Power grid problem tests Quantum Inspire’s link to DelftBlue

Quantum‑Enhanced Power‑Flow Calculation: A First End‑to‑End Demo

A team led by TU Delft researcher Zeynab Kaseb has completed the first full‑stack test of a power‑grid algorithm that moves from classical high‑performance computing (HPC) on DelftBlue, through quantum‑cloud access on Quantum Inspire, to execution on QuTech’s 17‑qubit Tuna‑17 superconducting processor. The work showcases how classical preparation, quantum execution, and post‑processing can be linked in a single, reproducible workflow.

From Classical Preparation to Quantum Execution

1. Classical pre‑processing on DelftBlue – The power‑flow problem (determining bus voltages in a small network) was encoded into a binary optimisation form. DelftBlue performed the heavy‑lifting of data preparation and circuit optimisation.
2. Quantum submission via Quantum Inspire – The prepared circuits were sent to the Quantum Inspire cloud platform, which schedules jobs on the Tuna‑17 processor.
3. Execution on Tuna‑17 – Two QAOA‑style circuits were run: a 4‑qubit encoding (30‑layer circuit) and a 6‑qubit encoding (165‑layer circuit). Measurements were returned to Quantum Inspire and fed back to DelftBlue for analysis.

Results and What They Reveal

• 4‑qubit run: The best candidate solution appeared 215 times out of 1 024 measurements (~21 %).
• 6‑qubit run: The deeper circuit produced a lower residual error (0.003262 vs. 0.01053) but the top candidate was observed only 16 times (~1.6 %). This frequency matches the uniform‑sampling expectation for 64 possible bit‑strings, highlighting the impact of circuit depth on outcome quality.
• Benchmarking – An exhaustive classical calculation provided a reference solution, allowing the team to quantify how close the quantum results came to the optimal answer.

Why This Matters

The demonstration moves beyond pure simulation by exercising real quantum hardware for a practical engineering task. It proves that:

• Classical HPC resources can prepare and validate quantum‑ready problem instances.
• Cloud‑based quantum processors can be accessed seamlessly from existing HPC workflows.
• Even modest‑size devices (17 physical qubits) can tackle non‑trivial optimisation problems, though circuit depth remains a limiting factor.

Future work will likely explore larger grid models, error‑mitigation techniques, and tighter integration with other quantum‑software stacks.

---

The integration of DelftBlue, Quantum Inspire, and Tuna‑17 marks a concrete step toward hybrid quantum‑classical pipelines for industry‑relevant problems, setting a baseline for more ambitious quantum‑enhanced simulations.