This document constitutes Deliverable D2.3 of the project FR8RAIL III Work Package (WP) 2 “Real-Time Network Management” within the framework of the Technology Demonstrator - TD5.2 “Digital Transport Management” of IP5 “Technologies for Sustainable & Attractive European Rail Freight” in Shift2Rail. The report documents the conducted work and results from task T2.3 and is the final deliverable from WP2. 

Three methods and tools have been developed, analysed, and demonstrated on real-world operational data. These tools address three different but equally important aspects for improving the efficiency of real-time network management of railways and contribute to closing the gap between planning and operations by enabling the traffic management to have a more proactive way of working . The focus of the work is on the coordination of freight operations at lines and marshalling yards, namely: • Coordination of all operational activities taking place at arrival/departure yards. • Replanning of timetables for line traffic. • Prediction of system effects and the combined operations of yards and lines. 

Firstly, an integrated demonstration platform for planning operational activities at a marshalling yard is studied. The developed Yard Coordination System (YCS) itself is described as well as how it has been applied and demonstrated in a workshop with experienced participants from the three principally involved stakeholders. The demonstration has shown that a tool like YCS can improve transparency and enable cooperative and pro-active planning. The practitioners reckoned that the tool could prevent and alleviate departure delays, and they expressed a strong wish for continued development of such support. An extensive list of experiences, development suggestions, potentials and risks are reported. 

Secondly, a timetable modification module (TIMO) for short-term replanning of line traffic is evaluated. The method uses a heuristic approach that aims at achieving a high bottleneck robustness, which together with algorithm runtime and several other criteria (train path deviation, change in departure time etc) are used in the evaluation. The effect of several parameters in TIMO are studied, such as iteration settings, size of allowed time windows and share of other train paths that may be adjusted—both for peak and off-peak traffic. Furthermore, how TIMO can be used in an iterative procedure to solve the replanning problem on the line in case of ad-hoc maintenance at the departure marshalling yard is demonstrated. The results show that TIMO’s performance depends greatly on various parameter settings, which delimits the (current) use cases for TIMO. 

Thirdly, a proof-of-concept model framework for increasing the predictability of yard departures and arrivals is evaluated. The model framework incorporates a machine learning-based yard departure deviation prediction model (YPM) into a macroscopic network simulation model (Proton).  Both the infrastructure manager and the yard operator can benefit from this model framework; the former by getting a more realistic picture of the train that runs along the line, the latter by improved yard arrival estimations.  Finally, the possibilities for real-time usage of these methods and tools are discussed along with their respective impact on the three system level performance indicators load factor, punctuality and average (transportation) speed.