Optimal G2V/V2G fleet scheduling under uncertainty using a quadratic interpolation supercell thunderstorm algorithm

This paper develops and applies a Quadratic Interpolation Supercell Thunderstorm Algorithm to electric vehicle fleet charging scheduling optimization under uncertainties over a 24-h horizon, through smart coordination between grid-to-vehicle and vehicle-to-grid operations. Validated on the Congress on Evolutionary Computation 2022 benchmark, Quadratic Interpolation Supercell Thunderstorm balances global exploration and local exploitation through a supercell-inspired population structure combined with quadratic interpolation, navigating multi-modal solution spaces and accelerating convergence toward locally smooth cost regions without requiring gradient information. The addressed electric vehicle scheduling problem is an inherently non-convex, high-dimensional, due to coupled fleet constraints, stochastic operating conditions, and battery degradation dynamics, which limits the applicability of classical deterministic optimization methods. In addition, a state of health based adaptive C-rate control mechanism is introduced, in which charging and discharging rates dynamically adjust according to battery health and state of charge conditions, keeping the operation electrochemically safe and limiting accelerated wear. Five case studies are investigated, including a deterministic baseline, thermal excursion battery degradation uncertainty, day-ahead price uncertainty, arrival/departure uncertainty, and scalability testing. The Quadratic Interpolation Supercell Thunderstorm Algorithm is benchmarked against five competing metaheuristic and two deterministic approaches under equal computational budgets across 30 independent runs. Comparative results show that the improved approach achieves up to 20% lower operational cost compared with the baseline approach, while also obtaining the lowest cumulative cost across all scalability scenarios. Statistical validation further confirms its superiority, with significant differences in 50 out of 56 Wilcoxon pairwise comparisons and the lowest average Friedman rank (1.71). In addition, the Quadratic Interpolation Supercell Thunderstorm Algorithm consistently satisfies target state of charge constraints with zero violations and demonstrates strong scheduling robustness across all investigated scenarios.