Within the European Shift2Rail research project, a macroscopic simulation tool was developed in the sub-projects Plasa and Plasa2. Development was done by DB Analytics,which is part of DB (Deutsche Bahn) and one of the main goals was that the tool shouldbe able to simulate large networks in short computation time. Both DB and the SwedishTransport Administration (Trafikverket), along with several others participated in the Plasaand Plasa2 projects. The tool is named PROTON (Punctuality and Railway OperationSimulaTiON), it was formerly known under the name PRISM (Plasa Railway InteractionSimulation Model).

Trafikverket has an intention of introducing PROTON as an in-house tool and to increasethe use of simulation as a method for analysing for example proposed future timetables.Trafikverket is currently using RailSys which is a microscopic timetable and simulationsoftware. The intention is to use macrosimulation in applications where microsimulation isimpractical or infeasible to use, typically in large area or even network wide simulations.

SIMPOR is a project with the aim of using PROTON in different types of Swedish applications and it is carried out within KAJT (Kapacitet i järnvägstrafiken), Capacity in theRailway Traffic System which is a research program for improved railway system performance financed by Trafikverket. Most of the applications where PROTON has been usedhas been carried out in combination with other projects, such as FR8Rail2 and FR8Rail3,in which the use cases have been formulated. This report describes briefly PROTON andthe input data needed for running simulations, and summarizes results from the differentother projects where PROTON was used.

DB has further developed PROTON into a microscopic simulation tool, and it is still underdevelopment. However, no microscopic application with PROTON has yet been performedor tested in Sweden. Consequently, this report deals only with the macroscopic part ofPROTON. The microscopic development of PROTON is briefly mentioned in section 5.

This thesis covers applications and proposes methods for using simulation in a more effectiveway and also in a wider context than normally used. One of the proposed methods deals withdelay modeling that can be used in a calibration process. Furthermore, a method is presentedthat facilitates the management of having timetables, infrastructure scenarios and delays asvariables in simulation studies. The simulation software used in this thesis is RailSys, whichuses a microscopic formulation to describe the infrastructure and train movements.Timetable changes with respect to allowances and buffer times are applied on a real case(Western Main Line) in Sweden in order to analyze how the on-time performance is affectedfor high-speed passenger trains. The potential benefit is that increased allowances and buffertimes will decrease the probability of train interactions and events where the scheduled trainsequence is changed. The on-time performance improves when allowances are increasedand when buffer times concerning high-speed trains are adjusted to at least five minutes inlocations with potential conflicts. One drawback with this approach is that it can consumemore space in a timetable at certain locations, hence other trains may need adjustments inorder reach these buffer times.Setting up simulations, especially in large networks, can take significant amount of timeand effort. One of the reasons is that different types of delay distributions, representingprimary events, are required in order to obtain conformity with reality if a real timetable andnetwork is modeled. Considering train registration data in Sweden, the separation in primaryand secondary delays is not straightforward. The presented method uses the basic trainregistration data to compile distributions of run time deviations for different train groups ina network. The results from the Southern Main Line case study show that a reasonable goodfit was obtained, both for means and standard deviations of delays. A method for capturingthe variance in freight train operations is proposed, partly based on the findings from theaforementioned study. Instead of modeling early freight trains on time, the true initiationdistributions are applied on time-shifted freight trains.In addition to the already mentioned methods, which are applied on real networks, a methodfor reducing the uncertainties coming from assumptions of future conditions is proposed. It isbased on creating combinatorial departure times for train groups and locations and formulatingthe input as nominal timetables to RailSys. The dispatching algorithm implementedin the software can then be utilized to provide feasible, conflict-managed, timetables whichcan be evaluated. This can be followed by operational simulations with stochastic delays ona subset of the provided timetables. These can then consequently be evaluated with respectto mean delays, on-time performance etc.To facility the use of the infrastructure as a variable in these type of studies, an infrastructuregenerator is developed which makes it relatively easy to design different station layouts andproduce complete node-link structures and other necessary definitions. The number, locationand type of stations as well as the linking of stations through single-track or multi-tracksections can be done for multiple infrastructure scenarios. Although the infrastructure canbe defined manually in RailSys, a considerably amount of time and effort may be needed.In order to examine the feasibility of this method, case studies are performed on fictive linesconsisting mostly of single-track sections. This shows that the method is useful, especiallywhen multiple scenarios are studied and the assumptions on timetables consist of departureintervals for train groups and their stop patterns.