Investigating Energy Consumption in Electric Vehicles

Abstract

As part of the fight to limit climate change, the European Union has set a target of zero-emissions from the transport sector in European cities by 2035. The main component of reducing emissions from transport is thought to be a complete transition from traditional internal combustion engine vehicles (ICEVs) to fully electric vehicles (EVs). However, EV sales are not on track to meet the ambitious goals set by government agencies like the European Union. Only 17% of new vehicles registered in Ireland in 2025 were EVs, but the goal is for that to reach 100% by 2030. One reason for the current scepticism around EVs is the phenomenon known as range anxiety, where drivers fear EVs lack the necessary range to complete desired trips. An important piece of combatting range anxiety is developing better estimations of EV range. EV range is estimated both by the manufacturer to provide a selling point for the vehicle and by in-vehicle systems to advise drivers on how far the vehicle can travel before requiring a charge. However, these estimations can differ significantly from real-world driving scenarios due to a multitude of factors. Advances in the state-of-the-art literature, through developing models of energy consumption (EC) and understanding how different factors influence EC, can help to produce better estimations of EV range. To fill gaps identified in the literature, several objectives are set. The first objective is to build an EC model at a trip level. The accuracy of a macroscopic and a mesoscopic EC model are compared, and the factors influencing the model at each scale are also compared. The next objective is to find a method that relates driving behaviour to EC and determine how much driving behaviour can affect EC. The final objective is to model EC for realistic trips using a road network model and to analyse with these models how speed limits, route choice, and driving behaviour affect EC. Data was collected from 36 EV drivers in Ireland to build an EC model. The EC model was built at the mesoscopic scale to model driving segments, which is shown to decrease error by 53% compared to a macroscopic model when predicting EC of a trip. Additionally, the mesoscopic model is shown to be more comprehensive in modelling driving behaviour parameters, whereas the macroscopic model predicted EC mainly by distance travelled. At both scales, the most accurate models were artificial neural networks (ANNs) when compared to other machine learning (ML) methods. The model includes driving behaviour, environmental, and driver and vehicle parameters. A sensitivity analysis (SA) is used to determine to what degree each modelled parameter affects EC. The results show that, among all driving segments, speed, gradient, time spent accelerating, and time spent decelerating have larger effects than any other parameter. It is shown that in both modelled and actual segments, the most energy efficient driving occurs at speeds between 20-60 km/h. Speeds above 100 km/h cause drastic increases in EC over the same distance travelled. As a result of the SA, a further investigation is conducted to determine how rates of acceleration, rates of deceleration, and speed affect EC at high and low speeds. The results show how moderate acceleration and deceleration rates can lead to improved energy efficiency at all speeds. The most energy efficient driving behaviour at high speeds can reduce EC by 71%. Additionally, it is demonstrated that common classification of driving behaviour into passive, moderate, and aggressive categories fails to sufficiently correlate with EC across different driving segments. Finally, the EC model is applied to routes found through a model of the Ireland road network. Realistic drive cycles (DCs) are developed for routes found in the network model, and these DCs are used to estimate EC under different conditions such as reduced speed limits. Reduced speed limits have a significant effect on EC, especially on interurban trips. With the combined models, it is shown that drivers can save 10% of EC by slowing down, which results in an increase in travel time of 9%. In other words, increasing a journey time by about 15 minutes can save EV drivers about 5 kWh of battery power. The effect of eco-driving behaviours and route choice on EC is also analysed by combining the EC and road network models.