Lithofacies Prediction

Lithofacies have been recognized and described from 15 cored wells. Five key core wells (red triangles in the figure) in Big Bow and Sand Arroyo Creek Fields were chosen to build the Neural Network model.

Determined predictor variables (GR, IDID, IDIM, PE, Neutron/Density Porosity and lithofacies categories were inputted into Kipling to build Neural Network models. The size of network and dampling parameters were optimized using cross-validation and repeatedly tested with whole training data and randomly chosen partial of whole data set.