The design of physics-augmented neural networks (PANNs) for the purposes of constitutive modeling has received considerable attention as of late for a variety of material behaviors. Here, we revisit the classical framework of isotropic incompressible hyperelasticity in light of recent advances in the study of constitutive inequalities. We show that polyconvexity implies true-stress-true-strain monotonicity for a large class of incompressible strain-energy functions. The resulting elastic law obeys the physically reasonable Legendre-Hadamard (or ellipticity) condition as well as the notion of increasing stress with increasing strain. These results then inform the architecture of four distinct PANNs which are subsequently calibrated to three different sets of experimental data each. We show that different PANN parametrizations - satisfying the same constitutive constraints a priori - have varying approximation power for the description of material behavior. Moreover, even when distinct parametrizations perform comparatively well within the calibration regime, they show pronounced differences in extrapolation. This observation motivates a critical discussion about the predictive power of PANNs which also has implications for the modeling of more complex material behavior by virtue of neural networks.
Time series aggregation (TSA) aims to construct temporally aggregated optimization models that accurately represent the output space of their full-scale counterparts while using a significantly reduced temporal dimensionality. This paper presents a theoretical approach that achieves exact temporal aggregation of full-scale power system models-even in the presence of energy storage time-coupling constraints-by leveraging active constraint sets and dual information. This advances the state of the art beyond existing TSA methods, which typically cannot guarantee solution accuracy or rely on iterative procedures to determine the required number of representative periods. To bridge the gap between this theoretical analysis and practical application, we employ machine learning, i.e., classification and clustering, to inform TSA in models that co-schedule variable renewable energy sources and energy storage. Numerical results show substantially improved computational performance relative to the full-scale model, while maintaining a favorable trade-off between solution accuracy and complexity.
Cross-border electricity exchanges are crucial for operating and planning highly renewable power systems. Many studies reduce spatial granularity to keep models tractable and prescribe cross-border exchanges exogenously, often by reusing historical import/export time series. Such assumptions become inconsistent as renewable penetration changes the magnitude and timing of flows. The core contribution of this work is a machine-learning (ML) surrogate framework that maps nodal time-series data (e.g., hourly demand and renewable generation) to synthetic interconnector-level flow time series. The resulting flows serve as fixed boundary conditions in reduced power system optimization models (PSOMs). This addresses studies focused on a subset of an interconnected system, for which repeatedly solving the full model across multiple climate years or scenarios is computationally expensive. We demonstrate the framework on a pan-European single-node-per-country DC optimal power flow model. We benchmark k-nearest neighbors (KNN), ridge regression, random forests, Extra Trees, gradient boosting, XGBoost, and feedforward neural-network (SQU) surrogates. As an additional contribution, we introduce a novel custom neural-network loss that softly penalizes physically implausible flow patterns. This soft penalty is not a guarantee of full power-flow feasibility. Across benchmarks, the SQU variants provide the most consistent out-of-sample and downstream optimization performance while substantially outperforming scaled historical profiles. The ML-generated flows closely reproduce country-level results of the full European model while greatly reducing computation time (up to ∼500×). The proposed framework enables scenario-consistent reduced PSOM studies across climate years after a single full-model run, eliminating repeated pan-European simulations and supporting more robust climate-year analyses under fixed system assumptions.
Ammonia-fueled engines are emerging as a promising solution for decarbonizing global shipping and power generation. High-pressure liquid ammonia injection has demonstrated potential to reduce harmful emissions primarily in experimental set-ups such as rapid compression machines and constant volume combustion chambers, with limited data available for large-bore engine applications. This study investigates the impact of different ammonia injection strategies on engine performance and emissions, including nitrogen oxides (NOx), nitrous oxide (N2O) and unburnt ammonia, in a large-bore, medium-speed engine. Investigations were conducted using a fuel-actuated common rail injector prototype developed by OMT in a dual-fuel configuration with a diesel pilot injector. Based on initial fuel spray characterization in an optically accessible constant volume chamber, investigations on injection strategies were performed in a single-cylinder research engine for a part load operating point with 12 bar IMEP. The combustion concept demonstrated stable operation with an ammonia fuel fraction of up to 95 %. Due to increasing N2O emissions at the lowest diesel fractions, the lowest CO2-equivalent emissions are achieved at substitution rates of 85 to 90 %, with reductions of up to 70 % compared to conventional diesel combustion. This reduction was achieved at an injection pressure of 600 bar, whereas higher injection pressures resulted in increased emissions. Additionally, a reduction in NOx emissions of up to 33 % can be achieved at similar combustion phasing. These findings deepen the understanding of the combustion process and lay the foundation for future research at higher loads and in connection with exhaust gas aftertreatment systems.
ENTSO-E technical reports on grid-forming capability describe converters as an internal voltage source behind an effective impedance zeff at the port. This paper examines the uniqueness and testability of the reported effective impedance in a Thevenin-type compliance representation. Since the internal voltage e is generally inaccessible, the transient difference method is often considered. The active power–frequency outer loop changes the internal-voltage phase during the early response. A sufficient error bound relates extraction accuracy to source motion, current excitation and control tuning. Even when e is exposed for direct back-calculation, the result depends on the dynamic compensation mechanism. Vector filtering removes the steady-state virtual-impedance contribution, while scalar compensation changes the reported impedance according to the internal-voltage extraction point and the operating point. The remaining inner-loop and physical-filter contributions are retained. Controller hardware-in-the-loop experiments quantify the impedance differences together with the port-level power and energy responses. Changing only the internal-voltage extraction point changes the reported impedance without changing the implemented control law. The reported zeff therefore requires a declared internal-voltage extraction point and test conditions, together with port-level support metrics.
