This thesis is part of a project between the PoliMi and Trenitalia to develop a DDS to optimize the integration of the Italian railway network with HS services, through transport simulations and «what if» scenarios. The work will focus on the development of a Travel Demand model with a schedule-based approach, capable of a path-based modelling. To this aim, an extensive literature review has been carried out, highlighting why it is important to use a, not so common, Schedule-based approach on low-frequency services. Among the others, one of the main innovations is the use of new mobile network data as main data source to build the Demand Model, offering the opportunity to explore travel demand in ways impossible with conventional surveys since this type of approach is new to the literature in this field on such a large scale. The concept of time-expanded network (Diachronic Graph) is used to model the supply system of the entire nation, also considering a road auxiliary graph, other than the railway infrastructure. A new hybrid method has been developed for the demand side of the model, able to estimate HSR OD flows by re-weighting Station-Station observed data. To run a SNL assignment, a new complete procedure has been developed, requiring the simultaneous interaction of multiple software. Flows have been hierarchically analyzed, starting from an aggregated view, and deepening more and more on a disaggregated level, always looking at the level of correlation between modelled and observed flows. The results show a strong correlation at aggregated level and for station-station pairs, with R2 values of 0.98 for the former and 0.94 for the latter. The coefficient remains still high mainly for important breakdowns: peak-hour and medium-long distance classes, with an average value of 0.71. At train link level it drops at 0.12, symptom of too much noise in the model, which leaves room for improvement. Together with the overall complexity of the model, calculation times have unexpectedly become one of the many challenges to overcome, since the work aims to obtain new theories for modelling public transport that can make a contribution to the scientific community.
Questa tesi fa parte di un progetto tra PoliMi e Trenitalia per lo sviluppo di un DDS per ottimizzare l'integrazione della rete ferroviaria italiana con i servizi AV, con simulazioni e scenari «what if». Il lavoro si concentra sullo sviluppo di un modello Travel Demand con un approccio ad orario, in grado di modellare l'intero percorso. A tal fine, è stata condotta un'ampia revisione della letteratura, mettendo in evidenza i motivi per cui è importante utilizzare un approccio ad orario per i servizi a bassa frequenza. Una delle principali innovazioni è l'utilizzo dei dati di reti cellulari come fonte principale, per costruire il modello di domanda, offrendo l'opportunità di esplorare la domanda di viaggio in modi impossibili con le indagini convenzionali. Il concetto di grafo diacronico è utilizzato per modellare la rete di trasporto dell'intera nazione, considerando anche un grafo ausiliario stradale, oltre all'infrastruttura ferroviaria. Lato domanda, il modello è stato sviluppato con un nuovo metodo ibrido, in grado di stimare i flussi OD HSR attraverso la riponderazione dei dati stazione-stazione osservati. Viene sviluppata una nuova procedura, che richiede l'interazione simultanea di più software per eseguire un'assegnazione stocastica. I flussi sono stati analizzati in modo gerarchico, partendo da una visione aggregata e approfondendo sempre più a livello disaggregato, guardando sempre il livello di correlazione tra flussi modellati e osservati. I risultati mostrano una forte correlazione a livello aggregato e per coppie stazione-stazione, con valori di R2 pari a 0,98 per il primo e 0,94 per il secondo. Il coefficiente rimane elevato soprattutto per le ripartizioni importanti: fasce orarie di punta e distanze medio-lunghe, con un valore medio di 0,71. A livello di segmento ferroviario scende a 0,12, sintomo di un eccessivo rumore nel modello, che lascia spazio a miglioramenti. Insieme alla complessità generale del modello, i tempi di calcolo sono diventati inaspettatamente una delle tante sfide da superare, ma il lavoro punta ad ottenere nuove teorie per la modellazione del trasporto pubblico che possano dare un contributo alla comunità scientifica.
Schedule-based high speed rail assignment model : application to the italian case study
Mariano, Pietro
2021/2022
Abstract
This thesis is part of a project between the PoliMi and Trenitalia to develop a DDS to optimize the integration of the Italian railway network with HS services, through transport simulations and «what if» scenarios. The work will focus on the development of a Travel Demand model with a schedule-based approach, capable of a path-based modelling. To this aim, an extensive literature review has been carried out, highlighting why it is important to use a, not so common, Schedule-based approach on low-frequency services. Among the others, one of the main innovations is the use of new mobile network data as main data source to build the Demand Model, offering the opportunity to explore travel demand in ways impossible with conventional surveys since this type of approach is new to the literature in this field on such a large scale. The concept of time-expanded network (Diachronic Graph) is used to model the supply system of the entire nation, also considering a road auxiliary graph, other than the railway infrastructure. A new hybrid method has been developed for the demand side of the model, able to estimate HSR OD flows by re-weighting Station-Station observed data. To run a SNL assignment, a new complete procedure has been developed, requiring the simultaneous interaction of multiple software. Flows have been hierarchically analyzed, starting from an aggregated view, and deepening more and more on a disaggregated level, always looking at the level of correlation between modelled and observed flows. The results show a strong correlation at aggregated level and for station-station pairs, with R2 values of 0.98 for the former and 0.94 for the latter. The coefficient remains still high mainly for important breakdowns: peak-hour and medium-long distance classes, with an average value of 0.71. At train link level it drops at 0.12, symptom of too much noise in the model, which leaves room for improvement. Together with the overall complexity of the model, calculation times have unexpectedly become one of the many challenges to overcome, since the work aims to obtain new theories for modelling public transport that can make a contribution to the scientific community.File | Dimensione | Formato | |
---|---|---|---|
2023_05_Mariano_Tesi_01.pdf
accessibile in internet per tutti
Descrizione: Testo della Tesi
Dimensione
54.01 MB
Formato
Adobe PDF
|
54.01 MB | Adobe PDF | Visualizza/Apri |
2023_05_Mariano_Executive_Summary_02.pdf
accessibile in internet per tutti
Descrizione: Executive Summary
Dimensione
6.94 MB
Formato
Adobe PDF
|
6.94 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/210344