Data-Driven Quantum Approximate Optimization Algorithm for Cyber-Physical Power Systems

TL;DR

This paper introduces a data-driven approach to enhance the Quantum Approximate Optimization Algorithm (QAOA) for power systems, enabling efficient parameter transfer across weighted graphs and demonstrating competitive performance without extensive optimization.

Contribution

The paper proposes a novel data-driven QAOA method that transfers parameters based on graph density, improving practical applicability in power system optimization.

Findings

Comparable performance to Goemans-Williamson algorithm without parameter tuning

Verified strategy on nearly 40,000 instances

Advances QAOA for noisy intermediate-scale quantum devices

Abstract

Quantum technology provides a ground-breaking methodology to tackle challenging computational issues in power systems, especially for Distributed Energy Resources (DERs) dominant cyber-physical systems that have been widely developed to promote energy sustainability. The systems' maximum power or data sections are essential for monitoring, operation, and control, while high computational effort is required. Quantum Approximate Optimization Algorithm (QAOA) provides a promising means to search for these sections by leveraging quantum resources. However, its performance highly relies on the critical parameters, especially for weighted graphs. We present a data-driven QAOA, which transfers quasi-optimal parameters between weighted graphs based on the normalized graph density, and verify the strategy with 39,774 instances. Without parameter optimization, our data-driven QAOA is comparable with the Goemans-Williamson algorithm. This work advances QAOA and pilots the practical application of quantum technique to power systems in noisy intermediate-scale quantum devices, heralding its next-generation computation in the quantum era.