Advancing the Construction of Representative Drive Cycles for Electric Vehicle

Abstract

The transition to electric vehicles (EVs) demands accurate tools for evaluating energy consumption, range, and performance under real-world conditions. Drive cycles (DCs) are central to this task, yet widely used international standards such as the Worldwide Harmonised Light Duty Test Procedure (WLTP) and the New European Drive Cycle (NEDC) lack the fidelity to capture local driving behaviour, infrastructure, and environmental influences. This thesis advances methodologies for DC development, with a focus on creating representative Irish DCs that integrate traffic, road, and weather conditions. The research begins with a systematic review of existing DC approaches, identifying gaps in dataset representativeness, slope integration, and the consideration of environmental factors. To address these, a large-scale dataset of Irish driving behaviour was compiled and complemented with secondary sources such as GIS-based area types, digital elevation models, and weather records. Two novel sampling methods�weighted random sampling and uniqueness random sampling�were introduced to ensure balanced participant contributions and representative coverage. The vehicle speed modelling was undertaken using ensemble machine learning, and a new Cascading Boosted Forest (CBForest) model was proposed by combining random forest and extreme gradient boosting in a layered structure. This model improved predictive accuracy and provided transparent feature importance, identifying road type, traffic density, speed limits, and driver behaviour as the dominant factors influencing speed profile of DCs. Building on these insights, a hybrid framework was developed that combined microtrip (MT) based speed synthesis with Markov Chain Monte Carlo (MCMC) slope generation. A two-stage evaluation process was introduced, combining characteristic parameters (CPs) with SHAP-based battery energy demand metrics to ensure statistical fidelity and energy consumption relevance. Using this framework, the Irish Representative Drive Cycle (IRDC) and a suite of Condition-Based Drive Cycles (CBDCs) were constructed. Simulation with the Future Automotive Systems Technology Simulator (FASTSim) demonstrated that slope reduced energy consumption by 18�35% across vehicles, urban cycles were the most energy-intensive, and auxiliary heating and cooling loads significantly impacted range under temperature extremes. Novel environmental CBDCs for wind and rain were also developed for the first time, offering insights into behavioural sensitivity under these conditions. Finally, an advanced MCMC�reinforcement learning (RL) approach was introduced to address the stochasticity and computational complexity of conventional MCMC methods. The process was framed as a sequential decision problem. In the first phase, speed and slope trajectories were generated using MCMC approach, and in the second, the RL component refined these trajectories to produce realistic speed�slope transitions while preserving representativeness. Together, the findings provide the first set of Irish-specific DCs that reflect local conditions and methodological contributions that extend the international literature on DC development. The research demonstrates how local datasets, when combined with advanced modelling, can generate cycles that are both statistically robust and directly applicable to EV energy analysis. Academically, the work contributes novel approaches to dataset representativeness, hybrid synthesis, and speed modelling, establishing a foundation for future studies on condition-based DCs. For industry, the framework provides a pathway to enhance simulation, calibration, and consumer communication, enabling the delivery of more reliable energy consumption and range estimates. At the policy level, the outputs offer evidence to set realistic efficiency benchmarks, inform charging infrastructure and grid planning, and align national actions with EU climate and energy commitments. By linking methodological innovation with practical application, the thesis shows how cycle research can advance scientific knowledge while supporting the transition to sustainable transport systems.