BỘ KHOA HỌC VÀ CÔNG NGHỆ
HỆ THỐNG THÔNG TIN KHOA HỌC VÀ CÔNG NGHỆ

Lọc theo danh mục
  • Năm xuất bản
    Xem thêm
  • Lĩnh vực
liên kết website
Lượt truy cập
 Lượt truy cập :  14,856,450

BB

Xuan Nam Bui, Hoang Nguyen, Qui Thao Le , Tran Quang Hieu; Trần Quang Hiếu(1); 

Application of artificial neural network with fine–tuning parameters for forecasting PM2.5 in deep open–pit mines: A case study

Application of artificial neural network with fine–tuning parameters for forecasting PM2.5 in deep open–pit mines: A case study

Vietnam Journal of Hydrometeorology

2022

No-10

64-72

In this paper, an artificial neural network (ANN) model was applied to forecast PM2.5 at the Coc Sau open–pit coal mine (Northern Vietnam) with fine–tuning parameters. It aims to provide the feasibility and insights into controlling air quality in open–pit mines using artificial intelligence techniques. Accordingly, an air quality monitoring system was established to monitor hourly PM2.5 datasets for more than three months. Subsequently, 80% of the whole data was used to design and tune the ANN model, and the remaining 20% was used for testing the PM2.5 predictions. An ANN model with a single hidden layer and ten nodes was developed for this aim. The stochastic gradient descent algorithm was applied to train the ANN model under the learning rate of 0.001 to avoid the overfitting of the model. In addition, 10 time steps (multi–step forecasting model) were applied to forecast the next time step. The results indicated that ANN is a potential model for forecasting PM2.5 in open–pit mines with high accuracy (RMSE = 2.000), and it can be used to control real–time air quality in open–pit mines.

In this paper, an artificial neural network (ANN) model was applied to forecast PM2.5 at the Coc Sau open–pit coal mine (Northern Vietnam) with fine–tuning parameters. It aims to provide the feasibility and insights into controlling air quality in open–pit mines using artificial intelligence techniques. Accordingly, an air quality monitoring system was established to monitor hourly PM2.5 datasets for more than three months. Subsequently, 80% of the whole data was used to design and tune the ANN model, and the remaining 20% was used for testing the PM2.5 predictions. An ANN model with a single hidden layer and ten nodes was developed for this aim. The stochastic gradient descent algorithm was applied to train the ANN model under the learning rate of 0.001 to avoid the overfitting of the model. In addition, 10 time steps (multi–step forecasting model) were applied to forecast the next time step. The results indicated that ANN is a potential model for forecasting PM2.5 in open–pit mines with high accuracy (RMSE = 2.000), and it can be used to control real–time air quality in open–pit mines.

Xem toàn văn