Preliminary Study of Agent-Based Simulation to Predict Charging Behavior of Electric Vehicles in New Delhi, India

Agent-based simulations have gained popularity in travel demand modeling because they exhibit greater power in explaining dependencies within activity chains and are based on a dynamic decision-making procedure. These features become even more critical while modeling electric vehicles because variations in charging behavior can significantly affect the spatio-temporal distribution of load on a power grid. Given India’s aggressive electric vehicle deployment targets (6 – 7 million hybrid and electric vehicles by 2020), there are great concerns about the country’s ability to meet future BEV charging demand. A first step towards modeling the spatial and temporal effects of increased EV penetration on the grid is to generate accurate trip profiles for the study region. As a pilot study, agent-based simulation is performed for the city of New Delhi. A detailed traffic network is extracted from Open Street Maps (OSM) and processed through the java application Osmosis. Activity chains for cars, two-wheelers and electric vehicles (assuming 15% market penetration by 2025) are generated from household travel survey data and Origin-Destination (O-D) data using Non-Parametric Resampling. Finally, the base network and activity demands are served as input for agent-based traffic simulation software MATSim to generate equilibrium routes. To the knowledge of the authors, this is the first attempt to produce an agent-based traffic simulation for the city of New Delhi and provides important insights into impacts of EV introduction in India.