Pengembangan IoT dan AI untuk Optimalisasi Bioremediasi Mikroalga pada Pengolahan Limbah Cair Kelapa Sawit Skala Laboratorium

Mirda, Irfan (2026) Pengembangan IoT dan AI untuk Optimalisasi Bioremediasi Mikroalga pada Pengolahan Limbah Cair Kelapa Sawit Skala Laboratorium. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025241064-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Pencemaran lingkungan akibat Palm Oil Mill Effluent (POME) menjadi permasalahan lingkungan. Salah satu alternatif penanganan masalah ini adalah penurunan kontaminasi POME menggunakan mikroalga. Meskipun efisien dalam menurunkan kontaminan dan memiliki nilai tambah, kultivasi mikroalga sangat bergantung pada kondisi lingkungan. Penelitian ini mengembangkan sistem terintegrasi berbasis Internet of Things (IoT) dan Artificial Intelligence (AI) untuk optimalisasi bioremediasi POME menggunakan mikroalga Chlorella vulgaris pada fotobioreaktor skala laboratorium. Penelitian ini terdiri atas empat tahapan, yaitu kultivasi mikroalga dengan pemantauan real-time, pengembangan model klasifikasi fase pertumbuhan mikroalga berbasis citra, pengembangan model prediksi pencapaian baku mutu air limbah dan treatment mikroalga, serta analisis Multi-Criteria Decision Making (MCDM) untuk mengevaluasi konfigurasi kultivasi paling optimal. Hasil penelitian menunjukkan sistem kultivasi mikroalga skala laboratorium berhasil diimplementasikan dengan integrasi IoT mencakup 9 sensor dan 16 variabel dengan rata-rata keberhasilan akuisisi data 95,59%. Model XGBOOST-BO dan fitur RGB mencapai akurasi tertinggi 97,83% untuk klasifikasi fase hidup mikroalga dengan keunggulan inference time 1-5 ms dan model size 1-3 MB yang memungkinkan deployment pada perangkat edge computing. Untuk prediksi kebutuhan treatment, XGBoost-GWO optimal pada resolusi harian untuk Molase (R²=0,99) dan Sodium Bikarbonat (R²=0,96), sedangkan parameter kontrol lingkungan seperti Intensitas Cahaya (R²=0,99) dan Aerasi (R²=0,98) lebih optimal pada resolusi jam. Untuk prediksi baku mutu air limbah, model XGBoost dengan optimasi metaheuristik pada resolusi jam memberikan performa optimal dengan R² mencapai 0,9937 untuk NH₃, 0,9840 untuk TSS, 0,9769 untuk COD, 0,9757 untuk BOD, 0,9776 untuk pH, dan 0,9884 untuk biomassa pada resolusi harian. Analisis MCDM menggunakan metode TOPSIS, VIKOR, dan DEA mengidentifikasi Kultivasi K5 sebagai konfigurasi optimal dengan keseimbangan terbaik antara waktu pencapaian baku mutu 14 hari, produksi biomassa tertinggi 1,1794 g/L, serta biaya operasional Rp 27.913,71. Penelitian ini berkontribusi pada pengembangan sistem bioremediasi POME yang dapat dipantau dan dioptimalkan secara presisi berbasis data real-time untuk penerapan di industri kelapa sawit Indonesia.
=======================================================================================================================================
Environmental pollution caused by Palm Oil Mill Effluent (POME) is a critical issue in Indonesia. One alternative approach to address this problem is the reduction of POME contamination using microalgae. Although efficient in reducing contaminants and offering added value, microalgae cultivation is highly dependent on environmental conditions. This study develops an integrated system based on Internet of Things (IoT) and Artificial Intelligence (AI) for the optimization of POME bioremediation using microalgae Chlorella vulgaris in a laboratory-scale photobioreactor. This research comprises four stages: microalgae cultivation with real-time monitoring, development of an image-based microalgae growth phase classification model, development of a predictive model for wastewater quality standard achievement and microalgae treatment, and Multi-Criteria Decision Making (MCDM) analysis to evaluate the most optimal cultivation configuration. The research results demonstrate that a laboratory-scale microalgae cultivation system was successfully implemented with IoT integration comprising 9 sensors and 16 variables, achieving an average data acquisition success rate of 95.59%. The XGBoost-BO with RGB features achieved the highest accuracy of 99.07% for microalgae life phase classification, with advantages including 1-5 ms inference time and 1-3 MB model size, enabling deployment on edge computing devices. For treatment requirement prediction, XGBoost-GWO was optimal at daily resolution for Molasses (R²=0.99) and Sodium Bicarbonate (R²=0.96), while environmental control parameters such as Light Intensity (R²=0.99) and Aeration (R²=0.98) were more optimal at hourly resolution. For wastewater quality standard prediction. XGBoost models with metaheuristic optimization at hourly resolution provided optimal performance with R² reaching 0.9937 for NH₃, 0.9840 for TSS, 0.9769 for COD, 0.9757 for BOD, 0.9776 for pH, and 0,9884 for biomassa at daily resolution. MCDM analysis using TOPSIS, VIKOR, and DEA methods identified Cultivation K5 as the optimal configuration with the best balance between 14-day compliance achievement time, highest biomass production of 1.1794 g/L, and operational cost of IDR 27,913.71. This research contributes to the development of a POME bioremediation system that can be monitored and optimized with precision based on real-time data for application in the Indonesian palm oil industry.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	AI, Bioremediasi, IoT, Mikroalga, dan POME
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QH Biology > QH301 Biology
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Irfan Mirda
Date Deposited:	30 Jan 2026 09:50
Last Modified:	30 Jan 2026 09:50
URI:	http://repository.its.ac.id/id/eprint/131270

Actions (login required)

View Item