Problem and scope
Unit commitment chooses which generators to operate while meeting demand and respecting operating constraints. QuantumGrid contains classical scheduling code and quantum-circuit experiments around this problem. The reviewed loader generates synthetic demand, including when an ENTSO-E API key is supplied; this is not a study on downloaded European grid data.
Questions to resolve
- Do all solvers optimize the same feasible schedules and physical objective?
- Does an independently checked feasible solution exist for each reported energy?
- How much do encoding size, penalty weights and simulator cost constrain the experiment?
Implementation and boundaries
- Classical path: OR-Tools MILP with continuous dispatch and binary commitment; heuristic solvers are separate baselines.
- Quantum path: binary QUBO to Ising mapping, QAOA and a VQE ansatz executed through PennyLane.
- Scope: VQE operates on the binary commitment encoding. A continuous optimal-power-flow solver is not implemented in this path.
- Data: synthetic hourly demand and generated fleets. No real-grid outcome or quantum advantage is established.
Comparison status
The existing QUBO and classical formulations are not yet a like-for-like benchmark. The QUBO penalizes distance from demand using binary full-capacity dispatch; the MILP allows continuous dispatch. QUBO energy contains penalty terms and omits constant offsets, so it is not a monetary generation cost. The scaling entry point also constructs different problem instances for its classical and quantum paths.
No scaling-results table is published here: the reviewed repository has no persisted benchmark outputs supporting one. A useful comparison must first share an instance, objective, feasibility check and cost units; solver status and optimality gaps must also be retained.
What a circuit result means
QAOA alternates cost and mixer layers, while VQE uses a parameterized ansatz to minimize the encoded Hamiltonian’s expectation. Both optimize the encoding supplied to them. Low energy alone does not demonstrate that electricity demand is met, that dispatch is economical, or that a quantum method outperforms MILP.
Paths that need a common evaluator
The shared evaluation stage is the next validation requirement, not a completed benchmark pipeline.
Next reproducible milestone
Use a tiny instance whose schedules can be exhaustively enumerated. Independently verify demand, generation limits, startup costs and objective offsets. Only after the encoding, feasibility checks and cost calculations agree with that oracle should experiments grow in size. Solver-quality gaps can then be measured against the exact solution.
Current scaling entry point · Classical solvers · Existing tests. Quantum-hardware experiments and real-data conclusions remain outside the evidence presented here.