Large-scale scheduling in heterogeneous EV depots: A sequential matheuristic with dynamic interaction-aware grouping

J. Guasch Albareda, M. Houchati, P. Rodilla, J.P. Chaves

Next Energy Vol. 13, pp. 101102

Resumen:

Electric vehicle (EV) adoption transforms fleet charging depots into complex energy systems where capacity limits, heterogeneous charging technologies, and state-of-charge (SoC)-dependent power constraints must be considered. Mixed-integer linear programming (MILP) models can capture these features but face rapid computational growth as fleet size increases. This paper proposes a sequential fix-and-optimize (F&O) framework for large-scale EV charging scheduling, retaining detailed operational modelling and reducing computational complexity. The formulation incorporates station power limits, SoC-dependent piecewise charging dynamics, charging continuity, and explicit vehicle-charger compatibility, enabling realistic representation of heterogeneous AC/DC infrastructures. To manage computational burden, the global problem is decomposed into a sequence of smaller subproblems using a Dynamic Interaction-Aware Grouping (DIAG) algorithm, which clusters vehicles based on resource conflict. The methodology is tested on congestion scaling synthetic scenarios (10–100 vehicles) and validated with real electric bus depot data. Under a two-hour time limit, monolithic MILP fails for medium and large fleets, delivering operationally unviable results with energy satisfaction dropping to near-zero. In contrast, the sequential matheuristic method consistently meets all infrastructure constraints and preserves objective performance. Results indicate that subproblem size mainly affects computational effort, with limited impact on the objective, and demonstrate controlled runtime growth as system size increases. The framework offers a scalable and tractable alternative to centralized MILP scheduling. Its short execution times make it possible to recompute schedules whenever operating conditions change, opening the door to intraday rescheduling in future work.

Resumen divulgativo:

Este artículo aborda la programación de carga a gran escala en flotas de vehículos eléctricos y cargadores heterogéneos. Se propone una optimización mateheurística secuencial que agrupa dinámicamente los vehículos y divide el problema en subproblemas que se resuelven uno tras otro.

Palabras Clave: EV charging Optimization; EV fleets; MILP; EV type Compatibility; Scalability

Índice de impacto JCR-JIF y cuartil WoS: 6,400 - Q2 (2025)

Referencia DOI: https://doi.org/10.1016/j.nxener.2026.101102

Publicado en papel: Octubre 2026.

Publicado on-line: Octubre 2026.

Cita:
J. Guasch Albareda, M. Houchati, P. Rodilla, J.P. Chaves, "Large-scale scheduling in heterogeneous EV depots: A sequential matheuristic with dynamic interaction-aware grouping", Next Energy, Vol. 13, pp. 101102, Octubre 2026. [Online: Octubre 2026] doi: 10.1016/j.nxener.2026.101102