CASE STUDY 14 Quantum computing Scheduling vehicle-to-grid fleets using hybrid learning and quantum kernels Modern power systems are increasingly treating electric vehicles as distributed storage that can provide grid services, but turning a large, heterogeneous EV fleet into a reliable vehicle-to-grid “virtual battery” requires real-time scheduling under tight operational constraints (availability windows, state-of-charge targets, site limits, user requirements and volatile prices). As fleet size and variability grow, the scheduling and control problem becomes high-dimensional and time-sensitive, which pushes classical optimization and forecasting stacks toward higher complexity and shorter decision cycles. E.ON and IBM tackled the practicality gap in vehicle-to-grid (V2G) scheduling by training a learning-based model to quickly predict near-optimal EV charging and discharging actions from simulated scenarios. Their approach derives policies using approximate dynamic programming and then learns them with kernel methods, so decisions can be generated fast while still respecting real constraints. They report that the learned model achieves objective values comparable to those obtained by approximate dynamic programming with CPLEX on a classical computer, but with lower runtimes, enabling more frequent re-optimization. In parallel, they investigated quantum machine learning for high-dimensional prediction tasks linked to scheduling uncertainty. Because fidelity-based quantum kernels can suffer from “exponential concentration”, they introduced an error-mitigation method called bit flip tolerance (BFT) in the scientific journal npj Quantum Information. They reported that BFT significantly improved results on 40+ qubits (about 80% accuracy with mitigation versus ~33% without) and demonstrated experiments up to 156 qubits with performance close to classical baselines. Key reported benefits included faster operational decision- making through learning-guided scheduling that reduces solver runtimes while retaining competitive solution quality, a practical path to frequent re-optimization for large fleets as grid conditions vary, and a demonstrated mitigation strategy for quantum kernels that improves robustness on real EV scheduling-related data and scales up experiments into the 100+ qubit regime.19,20 Quantum for Energy and Utilities: Key Opportunities for Energy Transition 28

Quantum for Energy and Utilities 2026