Simulazione, ottimizzazione e gestione delle reti ferroviarie

Chapter 1: Introduction ................................................................................... 1

1.1 Research relevance and contributions ...................................................... 1

1.2 Thesis outline ........................................................................................... 5

Chapter 2: Literary review .............................................................................. 7

2.1 Simulation models for transport systems ................................................. 7

2.1.1 Delimitation of the study area ........................................................... 8

2.1.2 Zoning ................................................................................................ 8

2.1.3 Relevant infrastructures and services ............................................. 10

2.1.4 The supply model ............................................................................. 11

2.1.5 The demand model ........................................................................... 18

2.1.6 The assignment models .................................................................... 23

2.2 Rail simulation models ........................................................................... 29

2.3 Estimation techniques for travel demand flows ..................................... 42

2.4 Simulation algorithms ............................................................................ 51

2.5 Optimisation models for transport systems ............................................ 61

2.6 Rail optimisation models ........................................................................ 63

2.6.1. The timetabling phase .................................................................... 64

2.6.1.1 Dwell times estimation techniques ........................................... 68

2.6.2. The rescheduling problem .............................................................. 73

2.7 Optimisation algorithms ......................................................................... 83

2.8 Energy issues related to rail systems ...................................................... 95

Chapter 3: Proposed methodology for managing rail systems both in

ordinary and disruption conditions ............................................................ 104

3.1 Optimisation framework ....................................................................... 106

3.2 Basic simulation architecture ............................................................... 109

I

3.3 Extended simulation architecture ......................................................... 111

3.3.1 Stochastic simulation framework .................................................. 112

3.3.2 Decision support system for implementing energy saving strategies

................................................................................................................ 113

3.3.3 Modelling of the snowball effect ................................................... 121

3.3.4 Travel demand estimation ............................................................. 131

3.3.4.1 Analytical methodology for extending passenger counts ....... 133

3.3.4.2 A long-term evaluation of travel demand ............................... 140

3.4 Concluding remarks ............................................................................. 152

Chapter 4: Applications to real network contexts of the proposed approach

........................................................................................................................ 155

4.1 Case studies .......................................................................................... 155

4.2 Rescheduling applications .................................................................... 162

4.2.1 Evaluation of unconventional rescue strategies for managing

disruption conditions .............................................................................. 178

4.3 Energy saving policies applications ..................................................... 183

4.4 Planning tasks: estimation of dwell times as flow-dependent factors .. 189

4.4.1 A comparison between FIFO and RIFO queuing rules................. 202

4.5 Travel demand estimation applications ................................................ 210

4.5.1 Calibration and validation of space-time relations representative of

passenger flow data ................................................................................ 210

4.5.2 A cost-benefit analysis relative to the implementation of an innovative

signalling system in a regional context .................................................. 216

Chapter 5: Conclusion ................................................................................. 227

5.1 Resume of the main achievements ....................................................... 227

5.2 Research prospects ............................................................................... 229

II

Dissemination of the main research achievements .................................... 232

References ..................................................................................................... 234

III

C 1: I

HAPTER NTRODUCTION

1.1 Research relevance and contributions

The proposed research is situated in the field of design, management and

optimisation in railway network operations. Rail transport has in its favour several

specific features which make it a key factor in public transport management,

above all in high-density contexts. Indeed, such a system is environmentally

friendly (reduced pollutant emissions), high-performing (high travel speeds and

low values of headways), competitive (low unitary costs per seat-km or carried

passenger-km) and presents a high degree of adaptability to intermodality.

However, it manifests high vulnerability in the case of breakdowns. This occurs

because a faulty convoy cannot be easily overtaken and, sometimes, cannot be

easily removed from the line, especially in the case of isolated systems (i.e.

systems which are not integrated into an effective network) or when a breakdown

occurs on open tracks. Thus, re-establishing ordinary operational conditions may

require excessive amounts of time and, as a consequence, an inevitable increase

in inconvenience (user generalised cost) for passengers, who might decide to

abandon the system or, if already on board, to exclude the railway system from

their choice set for the future. It follows that developing appropriate techniques

and decision support tools for optimising rail system management, both in

ordinary and disruption conditions, would consent a clear influence of the modal

split in favour of public transport and, therefore, encourage an important reduction

in the externalities caused by the use of private transport, such as air and noise

pollution, traffic congestion and accidents, bringing clear benefits to the quality

of life for both transport users and non-users (i.e. individuals who are not system

users).

Managing to model such a complex context, based on numerous interactions

among the various components (i.e. infrastructure, signalling system, rolling stock

and timetables) is no mean feat. Moreover, in many cases, a fundamental element,

which is the inclusion of the modelling of travel demand features in the simulation

of railway operations, is neglected. Railway transport, just as any other transport

1

system, is not finalised to itself, but its task is to move people or goods around,

and, therefore, a realistic and accurate cost-benefit analysis cannot ignore involved

flows features. In particular, considering travel demand into the analysis

framework presents a two-sided effect.

Primarily, it leads to introduce elements such as convoy capacity constraints and

the assessment of dwell times as flow-dependent factors which make the

simulation as close as possible to the reality. Specifically, the former allows to

take into account the eventuality that not all passengers can board the first arriving

train, but only a part of them, due to overcrowded conditions, with a consequent

increase in waiting times. Due consideration of this factor is fundamental because,

if it were to be repeated, it would make a further contribution to passengers’

discontent. While, as regards the estimate of dwell times on the basis of flows, it

becomes fundamental in the planning phase. In fact, estimating dwell times as

fixed values, ideally equal for all runs and all stations, can induce differences

between actual and planned operations, with a subsequent deterioration in system

performance. Thus, neglecting these aspects, above all in crowded contexts, would

render the simulation distorted, both in terms of costs and benefits.

The second aspect, on the other hand, concerns the correct assessment of effects

of the strategies put in place, both in planning phases (strategic decisions such as

the realisation of a new infrastructure, the improvement of the current signalling

system or the purchasing of new rolling stock) and in operational phases

(operational decisions such as the definition of intervention strategies for

addressing disruption conditions). In fact, in the management of failures, to date,

there are operational procedures which are based on hypothetical times for

re-establishing ordinary conditions, estimated by the train driver or by the staff of

the operation centre, who, generally, tend to minimise the impact exclusively from

the company’s point of view (minimisation of operational costs), rather than from

the standpoint of passengers. Additionally, in the definition of intervention

strategies, passenger flow and its variation in time (different temporal intervals)

and space (different points in the railway network) are rarely considered. It

2

appears obvious, therefore, how the proposed re-examination of the dispatching

and rescheduling tasks in a passenger-orientated perspective, should be

accompanied by the development of estimation and forecasting techniques for

travel demand, aimed at correctly taking into account the peculiarities of the

railway system; as well as by the generation of ad-hoc tools designed to simulate

the behaviour of passengers in the various phases of the trip (turnstile access,

transfer from the turnstiles to the platform, waiting on platform, boarding and

alighting process, etc.).

The latest workstream in this present study concerns the analysis of the energy

problems associated to rail transport. This is closely linked to what has so far been

described. Indeed, in order to implement proper energy saving policies, it is, above

all, necessary to obtain a reliable estimate of the involved operational times

(recovery times, inversion times, buffer times, etc.). Moreover, as the adoption of

eco-driving strategies generates an increase in passenger travel times, with

everything that this involves, it is important to investigate the

trade-off between energy efficiency and increase in user generalised costs.

Within this framework, the present study aims at providing a DSS (Decision

Support System) for all phases of planning and management of rail transport

systems, from that of timetabling to dispatching and rescheduling, also

considering space-time travel demand variability as well as the definition of

suitable energy-saving policies, by adopting a passenger-orientated perspective.

Therefore, the provided contributions can be outlined as follows.

• Creating a dynamic database representing a decision-making tool for

assisting dispatchers in handling both ordinary and disruption conditions.

In particular, for each possible intervention strategy, related or not to a

specific failure event, such database provides the identification and the

quantification of relevant impacts on each part of the analysed system. In

this way, dispatchers can be fully aware of the consequences of their own

decisions and, thus, face the perturbed conditions in an appropriate

manner, never opting again for the non-intervention strategy; moreover,

3

response times can be made comparable with real-time rescheduling

approaches, without, however, the computational effort they require.

• Developing an analytical framework which allows an accurate estimation

of operational times within timetable as a support tool for the

implementation of eco-driving strategies. Indeed, such policies imply an

increase in travel times and, therefore, result feasible exclusively in the

event of extra time rates available, which have to be suitably designed

during the timetabling process.

• Defining a simulation-based methodology for computing dwell times as

flow-dependent factors, rather than as fixed values. This task is

fundamental in order to design a robust timetable, with a high degree of

resilience to delays, and grows in importance in overcrowded contexts.

Indeed, the dynamic interaction between rail service and passengers flows,

which occurs on the interface platform-train, gives rise to the so called

snowball effect: the number of passengers on the platform influences the

dwell times of trains at stations, which may cause delays; these, in turn,

produce an increase in headways which generates more passenger flows

on the platform providing a further extension of dwell times and, therefore,

additional delays. In particular, two different boarding behavioural

patterns (i.e. FIFO and RIFO) are modelled and compared in terms of

effects on rail service and passenger satisfaction.

• Customising travel demand estimation and forecasting techniques

proposed in the literature to the specific features of rail transport, related

to the discontinuous fruition in space and time which it offers. The

relevance of this lies in the fact that, each planning task, both in the case

of short and long term policies, requires an estimation of involved

passenger flows as input information.

1.2 Thesis outline

This section provides a brief foreword to each chapter of the presented work. 4

Chapter 2 is focused on the literary review of research fields of concern.

Specifically, the comprehensive nature of the proposed approach gives rise to the

necessity of investigating a wide range of operational issues related to planning

and management tasks in rail transport. Therefore, after a general analysis of

simulation and optimisation models adopted for transportation systems, a focus

on such techniques in the case of rail systems is provided. Additionally, both

simulation and optimisation algorithms are described. Moreover, the estimation

and forecasting techniques for travel demand are evaluated, with the aim of

adapting them to the peculiarities of rail systems. Finally, an analysis of the main

issues related to the application of energy savings policies in the rail field is given,

with a focus on the existing deep relationship between eco-driving strategies and

operational parameters within the planned timetable.

Chapter 3 describes the developed decision support tool which is based on suitable

simulation models, properly integrated into an optimisation layout. In particular,

it is possible to define a basic simulation structure which is improved and made

more accurate by means of the development of methodological frameworks

enabling the modelling of crucial operational factors, such as stochasticity of rail

operations, the interaction between rail service and travel demand as well as

energy saving issues. The adopted perspective is

passenger-centric which means that the goal is to improve service quality so as to

drive the modal split towards systems based on railway technology which is

sustainable and high-performing.

Chapter 4 aims at pointing out the effectiveness of the proposed methodology, by

applying it to real network contexts. In particular, most of the presented

applications are focused on metro systems which, generally, operate in

high-density conditions and, frequently, have to address overcrowded situations.

Therefore, in such circumstances, the necessity of properly modelling the

interaction between rail service and travel demand, as well as the need for ensuring

a certain service quality, grow in importance. The second case-study is

represented by a regional rail line, with the aim of showing the capacity of the

5

proposed approach of dealing with different network contexts. Clearly, the

differences between the two analysed systems have been duly taken into account.

Indeed, a metro service is affected by urban user flows, while a regional network

has to deal with extra-urban (i.e. rural) trips; moreover, the former are frequency-

based, while the latter operates according to specific departure/arrival times at

each station dictated by the planned timetable.

Finally, concluding remarks and research prospects are provided in chapter 5. 6

C 2: L R

HAPTER ITERARY EVIEW

The proposed framework for managing railway systems is characterised by a

simulation-optimisation integrated approach and, therefore, in this chapter both

simulation and optimisation models presented in the literature, with related

resolutions methods, are investigated. In particular, after an analysis concerning

transportation systems in general, a deepening of railway contexts is carried out.

Moreover, given the crucial role played in this work by the travel demand, related

estimation techniques are assessed with the aim of customising them to the railway

case. Finally, environmental issues relative to railway systems are described with

particular attention to energy saving strategies involving the design of eco-driving

profiles and the adjustment of operational times within the planned timetable.

2.1 Simulation models for transport systems

Transport systems are made up of physical and organisational elements which

interact with each other to produce transport opportunities and satisfy travel

demand which, in turn, is the result of the interactions among the various social

and economic activities localised in a specific area.

Mathematical models concerning transport systems aim at simulating the

interaction between demand flows and supply performance, both for existing

contexts (operational phase) and hypothetical ones (planning phase). Therefore,

such models, and the different techniques which they make use of, are

fundamental tools for the assessment and/or design of interventions concerning

physical (e.g. a new railway line) and/or functional elements (e.g. a new railway

timetable) of a transport system. According to the analysed context, the elements

considered relevant to the problem ar