Coasting point optimisation for mass rail transit lines using artificial neural networks and genetic algorithms


Acikbas S., Soylemez M. T.

IET Electric Power Applications, vol.2, no.3, pp.172-182, 2008 (Journal Indexed in SCI Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 2 Issue: 3
  • Publication Date: 2008
  • Doi Number: 10.1049/iet-epa:20070381
  • Title of Journal : IET Electric Power Applications
  • Page Numbers: pp.172-182

Abstract

Energy consumption of a rail transit system depends on many parameters. One of the most effective methods of reducing energy consumption in a rail transit system is optimising the speed profile of the trains along the route. A new efficient method will be presented for the optimisation of the coasting points for trains in a global manner. The proposed approach includes realistic system modelling using multi-train, multi-line simulation software and application of artificial neural networks (ANN) and genetic algorithms (GA). The simulation software used can model regenerative braking and train performance at low voltages. Using ANN and GA together, optimal coasting points for long line sections covering five stations and two lines are achieved. Simulation software is used for creating training and test data for the ANN. These data are used for training of the ANN. Trained ANNs are then used for estimating energy consumption and travel time for new sets of coasting points. Finally, the outputs of the ANN are optimised to find optimal train coasting points. For this purpose, a fitness function with target travel time, energy consumption and weighting factors is proposed. An interesting observation is that the use of ANN increases the speed of optimisation. The proposed method is used for optimising coasting points for minimum energy consumption for a given travel time on the first 5 km section of Istanbul Aksaray-Airport metro line, where trains operate every 150 s. The section covers five passenger stations, which means four coasting points for each line. It has been demonstrated that an eight input ANNs can be trained with acceptable error margins for such a system. © The Institution of Engineering and Technology 2008.