Abstract:
The evaluation of highway projects where traffic congestion is a significant problem
requires careful analysis to accurately identify the true costs, and therefore potential
benefits, of operational changes. Models have previously been derived that employ
either simplistic speed-flow approaches or highly detailed micro-simulation
techniques. Another approach, and the goal for this research, is to develop a midrange
modelling framework whereby the predictive capability could be enhanced in
comparison to the simple speed-flow models but without the onerous data issues of
micro-simulation, for use in highway evaluation projects.
The approach adopted was based on modelling of acceleration noise defined as the
standard deviation of the accelerations. During periods of high traffic congestion
there is a greater variability in speed resulting in higher acceleration noise levels.
Once the acceleration noise level is estimated, the impact on fuel consumption and
vehicle emissions can then be determined.
The framework consists of a series of discrete sub-models that firstly estimate the
base (steady-speed) fuel consumption and then the acceleration noise level value.
From these data, the additional fuel consumption from congestion is calculated and
vehicle emissions are finally estimated. In this way, the model can be made more
flexible for application to future research because it allows replacement of any portion
of the model sub-routines without the need for a complete restructuring of the model
framework.
Data collection and analysis of the regimes for modelling were undertaken on
highways in Auckland (New Zealand), Kuala Lumpur (Malaysia) and Bangkok
(Thailand). The new modelling approach and the results of the work have been
integrated into with the International Study of Highway Development and
Management Tools (ISOHDM) in order to provide a model for inclusion into the
Highway Development and Management Model version 4 (HDM-4).
Fuel consumption predictions were tested both with and without the impact of a
simulated acceleration noise level. For the latter of the two (i.e. a given drive cycle)
the predictions were within 0.25 per cent. When including the impacts of a generated
drive cycle from a known level of acceleration noise, the results show a consistent
under-prediction of some 25 percent when compared with on-road test results. It is
believed that this under-prediction is largely due to the assumption of the
acceleration noise data conforming to a Normal Distribution.
The research results of the new modelling approach for emissions have been tested
against an existing data set of seven vehicles sourced from independent research on
Auckland’s motorways. Vehicle emissions of carbon dioxide and hydrocarbons were
under-predicted by some 50 per cent in relation to the average tested vehicle, but
were still well within the range of the seven observed results. Carbon monoxide and
oxides of nitrogen were grossly under-predicted and were below the minimum
observed values.
The new modelling approach, even with the above limitations, still has wide
application in improving the prediction of vehicles operating on highways in
congested conditions. The patterns of fuel consumption and emissions are showing
the appropriate changes in relation to traffic congestion, and therefore further
calibration is required. Furthermore, the model framework readily lends itself to
enhancement via adoption of some new sub-models.