Identifikasi Recharge Area Aplikasi Algoritma Machine Learning Random Forest (Studi Kasus: Lapangan Panas Bumi Patuha, Kabupaten Bandung)

Indriani, Rista Fitri (2023) Identifikasi Recharge Area Aplikasi Algoritma Machine Learning Random Forest (Studi Kasus: Lapangan Panas Bumi Patuha, Kabupaten Bandung). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6016211003-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 September 2025. Download (6MB) | Request a copy

Abstract

Keberlangsungan sistem panas bumi sangat bergantung pada volume fluida pada reservoir panas bumi. Volume fluida pada reservoir erat kaitannya dengan recharge area sebagai zona infiltrasi air ke reservoir panas bumi. Tindakan pengelolaan, pengembangan, dan monitoring terkait sumber daya air pada recharge area perlu dilakukan. Penelitian ini bertujuan menganalisis potensi recharge area Lapangan Panas Bumi Patuha Kabupaten Bandung melalui Aplikasi Algoritma Machine Learning Random Forest. Potensi recharge area pada penelitian ini ditentukan berdasarkan 10 data parameter meliputi elevasi, kemiringan lereng, tata guna lahan, jenis tanah, NDVI, kerapatan sungai, curah hujan, complete bouguer anomaly (CBA), jenis batuan, dan lineament density. Model dibangun berdasarkan data training infiltrasi batuan dan tanah dengan splitting ratio 70:30. Setiap model dilakukan evaluasi diantaranya Confusion Matrix (overall accuracy dan user accuracy), ROC-AUC, dan Cohen’s Kappa. Evaluasi model diperoleh hasil terbaik pada model data training infiltrasi batuan dengan nilai overall accuracy 1, user accuracy (non resapan 0,98 dan resapan 0,97), AUC 1, dan Cohen’s Kappa 1. Sementara model data training infiltrasi tanah memiliki nilai overall accuracy 0,98, user accuracy (non resapan 0,98 dan resapan 0,96), AUC 0,99, dan Cohen’s Kappa 0,94. Hasil evaluasi kedua model menunjukkan konsistensi yang hampir sempurna pada kedua model. Model data training infiltrasi batuan diperoleh kontribusi parameter tertinggi senilai 100 % pada jenis batuan, sementara jenis tanah dengan kontribusi 100 % untuk model data training infiltrasi tanah. Hasil dari korelogram model data training infiltrasi batuan menunjukkan korelasi positif tertinggi antara parameter elevasi dan CBA senilai 0,77. Sementara parameter elevasi dan curah hujan menunjukkan nilai korelasi positif sebesar 0,84 pada model data training infiltrasi tanah. Nilai-nilai korelasi tersebut menunjukkan korelasi antara dua parameter kuat. Hal tersebut menunjukkan semakin tinggi nilai parameter maka semakin besar nilai parameter lainnya. Luas lahan wilayah dengan potensi resapan air yaitu 163,17 km2 yang dijumlahkan dari wilayah dengan potensi resapan air sangat tinggi dan tinggi. Berdasarkan hasil yang diperoleh dari penelitian ini diharapkan dapat digunakan sebagai dasar pengelolaan, pengembangan, dan monitoring terkait sumber daya air pada recharge area wilayah kajian.
======================================================================================================================================
Geothermal system sustainability depends on reservoir fluid volume. As a water infiltration zone into the geothermal reservoir, the recharge region strongly affects reservoir fluid volume. Manage, develop, and monitor water resources in the recharge area. This study aims to analyze the recharge potential of the Patuha Geothermal Field in Bandung Regency through the Application of the Random Forest Machine Learning Algorithm. The potential recharge area in this study determines based on 10 data parameters, including elevation, slope, land use, soil type, NDVI, river density, rainfall, complete bouguer anomaly (CBA), rock type, and lineament density. The model builds based on rock and soil infiltration training data with a splitting ratio of 70:30. Each model evaluates, including Confusion Matrix (overall accuracy and user accuracy), ROC-AUC, and Cohen's Kappa. Model evaluation obtained the best results on the rock infiltration-training data model with an overall accuracy value of 1, user accuracy (non-recharge 0.98 and recharge 0.97), AUC 1, and Cohen's Kappa 1. Meanwhile, the soil infiltration-training data model has an overall accuracy value of 0.98, user accuracy (non-recharge 0.98 and recharge 0.96), AUC 0.99, and Cohen's Kappa 0.94. The evaluation results of the two models show almost perfect consistency in both models. The rock infiltration-training data model obtained the highest parameter contribution the rock type (100%), while the soil type (100%) contributed for the soil infiltration-training data model. The correlogram of the rock infiltration-training data model shows the highest positive correlation between elevation-CBA of 0.77. In contrast, the elevation-rainfall parameters show a positive correlation value of 0.84 in the soil infiltration-training data model. This data shows a significant link between the two factors. Correlation demonstrates that the greater the value of a parameter, the greater the other values. The land area with the potential for water recharge is 163.17 km2 which is the sum of the areas with very high and high water recharge potential. The management, development, and monitoring of water resources in the study region's recharge area can be recommended as a result of the research's findings.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	infiltrasi, machine learning, Random Forest, panas bumi, recharge area.
Subjects:	G Geography. Anthropology. Recreation > G Geography (General) > G70.5.I4 Remote sensing G Geography. Anthropology. Recreation > GB Physical geography > GB1003.2 Groundwater. Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geomatics Engineering > 29101-(S2) Master Thesis
Depositing User:	Rista Fitri Indriani
Date Deposited:	03 Aug 2023 05:44
Last Modified:	03 Aug 2023 05:44
URI:	http://repository.its.ac.id/id/eprint/100987

Actions (login required)

View Item