--> --> Evaporite Facies Characterization Based on an Unsupervised Artificial Neural Network in Northern Saudi Arabia

2018 AAPG International Conference and Exhibition

Datapages, Inc.Print this page

Evaporite Facies Characterization Based on an Unsupervised Artificial Neural Network in Northern Saudi Arabia

Abstract

The late Jurassic Arab Formation, a typical carbonate-evaporite sequence, is the most prolific oil producing interval in the world and forms many super giant fields such as Ghawar. The Arab-A is the top member of four upward-shoaling carbonate-anhydrite cycles in the Arab Formation, and is overlain by the Hith anhydrite. To characterize the depositional environment of the Arab-A in northern Saudi Arabia, multiple seismic attributes are classified using an unsupervised artificial neural network (ANN) to recognize the seismic facies and characterize sedimentary facies. Unsupervised ANNs are powerful classification techniques, which are implemented by a single layer perceptron network. The network is trained by comparing the neurons to the input vectors using competitive-learning techniques. The most similar neuron is “the winner” and moves towards the class. Once the neuron migrates to the center of the class, the network stabilizes, training is finished and the neuron is assigned to a representative class. Without a priori information, the unlabeled classes are calibrated and analyzed by lithofacies generated from log and core data; and further sedimentary facies are recognized by integrating local geological information. Three lithofacies comprising tight limestone, grain-dominated packstone and oolitic grainstone are identified based on wireline logs and core data. The unlabeled classes generated from unsupervised ANNs are interpreted based on the integration of lithofacies and geological knowledge. Generally, lagoonal deposits of the inner ramp, ramp crest shoal and proximal deposits of the middle ramp are recognized in the study area. The proximal middle ramp with tight limestone develops in the northeast portion of the study area near Gotnia basin. Ramp crest shoals with peloid and oolitic grainstone are widely distributed in the western area and are semi-parallel to the coastline. The widely developed ramp crest shoal provides good quality reservoir in proven stratigraphic traps in the study area. The lagoonal facies between shorelines which are dominated by tight limestone constitute the lateral seal. Facies recognition based on unsupervised ANNs is shown to characterize the environment of deposition and identify potential stratigraphic traps in northern Saudi Arabia.