Rancang Bangun Sistem Pengendalian Level pada Mini Plant Separator Tiga Fasa Menggunakan Deep Reinforcement Learning-PID

Negoro, Muhamamd Satrio Pinoto (2023) Rancang Bangun Sistem Pengendalian Level pada Mini Plant Separator Tiga Fasa Menggunakan Deep Reinforcement Learning-PID. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 02311940000108-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Industri oil & gas sering kali ditemui mengenai proses separasi minyak mentah yang dil-akukan pada instrument separator tiga fasa. Instrument separator diperlukan pengendalian, yang salah satunya adalah pengendalian level agar produksi minyak dapat terkontrol. Pada penelitian ini dilakukan perancangan mengenai sistem pengendalian level PID yang dikom-binasikan dengan Deep Reinforcement Learning yang akan melakukan tuning gain PID secara adaptif selama proses pengontrolan level berlangsung. Berdasarkan penelitian yang telah dilakukan diperoleh bahwa sistem pengendalian menggunakan DRL-PID dapat melakukan kontrol dengan cukup baik pada sistem yang non-linear. Dan diperoleh bahwa performansi terkait pengontrolan menggunakan DRL-PID pada level campuran memen-ghasilkan rise time sebesar 241,1 detik, settling time sebesar 261,71 detik, overshoot sebesar 1,4286%, dan error steady state sebesar 1,07%. Diperoleh pula performansi terkait pen-gontrolan DRL-PID pada level minyak menghasilkan rise time sebesar 76,36 detik, settling time sebesar 80,68 detik, overshoot sebesar 1,42%, dan error steady state sebesar 2,42%. Selain itu, penggunaan Nb-IoT dengan menggunakan module SIM7000E sebagai media komunikasi pada mini plant separator sangat cocok digunakan dengan kebutuhan transfer data yang kecil, yaitu hanya data teks berupa perubahan level fluida. Sistem monitoring pada mini plant separator tiga fasa mampu mengirimkan data dengan delay time rata-rata 120 ms.
========================================================================================================================
The oil & gas industry is often encountered regarding the crude oil separation process carried out in a three-phase separator instrument. The separator instrument requires control, one of which is level control so that oil production can be controlled. In this research, the design of a PID level control system combined with Deep Reinforcement Learning will adaptively tune the PID gain during the level control process. Based on the research that has been done, it is obtained that the control system using DRL-PID can control quite well on non-linear systems. And it is obtained that the performance related to controlling using DRL-PID at the mixed level produces a rise time of 241.1 seconds, settling time of 261.71 seconds, overshot of 1.4286%, and steady-state error of 1.07%. The performance related to DRL-PID control of the oil level resulted in a rise time of 76.36 seconds, a settling time of 80.68 seconds, an overshot of 1.42%, and a steady state error of 2.42%. In addition, the use of Nb-IoT using the SIM7000E module as a communication medium on the mini plant separator is very suitable for use with small data transfer requirements, namely only text data in the form of fluid level changes. The monitoring system on the three-phase mini plant separator can send data with an average delay time of 120 ms.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Control, Deep Reinforcement Learning, Level, Separator, PID Control, Deep Reinforcement Learning, Level, Separator, PID
Subjects:	T Technology > TJ Mechanical engineering and machinery > TJ213 Automatic control. T Technology > TJ Mechanical engineering and machinery > TJ217 Adaptive control systems T Technology > TJ Mechanical engineering and machinery > TJ223 PID controllers
Divisions:	Faculty of Industrial Technology > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Satrio Pinoto Negoro
Date Deposited:	04 Aug 2023 04:08
Last Modified:	04 Aug 2023 04:08
URI:	http://repository.its.ac.id/id/eprint/101573

Actions (login required)

View Item