Q-Site

The planning problem

Q-Site studies where to install two battery energy storage systems on the IEEE 14-bus test grid after adding a synthetic 80 MW / 25 MVAr AI data-centre load at Bus 9. The aim is to reduce unsafe-voltage incidents during individual transmission-line outages.

The study evaluates 15 N-1 line-outage scenarios using pandapower AC power flow. Six candidate buses - 9, 10, 5, 8, 11 and 13 - are encoded as binary storage-placement decisions.

A hybrid quantum approach

A six-variable QUBO combines under-voltage vulnerability, an exactly-two-batteries budget penalty and a small adjacency penalty. It is converted to an Ising Hamiltonian and solved with depth-2 QAOA, using Qiskit and the classical COBYLA optimiser. Exhaustive enumeration of all 15 valid pairs provides the exact classical baseline.

The simulator used 100 optimiser iterations and 4,096 final shots. The report records real-hardware execution on IBM Quantum ibm_kingston through qBraid, with 11 completed hardware evaluations and 2,048 final shots. Classical search, simulation and the hardware run all selected Buses 9 and 10, at a QUBO cost of -20.1027. This small-instance agreement is not evidence of quantum advantage.

My contribution

As Point of Contact, I was responsible for Qiskit integration and QAOA execution. The project connects quantum optimisation with a classical grid model and independently checks the proposed siting using AC power-flow calculations.

Results and validation

The selected sites were evaluated with 15 MVAr reactive-power support at each of Buses 9 and 10. Across the 13 converged scenarios within the 15-scenario test set, unsafe-voltage bus incidents fell from 30 to 27, a 10% reduction.

The outage of line 5-10 improved from four to two unsafe buses; line 9-10 improved from four to three. The two non-convergent collapse scenarios, involving lines 0-1 and 8-9, were not resolved. The 10% improvement comes from AC power-flow simulation, not an independent physical-grid or quantum-hardware voltage measurement.

Scope and limitations

This is a planning prototype with reactive-power support only: no real-power discharge, state-of-charge dynamics or time-series dispatch. Voltage safety is defined as 0.94-1.06 per unit, and outages are evaluated independently, without cascading or N-2 analysis.

The objective scores under-voltage only because the modelled inverter support raises local voltage. The report does not establish the same outcome at larger grid sizes or under different storage-control strategies.

The challenge

Developed for the Global Industry Challenge 2026, U.S. DOE Office of Technology Commercialization, Energy Infrastructure track. The project reached Phase 3, the final stage, and finished as a finalist. The complete submitted technical report is available above.