Search NASA⌕ Search

SEARCH · Search NASA

Results for “load pattern modeling”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 181 records · Page 10
All sourcesNASA NTRSDOE OSTI
DOE OSTI2019

Coordinated PEV Charging for Distribution System Management

Plug-in electric vehicles (PEVs) can help to reduce worldwide dependence on petroleum and carbon emissions. In the past a few years, PEVs have received considerable attention as an eco-friendly and cost-effective alternative to conventional gasoline vehicles. PEVs consume higher power from the grid during charging compared to conventional residential loads. Therefore, the emerging fleet of PEVs will introduce a considerable amount of additional load on power system. On the other hand, majority of PEVs are parked for more than 90% of the time, making them ideal for providing various services through Electric Vehicle Supply Equipment (EVSE), a.k.a. charging stations which are PEVs’ connection points to the power system. In this project, coordinated PEV charging are developed to provide distribution system services, considering PEV characteristics, realistic travel pattern, charging behaviors, and EVSE power rating and availability. The main efforts and contributions are summarized as follows. • Mobility models and EVSE charging capability/availability are indispensable for evaluating PEV charging coordination strategies and developing PEV charging control. Many existing studies utilize simplified mobility models assuming that the entire vehicle fleet returns home in the evening and is parked at home until the next morning. Some other PEV models better represent diversified home arrival and departure time, but cannot capture varying charging flexibility and capability at different locations, and therefore are not appropriate to study the impacts of public EVSE on PEV load and utilization. In this work, a mobility and charging flexibility model is proposed to better represent the temporal availability and varying charging capability from PEV onboard charger and EVSE. • Based on the mobility model, a generalized optimization method is proposed to evaluate different PEV charging coordination strategies. In existing literature, different algorithms and methods need to be designed to evaluate each charging strategy from both vehicle owners’ and system’s perspective. With the optimization method proposed in this work, different charging control strategies can be studied and compared by only updating objective functions. Moreover, optimization tricks are provided to convert the optimization problems to equivalent linear programming problems, which can be efficiently solved with existing solvers. • An innovative scheduling and control framework is proposed to enable smart PEV charging for grid services while meeting PEV owners’ travel needs. In the proposed framework, each set of EVSE is equipped with a controller that estimates charging power and energy flexibility based on vehicle characteristics, EVSE power rating, battery energy state, and upcoming trip information. With the simplified charging flexibility model received from each EVSE controller, the central coordinator determines the optimal power allocation for a look-ahead time window for the given grid services. The proposed charging control can help reduce the computational complexity and communication requirement compared with existing methods. It is also scalable to the expanding PEV fleet and robust to uncertainties in upcoming vehicle trips and future system condition. • The proposed charging strategy evaluation and smart charging control methods are applied to one of the prototypical feeders developed by the Pacific Northwest National Laboratory (PNNL) for case studies. Detailed trip information extracted from the National Household Travel Survey (NHTS) is used to represent travel and parking patterns. Different scenarios are simulated to evaluate the performance of the proposed methods and understand how PEV mobility and public EVSE availability affect PEV load and potential grid services.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2019

Model Study of Blast Furnace Operation with Central Coke Charging

Blast furnace (BF) remains the dominant ironmaking process worldwide. Central coke charging (CCC) operation is a promising technology for stabilizing BF operations, but it needs reliable and quantified process design and control. In this work, a multi-fluid BF model is further developed for quantitatively investigating flow-thermal-chemical phenomena of a BF under CCC operation. This model features the respective chemical reactions in the respective coke and ore layers, and a specific sub-model of layer profile for the burden structure for the CCC operation. The simulation results confirm that the gas flow patterns and cohesive zone’s shape and location under the CCC operation are quite different from the non-CCC operation. Under the CCC operation, the heat is overloaded at the furnace center while the reduction load is much heavier at the periphery regions; the profiles of top gas temperature and gas utilization show bell-shape and inverse-bell-shape patterns, respectively. More importantly, these differences are characterized quantitatively. In this given case, when the CCC opening radius at the throat is 0.35 m, the cohesive zone top opening radius is around 0.50 m, and the isotherms of CCC operation become much steeper (~ 80 deg) than those of non-CCC operation (~ 60 deg) near BF central regions. In addition, it is confirmed that carbon solution-loss reaction rate can be decreased significantly at BF central regions under CCC operation. The model helps to understand CCC operation and provides a cost-effective method for optimizing BF practice.

36 MATERIALS SCIENCE↗
NASA NTRS1985

Gravity and lithospheric stress on the terrestrial planets with reference to the Tharsis region of Mars

On Mars and Venus, a strong positive correlation is found between geoid height and topography. The Tharsis region of Mars provides an exhibition of this correlation. Several hypotheses have been proposed regarding the origin of Tharsis. For purposes of explanation, three end-member dynamic hypotheses are considered. A hypothesis that the flexural doming of Tharsis resulted from uplift caused by some force acting on the base of the lithosphere can be rejected. According to another hypothesis, Tharsis is associated with a lithospheric load, while a third one considers that Tharsis is primarily isostatically compensated. In the present study, improved stress models for isostatic compensation on Mars are obtained. The strains inferred from fracture patterns on Mars are compared with the stresses predicted by the isostatic theory. It is found that the computed stresses are in reasonable agreement with tectonic features on Mars.

Sleep, N. H.↗
DOE OSTI2015

Partial Shade Stress Test for Thin-Film Photovoltaic Modules: Preprint

Partial shade of monolithic thin-film PV modules can cause reverse-bias conditions leading to permanent damage. In this work, we propose a partial shade stress test for thin-film PV modules that quantifies permanent performance loss. We designed the test with the aid of a computer model that predicts the local voltage, current and temperature stress that result from partial shade. The model predicts the module-scale interactions among the illumination pattern, the electrical properties of the photovoltaic material and the thermal properties of the module package. The test reproduces shading and loading conditions that may occur in the field. It accounts for reversible light-induced performance changes and for additional stress that may be introduced by light-enhanced reverse breakdown. We present simulated and experimental results from the application of the proposed test.

14 SOLAR ENERGY↗
NASA NTRS2013

Enhanced Contact Graph Routing (ECGR) MACHETE Simulation Model

Contact Graph Routing (CGR) for Delay/Disruption Tolerant Networking (DTN) space-based networks makes use of the predictable nature of node contacts to make real-time routing decisions given unpredictable traffic patterns. The contact graph will have been disseminated to all nodes before the start of route computation. CGR was designed for space-based networking environments where future contact plans are known or are independently computable (e.g., using known orbital dynamics). For each data item (known as a bundle in DTN), a node independently performs route selection by examining possible paths to the destination. Route computation could conceivably run thousands of times a second, so computational load is important. This work refers to the simulation software model of Enhanced Contact Graph Routing (ECGR) for DTN Bundle Protocol in JPL's MACHETE simulation tool. The simulation model was used for performance analysis of CGR and led to several performance enhancements. The simulation model was used to demonstrate the improvements of ECGR over CGR as well as other routing methods in space network scenarios. ECGR moved to using earliest arrival time because it is a global monotonically increasing metric that guarantees the safety properties needed for the solution's correctness since route re-computation occurs at each node to accommodate unpredicted changes (e.g., traffic pattern, link quality). Furthermore, using earliest arrival time enabled the use of the standard Dijkstra algorithm for path selection. The Dijkstra algorithm for path selection has a well-known inexpensive computational cost. These enhancements have been integrated into the open source CGR implementation. The ECGR model is also useful for route metric experimentation and comparisons with other DTN routing protocols particularly when combined with MACHETE's space networking models and Delay Tolerant Link State Routing (DTLSR) model.

Segui, John S.↗
DOE OSTI2026

Roadmap and Benchmarking: Privacy in Federated Load Forecasting

Data-driven techniques for energy demand forecasting continue to emerge with promising impacts on distribution grid planning. However, the development of robust and generalizable machine learning models requires that representative high quality training data are available. Distributed energy resources have begun to embed intelligence, gathering large amounts of data on customer demand, behavior, and household devices that are connected to the grid. Though utilities aggregate meter-level demand data for load shaping, demand response, outage management, reliability planning, and billing applications, there lies an inherent privacy concern in sharing consumption data that may identify individual consumer behavioral patterns. Hence, while sharing the data is crucial, the private sensitive customer data must be safeguarded from being exposed or manipulated. In this study, we propose a roadmap for implementing a based privacy preserving framework to support the advancement of data-driven analytics in data-sensitive distributed energy resources environments. The roadmap incorporates federated learning–a distributed training framework, differential privacy–a statistical framework that provides guarantees to safeguard the leakage of sensitive data, secure multiparty computation and homomorphic encryption– techniques for encrypting model gradients and applying secure aggregation on the server. Moreover, we perform baseline experiments on the federated short-term load forecasting (STLF) task using open-source residential load profile datasets, offering insights into the challenges of integrating differential privacy into federated learning.

Abebe, Waqwoya [Oak Ridge National Laboratory (ORN↗
NASA NTRS1974

Aperture excited dielectric antennas

The results of a comprehensive experimental and theoretical study of the effect of placing dielectric objects over the aperture of waveguide antennas are presented. Experimental measurements of the radiation patterns, gain, impedance, near-field amplitude, and pattern and impedance coupling between pairs of antennas are given for various Plexiglas shapes, including the sphere and the cube, excited by rectangular, circular, and square waveguide feed apertures. The waveguide excitation of a dielectric sphere is modeled using the Huygens' source, and expressions for the resulting electric fields, directivity, and efficiency are derived. Calculations using this model show good overall agreement with experimental patterns and directivity measurements. The waveguide under an infinite dielectric slab is used as an impedance model. Calculations using this model agree qualitatively with the measured impedance data. It is concluded that dielectric loaded antennas such as the waveguide excited sphere, cube, or sphere-cylinder can produce directivities in excess of that obtained by a uniformly illuminated aperture of the same cross section, particularly for dielectric objects with dimensions of 2 wavelengths or less. It is also shown that for certain configurations coupling between two antennas of this type is less than that for the same antennas without dielectric loading.

Crosswell, W. F.↗
NASA NTRS1959

Experimental Influence Coefficients and Vibration Modes

Test results are presented for both symmetrical and antisymmetrical static loading of a wing model mounted on a three-point support system. The first six free-free vibration modes were determined experimentally. A comparison is made of the symmetrical nodal patterns and frequencies with the symmetrical nodal patterns and frequencies calculated from the experimental influence coefficients.

Weidman, Deene J.↗
NASA NTRS2011

Quantifying the Aerosol Semi-Direct Effect in the NASA GEOS-5 AGCM

Aerosols such as black carbon, dust, and some organic carbon species both scatter and absorb incoming solar radiation. This direct aerosol radiative forcing (DARF) redistributes solar energy both by cooling the surface and warming the atmosphere. As a result, these aerosols affect atmospheric stability and cloud cover (the semi-direct effect, or SDE). Furthermore, in regions with persistent high loadings of absorbing aerosols (e.g. Asia), regional circulation patterns may be altered, potentially resulting in changes in precipitation patterns. Here we investigate aerosol-climate coupling using the NASA Goddard Earth Observing System model version 5 (GEOS-5) atmospheric general circulation model (AGCM), in which we have implemented an online version of the Goddard Chemistry, Aerosol, Radiation and Transport (GOCART) model. GOCART includes representations of the sources, sinks, and chemical transformation of externally mixed dust, sea salt, sulfate, and carbonaceous aerosols. We examine a series of free-running ensemble climate simulations of the present-day period (2000-2009) forced by observed sea surface temperatures to determine the impact of aerosols on the model climate. The SDE and response of each simulation is determined by differencing with respect to the control simulation (no aerosol forcing). In a free-running model, any estimate of the SDE includes changes in clouds due both to atmospheric heating from aerosols and changes in circulation. To try and quantify the SDE without these circulation changes we then examine the DARF and SDE in GEOS-5 with prescribed meteorological analyses introduced by the MERRA analysis. By doing so, we are able to examine changes in model clouds that occur on shorter scales (six hours). In the GEOS-5 data assimilation system (DAS), the analysis is defined as the best estimate of the atmospheric state at any given time, and it is determined by optimally combining a first-guess short-term GCM forecast with all available observations. The Incremental Analysis Update (IAU) is added to the model forecast tendencies to align them with the analysis every six hours, thus preventing longer timescale feedbacks due to the aerosol forcing. We calculate the SDE by comparing model runs with and without aerosols, and the difference in the IAU between these runs is a useful metric with which to evaluate the impact of the SDE on the model atmosphere and clouds. Decreasing the IAU indicates that the aerosol direct and semi-direct effects act to reduce the bias between the model and observations and vice versa.

Randles, Cynthia A.↗
DOE OSTI2019

A depth-averaged non-cohesive sediment transport model with improved discretization of flux and source terms

Here, this paper presents novel flux and source term treatments within a Godunov-type finite volume framework for predicting the depth-averaged shallow water flow and sediment transport with enhanced the accuracy and stability. The suspended load ratio is introduced to differentiate between the advection of the suspended load and the advection of water. A modified Harten, Lax and van Leer Riemann solver with the contact wave restored (HLLC) is derived for the flux calculation based on the new wave pattern involving the suspended load ratio. The source term calculation is enhanced by means of a novel splitting-point implicit discretization. The slope effect is introduced by modifying the critical shear stress, with two treatments being discussed. The numerical scheme is tested in five examples that comprise both fixed and movable beds. The model predictions show good agreement with measurement, except for cases where local three-dimensional effects dominate.

58 GEOSCIENCES↗
DOE OSTI2025

Probabilistic Predictions for Fastener Failure in the Sandia Mechanics Challenge Using the Discrete-Direct Uncertainty Quantification Approach

This paper documents the blind and post-blind analysis predictions for the 2023 Sandia Mechanics Challenge (SMC), which involved predicting the behavior of a threaded fastener joint structure subjected to shock loading. Utilizing repeat sets of fastener calibration data from various experimental configurations including tension, double shear, and joint tension, we developed a library of calibrated models which were propagated through the application model using the Discrete-Direct (DD) uncertainty quantification (UQ) approach. Although the initial blind predictions did not incorporate spare-sample processing to quantify fastener failure probabilities, the analyses yielded reasonable conclusions aligned with experimental results. In the post-blind analysis phase, we focused on enhancing the fidelity of the aluminum constitutive model and innovating the DD approach to obtain probabilistic predictions for fastener failure, particularly when quantities of interest (QoIs) approach their bounds. The improved aluminum model captures the behavior of the cantilever under shock loading more accurately, predicting both partial and complete cracks, although it tends to underpredict failure propagation. The enhanced DD approach facilitates probabilistic predictions that reflect the interdependent failure mechanisms of the fasteners and the cantilever, revealing that while certain fasteners are more likely to fail, the failure does not necessarily follow a progressive pattern. Overall, the post-blind analyses significantly improved the predictive capabilities of the model, providing valuable insights into the SMC application and establishing a robust foundation for informed engineering decisions. The methodology demonstrates a cost-effective and extensible approach suitable for a wide range of applications, highlighting the importance of uncertainty quantification to provide context for engineering decision making.

42 ENGINEERING↗
NASA NTRS1982

Detection of regional air pollution episodes utilizing satellite digital data in the visual range

Digital analyses of satellite visible data for selected high-sulfate cases over the northeastern U.S., on July 21 and 22, 1978, are compared with ground-based measurements. Quantitative information on total aerosol loading derived from the satellite digitized data using an atmospheric radiative transfer model is found to agree with the ground measurements, and it is shown that the extent and transport of the haze pattern may be monitored from the satellite data over the period of maximum intensity for the episode. Attention is drawn to the potential benefits of satellite monitoring of pollution episodes demonstrated by the model.

Burke, H.-H. K.↗
DOE OSTI2012

WIMS/PANTHER analysis of UO{sub 2}/MOX cores using embedded super-cells

This paper describes a method of analysing PWR UO{sub 2}MOX cores with WIMS/PANTHER. Embedded super-cells, run within the reactor code, are used to correct the standard methodology of using 2-group smeared data from single assembly lattice calculations. In many other codes the weakness of this standard approach has been improved for MOX by imposing a more realistic environment in the lattice code, or by improving the sophistication of the reactor code. In this approach an intermediate set of calculations is introduced, leaving both lattice and reactor calculations broadly unchanged. The essence of the approach is that the whole core is broken down into a set of 'embedded' super-cells, each extending over just four quarter assemblies, with zero leakage imposed at the assembly mid-lines. Each supercell is solved twice, first with a detailed multi-group pin-by-pin solution, and then with the standard single assembly approach. Correction factors are defined by comparing the two solutions, and these can be applied in whole core calculations. The restriction that all such calculations are modelled with zero leakage means that they are independent of each other and of the core-wide flux shape. This allows parallel pre-calculation for the entire cycle once the loading pattern has been determined, in much the same way that single assembly lattice calculations can be pre-calculated once the range of fuel types is known. Comparisons against a whole core pin-by-pin reference demonstrates that the embedding process does not introduce a significant error, even after burnup and refuelling. Comparisons against a WIMS reference demonstrate that a pin-by-pin multi-group diffusion solution is capable of capturing the main interface effects. This therefore defines a practical approach for achieving results close to lattice code accuracy, but broadly at the cost of a standard reactor calculation. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
NASA NTRS1990

Unsteady aerodynamic loading of delta wings for low and high angles of attack

Experimental and theoretical investigations dealing with unsteady flow phenomena are surveyed, with the emphasis on the pattern of vortices which originate from flow separation at sharp leading edges. It is concluded that these vortices exhibit quasi-steady behavior when the alpha-vibrations are such that bursting instability does not occur above the wing surface. A selection of test results from Jarrah (1988) is presented and discussed. For sharp-edged delta models at low speeds, the aerodynamic loads which are plotted quantify the role of parameters AR and K for three ranges of alpha-variation. An extremely approximate and empirical 'theory' is offered, with data on crossflow drag and burst location, to reproduce the behavior of these airloads up to 90 deg. Recent attempts to apply the more sophisticated tools of computational fluid dynamics to the combination of unsteadiness and very high alpha are shown to be deficient.

Ashley, H.↗
DOE OSTI2020

Analyzing the Validity of Brazilian Testing Using Digital Image Correlation and Numerical Simulation Techniques

Characterizing the mechanical behavior of rocks plays a crucial role to optimize the fracturing process in unconventional reservoirs. However, due to the intrinsic anisotropy and heterogeneity in unconventional resources, fracture process prediction remains the most significant challenge for sustainable and economic hydrocarbon production. During the deformation tracking under compression, deploying conventional methods (strain gauge, extensometer, etc.) is insufficient to measure the deformation since the physical attachment of the device is restricted to the size of the sample, monitoring limited point-wise deformation, producing difficulties in data retrieval, and a tendency to lose track in failure points, etc. Where conventional methods are limited, the application of digital image correlation (DIC) provides detailed and additional information of strain evolution and fracture patterns under loading. DIC is an image-based optical method that records an object with a camera and monitors the random contrast speckle pattern painted on the facing surface of the specimen. To overcome the existing limitations, this paper presents numerical modeling of Brazilian disc tests under quasi-static conditions to understand the full-field deformation behaviors and finally, it is validated by DIC. As the direct tensile test has limitations in sample preparation and test execution, the Brazilian testing principle is commonly used to evaluate indirectly the tensile strength of rocks. The two-dimensional numerical model was built to predict the stress distribution and full-field deformation on Brazilian disc under compression based on the assumptions of a homogenous, isotropic and linear elastic material. The uniaxial compression test was conducted using the DIC technique to determine the elastic properties of Spider Berea sandstone, which were used as inputs for the simulation model. The model was verified by the analytical solution and compared with the digital image correlation. The numerical simulation results showed that the solutions matched reasonably with the analytical solutions where the maximum deviation of stress distribution was obtained as 14.59%. The strain evolution (normal and shear strains) and displacements along the central horizontal and vertical planes were investigated in three distinguishable percentages of peak loads (20%, 40%, and 90%) to understand the deformation behaviors in rock. The simulation results demonstrated that the strain evolution contours consistently matched with DIC generated contours with a reasonable agreement. The changes in displacement along the central horizontal and vertical planes showed that numerical simulation and DIC generated experimental results were repeatable and matched closely. In terms of validation, Brazilian testing to measure the indirect tensile strength of rocks is still an issue of debate. The numerical model of fracture propagation supported by digital image correlation from this study can be used to explain the fracturing process in the homogeneous material and can be extended to non-homogeneous cases by incorporating heterogeneity, which is essential for rock mechanics field applications.

58 GEOSCIENCES↗
DOE OSTI2025

RingX: Scalable Parallel Attention for Long-Context Learning on HPC

The attention mechanism has become foundational for remarkable AI breakthroughs since the introduction of the Transformer, driving the demand for increasingly longer context to power frontier models such as large-scale reasoning language models and high-resolution image/video generators. However, its quadratic computational and memory complexities present substantial challenges. Current state-of-the-art parallel attention methods, such as ring attention, are widely adopted for long-context training but utilize a point-to-point communication strategy that fails to fully exploit the capabilities of modern HPC network architectures. In this work, we propose ringX, a scalable family of parallel attention methods optimized explicitly for HPC systems. By enhancing workload partitioning, refining communication patterns, and improving load balancing, ringX achieves up to 3.4 × speedup compared to conventional ring attention on the Frontier supercomputer. Optimized for both bi-directional and causal attention mechanisms, ringX demonstrates its effectiveness through training benchmarks of a Vision Transformer (ViT) on a climate dataset and a Generative Pre-Trained Transformer (GPT) model, Llama3 8B. Our method attains an end-to-end training speedup of approximately 1.5 × in both scenarios. To our knowledge, the achieved 38% model FLOPs utilization (MFU) for training Llama3 8B with a 1M-token sequence length on 4,096 GPUs represents one of the highest training efficiencies reported for long-context learning on HPC systems. Our code implementation is available at https://github.com/jqyin/ringX-attention.

Yin, Junqi [ORNL] (ORCID:0000000338435520)↗
NASA NTRS2017

Environmental Barrier Coating Fracture, Fatigue and High-Heat-Flux Durability Modeling and Stochastic Progressive Damage Simulation

Advanced environmental barrier coatings will play an increasingly important role in future gas turbine engines because of their ability to protect emerging light-weight SiC/SiC ceramic matrix composite (CMC) engine components, further raising engine operating temperatures and performance. Because the environmental barrier coating systems are critical to the performance, reliability and durability of these hot-section ceramic engine components, a prime-reliant coating system along with established life design methodology are required for the hot-section ceramic component insertion into engine service. In this paper, we have first summarized some observations of high temperature, high-heat-flux environmental degradation and failure mechanisms of environmental barrier coating systems in laboratory simulated engine environment tests. In particular, the coating surface cracking morphologies and associated subsequent delamination mechanisms under the engine level high-heat-flux, combustion steam, and mechanical creep and fatigue loading conditions will be discussed. The EBC compostion and archtechture improvements based on advanced high heat flux environmental testing, and the modeling advances based on the integrated Finite Element Analysis Micromechanics Analysis Code/Ceramics Analysis and Reliability Evaluation of Structures (FEAMAC/CARES) program will also be highlighted. The stochastic progressive damage simulation successfully predicts mud flat damage pattern in EBCs on coated 3-D specimens, and a 2-D model of through-the-thickness cross-section. A 2-parameter Weibull distribution was assumed in characterizing the coating layer stochastic strength response and the formation of damage was therefore modeled. The damage initiation and coalescence into progressively smaller mudflat crack cells was demonstrated. A coating life prediction framework may be realized by examining the surface crack initiation and delamination propagation in conjunction with environmental degradation under high-heat-flux and environment load test conditions.

Mechanical creep and fatigue loading condition↗
DOE OSTI2026

Knowledge-guided graph machine learning for spatially distributed prediction of daily discharge and nitrogen export dynamics

Spatially distributed prediction of streamflow and nitrogen export dynamics is essential for precision management of agricultural watersheds. While temporal deep learning models such as Long Short-Term Memory (LSTM) have shown strong performance at basin scales, their ability to generalize spatially is limited by insufficient representation of spatial dependencies and flow paths, particularly under data-scarce conditions. To address this gap, we propose HydroGraphNet, a knowledge-guided graph machine learning framework that integrates process-based knowledge and explicit spatial learning into temporal modeling. This framework incorporates directed graph topology to encode watershed connectivity and upstream inflows, with mass balance constraints to improve physical consistency. To enhance generalization in sparsely monitored regions, HydroGraphNet is pretrained on synthetic data generated by the SWAT+ (Soil and Water Assessment Tool Plus) model. We evaluated HydroGraphNet in the Upper Sangamon River Basin (44 HUC-12 subwatersheds, 2001–2020) against two LSTM baselines: a lumped basin-level model and a distributed variant. When benchmarked on SWAT+ simulations in pretraining, HydroGraphNet improved test NSEs by 8.9% (discharge) and 13.7% (NO₃–N load) in temporal extrapolation, and by 27.1% and 34.7% in spatial extrapolation, relative to the Lumped LSTM baseline. After fine-tuning with USGS monitoring data, the model achieved mean test NSE (KGE) scores of 0.768 (0.861) for discharge and 0.626 (0.664) for NO₃–N load, substantially outperforming baselines. Attribution analysis further highlighted the importance of upstream inflow representation and graph-based spatial learning in capturing cross-subwatershed dependencies. The model also reproduced seasonal hydrological and biogeochemical patterns consistent with known processes, demonstrating its robustness and process fidelity for spatially distributed prediction. Altogether, HydroGraphNet advances the integration of physical knowledge and spatially explicit learning in hydrological modeling, offering a generalizable framework for distributed modeling to support spatially targeted water quality management in data-scarce watersheds.

54 ENVIRONMENTAL SCIENCES↗