Long sequence time-series forecasting (LSTF) is required in practical application scenarios of early warning for key equipment failures of railway high-speed trains. Such as predicting the temperature of the stator of a traction motor to issue early warning of failures, in order to guide the maintenance and repair of key equipment. Recent research in the field of artificial intelligence shows that Convolutional Neural Network (CNN) and Recurrent neural network (RNN) have certain time series prediction capabilities, but there are also problems with insufficient accuracy and excessive computing power when predicting LSTF. This paper adopts a self-attention model that includes multi-head self-attention, position coding, encoder-decoder, and adjusts the hyperparameters to make the model perform better. Compares it with other models based on real data from the high-speed railway operation site. The results show that the modsel used in this paper can predict changes of sequence data in the future, and has better parallel computing and computing cost.
A Self-attention Model for High-Speed Railway Trains Key Equipment Long Sequence Time-Series Data Fault Warning Based on Data-Driven
Lect. Notes Electrical Eng.
International Conference on Electrical and Information Technologies for Rail Transportation ; 2021 October 21, 2021 - October 23, 2021
Proceedings of the 5th International Conference on Electrical Engineering and Information Technologies for Rail Transportation (EITRT) 2021 ; Chapter : 63 ; 629-645
2022-02-19
17 pages
Article/Chapter (Book)
Electronic Resource
English
British Library Conference Proceedings | 2022
|High-speed railway trains in Germany
Engineering Index Backfile | 1905
Railway Noise Generated by High-Speed Trains
Springer Verlag | 2015
|Bogie for high-speed railway freight trains
Tema Archive | 1995