Accurate long-term dust concentration prediction in open-pit mines: A novel machine learning approach integrating meteorological conditions and mine production intensity

Abstract

Amidst the transition from rapid growth to high-quality development in the surface mining industry, mine dust remains a severe public health and safety issue. This study introduces a novel approach that combines mine production intensity with meteorological conditions to enable accurate daily prediction of dust concentration in open-pit mines. To achieve this, the study collected meteorological factors and production intensity parameters spanning 365 days from the Haerwusu open-pit coal mine in China. To enhance the accuracy of predictions, six heuristic algorithms were employed to optimize the hyperparameters of the random forest algorithm. During the prediction process, the results of PM2.5, PM10, and TSP were compared, and the model performance with and without mine operational parameters was analyzed. The findings demonstrate significant advancements in predicting three distinct indicators through the proposed method. Specifically, the research successfully predicted mine dust concentration on a daily basis over a long-term period. The introduction of production intensity parameters ensures the accuracy and reliability of dust concentration prediction, and the availability of future meteorological factors enables a realistic prediction of dust concentration in open-pit mines. These outcomes carry substantial practical implications, including safeguarding worker health, optimizing mine operation plans, and enhancing coal production efficiency.