International Conference on Power System and Energy Technology (ICPSET)
A Hybrid Classical-Quantum Newton-Raphson Framework for Power Flow Computation
Xin Jin, Jun Lu, Lanlan Zhang, Bofan Wu, Heng Fan
Abstract
The Newton-Raphson method, despite its quadratic convergence, faces a growing performance bottleneck in solving high-dimensional sparse linear systems as power grid scale and complexity increase. This paper proposes a hybrid classical-quantum Newton-Raphson framework for power flow computation. On the classical side, the polar-coordinate Newton-Raphson iteration and Jacobian assembly are retained; on the quantum side, Krylov subspace reduction is coupled with a variational quantum linear solver (VQLS) to solve the projected low-dimensional linear system. A residual-driven nested iterative refinement strategy and the parameter shift rule are introduced to ensure solution accuracy and gradient efficiency, respectively. Due to current quantum chip hardware constraints, five physical qubits with relatively high fidelity are selected for validation on the domestically developed Origin Wukong superconducting quantum processor. The IEEE 57-bus system (106-dimensional Jacobian) and the Hefei 81-bus regional grid system (141-dimensional) both converge within three Newton iterations with maximum power mismatch below$1 \times 10^{-3}$; voltage relative errors are under$1 \times 10^{-5}$for the 57 -bus case and under$1 \times 10^{-3}$for the 81-bus case. These results demonstrate that the hybrid architecture can deliver accurate power flow solutions under NISQ-era hardware constraints.