Research on Short- and Medium-Term Airport Visibility Prediction Based on the TimeGAN-GA-LSTM Model

Pengcheng Shen

doi:10.54097/6nt94k37

Authors

Pengcheng Shen

DOI:

https://doi.org/10.54097/6nt94k37

Keywords:

Visibility Prediction, TimeGAN, GA-LSTM, Feature Selection.

Abstract

Accurate airport visibility prediction is crucial for flight safety, as reduced visibility may lead to flight delays, cancellations, or even accidents, which impact the efficiency of air transportation. To improve prediction accuracy, this paper proposes a medium-to-short-term visibility prediction model based on TimeGAN-GA-LSTM. Using nearly 10 years of daily meteorological and pollutant data from Xiaoshan International Airport, feature selection is conducted through Spearman correlation analysis. The training dataset is enhanced using TimeGAN, and the LSTM hyperparameters are optimized using a Genetic Algorithm (GA). The results show that under the condition of a 1-day sliding window, the TimeGAN-GA-LSTM model achieves the best prediction accuracy for airport visibility. For a 580-day prediction dataset, the RMSE is 2.01, MAE is 1.49, and MAPE is 16%, which are reductions of 1.01, 0.72, and 15.2%, respectively, compared to the LSTM model, and the R^2 increases to 0.83. This study provides an effective method for intelligent airport visibility prediction, contributing to flight schedule optimization and safety enhancement

References

[1]Zixuan W, Qingchi Y, Zhihong Y, et al. Visibility prediction of plateau airport based on lstm[C]//2021 IEEE 5th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC). IEEE, 2021: 1886-1891.

[2]Kara A, Pekel E, Ozcetin E, et al. Genetic algorithm optimized a deep learning method with attention mechanism for soil moisture prediction[J]. Neural Computing and Applications, 2024, 36(4): 1761-1772.

[3]Chen Y, Dong Z, Wang Y, et al. Short-term wind speed predicting framework based on EEMD-GA-LSTM method under large scaled wind history[J]. Energy Conversion and Management, 2021, 227: 113559.

[4]Wang K, Hua Y, Huang L, et al. A novel GA-LSTM-based prediction method of ship energy usage based on the characteristics analysis of operational data[J]. Energy, 2023, 282: 128910.

[5]Croux C, Dehon C. Influence functions of the Spearman and Kendall correlation measures[J]. Statistical methods & applications, 2010, 19: 497-515.

[6]Zhang Y, Zhou Z, Liu J, et al. Data augmentation for improving heating load prediction of heating substation based on TimeGAN[J]. Energy, 2022, 260: 124919.

[7]Fochesato M, Khayatian F, Lima D F, et al. On the use of conditional TimeGAN to enhance the robustness of a reinforcement learning agent in the building domain[C]//Proceedings of the 9th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation. 2022: 208-217.

[8]Han W, Ying W, Hu W, et al. Time series anomaly detection based on TimeGAN and LSTM neural network[C]//2023 Eleventh International Conference on Advanced Cloud and Big Data (CBD). IEEE, 2023: 122-127.

[9]Liu Y, Liang Z, Li X. Enhancing short-term power load forecasting for industrial and commercial buildings: A hybrid approach using TimeGAN, CNN, and LSTM[J]. IEEE Open Journal of the Industrial Electronics Society, 2023, 4: 451-462.

[10]Kassa Y, Zhang J H, Zheng D H, et al. A GA-BP hybrid algorithm based ANN model for wind power prediction[C]//2016 IEEE Smart Energy Grid Engineering (SEGE). IEEE, 2016: 158-163.