than other sedimentary layers. Therefore wire-line logs have been used to identify source rocks and serve as an indicator for their potentiality. It is usually done using intelligent systems such as artificial neural network (ANN) and ΔLogR techniques. Shaly-lime Pabdeh Formation with variable lithology and TOC has been used to make a comparison between results of these techniques and evaluate their validity. Regression analysis shows that correlation of ANN results with Rock-Eval pyrolysis outputs (99%) is more appropriate than ΔLogR results (60%). Calculation of mean square error (MSE) for mentioned procedures (used because MSE method has a better efficiency to determine real error) is in accordance with the said result. Here the MSE of ANN method (0. 07) is much lower than that of ΔLogR technique (0. 98). With an increase in TOC and clay content، ΔLogR accuracy will be increased. In this study، MSE of ΔLogR technique has been increased from 0. 27 to 1. 4 from shale to limestone lithology. TOC content of this formation vary from 0. 5 to 4 wt. % based on ANN results. Pabdeh Formation can be divided into three members: A and C with lower than 1% and B with higher than 1% total organic carbon (TOC) values. Increase in formation thickness، clay percentage and TOC content toward the south-east of oilfield demonstrate that paleo-sedimentary basin had been deeper in this direction. Finally، since rush undulation response of gama-ray log with top of B member، therefore، this top can be used as an indicator of Eocene-Oligocene boundary and Pyrenean orogeny.