Prediksi Porositas Menggunakan Atribut Seismik Dengan Machine Learning Dan Uncertainty Quantification: Studi Kasus Di Lapangan Inas, Cekungan Malay Utara

Harahap, Almer Nawfal Zahranvi (2025) Prediksi Porositas Menggunakan Atribut Seismik Dengan Machine Learning Dan Uncertainty Quantification: Studi Kasus Di Lapangan Inas, Cekungan Malay Utara. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5017201088-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Prediksi porositas yang akurat merupakan hal yang penting untuk eksplorasi hidrokarbon dan karakterisasi reservoir. Namun, metode konvesional memiliki berbagai keterbatasan ataupun kekurangan, seperti biaya yang tinggi, resolusi rendah, inherent uncertainities, dan faktor lain yang terkait. Studi ini memanfaatkan machine learning (ML) untuk memprediksi porositas berdasarkan atribut seismik di Lapangan Inas, Cekungan Malay Utara, dengan mengkuantifikasi uncertainty. Metodologi yang digunakan merupakan integrasi data seismik dan log sumur, dengan pemilihan fitur (feature selection) menggunakan matriks korelasi heatmap yang bertujuan untuk mengidentifikasi atribute seismik paling relavan. Lima model ML dievaluasi, yaitu Stochastic XGBoost (SXGB), Bayesian Optimization untuk CatBoost (CB), Quantile Regression Forest (QRF), Gradient Boosting (GB), dan Random Forest (RF). SXGB terbukti paling efektif, dengan nilai root mean square error (RMSE) sebesar 0.0196, koefisisen korelasi Pearson (PCC) sebesar 0.83, dan mean stardardized log likelihood (MSLL) sebesar 0.8713. Prediksi model divalidasi menggunakan penampang impedansi hasil inversi, yang menunjukkan hubungan terbalik antara porositas dan impedansi yang sejalan dengan prinsip geologi. Uncertainty quantification (UQ) melalui metode bootstrap resampling menghasilkan standard deviation antara 0.002 hingga 0.007, dengan nilai porositas yang diprediksi berkisar antara 2% hingga 32%. Peta horizon dan penampang seismik mengidentifikasi zona reservoir potensial, terutama pada interval B100 hingga D34. Studi ini menegaskan efektivitas ML dalam mengoptimalkan akurasi dan reliable prediksi porositas, serta memberikan alternatif yang hemat biaya dibandingkan metode tradisional.
===================================================================================================================================
Accurate porosity prediction is crucial for hydrocarbon exploration and reservoir characterization, yet conventional methods face limitations due to cost, resolution, inherent uncertainties, and many more that related to it. This study leverages machine learning (ML) to predict porosity from seismic attributes in the Inas Field, North Malay Basin, while quantifying prediction uncertainty. The methodology integrates seismic and well log data, employing a heatmap matrix correlation for feature selection to identify the most relevant seismic attributes. Five ML models: Stochastic XGBoost (SXGB), Bayesian Optimization for CatBoost (CB), Quantile Regression Forest (QRF), Gradient Boosting (GB), and Random Forest (RF) were evaluated. SXGB experienced as the most effective, achieving a root mean square error (RMSE) of 0.0196, a Pearson correlation coefficient (PCC) of 0.83, and a mean standardized log likelihood (MSLL) of 0.8713. The model's predictions were validated by inverted impedance sections, demonstrating an inverse relationship between porosity and impedance, consistent with geological principles. Uncertainty quantification (UQ) via bootstrap resampling revealed standard deviations ranging from 0.002 to 0.007, with predicted porosity values ranging from 2% to 32%. Horizon maps and seismic sections highlighted potential reservoir zones, particularly within the B100 to D34 interval. This study underscores the efficacy of ML in optimizing porosity prediction accuracy and reliability, offering a cost-effective alternative to traditional methods.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Prediksi porositas, atribut seismik, machine learning, kuantifikasi uncertainty, Lapangan Inas, Cekungan Malay Utara; feature selection, bootstraping, stochastic XGBoost, Porosity prediction, seismic attributes, machine learning, uncertainty quantification, Inas Field, North Malay Basin, feature selection, bootstraping, stochastic XGBoost
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geophysics Engineering > 33201-(S1) Undergraduate Thesis
Depositing User:	Almer Nawfal Zahranvi Harahap
Date Deposited:	06 Aug 2025 02:45
Last Modified:	06 Aug 2025 02:49
URI:	http://repository.its.ac.id/id/eprint/127681

Actions (login required)

View Item