Optimization methods for developing electric vehicle charging strategies

Electric vehicles (EVs) are considered to be a crucial and proactive player in the future for transport electrification, energy transition, and emission reduction, as promoted by policy-makers, relevant industries, and the academia. EV charging would account for a non-negligible share in the future electricity demand. The integration of EV brings both challenges and opportunities to the electricity system, mainly from their charging profiles. When EV charging behaviors are uncontrolled, their potentially high charging rate and synchronous charging patterns may result in the bottleneck of the grid capacity and the shortage of generation ramping capacity. However, the promising load shifting potential of EVs can alleviate these problems and even bring additional flexibilities to the demand side for further applications, such as peak shaving and the integration of renewable energy. To grasp these opportunities, novel controlled charging strategies should be developed to help integrate electric vehicles into energy systems. However, corresponding methods in current literature often have customized assumptions or settings so that they might not be practically or widely applied. Furthermore, the attention of literature is more paid to explaining the results of the methods or making consequent policy recommendations, but not sufficiently paid to demonstrating the methods themselves. The lack of the latter might undermine the credibility of the work and hinder readers’ understanding. Therefore, this thesis serves as a methodological framework in response to the fundamental and universal challenges in developing charging strategies for integrating EV into energy systems. The discussions aim to raise readers’ awareness of the essential but often unnoticed concerns in model development and hopefully would enlighten future researchers into this topic. Specifically, this cumulative thesis comprises four papers and analyzes the research topic from two perspectives. With Paper A and Paper B, the micro perspective of the thesis is more applied and focuses on the successful implementation of charging scheduling solutions for each EV individually. Paper A proposes a two-stage scenario-based stochastic linear programming model to schedule EV charging behaviors and considers the uncertainties from future EVs. The model is calculated in a rolling window fashion with updated parameters. Scenario generation for future EVs is simulated by inhomogeneous Markov chains, and scenario reduction is achieved by a fast forward selection method to reduce the computational burden. The objective function is formulated as variance minimization so that the model can be flexibly implemented for various applications. Paper B applies the model proposed in Paper A to investigate how the generation of a wind turbine could be correlated with the EV controlled charging demand. An empirical controlled charging strategy is designed for comparison where EVs would charge as much as possible when wind generation is sufficient or would postpone charging otherwise. Although these two controlled charging strategies perform similarly in terms of wind energy utilization, the solutions from the proposed model could additionally alleviate the volatility of wind energy generation by matching the EV charging curve to the electricity generation profile. With Paper C and Paper D, the macro perspective of the thesis is more explorative and investigates how EVs as a whole would contribute to energy transition or emission reduction. Paper C investigates the greenhouse gas emissions of EVs under different charging strategies in Europe in 2050. Methodologically, the paper proposes an EV module that enables different EV controlled charging strategies to be endogenously determined by energy system models. The paper concludes that EVs would contribute to a 36% emission reduction on the European level even under an uncontrolled charging strategy. Unidirectional and bidirectional controlled charging strategies could further reduce emissions by 4% and 11%, respectively, compared with the original level. As a follow-up study of Paper C, Paper D develops, demonstrates, improves, and compares three different types of EV aggregation methods for integrating an EV module into energy system models. The analysis and demonstration of these methods are achieved by having a simplified energy system model as a testbed and the results from the individual EV modeling method as the benchmark. As different EV aggregation methods share the same data set as for the individual EV modeling method, the disturbance from parameters is minimized, and the influence from mathematical formulations is highlighted. These EV aggregation methods are compared from multiple aspects