power consumption modeling

This work proposes a physics-based, linearly parameterized electrical power consumption model to enable energy-aware motion planning and thermal management for humanoid robot arms. The model explicitly accounts for actuator losses, gravity-compensated payload effects, and inter-joint coupling dynamics, with its parameters identified via linear regression using onboard power measurements. Trained on 897 trajectories, the model achieves an R² of 0.933 and an RMSE of 1.07 W, and demonstrates strong generalization with an R² of 0.965 on 46 unseen validation trajectories featuring novel velocity profiles. The results not only confirm high predictive accuracy but also reveal distinct dominant loss mechanisms across individual joints.

This work addresses the challenge of accurately predicting battery state of charge (SOC) in autonomous robots to prevent task interruption, a problem exacerbated by the complex nonlinear dynamics between SOC, time, and motor control signals (PWM). To overcome the limitations of existing methods, the authors propose a hybrid modeling framework that integrates physical principles with data-driven techniques. Leveraging the electrical and mechanical dynamics of DC motors, they generate high-fidelity simulated SOC time-series data for a four-wheel Arduino robot across PWM inputs ranging from 1% to 100%. For the first time, they apply Sparse Identification of Nonlinear Dynamics (SINDy) combined with least-squares regression to derive an interpretable and scalable unified model, SOC(t, p), capable of adapting to arbitrary initial SOC levels and environmental conditions. This model provides a foundation for energy-aware path planning through precise power consumption forecasting and demonstrates strong potential for transfer to real-world robotic systems.

This study addresses the challenge of modeling low-inertia power grids, where accurate state prediction and physically consistent sensitivities are essential for reliable control, yet the relative merits of existing differentiable modeling paradigms remain unclear. Focusing on a single-machine infinite-bus system, the work systematically evaluates physics-informed neural networks (PINNs), neural ordinary differential equations (NODEs), and differentiable programming (DP) in terms of trajectory extrapolation, parameter identification, and LQR controller synthesis. The findings reveal a fundamental trade-off between data-driven flexibility and physical fidelity, and for the first time delineate clear applicability boundaries of these methods in power system modeling and control: NODEs excel in trajectory extrapolation, DP achieves faster convergence in parameter identification and near-optimal closed-loop stability, while PINNs exhibit limited generalization capability.

Existing robotic arm energy consumption models are predominantly derived from traditional industrial robots, exhibiting poor generalizability and insufficient accuracy. To address this, we propose the first open-source, MATLAB-based, data-driven energy modeling toolkit—eliminating reliance on theoretical kinematic and dynamic parameters and enabling accurate electrical power consumption estimation across diverse lightweight robotic arms from multiple manufacturers. Our method integrates Denavit–Hartenberg parameters, link masses, and center-of-mass information, using joint positions, velocities, accelerations, timestamped measured power, and time as inputs to construct a cross-platform, data-driven model. Evaluated on four distinct lightweight robotic arms from different vendors, the model achieves training-set RMSEs of 1.42–2.80 W and test-set RMSEs of 1.45–5.25 W. This represents a substantial improvement in both prediction accuracy and generalization capability, establishing a reliable foundation for energy-aware robot design and task scheduling in low-power applications.

Addressing the challenge of high-accuracy, high-efficiency, and interpretable modeling for energy systems (e.g., lithium-ion batteries) under data scarcity, this paper proposes the Model-Integrated Neural Network (MINN)—the first method to directly embed differential-algebraic equation (DAE)-based physical structural priors into a neural network architecture. MINN explicitly encodes DAE constraints and employs a hybrid data–physics joint training paradigm to enable end-to-end learning of system-level physical dynamics. Compared to conventional first-principles models, MINN achieves comparable global output accuracy and local electrochemical behavior prediction using only a minimal amount of training data, while accelerating computation by two orders of magnitude. This work uniquely unifies physical interpretability, numerical fidelity, and computational scalability, establishing a new paradigm for control-oriented modeling of sustainable energy systems.

This study addresses the limited engineering applicability of machine learning surrogate models in power system dynamic simulation due to their lack of physical interpretability. For the first time, neural tangent kernel (NTK) theory is introduced into this domain, integrated with small-signal eigenvalue analysis to establish a formal connection between system physical stiffness and neural network optimization stiffness. This linkage reveals the evolution mechanism of error modes during training, enabling the development of an adaptive loss weighting strategy. The proposed approach not only provides modal-level physical explanations for performance differences across network architectures—such as ActNet—but also significantly enhances training convergence and reliability. Consequently, this work lays a theoretical foundation for designing interpretable, structure-aware surrogate models tailored to power system dynamics.

This study addresses the degradation of accuracy observed when physics-informed surrogate models—despite performing well in isolated evaluations—are integrated into dynamic power system simulators, particularly under high-stress operating conditions. Framing this challenge as a verification and validation (V&V) problem, the work proposes a theoretical framework that combines algebraic coupling sensitivity, dynamic error amplification mechanisms, and finite-time simulation horizons to bound prediction errors. The approach integrates model-based verification with data-driven conformal calibration. Using a physics-informed neural network as a surrogate for synchronous machine dynamics, the authors conduct residual analysis and generate conformal predictions within a differential-algebraic equation simulation framework. Their findings reveal that small residuals in governing equations do not necessarily guarantee small trajectory errors, thereby underscoring the critical need—and a new paradigm—for validating surrogate models in coupled dynamical environments.

This work addresses the challenge of accurately predicting dynamic trajectories in power systems dominated by high-penetrating renewable energy and inverter-based resources, where traditional methods struggle under time-varying parameters, data privacy constraints, and diverse operating conditions. The authors propose LASS-ODE-Power, a novel framework that introduces the foundation model paradigm to power system dynamics for the first time. Leveraging over 40 GB of trajectory data generated by ordinary differential equations (ODEs), the framework employs large-scale pretraining combined with a parallelized linearized neural network architecture and a tailored fine-tuning strategy. It achieves highly accurate and efficient cross-system trajectory prediction in a zero-shot setting. Experimental results demonstrate that the proposed method significantly outperforms existing learning-based approaches across multiple scenarios, exhibiting superior generalization capability and computational efficiency.

This study addresses the insufficient accuracy and reliability of energy consumption prediction for electric trucks by proposing a physics-informed modeling framework that integrates first-principles physical knowledge with data-driven techniques. The approach embeds fundamental energy loss mechanisms into machine learning architectures and leverages an ensemble of models—including Bayesian linear regression, neural networks, and gradient-boosted regression trees—to achieve both high-fidelity point predictions and robust uncertainty quantification. Experimental results demonstrate that the proposed method significantly outperforms conventional purely data-driven models in both predictive accuracy and uncertainty estimation, thereby validating the effectiveness and superiority of physics-guided feature modeling for energy consumption forecasting in electric freight transport.

This study addresses the challenge of quantifying the performance gap between designed energy system architectures and their actual operational outcomes, which arises from mismatches among multi-fidelity models. To bridge this gap, the authors propose an online machine learning–accelerated, multi-resolution optimization framework that dynamically adjusts solution fidelity by integrating multi-objective architectural optimization with ML-guided receding horizon control. The approach leverages low-fidelity elite solutions to warm-start high-fidelity solvers, thereby significantly reducing the number of expensive high-fidelity model evaluations while closely approximating the theoretical performance upper bound achievable under a given architecture. Validated on a 1 MW industrial thermal load system, the method reduces the performance gap by 42% compared to a rule-based controller and decreases high-fidelity evaluations by 34% relative to a non-ML-guided multi-fidelity approach.