Accurate prediction of lithium-ion battery capacity fade across temperature extremes is critical for energy storage reliability, yet classical machine learning methods face challenges with the small, nonlinear datasets typical of accelerated aging studies. Quantum machine learning offers a potential solution through enhanced feature representation, but demonstrations on real quantum hardware remain limited. Here we benchmark quantum kernel ridge regression (Q-KRR) against classical baselines for battery capacity-fade prediction—a representative small-data problem with nonlinear, physics-driven features. Using six physics-informed descriptors, Q-KRR on a noiseless simulator achieved R 2 = 0.739, an 8.7% relative improvement over classical kernel ridge regression (R 2 = 0.680). On IBM's 127-qubit superconducting processor, systematic circuit optimization and targeted readout mitigation yielded R 2 = 0.770, exceeding both the classical baseline (13% relative gain) and the ideal simulator performance (4% absolute gain). This hardware advantage resulted from reducing circuit depth and qubit count while applying selective error mitigation matched to the dominant noise sources. The results demonstrate that quantum kernels can outperform classical methods on current noisy hardware when algorithms are co-designed with device constraints, offering a practical pathway for quantum-enhanced prediction in materials science and engineering domains characterized by limited training data and structured feature spaces.
Accurate prediction of lithium-ion battery capacity fade across temperature extremes is critical for energy storage reliability, yet classical machine learning methods face challenges with the small, nonlinear datasets typical of accelerated aging studies. Quantum machine learning offers a potential solution through enhanced feature representation, but demonstrations on real quantum hardware remain limited. Here we benchmark quantum kernel ridge regression (Q-KRR) against classical baselines for battery capacity-fade prediction—a representative small-data problem with nonlinear, physics-driven features. Using six physics-informed descriptors, Q-KRR on a noiseless simulator achieved R 2 = 0.739, an 8.7% relative improvement over classical kernel ridge regression (R 2 = 0.680). On IBM's 127-qubit superconducting processor, systematic circuit optimization and targeted readout mitigation yielded R 2 = 0.770, exceeding both the classical baseline (13% relative gain) and the ideal simulator performance (4% absolute gain). This hardware advantage resulted from reducing circuit depth and qubit count while applying selective error mitigation matched to the dominant noise sources. The results demonstrate that quantum kernels can outperform classical methods on current noisy hardware when algorithms are co-designed with device constraints, offering a practical pathway for quantum-enhanced prediction in materials science and engineering domains characterized by limited training data and structured feature spaces.
Battery management systems need accurate lifetime predictions to prevent failures and optimize replacement schedules, but acquiring training data requires months of controlled aging experiments per cell. We demonstrate that quantum computers can learn battery degradation patterns from far less data than classical methods. Using quantum kernel regression on 64 lithium-iron-phosphate cells, we achieved 2.1 % better prediction accuracy than classical approaches, but more importantly, required 6.6 times fewer training samples to reach equivalent performance. Real-world validation on IBM's 156-qubit quantum processor revealed severe noise challenges—performance dropped 54 %—but advanced error mitigation recovered 38 % of this loss. The quantum advantage emerges specifically in data-limited regimes where each additional cell requires months of testing. This positions quantum machine learning as a practical accelerator for battery research, materials discovery, and other domains where data acquisition time dominates development cycles.
Developing high-rate electrodes is essential for Na-ion pseudocapacitors, next-generation devices that bridge the performance gap between high-energy density batteries and high-power density capacitors at a fraction of the cost of Li-ion analogs. Ultra-short laser pulses for in-situ nanostructure generation (ULPING) enables one-step fabrication of 3D-architected TiO₂ electrodes that minimize solid electrolyte interface (SEI) formation, but mapping laser parameters to electrochemical lifetime remains a significant data-driven challenge, particularly under sparse experimental sampling conditions. This study evaluates quantum machine learning (QML) methods such as quantum support vector regression (QSVR), quantum kernel ridge regression (QKRR), and quantum neural networks (QNN) for electrode performance prediction under noisy intermediate-scale quantum (NISQ) era constraints, with classical machine learning as a benchmark. Using 5026 unique configurations (26 experimentally realized + 5000 synthetic generated with a Conditional Variational Autoencoder, CVAE), noiseless quantum simulation achieved R² = 0.900, compared to a classical kernel ridge regression benchmark of R² = 0.9919. Introducing realistic device noise reduced performance substantially (≈60% with mitigation; ≈24% without). Real hardware experiments on IBM Quebec using four features and a 25/10 train-test split yielded R² = 0.150 for QSVR versus R² = 0.844 for classical SVR. However, mitigation strategies showed promise: zero-noise extrapolation improved quantum performance to R² = 0.580 (+287%), while combined readout error mitigation and ZNE reached R² = 0.696 (+364%). These results demonstrate the feasibility of an end-to-end framework for QML prediction models on noisy quantum hardware, quantify resource-accuracy trade-offs, and highlight the calibrated mitigation and scaling conditions under which quantum kernel methods may complement classical approaches as quantum processors mature, laying a methodological foundation for quantum-assisted materials informatics.
Accurately predicting the lifetime of lithium-ion batteries during early charge–discharge cycles remains a significant challenge due to the nonlinear and weakly expressed degradation dynamics in the initial stages of operation. Classical machine learning (ML) models—although effective in pattern recognition—often face limitations in modeling complex correlations within small, high-dimensional datasets. To address these challenges, this study proposes a Hybrid Quantum–Classical Machine Learning (HQML) framework that integrates a Variational Quantum Circuit (VQC) as a quantum feature encoder with a Gradient Boosting Regressor (GBR) as the classical learner. The proposed approach is implemented using the Qiskit Aer simulator on the MIT Battery Degradation Dataset (124 cells, 42 engineered features). By encoding multi-source degradation descriptors (voltage, capacity, temperature, internal resistance) into Hilbert space via amplitude and angle encoding, the HQML model captures intricate nonlinear feature interactions that are inaccessible to conventional kernels. Experimental results demonstrate that the hybrid model achieves an RMSE of 93 cycles and an R² of 0.94, outperforming the best classical baseline (SVM + Wrapper selection, RMSE = 115, R² = 0.90). Furthermore, quantum observables analysis reveals interpretable correlations between entanglement strengths and physical degradation indicators. These results highlight the potential of quantum machine learning as a powerful paradigm for high-fidelity battery prognostics in the early-life regime.
Energy storage devices such as lithium-ion batteries and supercapacitors are essential for portable electronics, electric vehicles, and renewable energy systems, yet their long-term reliability is limited by capacity degradation. Accurate prediction of state-of-health (SOH), state-of-charge (SOC), and capacity retention are therefore critical for extending device lifetimes and improving safety. Classical machine learning (ML) methods including LSTMs, CNNs, and ensemble approaches have achieved success in forecasting degradation trends, but they face limitations in scalability, data requirements, and capturing complex electrochemical behaviors. Quantum computing, and in particular quantum machine learning (QML), offers new opportunities by exploiting superposition and entanglement to process information more efficiently and compactly. This review surveys recent advances in applying QML to energy storage, with a focus on CQ approaches (classical data–quantum processing), distinguishing Hybrid-CQ architectures. Comparative analyses highlight trade-offs: CQ and Hybrid-CQ models achieve higher accuracy and parameter efficiency but are constrained by noise, limited qubits, and slower runtimes on real devices. Looking forward, integrating error-mitigation strategies, benchmarking on actual quantum hardware, and embedding physics-informed modeling are critical to closing the gap between theoretical promise and practical deployment. Beyond prediction tasks, quantum approaches hold potential for dynamic charging optimization, materials discovery, and quantum battery concepts. Collectively, these developments underscore how QML can complement and eventually surpass classical ML, paving the way toward more accurate, sustainable, and efficient energy storage systems.
Quantum Support Vector Regression (QSVR) has emerged as one of the most promising quantum machine learning algorithms for nonlinear regression on Noisy Intermediate-Scale Quantum (NISQ) devices. However, despite numerous simulation-based studies, experimental evaluations on physical quantum hardware remain relatively scarce. This work presents a comparative analysis of three regression approaches: classical Support Vector Regression (SVR) with a Radial Basis Function (RBF) kernel, classical SVR utilizing precomputed quantum kernels, and the native QSVR implementation available in the qiskit-machine-learning framework. Experiments were conducted on three benchmark regression datasets using three execution environments: a noiseless Qiskit Aer simulator, the IQM Adonis hardware-aware simulator, and the physical 5-qubit IQM Spark superconducting quantum computer (Odra 5) located at Wrocław University of Science and Technology. Multiple quantum feature maps and circuit depths were investigated, while model hyperparameters were optimized using grid search. The obtained results demonstrate that QSVR can be successfully executed on contemporary superconducting quantum hardware and, for selected datasets, achieves prediction accuracy comparable to or exceeding that of a classical RBF-based SVR. The experiments further reveal the significant influence of quantum hardware noise, feature map selection, and circuit depth on regression quality. Moreover, the implementation based on precomputed quantum kernels generally exhibits greater robustness on physical hardware than the native QSVR implementation. The presented benchmark provides an experimental assessment of QSVR under realistic execution conditions and contributes to the ongoing evaluation of quantum machine learning algorithms on current NISQ-era quantum computers.
Abstract Quantum kernel methods have emerged as a powerful paradigm for extending classical machine learning into high-dimensional quantum Hilbert spaces, offering potential advantages in complex, nonlinear pattern recognition. However, their performance on near-term quantum devices remains constrained by circuit depth, noise sensitivity, and suboptimal parameterization of quantum feature maps. This paper introduces an Adaptive Quantum Kernel Optimization (AQKO) framework that integrates circuit-level adaptability, hybrid optimization, and regularization to enhance both expressivity and robustness in quantum kernel learning. The proposed method employs Adaptive Quantum Feature Maps (AQFM) that dynamically adjust rotation and phase parameters through a hybrid gradient-based optimization routine, reducing overfitting and improving generalization. Furthermore, a regularized quantum kernel objective is formulated to stabilize eigenvalue spectra under Noisy Intermediate-Scale Quantum (NISQ) conditions. Extensive experiments on nonlinear benchmark datasets — including Moons , Circles , and Iris — demonstrate that AQKO achieves up to 8.5% higher classification accuracy and 32% lower circuit depth compared to static quantum kernels. Hardware validation using IBM’s ibmq_quito device confirms AQKO’s resilience under depolarizing noise (p = 0.02) with a 41% error mitigation efficiency gain. These findings highlight AQKO’s potential as a scalable and noise-tolerant foundation for real-world quantum-enhanced learning, bridging the gap between theoretical quantum kernel models and practical NISQ implementations.