Evaluation of the Results of Passenger Trip Demand Model Calculation

Abstract

The paper addresses the problem of evaluating the results of passenger travel demand modelling in suburban and interregional public transport systems. Reliable origin–destination (O–D) matrices are a fundamental element of transport planning; however, their practical application requires systematic validation under conditions of incomplete and heterogeneous empirical data. The study aims to assess the accuracy and practical applicability of a methodology for modelling passenger correspondence matrices using a real-world case study of the suburban and intra-regional transport network surrounding Kharkiv. The research methodology combines limited statistical observations with a model-based estimation of travel demand. A transport supply model was developed in the PTV VISUM environment, including the representation of nodes, links, stop locations, transport zones, and access connections. Transport zones were defined at the level of settlements, taking into account spatial distribution, population size, distance from the city centre, and the availability of alternative modes, particularly railway services. The capacity of origin zones was estimated as a function of demographic characteristics and the influence intensity of the urban core. The O–D matrix was generated using a distance-based exponential distribution model reflecting empirically observed trip length patterns. A system of constraints ensured consistency between total departures from each transport zone and the probabilistic distribution of trips across distance intervals. Model validation was performed in two stages: comparison of calculated and observed trip distributions by distance groups, and comparison of simulated and reported passenger flows derived from assignment procedures within the network model. The results demonstrate that the average deviation between calculated and observed trip distributions equals 6.3%, with a maximum deviation of 19.62% for long-distance intervals exceeding 120 km. For passenger flows, the mean relative error is 5.4%, while the maximum deviation (18.11%) occurs in the 20–40 km distance range. The accuracy improves with increasing distance from the city, whereas larger discrepancies near the urban core are attributed to higher network density and flow redistribution effects. The findings confirm the adequacy of the proposed methodology for practical applications in regional transport planning, service level assessment, and route network optimisation.