Visibility is one of the most important standards for airport flight operations.To accurately predict visibility and ensure the safe as well as normal operation of air traffic,a new visibility prediction model based on Variance Inflation Factor(VIF),Principal Components Analysis(PCA),and Informer was constructed using the meteorological and air quality monitoring data of Tianjin Airport in 2020.A random forest method was used to fill in the missing data,VIF was used for feature screening,PCA was used for feature dimension reduction,and informer was used for prediction.Root Mean Square Error(RMSE),Mean Absolute Error(MAE),and Mean Absolute Percentage Error(MAPE)were used to evaluate the predicted visibility.The results show that the data based on VIF and PCA feature screening and dimension reduction can largely retain the information of the original data.The VIF-PCA-Informer model outperformed single Informer and simple combined models in predicting visibility.The RMSE,MAE,and MAPE of the VIF-PCA-Informer model are 0.568 8,0.505 4,and 0.787 0,respectively.Compared with the Informer model without feature screening and dimension reduction,the RMSE,MAE,and MAPE of the VIF-PCA-Informer model were decreased by about 0.214 1,0.184 2,and 0.322 4,respectively.The trend of the prediction curve is closer to the actual value.Compared with the Long Short-Term Memory(LSTM)model,the VIF-PCA-Informer model can better capture the relationship between long-time series features.The RMSE,MAE,and MAPE of the VIF-PCA-Informer model are about 0.348 6,0.275 3,and 0.527 0 lower than those of the VIF-PCA-LSTM model,respectively.Compared with the Gated Recurrent Unit(GRU)model,the RMSE,MAE,and MAPE of the VIF-PCA-Informer model were reduced by about 0.297 8,0.233 2,and 0.358 0 respectively.The VIF-PCA-Informer model was used to predict visibility from 1 to 4 hours respectively.It was found that the error increased as the time interval increased time.The VIF-PCA-Informer,which predicted visibility for 4 hours,was still better than the VIF-PCA-LSTM and VIF-PCA-GRU,which predicted visibility for 1 hour.The efficient airport visibility prediction model will play a great supporting role in ensuring the safe and efficient operation of flights.
NETL
NSTL
万方数据
维普
