arXiv:2609.36999
Towards Scalable Quantum Unit Commitment in Power Systems: Encoding and Complexity Reduction Techniques
Marc Carrillo-MuÑoz, Sergio MartÍnez-Hermida, Sara Barja-MartÍnez, Antonio E. SaldaÑa GonzÁlez, MÒnica AragÜÉs-PeÑalba
Abstract
This paper develops a Quantum optimisation Unit Commitment (UC) framework that benefits from Pauli Correlation Encoding (PCE) as a qubit reduction preprocessing technique and Sample-based Quantum Diagonalisation (SQD) as a hybrid post-processing refinement method. The UC problem is formulated in binary form, then transformed into a QUBO problem and mapped onto an Ising representation suitable for adiabatic quantum evolution within the Qibo framework. This study analyses how PCE and SQD techniques affect qubit scalability, computational effort, feasibility, and probability of recovering the optimal solution as the number of generators increases. The results show that PCE reduces the effective size of the encoded problem, SQD improves the concentration of probability on feasible and optimal solutions, and their combined application provides the best overall behaviour in several scalable case studies, achieving more reliable solutions with lower effective computational cost than the unreduced formulation. The repository is available open source in Github (link is not yet available until work is published. However, if requested by the reviewers, we can make it open for a scific time window, if requested).
AI Summary
The paper asks whether qubit compression and sample-based post-processing can improve a quantum-optimisation workflow for unit commitment. It studies a simplified, single-period model: minimise generator commitment costs subject to an exact demand-balance constraint, with fixed power per active generator. The binary problem is mapped to a penalty QUBO and an Ising Hamiltonian. The workflow combines Pauli Correlation Encoding (PCE), simulated adiabatic evolution in Qibo, and optional Sample-based Quantum Diagonalisation (SQD). PCE replaces logical $Z_i$ operators with products of reduced-register $Z$ operators, restricting the representable schedules and potentially increasing interaction order. Encodings are selected using cost-weighted coverage of known feasible configurations. Penalty calibration also uses the optimal feasible cost. SQD retains frequently sampled bitstrings and diagonalises a projected auxiliary Hamiltonian; the reported setting, $\beta=1$, makes this Hamiltonian diagonal, selecting minimum-energy retained configurations. Numerical tests cover constructed instances with $N=4$–$20$ generators; unreduced comparisons stop at $N=12$. Selected PCE encodings reduce qubit counts by $40\%$–$60\%$, with optimal-solution probabilities of $0.343$–$0.906$. At $N=20$, ten qubits retain two of 28 feasible schedules and yield optimal probability $0.906$. SQD raises reported feasible and optimal probabilities to one in every tested direct and compressed case, with overhead below $0.02$ seconds for direct cases. The stated limitations are classical-emulator resource constraints, instance-dependent compression performance, and the simplified operational model. Hardware resource benefits, sensitivity to feasible-space structure, and extensions to multi-period and more realistic generator constraints remain for future study.
Central claim may not hold: The claimed scalability is not established end to end because penalty calibration uses the optimal feasible cost and PCE selection uses cost-weighted feasible configurations without a demonstrated scalable method for obtaining that information. All reported SQD runs use beta=1, making post-processing equivalent to minimum-energy selection among retained samples, so unit output optimal probability is conditional on already retaining an optimum and demonstrates no advantage over ordinary best-sample selection. Table 2 reports eight retained states for both N=4 and N=5, although the stated rho_SQD=1/8 subspace-size rule permits at most two and four states, respectively. Selected PCE runs are slower than direct evolution at N=10 and N=12, and the reported timing does not establish the full cost of testing and selecting reduced dimensions.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.