Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2609.40327

Quantum Enhanced Deep Reinforcement Learning for Distribution Network Reconfiguration

Daniel Germain, Mason Blanchard, Audrey Versteegen, Nora Bauer, Jonathan Mei, Claudio Girotto, Paul Smith, Martin Roetteler

quant-phSubmitted Sep 30, 2026• Announced Oct 1, 2026
arXivHTMLPDF

Abstract

Modern smart grids generate increasingly rich operational data and deploy greater numbers of remotely controlled switching assets than ever before, creating new opportunities for data-driven distribution network reconfiguration (DNR). DNR optimizes grid topology to improve efficiency, reduce power losses, and maintain operational constraints, but finding optimal switch configurations remains a computationally demanding combinatorial problem. In this paper, we propose a deep reinforcement learning (DRL) framework for static DNR to systematically investigate how the Q-network architecture affects both solution quality and learning performance. We evaluate multiple architectures within a controlled DRL framework, varying only the Q-network to isolate the contribution of each architectural choice. Our results show that quantum-enhanced Q-networks can achieve higher rewards with fewer parameters than their classical counterpart, suggesting that such architectures are a promising direction for combinatorial power systems optimization.

AI Summary

The paper asks how replacing a classical Q-network backbone with parameterized quantum circuits affects reinforcement learning for static distribution-network reconfiguration. It uses Double DQN with prioritized replay to select branch-exchange operations that preserve radial topology, rewarding voltage quality, reduced line losses, and distance from the initial topology. The quantum backbone comprises four sequential circuit-and-measurement stages, each with angle encoding and four variational layers, followed by classical decision heads. Simulations compare 4–16-qubit models with four classical MLP widths on the IEEE 33-bus system, using four random seeds per architecture. The 12-qubit model reaches the known minimum loss of $0.1396\,\mathrm{MW}$ in all four seeds, using 6,328 parameters versus 21,332 for the smallest classical model with the same hit rate—a $70.3\%$ reduction. The 16-qubit model reaches the optimum in three seeds. Mean final rewards are $13.26\pm0.39$ for 12 qubits and $13.12\pm0.75$ for the corresponding classical comparator; reported convergence episodes overlap across architectures. These comparisons concern parameter counts and episode counts, not computational speedup. A single-seed 12-qubit model also finds the known IEEE 69-bus optimum in simulation. Inference on IonQ Forte Enterprise-1 with 2,000 shots per circuit recovers the 33-bus optimum and gives $0.0999\,\mathrm{MW}$ loss for 69 buses, approximately $0.30\%$ above its optimum. Stated limitations include the small seed count, simulation costs preventing a full 69-bus ablation, untested scaling beyond 16 qubits and larger configuration spaces, and an exploration reward not guaranteed to preserve the optimal policy. Noise-aware training and error mitigation remain future work.

Some signs of AI-generated textOverclaimsIncrementalMissing comparisonUnclear novelty

Error found: The reported quantum reward advantage depends on using final-episode means: the paper states that classical models instead have slightly higher average rewards over the final half of training.

Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.

Comments
14 pages, 8 figures

Related papers

Quantum Reinforcement Learning for Volt-VAR Control in Power Distribution Systems

Lin Ding, Jianhui Wang, Huan-Hsin Tseng +3 more

Volt-VAR control (VVC) is crucial in active distribution networks for optimizing voltage profiles and minimizing network losses. While traditional deep reinforcement learning (DRL) algorithms exhibit promise for VVC, they often require extensive computational resources to handle such a high-dimensional problem. As a potential solution, quantum reinforcement learning (QRL) algorithms integrate the computational capabilities of quantum computing into the DRL framework. However, existing QRL algorithms struggle with complex VVC problems due to the limitations of current quantum hardware. To bridge this gap, this paper proposes an innovative QRL algorithm featuring an end-to-end architecture that integrates a classical autoencoder, variational quantum circuits (VQCs), and classical post-processing layers. This design efficiently compresses high-dimensional grid states, enabling VQCs to leverage quantum advantages while producing multiple control device outputs, specifically tailored for VVC tasks. Numerical studies on three representative distribution systems verify the effectiveness and scalability of the proposed QRL algorithm, and demonstrate its enhanced performance over classical approaches with only approximately 1% of the parameters. Additionally, the robustness of our developed algorithm is validated through noisy quantum environments.

IEEE Open Access Journal of Power and Energy•Jan 1, 2025•3 citations

A Survey of Quantum Deep Reinforcement Learning for Resource Allocation in Future Wireless Networks

Ni Made Erma Pratiwi Astiti, Byung Moo Lee

The significant rise in wireless communication, driven by sixth-generation (6G) networks, underscores the urgent need for innovative resource allocation methods to achieve efficiency, scalability, and reliability. This survey explores the integration of Quantum Deep Reinforcement Learning (QDRL) as a strategy to address the challenges posed by dynamic and complex wireless environments. By leveraging quantum computing features such as superposition and entanglement, QDRL enhances computational efficiency and mitigates the scalability limitations of classical and deep reinforcement learning (DRL) in high-dimensional optimization. The study reviews recent QDRL approaches for key resource allocation tasks, including power distribution, user scheduling, and channel assignment, and examines quantum algorithms such as the Quantum Approximate Optimization Algorithm (QAOA) and Quantum Neural Networks (QNN) for improved decision-making in wireless systems. Furthermore, it emphasizes the connection between quantum computing and DRL, highlighting hybrid architectures that balance computational cost, energy efficiency, and scalability. Finally, the survey identifies research gaps and outlines future directions, paving the way for robust QDRL strategies to support next-generation wireless networks.

IEEE Communications Surveys & Tutorials•Nov 28, 2025•16 citations

Quantum-Inspired Reinforcement Learning for Edge-Driven Cyber-Physical Systems

Jagjit Singh Dhatterwal, Sandeep Kumar, Shilpa Choudhary

The integration of Quantum-Inspired Reinforcement Learning (QiRL) into Edge-Driven CyberPhysical Systems (CPS) presents a transformative approach to enhance real-time decision-making in distributed environments. While classical reinforcement learning (RL) struggles with scalability and computational complexity in edge-CPS, quantum-inspired Quantum-Inspired Reinforcement Learning for Edge-Driven Cyber-Physical Systems algorithms leverage superposition and entanglement principles to optimize policy exploration and convergence. This paper proposes a novel QiRL framework that combines quantum-inspired optimization with deep RL, enabling efficient resource allocation and adaptive control in latency-sensitive applications (e.g., smart grids, autonomous drones). We formalize the mathematical foundations of QiRL, design a hybrid neural-quantum architecture for edge deployment, and evaluate its performance under dynamic CPS constraints. Simulations on IoT-based case studies demonstrate $\mathbf{2 0}-\mathbf{3 5 \%}$ faster convergence and $\mathbf{1 5 \%}$ higher energy efficiency compared to classical Deep Q-Networks (DQN). Key challenges, such as quantum noise simulation and edge-hardware limitations, are discussed to guide future research. This work bridges quantum computing principles with edge intelligence, offering a scalable pathway for nextgeneration CPS.

International Conference on Mobile Networks and Wireless Communications (ICMNWC)•Dec 10, 2025•0 citations

QPLAN: Deep Reinforcement Learning Assisted Quantum Network Planning

Yuhang Gan, Shiyi Ling, Ruilin Zhou +2 more

Quantum networks attract increasing attention due to their unique advantages over classical networks. However, the challenge of efficiently and economically constructing reliable wide-area quantum networks remains unresolved. This problem, termed network planning, is typically modeled as an Integer Linear Programming (ILP) problem. Existing solver-based solutions for classical networks face scalability issues, while heuristic approaches that sacrifice optimality for scalability heavily rely on human expertise and lack generalizability. Recently, deep reinforcement learning (DRL) has demonstrated potential in adaptive decision-making for classical network planning. However, quantum networks introduce unique constraints due to their end-to-end entanglement distribution requirements. To address this, we propose a two-stage quantum network resource allocation framework, QPLAN, that combines multi-agent DRL with traditional optimization solvers. Our approach uses a novel multi-agent reinforcement learning strategy using graph neural networks (GNNs) and transformers for traffic flow and corresponding hardware resource allocation to minimize hardware cost while cooperatively satisfying problem constraints as much as possible. The output of the DRL algorithm is then used as an initial input to prune the solution space, "warm-starting" the ILP solver and further improving the solution. This hybrid approach effectively reduces manual effort and enhances scalability. Experimental results show that our method yields near-optimal solutions with better scalability than only using ILP solvers.

International Conference on Quantum Communications, Networking, and Computing (QCNC)•Mar 31, 2025•1 citation

Quantum Computing–Enhanced Intelligence for Sustainable Power Systems: Toward Next-Generation Smart Grid Architectures

Sahibzada Muhammad Ali, Abdul Jabbar Ehsan, Abdur Rahman +3 more

Transitioning to resilient, renewable, and EV-integrated power systems demands computational frameworks capable of handling high uncertainty, distributed decision-making, and large-scale dynamics. Classical AI and optimisation struggle with high-dimensional interactions among generators, storage, EV fleets, and real-time grid states. This paper proposes a quantum–AI smart grid architecture combining quantum machine learning, quantum-inspired combinatorial optimisation, and hybrid quantum–classical control. Grid states are encoded in variational quantum circuits and Quadratic unconstrained binary optimisation models, accelerating dispatch, scheduling, congestion management, and optimal power flow. A layered design integrates edge-level classical controllers with cloud-based quantum solvers for utility-compatible deployment. Evaluations on a high-fidelity quantum-assisted digital twin of a renewable- and EV-rich IEEE 14-bus network show 40–60% faster optimisation, improved stability, and enhanced solution robustness over classical baselines, demonstrating quantum–AI as a transformative enabler for next-generation power systems.

International Conference on Integrated Intelligence and Cognitive Engineering (ICIICE)•Apr 18, 2026•1 citation

A INTERPRETABLE QUANTUM-INSPIRED DEEP REINFORCEMENT LEARNING APPROACHES FOR LARGE-SCALE DATA ANALYTICS

G.S.Geethamani, D.Anitha

The rapid expansion of large-scale data analytics (LSDA) has presented significant issues concerning computing efficiency, scalability, adaptive learning, and model interpretability. In recent years, quantum-inspired deep reinforcement learning (QIDRL) has emerged as a promising research paradigm that integrates quantum-inspired optimization (QIO) concepts with deep reinforcement learning (DRL) approaches to enhance intelligent decision-making and complicated data processing. The need for transparent and reliable artificial intelligence (AI) systems has heightened the significance of interpretable learning frameworks that can elucidate model behavior and prediction results. This work offers a thorough examination of interpretable QIDRL methods for extensive data analysis. The review methodically examines current advancements in QIO models, interpretable RL architectures, policy explanation mechanisms, and scalable deep learning (DL) frameworks used for high-dimensional and dynamic datasets. The report examines the use of these methodologies across many sectors, including healthcare analytics, cybersecurity, smart transportation, industrial automation, financial forecasting, and intelligent Internet of Things (IoT) ecosystems. The study examines the benefits of QIDRL techniques in improving convergence speed, adaptive decision-making, feature optimization, and interpretability inside extensive analytical systems. Furthermore, significant research obstacles like computational complexity, training instability, scalability limitations, insufficient benchmark datasets, and the absence of defined criteria for interpretability assessment are thoroughly examined. The study emphasizes new research trends and prospects for creating robust, efficient, and transparent QIDRL frameworks for advanced intelligent DA applications.

The International Journal of Interdisciplinary Cultural Studies•Aug 3, 2026•0 citations
Modern smart grids generate increasingly rich operational data and deploy greater numbers of remotely controlled switching assets than ever before, creating new opportunities for data-driven distribution network reconfiguration (DNR). DNR optimizes grid topology to improve efficiency, reduce power losses, and maintain operational constraints, but finding optimal switch configurations remains a computationally demanding combinatorial problem. In this paper, we propose a deep reinforcement learning (DRL) framework for static DNR to systematically investigate how the Q-network architecture affects both solution quality and learning performance. We evaluate multiple architectures within a controlled DRL framework, varying only the Q-network to isolate the contribution of each architectural choice. Our results show that quantum-enhanced Q-networks can achieve higher rewards with fewer parameters than their classical counterpart, suggesting that such architectures are a promising direction for combinatorial power systems optimization.
The paper asks how replacing a classical Q-network backbone with parameterized quantum circuits affects reinforcement learning for static distribution-network reconfiguration. It uses Double DQN with prioritized replay to select branch-exchange operations that preserve radial topology, rewarding voltage quality, reduced line losses, and distance from the initial topology. The quantum backbone comprises four sequential circuit-and-measurement stages, each with angle encoding and four variational layers, followed by classical decision heads.
Simulations compare 4–16-qubit models with four classical MLP widths on the IEEE 33-bus system, using four random seeds per architecture. The 12-qubit model reaches the known minimum loss of 0.1396 MW0.1396\,\mathrm{MW}0.1396MW in all four seeds, using 6,328 parameters versus 21,332 for the smallest classical model with the same hit rate—a 70.3%70.3\%70.3% reduction. The 16-qubit model reaches the optimum in three seeds. Mean final rewards are 13.26±0.3913.26\pm0.3913.26±0.39 for 12 qubits and 13.12±0.7513.12\pm0.7513.12±0.75 for the corresponding classical comparator; reported convergence episodes overlap across architectures. These comparisons concern parameter counts and episode counts, not computational speedup.
A single-seed 12-qubit model also finds the known IEEE 69-bus optimum in simulation. Inference on IonQ Forte Enterprise-1 with 2,000 shots per circuit recovers the 33-bus optimum and gives 0.0999 MW0.0999\,\mathrm{MW}0.0999MW loss for 69 buses, approximately 0.30%0.30\%0.30% above its optimum.
Stated limitations include the small seed count, simulation costs preventing a full 69-bus ablation, untested scaling beyond 16 qubits and larger configuration spaces, and an exploration reward not guaranteed to preserve the optimal policy. Noise-aware training and error mitigation remain future work.
Volt-VAR control (VVC) is crucial in active distribution networks for optimizing voltage profiles and minimizing network losses. While traditional deep reinforcement learning (DRL) algorithms exhibit promise for VVC, they often require extensive computational resources to handle such a high-dimensional problem. As a potential solution, quantum reinforcement learning (QRL) algorithms integrate the computational capabilities of quantum computing into the DRL framework. However, existing QRL algorithms struggle with complex VVC problems due to the limitations of current quantum hardware. To bridge this gap, this paper proposes an innovative QRL algorithm featuring an end-to-end architecture that integrates a classical autoencoder, variational quantum circuits (VQCs), and classical post-processing layers. This design efficiently compresses high-dimensional grid states, enabling VQCs to leverage quantum advantages while producing multiple control device outputs, specifically tailored for VVC tasks. Numerical studies on three representative distribution systems verify the effectiveness and scalability of the proposed QRL algorithm, and demonstrate its enhanced performance over classical approaches with only approximately 1% of the parameters. Additionally, the robustness of our developed algorithm is validated through noisy quantum environments.
The significant rise in wireless communication, driven by sixth-generation (6G) networks, underscores the urgent need for innovative resource allocation methods to achieve efficiency, scalability, and reliability. This survey explores the integration of Quantum Deep Reinforcement Learning (QDRL) as a strategy to address the challenges posed by dynamic and complex wireless environments. By leveraging quantum computing features such as superposition and entanglement, QDRL enhances computational efficiency and mitigates the scalability limitations of classical and deep reinforcement learning (DRL) in high-dimensional optimization. The study reviews recent QDRL approaches for key resource allocation tasks, including power distribution, user scheduling, and channel assignment, and examines quantum algorithms such as the Quantum Approximate Optimization Algorithm (QAOA) and Quantum Neural Networks (QNN) for improved decision-making in wireless systems. Furthermore, it emphasizes the connection between quantum computing and DRL, highlighting hybrid architectures that balance computational cost, energy efficiency, and scalability. Finally, the survey identifies research gaps and outlines future directions, paving the way for robust QDRL strategies to support next-generation wireless networks.
The integration of Quantum-Inspired Reinforcement Learning (QiRL) into Edge-Driven CyberPhysical Systems (CPS) presents a transformative approach to enhance real-time decision-making in distributed environments. While classical reinforcement learning (RL) struggles with scalability and computational complexity in edge-CPS, quantum-inspired Quantum-Inspired Reinforcement Learning for Edge-Driven Cyber-Physical Systems algorithms leverage superposition and entanglement principles to optimize policy exploration and convergence. This paper proposes a novel QiRL framework that combines quantum-inspired optimization with deep RL, enabling efficient resource allocation and adaptive control in latency-sensitive applications (e.g., smart grids, autonomous drones). We formalize the mathematical foundations of QiRL, design a hybrid neural-quantum architecture for edge deployment, and evaluate its performance under dynamic CPS constraints. Simulations on IoT-based case studies demonstrate 20−35%\mathbf{2 0}-\mathbf{3 5 \%}20−35% faster convergence and 15%\mathbf{1 5 \%}15% higher energy efficiency compared to classical Deep Q-Networks (DQN). Key challenges, such as quantum noise simulation and edge-hardware limitations, are discussed to guide future research. This work bridges quantum computing principles with edge intelligence, offering a scalable pathway for nextgeneration CPS.
Quantum networks attract increasing attention due to their unique advantages over classical networks. However, the challenge of efficiently and economically constructing reliable wide-area quantum networks remains unresolved. This problem, termed network planning, is typically modeled as an Integer Linear Programming (ILP) problem. Existing solver-based solutions for classical networks face scalability issues, while heuristic approaches that sacrifice optimality for scalability heavily rely on human expertise and lack generalizability. Recently, deep reinforcement learning (DRL) has demonstrated potential in adaptive decision-making for classical network planning. However, quantum networks introduce unique constraints due to their end-to-end entanglement distribution requirements. To address this, we propose a two-stage quantum network resource allocation framework, QPLAN, that combines multi-agent DRL with traditional optimization solvers. Our approach uses a novel multi-agent reinforcement learning strategy using graph neural networks (GNNs) and transformers for traffic flow and corresponding hardware resource allocation to minimize hardware cost while cooperatively satisfying problem constraints as much as possible. The output of the DRL algorithm is then used as an initial input to prune the solution space, "warm-starting" the ILP solver and further improving the solution. This hybrid approach effectively reduces manual effort and enhances scalability. Experimental results show that our method yields near-optimal solutions with better scalability than only using ILP solvers.
Transitioning to resilient, renewable, and EV-integrated power systems demands computational frameworks capable of handling high uncertainty, distributed decision-making, and large-scale dynamics. Classical AI and optimisation struggle with high-dimensional interactions among generators, storage, EV fleets, and real-time grid states. This paper proposes a quantum–AI smart grid architecture combining quantum machine learning, quantum-inspired combinatorial optimisation, and hybrid quantum–classical control. Grid states are encoded in variational quantum circuits and Quadratic unconstrained binary optimisation models, accelerating dispatch, scheduling, congestion management, and optimal power flow. A layered design integrates edge-level classical controllers with cloud-based quantum solvers for utility-compatible deployment. Evaluations on a high-fidelity quantum-assisted digital twin of a renewable- and EV-rich IEEE 14-bus network show 40–60% faster optimisation, improved stability, and enhanced solution robustness over classical baselines, demonstrating quantum–AI as a transformative enabler for next-generation power systems.
The rapid expansion of large-scale data analytics (LSDA) has presented significant issues concerning computing efficiency, scalability, adaptive learning, and model interpretability. In recent years, quantum-inspired deep reinforcement learning (QIDRL) has emerged as a promising research paradigm that integrates quantum-inspired optimization (QIO) concepts with deep reinforcement learning (DRL) approaches to enhance intelligent decision-making and complicated data processing. The need for transparent and reliable artificial intelligence (AI) systems has heightened the significance of interpretable learning frameworks that can elucidate model behavior and prediction results. This work offers a thorough examination of interpretable QIDRL methods for extensive data analysis. The review methodically examines current advancements in QIO models, interpretable RL architectures, policy explanation mechanisms, and scalable deep learning (DL) frameworks used for high-dimensional and dynamic datasets. The report examines the use of these methodologies across many sectors, including healthcare analytics, cybersecurity, smart transportation, industrial automation, financial forecasting, and intelligent Internet of Things (IoT) ecosystems. The study examines the benefits of QIDRL techniques in improving convergence speed, adaptive decision-making, feature optimization, and interpretability inside extensive analytical systems. Furthermore, significant research obstacles like computational complexity, training instability, scalability limitations, insufficient benchmark datasets, and the absence of defined criteria for interpretability assessment are thoroughly examined. The study emphasizes new research trends and prospects for creating robust, efficient, and transparent QIDRL frameworks for advanced intelligent DA applications.