Simulating Electric Vehicle driving and charging behaviour using an agent-based approach

Abstract

Electric Vehicles (EVs) are an opportunity for governments to reduce the greenhouse gas emissions and to improve air quality. There is an increasing need to understand the processes behind EV development and use. For this study, a behavioural model is developed to explore the behaviour of electric motorists in relation to CP placement. A literature research on factors influencing EV driver behaviour is conducted to explore which parameters could be included in the EV driving and charging behaviour model. As a simulation is an abstract representation of the reality, simplifications of these parameters have been made, and the parameter conceptualisation is based on multiple assumptions. The appropriate method for developing this simulation model is agent-based modelling, which allows for modelling the individual behaviour of EV drivers. The model as developed includes the main simulation process of EV drivers leaving home and moving to their destination. When it is needed, the EV driver will search for a charging point nearby. The model further includes behavioural concepts found in literature: charge point hogging, range anxiety, disappointment when a charging point is occupied and the level of satisfaction of the agents. The conceptual model is implemented in GAMA using spatial datasets of the case study area in Amsterdam. The model verification and validation process is conducted to test the model on its plausibility. It is verified that the model behaves as it was designed and the final model is bug-free. A limited validation of the model is successfully conducted to test if the model represents the real-world. Due to lack of validation data, this is done by visual observation and comparing outputs to input variables. During the sensitivity analysis it was found that the share of EV ownership and the range anxiety have a notable impact on the model. These parameters could be further researched in order to improve the model. Additional data on these parameters will be needed to fully calibrate the model. Additionally, the usefulness of the model is tested by using two scenarios based on policies and technologies. This showed that the simulation model has clear potential for testing real-life scenarios and it can be used as a tool for policy makers. The ABM developed in this research proved itself as an adequate simulation model but it should be noted that the model is sensitive for parameters which are based on assumptions. There are many opportunities for further model improvement. When more extensive datasets are retrieved for validation and calibration, the quality of the model will increase.