Latency Compensation and Prediction for Wireless Train to Ground Communication Network Based on Hybrid LSTM Model
Jiajun Duan, Xiong Wei, Jiahua Zhou, Tingting Wang, Xuechao Ge, Zhi Wang
Abstract
This paper proposes a novel method compensate for the communication imperfection of wireless train-to-ground communication network (WT2GCN). WT2GCN has an unavoidable transmission delay, which wastes the expensive investment of wired network and impacts the economic operation of the energy management for the rail transit system. Instead of upgrading the WT2GCN, which will lead to unnecessary costs, this paper has developed an accurate prediction method to forecast the WT2GCN measurement based on the long short-term memory (LSTM) and high-degree polynomial linear regression (HPLR) technique. To capture different characteristics of physical parameters, multiple LSTM and HPLR models are specifically designed to predict various measurements. In addition, the attention mechanism is applied to help exact the key information of features. Then, a steady-state model is derived for a priori evaluation of WT2GCN measurement prediction with non-delayed electric measurements. A detailed implementation illustration of the proposed methodology is also presented. Finally, the developed platform is tested with the actual data and implemented in one practical rail transit project of China Metro. The real-time operating result has fully demonstrated the effectiveness of the proposed method.