Virtual Booth

Traditional methods of environmental management for the POME (palm oil mill effluent) in Malaysia are often slow to react and therefore may lead to regulatory breaches due to the lack of proactive monitoring of the discharge of palm oil mills. This project seeks to develop the "POME treatment optimisation and forecasting system", which will transform mill operations from reactive reporting to proactive intervention. It will use a custom-built Excel parser called Enhanced Excel Parser to collect, normalise and combine into one database eighteen industrial log files from 2024 and 2025 containing 495 records of monitoring data on POME. Due to the high volatility of data recorded at palm oil mills, it has been decided to implement median-based statistical threshold anomalies to detect anomalies in POME and thus provide a more robust baseline compared to the mean-based method used traditionally. In addition, the project will analyse four different machine learning algorithms to classify the degree of compliance with events related to the treatment process of POME, which are Random Forest (RF), XGBoost, Logistic Regression (LR), and Artificial Neural Network (ANN). Finally, a multi-model forecasting ensemble using Prophet, SARIMA, and LSTM will also be developed to forecast the trends of organic loads in the next thirty to ninety days. Results will be presented as compliance alerts, which will be dynamically aligned with the Environmental Quality Act 1974 and the Sabah State Government Policy 2006. Overall, the goal of this study is to design and develop a data-driven system for the identification of critical control points for the treatment of POME and to provide operational insights for mill operators on long-term predictive horizons to increase the environmental sustainability and regulatory compliance of the palm oil industry.