Forecasting PM_2.5 emissions in open-pit minesusing a functional link neural network optimized by various optimization algorithms

Abstract

PM2.5 air pollution is not only a significant hazard to human health in everyday life but also a dangerous risk to workers operating in open-pit mines OPMs), especially open-pit coal mines (OPCMs). PM2.5 in OPCMs can cause lung-related (e.g., pneumoconiosis, lung cancer) and cardiovascular diseases due to exposure to airborne respirable dust over a long time. Therefore, the precise prediction of PM2.5 is of great importance in the mitigation of PM2.5 pollution and improving air quality at the workplace. This study investigated the meteorological conditions and PM2.5 emissions at an OPCM in Vietnam, in order to develop a novel intelligent model to predict PM2.5 emissions and pollution. We applied functional link neural network (FLNN) to predict PM2.5 pollution based on meteorological conditions (e.g., temperature, humidity, atmospheric pressure, wind direction and speed). Instead of using traditional algorithms, the Hunger Games Search (HGS) algorithm was used to train the FLNN model. The vital role of HGS in this study is to optimize the weights in the FLNN model, which was finally referred to as the HGS-FLNN model. We also considered three other hybrid models based on FLNN and metaheuristic algorithms, i.e., ABC (Artificial Bee Colony)-FLNN, GA (Genetic Algorithm)- FLNN, and PSO (Particle Swarm Optimization)-FLNN to assess the feasibility of PM2.5 prediction in OPCMs and compare their results with those of the HGS-FLNN model. The study findings showed that HGS-FLNN was the best model with the highest accuracy (up to 94–95 % in average) to predict PM2.5 air pollution. Meanwhile, the accuracy of the other models ranged 87 % to 90 % only. The obtained results also indicated that HGS-FLNN was the most stable model with the lowest relative error (in the range of −0.3 to 0.5 %).