Research On Airport Visibility Prediction And Flight Delay Recovery Based On Improved BP Neural Network

With the rapid development of China’s civil aviation industry and the increasing demand for air travel,the problem of flight safety and punctuality rate has been widely concerned by airports,airlines and passengers.The appearance of low visibility will have an impact on the normal take-off and landing of flights.Therefore,establishing an accurate and effective visibility prediction model can better ensure the normal operation of flights and reduce the losses caused by delays.The main research contents of this thesis are as follows:(1)Conduct preprocessing on the obtained meteorological data and pollutant concentration data,fill in missing values,standardize processing,and perform correlation analysis.The principal component analysis method is used to reduce the dimensions of the data to simplify the model.Through the analysis of the low visibility situation of Jiangbei Airport from 2017 to2021,the yearly,quarterly,monthly,daily changes and duration are obtained.The results show that low visibility often occurs in autumn and winter,with significant seasonality.The occurrence time is concentrated at 0-3 o’clock and 20-23 o’clock,and the duration is mostly 0-2 hours.(2)Build a visibility prediction model.Aiming at the problem that the prediction accuracy of BP neural network is affected by initial weights and thresholds,Genetic Algorithm(GA)is used to optimize the initial weights and thresholds of BP neural network,and a visibility prediction model based on GA-BP neural network is established.Compared with BP neural network prediction model,the average absolute error(MAE)and the root mean square error(RMSE)are improved by 11% and 24% respectively.According to the different distribution of visibility,prediction models are established for autumn-winter and spring-summer to predict visibility,and the results show that the forecasting effect is better in different time periods.(3)In order to improve the prediction accuracy and stability of the GA-BP neural network prediction model,the Ada Boost algorithm is used to combine multiple GA-BP neural network models as weak predictors to construct a strong predictor model.Comparative analysis shows that the Ada Boost-GA-BP forecasting model has better performance than GA-BP forecasting model.(4)Analyze historical delay data of Jiangbei Airport,classify visibility levels based on relevant standards and actual operation conditions,and establish a visibility level prediction model.Statistically analyzes and calculates the time required for the flight delay to recover to normal operation levels after different levels of visibility appear in 2019-2021.Combining recovery time with visibility level prediction models can provide certain decision-making support for relevant departments such as airports and airlines to take corresponding measures before low visibility situations occur.