Penentuan Acuan Nilai Level Amine Surge Tank pada Sistem Amine di Gas Plant dengan Pendekatan Analisis Prediktif

Huda, Muhammad Miftahul (2024) Penentuan Acuan Nilai Level Amine Surge Tank pada Sistem Amine di Gas Plant dengan Pendekatan Analisis Prediktif. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032212007-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (13MB) | Request a copy

Abstract

Dalam perjanjian jual beli gas, gas asam merupakan impurity yang harus dikurangi kandungannya hingga memenuhi spesifikasi yang dipersyaratkan. Jika terjadi off-spec pada gas jual, maka pihak produsen berkewajiban membayar denda kepada pihak pembeli. Fasilitas yang berperan dalam mengurangi kandungan gas asam (H2S and CO2) pada Gas Plant XYZ adalah Fasilitas Amine Gas Treating melalui proses absorpsi menggunakan larutan amine. Larutan amine mempunyai parameter yang disebut amine strength (Y). Y mempunyai nilai normal operating limit (NOL) 50% hingga 52%. Jika nilai Y kurang dari NOL, maka akan menyebabkan lemahnya penyerapan gas asam yang mengakibatkan gas off-spec. Jika nilai Y lebih dari NOL maka menyebabkan korosi pada Fasilitas Amine Gas Treating. Nilai Y dipengaruhi oleh level amine surge tank (X). Selama ini penentuan nilai X hanya didasarkan pada perkiraan operator saja, sehingga jumlah nilai Y yang mengalami out of range masih tinggi. Dari 242 instance data mulai bulan Desember 2022 hingga Juli 2023 masih terdapat 45% nilai Y mengalami out of range. Oleh karena itu, keakurasian dalam memprediksi nilai Y yang didasarkan pada nilai X perlu ditingkatkan untuk mengurangi persentase out of range ini. Penelitian ini bertujuan untuk memberikan acuan nilai X dengan menggunakan pendekatan analisis prediktif, sehingga nilai Y dapat dijaga pada rentang NOL. Dua metode utama yang digunakan adalah regresi linier (ordinary least square dan regresi linier bergerak beserta pengembangannya) dan regresi non linier (regresi polinomial dan eksponensial). Metode yang berbasiskan regresi tersebut dipilih karena kepopulerannya dalam memprediksi variabel respons berdasarkan variabel predictor. Metode ini dapat dilakukan di berbagai perangkat lunak, termasuk MS Excel, yang tersedia dan diizinkan penggunaannya di Gas Plant XYZ. Nilai classification accuracy (CA) dari masing-masing metode akan dikomparasikan untuk memperoleh metode dengan ranking terbaik. Regresi linier bergerak dengan 1 variabel predictor X dan 2 pengkonversi nilai (RLB-X-2PN) mempunyai nilai CA tertinggi dibandingkan dengan metode lainnya. RLB-X-2PN mempunyai nilai CA training sebesar 81.21% dan testing sebesar 77.97%. Pada tanggal 19 Desember 2023 telah dilakukan uji coba dengan hasil yang memuaskan, yang mana dengan mengatur level amine surge tank (Y) berdasarkan rekomendasi yang dihasilkan oleh aplikasi level amine reference, dapat memperoleh nilai Y sebesar 50.63% atau masuk rentang NOL-nya. Penerapan metode terpilih (RLB-X-2PN) pada Gas Plant XYZ berpotensi meningkatkan nilai manfaat sebesar 18.64%, dimana nilai manfaat tersebut merupakan gabungan dari berkurangnya resiko terkenanya denda dari pembeli dan berkurangnya laju korosi pada Fasilitas Amine Gas Treating.
=================================================================================================================================
In the gas sales and purchase agreement, acid gas is an impurity that must be reduced to meet the required specifications. If there is an off-spec in the sales gas, the producer will be penalized by the buyer. The facility that plays a role in reducing the acid gas content (H2S and CO2) at the Gas Plant XYZ is the Amine Gas Treating Facility through an absorption process using an amine solution. The amine solution has a parameter called amine strength (Y). Y has a normal operating limit (NOL) value of 50% to 52%. If the Y value is less than NOL, it will cause weak absorption of acid gas resulting in off-spec gas. If the Y value is more than NOL it will cause corrosion at the Amine Gas Treating Facility. The Y value is influenced by the amine surge tank level (X). So far, determining the X value is only based on operator estimates, so the number of Y values that are out of range is still high. Out of 242 data instances from December 2022 to July 2023, there are still 45% of Y values that are out of range. Therefore, the accuracy in predicting the Y value based on the X value needs to be improved to reduce the out of range percentage.
This research aims to provide a reference for the X value using a predictive analysis approach, so that the Y value can be maintained within the NOL range. The two main methods used are linear regression (which consists of ordinary least squares, moving linear regression and its developments) and non-linear regression (which consists of polynomial regression and exponential regression). The regression-based method was chosen because of its popularity in predicting response variables based on predictor variables. Another reason is that the regression method can be carried out in various software, including MS Excel, which is available and permitted for use at Gas Plant XYZ. The classification accuracy (CA) value of each method will be compared to determine the best ranking. Moving linear regression with 1 predictor variable X and 2 value converters (RLB-X-2PN) has the highest CA value compared to other methods. RLB-X-2PN has a CA training value of 81.21% and CA testing of 77.97%. On December 19, 2023, a trial was carried out with satisfactory results, where by adjusting the level of amine surge tank (Y) based on recommendations resulted by the amine level reference application, a Y value of 50.63% was obtained or within NOL range. Application of the selected method (RLB-X-2PN) at Gas Plant XYZ will increase the benefit value by 18.64%, which the benefit value is a combination of reducing the risk of being penalized from buyers and reducing the rate of corrosion at the Amine Gas Treating Facility.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Amine strength, Eksponensial, Gas Plant, Polinomial, Prediksi, Regresi
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting T Technology > T Technology (General) > T58.8 Productivity. Efficiency T Technology > TP Chemical technology > TP350 Natural gas--Drying.
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Muhammad Miftahul Huda
Date Deposited:	08 Feb 2024 08:22
Last Modified:	08 Feb 2024 08:22
URI:	http://repository.its.ac.id/id/eprint/106683

Actions (login required)

View Item