Prediction model for dust concentration in open-pit coal mines based on HA-RF-SHAP
In order to effectively predict and control the coal mine dust concentration and protect the health of coal miners and environmental safety,the random forest was used to predict the dust concen-tration,and proposed four heuristic intelligent optimization algorithms were proposed to optimize the hy-perparameters of the random forest,based on the on-site dust monitoring datas of Baorixile open-pit mine,and the model was evaluated through the RMSE,MAE,and Pearson's correlation coefficient R,and the SHAP interpretable model was adopted to analyze the factors affecting dust concentration in open-pit mine.The results show that the optimal models for PM2.5,PM10,and TSP are GWO-RF,WOA-RF,and HHO-RF,respectively;the hyperparameter adjustment improves the model's overall RMSE metrics by about 1～3,the MAE by 1～2.5,and the R by about 4％～6％;the best prediction performance is achieved for PM2.5,with the training set and the test set together having an R of 0.946 3,MAE of 3.059 and RMSE of 4.919,followed by PM10 and TSP;humidity has the greatest effect on the dust concentration in this mine under a single factor,and humidity and barometric pres-sure have the greatest effect on the change of dust concentration under the simultaneous effect of two factors.The study provides an effective dust concentration prediction method,possible to predict the dust concentration accurately and determine the most influential factors of dust,which has an important reference value for mine dust control.
万方数据
维普
