| dc.contributor.advisor | Suroso, Arif Imam | |
| dc.contributor.advisor | Hasanah, Nur | |
| dc.contributor.author | Wicaksono, Arif | |
| dc.date.accessioned | 2025-05-23T06:44:42Z | |
| dc.date.available | 2025-05-23T06:44:42Z | |
| dc.date.issued | 2025 | |
| dc.identifier.uri | http://repository.ipb.ac.id/handle/123456789/161770 | |
| dc.description.abstract | Minyak dan gas bumi (migas) merupakan sumber daya alam strategis Indonesia, tidak hanya sebagai pemasok bahan bakar dan bahan baku industri, namun juga merupakan sumber pendapatan negara. Pertumbuhan ekonomi yang cepat, jumlah penduduk yang besar, dan faktor geografis akan menyebabkan peningkatan permintaan sumber daya migas di masa mendatang, meskipun dari sisi produksi minyak, Indonesia terus mengalami penurunan sejak tahun 2016.
Salah satu kegiatan penting dalam kegiatan hulu migas adalah pengukuran sumur produksi di mana semua jenis pengukuran atau pengujian merupakan bagian rutin dalam menjalankan operasi. Suatu ladang minyak yang sudah tua biasanya memerlukan pengukuran produksi yang lebih sering untuk mendeteksi penurunan produksi. Dengan melakukan pengujian sumur produksi, kinerja sumur produksi dapat dievaluasi dan dibandingkan dengan hasil simulasi. Namun, pengujian produksi harian yang ada di lapangan Langgak, Central Sumatra Basin, tidak dapat dilakukan secara rutin dan berkala karena beberapa tantangan ekonomi dan teknis.
Penelitian ini bertujuan untuk mengidentifikasi serta mengevaluasi kinerja pengujian produksi harian sumur minyak mentah yang saat ini diterapkan (existing production test), menerapkan metode Jaringan Saraf Tiruan (JST) Backpropagation untuk memprediksi laju produksi harian, dan menganalisis tingkat akurasi prediksi guna menentukan parameter serta arsitektur jaringan optimal. Sebanyak 17.394 data produksi harian yang dikumpulkan dari 26 sumur digunakan dalam studi ini. Data dibagi menjadi 80% untuk pelatihan dan validasi, serta 20% untuk pengujian. Selanjutnya, data pelatihan dibagi menjadi 10 subset dan divalidasi secara bergilir menggunakan teknik 10-fold cross-validation.
Hasil penelitian menunjukkan perlunya revisi dan perbaikan pada standar operasional prosedur (SOP) pengujian produksi harian, peningkatan pelatihan intensif dan sertifikasi bagi personil pengujian, dan penggunaan peralatan serta perangkat lunak analisis data yang lebih canggih untuk meningkatkan akurasi pengujian. Implementasi JST Backpropagation berhasil memberikan prediksi produksi yang sangat akurat dengan arsitektur jaringan optimal 3-20-1, hidden node sebanyak 20, learning rate 0,05, dan epoch 481. Dengan nilai koefisien korelasi (R) yang tinggi pada tahap pelatihan, validasi, dan pengujian, serta nilai mean square error (MSE) sangat rendah, menunjukkan efektivitas metode ini.
Penelitian ini memperkuat potensi JST Backpropagation sebagai pendekatan prediksi yang adaptif dan akurat dalam industri migas, selaras dengan penelitian terdahulu mengenai penerapan kecerdasan buatan untuk peramalan produksi migas. Temuan ini diharapkan dapat mendukung optimalisasi operasi hulu migas dan pengambilan keputusan berbasis data di lapangan Langgak, Central Sumatra Basin. | |
| dc.description.sponsorship | Oil and gas are among Indonesia's strategic natural resources, serving not only as fuel and industrial raw material suppliers but also as a vital source of national revenue. Rapid economic growth, a large population, and geographical factors are projected to drive increased demand for oil and gas resources in the future, although Indonesia's oil production has been in decline since 2016.
One of the critical activities in upstream oil and gas operations is production well testing, where all forms of measurement or testing are routine parts of operations. Mature oil fields typically require more frequent production measurements to detect production declines. By conducting production well tests, the performance of the production wells can be evaluated and compared to simulation results. However, existing daily production testing at the Langgak field, Central Sumatra Basin, cannot be routinely and periodically conducted due to several economic and technical challenges. Developing predictive models for crude oil production plays a pivotal role in managing mature oil fields as it enables early detection of losses due to declining production.
This study aims to identify and evaluate the performance of existing crude oil well daily production testing, apply the Backpropagation Artificial Neural Network (ANN) method to predict daily production rates, and analyze the prediction accuracy to determine the optimal network parameters and architecture. A total of 17,394 daily production data points collected from 26 wells were utilized in this study. The dataset was divided into 80% for training and validation, and 20% for testing. Furthermore, the training data were partitioned into 10 subsets and validated iteratively using the 10-fold cross-validation technique.
The findings indicate the necessity to revise and improve the standard operating procedures (SOPs) for daily production testing, enhance personnel capabilities through intensive training and certification programs, and adopt more advanced equipment and data analysis software to improve testing accuracy. The implementation of the Backpropagation ANN successfully yielded highly accurate production forecasts, with the optimal network architecture identified as 3-20-1, consisting of 20 hidden nodes, a learning rate of 0.05, and 481 epochs. The high correlation coefficient (R) values during the training, validation, and testing phases, along with the extremely low mean square error (MSE), demonstrate the effectiveness of the proposed method.
This research reinforces the potential of Backpropagation ANN as an adaptive and accurate predictive approach in the oil and gas industry, aligning with previous studies on the application of artificial intelligence for oil and gas production forecasting. These findings are expected to support the optimization of upstream operations and data-driven decision-making in the Langgak Field, Central Sumatra Basin. | |
| dc.language.iso | id | |
| dc.publisher | IPB University | id |
| dc.title | Model Prediksi Produksi Harian Sumur Minyak Mentah Menggunakan Jaringan Saraf Tiruan Backpropagation | id |
| dc.title.alternative | Daily Production Prediction Model of Oil Wells using Backpropagation Neural Network | |
| dc.type | Tesis | |
| dc.subject.keyword | pembelajaran mesin | id |
| dc.subject.keyword | Jaringan Saraf Tiruan | id |
| dc.subject.keyword | Artificial Neural Network | id |
| dc.subject.keyword | Backpropagation neural network | id |
| dc.subject.keyword | machine learning | id |
| dc.subject.keyword | Backpropagation | id |
| dc.subject.keyword | Peramalan Produksi Minyak | id |
| dc.subject.keyword | Prediksi Hasil | id |
| dc.subject.keyword | Oil Production Forecasting | id |
| dc.subject.keyword | Yield Prediction | id |
