Laporan Kerja Praktik Energy & Optimization Embeb Gen AI Petrokimia Gresik

Nadya, Zafira (2026) Laporan Kerja Praktik Energy & Optimization Embeb Gen AI Petrokimia Gresik. Project Report. s.n.. (Unpublished)

Text 5025231310-Project_Report.pdf Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

PT Petrokimia Gresik, sebagai salah satu produsen pupuk terbesar di Indonesia, menghadapi tantangan dalam pengelolaan konsumsi gas alam yang digunakan sebagai bahan baku utama produksi amonia. Sebelum pelaksanaan kerja praktik ini, pemantauan konsumsi energi masih dilakukan secara manual tanpa sistem yang terintegrasi, sehingga menghambat deteksi anomali, analisis tren, dan pengambilan keputusan berbasis data secara cepat. Untuk mengatasi permasalahan tersebut, Tim CAPS09 dari Departemen Teknik Informatika Institut Teknologi Sepuluh Nopember (ITS) mengembangkan sistem EnergyMonitor selama pelaksanaan kerja praktik di PT Petrokimia Gresik. EnergyMonitor merupakan platform pemantauan energi berbasis web yang terdiri atas tiga modul utama, yaitu Real-Time Energy Monitoring Dashboard, model Machine Learning untuk peramalan konsumsi gas alam harian, dan Energy Intelligence Chatbot berbasis Generative Artificial Intelligence dengan pendekatan Retrieval-Augmented Generation (RAG). Sistem dikembangkan menggunakan Next.js pada sisi frontend, FastAPI (Python) pada sisi backend, PostgreSQL sebagai basis data relasional, serta Qdrant sebagai vector database untuk penyimpanan indeks embedding dokumen. Model peramalan menggunakan XGBoost dan mencapai akurasi holdout sebesar 92,6% selama periode evaluasi 61 hari, sedangkan chatbot memanfaatkan Meta Llama 3.1 8B Instruct melalui HuggingFace Inference Router yang diorkestrasi menggunakan LlamaIndex. Hasil pengujian menunjukkan bahwa seluruh 87 skenario positive case berhasil dijalankan dengan status Pass (100%), yang mencakup fitur autentikasi, dashboard pemantauan real-time, analisis Machine Learning, chatbot RAG, dan manajemen unggah dokumen PDF. Sistem telah berhasil di-deploy secara penuh pada HuggingFace Spaces dan dapat diakses oleh pengguna internal PT Petrokimia Gresik. EnergyMonitor terbukti mampu meningkatkan visibilitas konsumsi energi secara real-time sekaligus menyediakan landasan data yang andal untuk mendukung pengambilan keputusan strategis dalam optimasi energi.
================================================================================================================================
PT Petrokimia Gresik, one of the largest fertilizer producers in Indonesia, faces challenges in managing natural gas consumption, which serves as the primary raw material for ammonia production. Prior to this internship project, energy consumption monitoring was performed manually without an integrated system, limiting the ability to detect anomalies, analyze consumption trends, and support timely data-driven decision-making. To address these challenges, Team CAPS09 from the Department of Informatics Engineering at Institut Teknologi Sepuluh Nopember (ITS) developed EnergyMonitor during the internship at PT Petrokimia Gresik. EnergyMonitor is a web-based energy monitoring platform consisting of three main modules: a Real-Time Energy Monitoring Dashboard, a Machine Learning model for daily natural gas consumption forecasting, and an Energy Intelligence Chatbot powered by Generative Artificial Intelligence using the Retrieval-Augmented Generation (RAG) approach. The system was developed using Next.js for the frontend, FastAPI (Python) for the backend, PostgreSQL as the relational database, and Qdrant as the vector database for storing document embedding indexes. The forecasting model employs XGBoost, achieving a holdout accuracy of 92.6% over a 61-day evaluation period, while the chatbot utilizes Meta Llama 3.1 8B Instruct through the HuggingFace Inference Router orchestrated by LlamaIndex. The evaluation results indicate that all 87 positive test scenarios passed successfully (100%), covering authentication, the real-time monitoring dashboard, machine learning analysis, the RAG chatbot, and PDF upload management features. The system was successfully deployed on HuggingFace Spaces and is accessible to internal users at PT Petrokimia Gresik. EnergyMonitor has proven effective in improving real-time visibility of energy consumption while providing a reliable data foundation for strategic decision-making related to energy optimization.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	Energy Monitoring, Generative AI, Retrieval Augmented Generation, Machine Learning, Large Language Model
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting T Technology > T Technology (General) > T58.6 Management information systems T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Information and Communication Technology > Informatics > 55201-(S1) Undergraduate Thesis
Depositing User:	Nadya Zafira
Date Deposited:	09 Jul 2026 06:51
Last Modified:	09 Jul 2026 06:51
URI:	http://repository.its.ac.id/id/eprint/134576

Actions (login required)

View Item