Prediksi Temperatur Bawah Permukaan Berdasarkan Resistivitas Magnetotelurik (MT) Menggunakan Artificial Neural Network (ANN) pada Lapangan Geotermal Mountain Home, Idaho

Suhardi, Sultan Daffa Khoirussyah (2025) Prediksi Temperatur Bawah Permukaan Berdasarkan Resistivitas Magnetotelurik (MT) Menggunakan Artificial Neural Network (ANN) pada Lapangan Geotermal Mountain Home, Idaho. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5017211035-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Suhu bawah permukaan merupakan parameter penting dalam menentukan kelayakan ekonomi lapangan geotermal, namun penurunan akurasi sensor pada kondisi suhu tinggi menjadi tantangan teknis yang signifikan. Metode magnetotellurik (MT) dengan penetrasi puluhan hingga ratusan kilometer menjadi alternatif untuk eksplorasi geotermal. Penelitian ini bertujuan menganalisis variasi input parameter resistivitas, kedalaman, dan kombinasi keduanya terhadap hasil prediksi temperatur menggunakan Artificial Neural Network (ANN), serta menganalisis sebaran temperatur hasil prediksi ANN pada lokasi penelitian. Dataset menggunakan data open source dari lapangan Mountain Home, Idaho, terdiri dari 6 stasiun MT dalam radius 2 km dari pengeboran dan 1 sumur bor kedalaman 0-1670 meter. Data MT diolah melalui pemfilteran spektra koherensi, analisis dimensionalitas, dan inversi Occam 1D menghasilkan resistivitas terhadap kedalaman dengan spasi 10 meter. Hubungan eksponensial antara resistivitas dan suhu sumur bor menghasilkan koefisien korelasi (R²) sebesar 0,45. Model ANN digunakan untuk memodelkan hubungan non-linear dengan tiga variasi input yaitu resistivitas, kedalaman, dan kombinasi keduanya. Resistivitas tunggal memberikan error (MAPE) 10,36-24,55%, kedalaman tunggal menghasilkan MAPE 7,28%, sedangkan kombinasi resistivitas-kedalaman memberikan hasil optimal dengan MAPE 2,40-3,01%. Penggunaan kedalaman saja menghasilkan prediksi terlalu homogen, sehingga kombinasi menjadi model optimal. Rata-rata gradien geotermal hasil prediksi sebesar 69,40°C/km dengan error relatif 4,27% terhadap gradien sumur bor (72,50°C/km). Hasil membuktikan ANN dapat memprediksi temperatur menggunakan kombinasi resistivitas-kedalaman dengan akurasi tinggi. Model dasar ini dapat dikembangkan untuk lapangan geotermal Indonesia dengan variasi litologi berbeda.
=================================================================================================================================
Subsurface temperature is a crucial parameter for determining the economic feasibility of geotermal fields, but sensor accuracy degradation at high temperatures poses significant technical challenges. Magnetotelluric (MT) methods with penetration depths of tens to hundreds of kilometers offer an alternative for geotermal exploration. This study aims to analyze the variation of input parameters including resistivity, depth, and their combination on temperature prediction results using Artificial Neural Network (ANN), and to analyze the temperature distribution from ANN predictions at the research location. The dataset utilizes open source data from the Mountain Home geotermal field, Idaho, consisting of 6 MT stations within a 2 km radius from drilling point and 1 borehole with depths of 0-1670 meters. MT data were processed through spectra coherencies filtering, dimensionality analysis, and Occam 1D inversion to generate resistivity versus depth with 10-meter spacing. The exponential relationship between resistivity and borehole temperature yields a correlation coefficient (R²) of 0.45. ANN models were used to model non-linear relationships with three input variations: resistivity, depth, and their combination. Single resistivity input produced errors (MAPE) of 10.36-24.55%, single depth input yielded MAPE of 7.28%, while the resistivity-depth combination provided optimal results with MAPE of 2.40-3.01%. Using depth alone resulted in overly homogeneous predictions, making the combination the optimal model. The average predicted geotermal gradient was 69.40°C/km with a relative error of 4.27% compared to the borehole gradient (72.50°C/km). Results demonstrate that ANN can predict temperature using resistivity-depth combination with high accuracy. This model has potential for development and application to Indonesian geotermal fields with different lithological variations.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Geotermal, Magnetotelurik, Artificial Neural Network, Temperatur; Geothermal, Magnetotelluric, Artificial Neural Network, Temperature
Subjects:	Q Science > QC Physics > QC271 Temperature measurements Q Science > QC Physics > QC610.3 Electric conductivity Q Science > QC Physics > QC 611.97.T46 Temperature effects. Including transition temperature
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geophysics Engineering > 33201-(S1) Undergraduate Thesis
Depositing User:	Sultan Daffa Khoirussyah Suhardi
Date Deposited:	01 Aug 2025 02:24
Last Modified:	01 Aug 2025 02:24
URI:	http://repository.its.ac.id/id/eprint/123904

Actions (login required)

View Item