RF-PM10Hybrid Model for Real-Time PM10 Forecasting in Open-Pit Copper Mine: A Case Study at the Sin Quyen Copper Mine

Tóm tắt

Air pollution in open-pit mining areas poses significant environmental and health risks, with particulate matter (PM₁₀) being one of the most critical pollutants. Accurate forecasting of PM₁₀ concentrations is essential for real-time air quality management and dust mitigation strategies. This study develops a machine learning-based framework for PM₁₀ prediction at the Sin Quyen open-pit copper mine, leveraging advanced feature engineering, Principal Component Analysis (PCA), and Synthetic Minority Over-sampling technique for Regression (SMOGN) to enhance model accuracy. Six forecasting models were evaluated, including Random Forest (RF-PM₁₀Hybrid), XGBoost, LightGBM, ARIMA, SARIMA, and Holt-Winters exponential smoothing. The results indicate that machine learning models significantly outperform traditional time-series models with RMSE of 5.791, 8.293, 6.172, 4.233, 11.070, 13.108; MAE of 3.518, 3.953, 3.770, 4.208, 8.800, 10.224; MAPE of 11.70%, 13.18%, 12.57%, 14.03%, 29.32%, 34.07% for the RF-PM₁₀Hybrid, XGBoost, LightGBM, ARIMA, SARIMA, Holt-Winters, respectively. RF-PM₁₀Hybrid achieved the best forecasting performance, with the lowest RMSE (5.791) and MAE (3.518) on the testing dataset, followed by LightGBM and XGBoost. Conversely, statistical models (ARIMA, SARIMA, and Holt-Winters) exhibited higher forecasting errors, making them less suitable for predicting PM₁₀ variations in open-pit mining environments. Key methodological advancements include the integration of lag features, rolling statistics, and interaction terms, which improved the ability of ML models to capture PM₁₀ dynamics. SMOGN was applied to balance the dataset, ensuring better representation of high- PM₁₀ events. The findings demonstrated that machine learning-based approaches, particularly RF-PM₁₀Hybrid, provide a reliable tool for real-time PM₁₀ forecasting, supporting proactive dust control, regulatory compliance, and sustainable mining operations.