Current state and prospects of development of energy-optimal control systems for 2ES6 electric locomotives

Abstract

Introduction. The research focuses on the current state and prospects of development of the systems of energyoptimal train driven by freight main line DC electric locomotives 2ES6. An analysis of current trends in energy saving and increasing the efficiency of traction energy resources in railway transport and their impact on haulage of trains shows that train guidance based on machine learning and artificial intelligence remains poorly researched. The study is primarily intended to determine the actual use of automation of goods train driving in the sections of the Ural-Siberian railway proving ground and its impact on the energy-optimal schedule of completed train operations.Materials and methods. The problem solving involved the basic provisions of the theory of haulage of trains, concepts of the theory of automated control and diagnostics of electric rolling stock, as well as statistical methods of data processing.Results. The authors hypothesise that a smart adaptive rolling stock control support system with machine learning and AI would reduce the specific power consumption of locomotives. The researchers show that the most feasible way to build real-time dynamic models of energy-optimal locomotive motion for such smart system is to use data from the automated workstation of a freight locomotives motion recorder and auto-drive, as this is the data that contains accurate geographic coordinates to synchronise measurements on trips in a particular section.Discussion and conclusion. A tunable artificial recurrent neural network on long short-term memory in new or existing improved methods for energy-efficient train rolling stock control would improve the motion recorders used on locomotives. The developed algorithm may form the basis of a fundamentally new smart adaptive rolling stock control support system with machine learning and AI. Further research would be focused on the development of technology for building dynamic models of energy-optimal real-time locomotive movement with train.

Keywords

• electric rolling stock
• auto-driving
• energy-optimal control
• motion recorders
• dynamic models
• machine learning
• artificial intelligence