Early Results of Comparison between K-Nearest Neighbor and Artificial Neural Network Method for Facies Estimation

Hadyan Pratama (Department of Geophysical Engineering University of Pertamina, Jl. Teuku Nyak Arief, Simprug, Kebayoran Lama, Jakarta 12220, Indonesia)
Loris Alif Syahputra (Department of Geophysical Engineering University of Pertamina, Jl. Teuku Nyak Arief, Simprug, Kebayoran Lama, Jakarta 12220, Indonesia)
Muhammad Fauzan Albany (Department of Geophysical Engineering University of Pertamina, Jl. Teuku Nyak Arief, Simprug, Kebayoran Lama, Jakarta 12220, Indonesia)
Agus Abdullah (Department of Geophysical Engineering University of Pertamina, Jl. Teuku Nyak Arief, Simprug, Kebayoran Lama, Jakarta 12220, Indonesia)
Sandy Kurniawan Suhardja (Department of Geophysical Engineering University of Pertamina, Jl. Teuku Nyak Arief, Simprug, Kebayoran Lama, Jakarta 12220, Indonesia)
Epo Kusumah (Department of Geological Engineering University of Pertamina, Jl. Teuku Nyak Arief, Simprug, Kebayoran Lama, Jakarta 12220, Indonesia)
Weny Astuti (Department of Petroleum Engineering University of Pertamina, Jl. Teuku Nyak Arief, Simprug, Kebayoran Lama, Jakarta 12220, Indonesia)
Bambang Mujihardi (Upstream Researh and Technology - RTC, PT Pertamina (Persero), Jakarta 10110, Indonesia)

Jurnal Geofisika
Vol 18 No 1 (2020): Jurnal Geofisika

Artificial Intelligence method has been widely used recently in many aspects to understand big data. Fundamentally, the purpose of Artificial Intelligence is to solve nonlinear problem. Most methods are trying to optimize an output from one or many inputs parameter by identifying any potential patterns that fit or using a statistical data. In Oil & Gas industry, one of the main challenges that can be solved by Artificial Intelligence is estimating facies from well log or seismic data. The main scope of this study is estimating lithofacies by analyzing well logs input using two different methods, K-Nearest Neighbor (KNN) and Artificial Neural Network (ANN). We employed various well log data such as gamma-ray, resistivity, neutron density porosity, and photoelectric effect from well log data at Panoma Council Grove Field, South West Kansas, United States. This study shows that using optimized parameters, KNN method faster than ANN method but, ANN give result better than KNN. Nevertheless, despite the fact this research could estimate lithologies, many aspect should be considered in order to reach optimum result such as insights from geological regional models.