| Advanced Technology Vehicles (ATV) have been advertised to help solve transportation's oil use and environmental problems. Current traffic and transportation models are not suitable to study fuel consumption implications of using these vehicles in real traffic networks. The objectives of this thesis research were to study fuel consumption characteristics of ATV's in realistic traffic networks and to develop fuel consumption models for typical traffic facilities using Artificial Neural Networks (ANN).; To generate data for the ANN process, a methodology was established for three different facilities namely urban arterial, urban freeway, and Central Business District network. Three types of vehicles were used. These are conventional Internal Combustion Engine (ICE), hybrid ICE, and hybrid Fuel Cell. To obtain instantaneous speed/acceleration profiles, traffic conditions were simulated using the Federal Highway Administration sponsored Traffic Software Integrated System (TSIS) 5.1. Using the speed/acceleration profiles as inputs, fuel consumption data were generated using the Advanced V&barbelow;ehi&barbelow;cle S&barbelow;imulat or (ADVISOR) 2002 developed by the U.S. National Renewable Energy Laboratory.; A Bayesian regularization approach was applied to train a number of ANN models as noted below. (1) A microscopic fuel consumption model was developed by using time, speed, acceleration, state of charge, and the battery system current as explanatory variables and, (2) For macroscopic analyses of fuel consumption, a model was developed based on an aggregation of information obtained at the microscopic level, using link average speed as the explanatory variable of fuel consumption.; Quantitative and Qualitative analyses showed that all ANN models have a strong to moderately strong relationship between the predicted fuel consumption values and the fuel consumption values provided by ADVISOR.; Using coefficients provided by the ANN models, an algorithm coded in Visual Basic was developed to implement the fuel consumption models. Results obtained from this algorithm showed that the difference between the estimated fuel consumption and fuel consumption values provided by the ADVISOR and TSIS combination were always less than 5%.; The algorithm could be used as a standalone tool or, could be integrated into traffic simulation models for the estimation of the energy consumption effects of policies and operations in an efficient manner. |
