Identifying optimal number of driving cycles to represent diverse driving conditions

Abstract

Driving cycle is one of the main inputs of vehicle emission modeling. However, the variability of driving cycles due to fluctuations in weather conditions is one of the primary sources of uncertainty in vehicle emission estimation. This study aims to identify and determine an optimal number of driving cycles that can correctly represent driving patterns in diverse weather conditions. First, a multivariate multiple regression model is developed to determine the most important weather factors affecting the driving patterns. Then, similar weather conditions are identified according to these factors using unsupervised machine learning. Next, two driving cycles are constructed for diverse weather types, one for weekdays and one for weekends. Afterward, descriptive analysis and a similarity matrix are employed to determine how similar the generated driving cycles are in different weather types. Finally, 15 driving cycles are identified to represent driving patterns in diverse driving conditions.