Optimization of coasting points in a mass rail transit system

Abstract

Energy consumption of a rail transit system depends on many parameters such as train weight, maximum speed, power supply system voltage level, and operation concepts. One of the most effective methods of reducing energy consumption in a rail transit system is optimizing the speed profile of the trains along the route: Trains consume the maximum energy during flat-out mode operation where they speed up to the maximum speed and keep that speed until it reaches to the braking point which is determined by deceleration rate. This type of operation gives the shortest journey time. A small trade-off from this journey time gives high saving in energy consumption. This subject poses an optimization problem which could be very complicated. In this study, a new efficient method will be presented for optimization of the coasting points for trains in a global manner. The new method suggest using Artificial Neural Networks (ANN) together with classical approach of simulator tool and genetic algorithms. Trains run along the line according to a timetable. Timetables define the travelling time for every train from every station to station. Timetables always include some slack time for an unexpected time loss which could be caused by faulty equipment, but mostly by passengers. Slack times and station dwell times are very important for providing a punctual service. Delays disturb the punctual operation as well as reducing energy efficiency by consuming the slack times which can be used in normal operation conditions for energy efficient driving. Simulation software is used for creating training and test data for ANNs. These data are used for training of ANNs. Test data are used to validate the outputs of trained ANNs. The trained ANNs are then used for estimating energy consumption and travel time for new sets of coasting points. Finally, the outputs of ANNs are optimized to find optimal train coasting points. For this purpose, a fitness function with target travel time, energy consumption and weighting factors is proposed. Genetic algorithms were used for this search purposes. An interesting observation is that the use of ANNs increases the speed of optimization, and gives researchers the ability to test different optimization with differing genetic algorithm parameters. Proposed method is used for optimizing coasting points for minimum energy consumption for a given travel time of first 10 km section of Istanbul Aksaray- Airport metro line, where trains operate every 150 seconds. The section covers 9 passenger stations, which means 8 coasting points for each line. It has been demonstrated that an 16 input ANN can be trained with acceptable error margins for such a system .The optimization method finds the near optimum points for different target travel times and weight factors of the fitness function. It was found for the given line configurations that the energy saving potential with coasting schemes for the same amount of time increase is far less in the multi-train case, where trains regenerate and feed each other. In the 3 station case, 4.81% increase in travel time with optimum coasting points compared to the flat-out case creates 30.85% energy saving, whereas in the 5 station case, 4.65% increase in travel time with optimal coasting scheme creates only 18.25% energy saving. Using ANN and GA in combination speeds up optimization, and bigger line segments with more stations can also be optimized with this proposed method. A simple way of application of the proposed method is using a Driver Information System (DIS). Aksaray – Airport LRT system has been recently equipped with such a system. It is planned to find the optimum coasting points for different headways, and enter these values into the DIS as time and location dependent values. The DIS will give a warning to the driver for start of coast at these pre-defined locations. It should be noted, however, that some operational parameters are changing dynamically in real life. The train weights and the station waiting times, for instance, affect the energy consumption and the travel time. Authors are aware of staggering difficulties in finding optimum coasting points online for such operational variants. Nevertheless, the paper reveals the advantages of using ANN, and its possible application to optimization of coasting points for the case of multiple stations and multiple lines, and hopefully paves the road for future research in this direction.

Özet

Bir raylı taşıma sisteminde enerji tüketimi birçok farklı parametreye bağlıdır. Ancak, bir raylı sistemde tren rotası boyunca hız profilini en uygun hale getirmek, enerji tüketimini azaltmada en etkili yöntemlerden biridir. Makalede trenler için boşta gitme noktalarının optimize edilebilmesi için yeni geliştirilmiş bir yöntem sunulmaktadır. Çalışmada önerilen yaklaşım GA (Genetik Algoritma), YSA (Yapay Sinir Ağları), ve çok hatlı ve çok trenli sistem simulasyon yazılımının birlikte kullanımıyla gerçekleştirilmiş olan gerçekçi sistem modellemesini içermektedir. Simulasyon yazılımı araçların regeneratif frenleme yapabilme ve düşük gerilimdeki performansını da modellemektedir. Simulasyon yazılımı, YSA için eğitim ve test verilerinin oluşturulmasında kullanılmıştır. Bu veriler YSA’ların eğitiminde ve bu eğitilmiş YSA’lar ise değişik boşta gitme konumları için yolculuk süresi ve enerji tüketimini tahmin etmede kullanılmıştır. Optimizasyon kısmında hedef yolculuk süresi, hedef enerji tüketimi ve ağırlık faktörleri içeren bir uygunluk fonksiyonu sunulmuş ve YSA’lar bu uygunluk fonksiyonunu hesaplayan optimizasyon prosedürünün hızını arttırmada kullanılmıştır. Uygunluk fonksiyonunu minimize eden en uygun boşta gitme noktalarının belirlenmesinde GA araştırma yöntemi kullanılmıştır. Gerek GA’nın gerekse kullanılan uygunluk fonksiyonunun değişik parametreleri için optimizasyon çalışmalarının tekrar edilmesi işlemi mevcut literatürde bulunan yöntemlerle çok fazla zaman almaktadır. Çalışmada önerilen yöntemde eğitilmiş olan YSA’ların kullanılması ile parametre değişiklikleri için tekrarlanan optimizasyon çalışmlarında simülatörün kullanılmasına gerek kalmamakta, dolayısı ile yeni parametreler için optimizasyon sonuçları çok hızlı elde edilebilmektedir.