IEEE International Symposium on the Application of Artificial Intelligence in Electrical Engineering (AAIEE)
A Generalizable Variational Quantum Oracle with Angle-Embedded Load Profiles for Grover Search Based Unit Commitment
Yubin Ma, Xiaodong Zheng
Abstract
Security-Constrained Unit Commitment (SCUC) is the core optimization problem for day-ahead clearing of electricity spot markets. With the decision-making dimensions and constraint complexity of the SCUC problem increasing exponentially, conventional algorithms are confronted with a severe curse of dimensionality. By constructing the Oracle based on quantum machine learning (QML), Grover search leverages the intrinsic properties of quantum superposition and entanglement to deliver a quadratic speedup for this class of NP-hard combinatorial optimization problems, which is unattainable with traditional machine learning methods. Existing Grover-search-based quantum algorithms for unit commitment typically rely on basis encoding to construct quantum oracles using variational quantum circuits (VQCs). However, binary basis states are mutually orthogonal and contain little structural information, which limits the expressive power of the VQC and restricts its ability to generalize across diverse load profiles in electricity markets. To address this issue, this paper proposes a hybrid embedding strategy for quantum oracle training within a Grover search framework. By introducing angle encoding to represent the temporal characteristics of load curves, structured load information is embedded into the VQC, improving the representation capacity and generalization capability of the quantum oracle. Numerical simulations and experiments on real quantum hardware demonstrate the effectiveness and generalization capability of the proposed variational quantum oracle.