Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “predicted”

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 649 records · Page 36
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

A deep dilated convolutional residual network for predicting interchain contacts of protein homodimers

Abstract Motivation Deep learning has revolutionized protein tertiary structure prediction recently. The cutting-edge deep learning methods such as AlphaFold can predict high-accuracy tertiary structures for most individual protein chains. However, the accuracy of predicting quaternary structures of protein complexes consisting of multiple chains is still relatively low due to lack of advanced deep learning methods in the field. Because interchain residue–residue contacts can be used as distance restraints to guide quaternary structure modeling, here we develop a deep dilated convolutional residual network method (DRCon) to predict interchain residue–residue contacts in homodimers from residue–residue co-evolutionary signals derived from multiple sequence alignments of monomers, intrachain residue–residue contacts of monomers extracted from true/predicted tertiary structures or predicted by deep learning, and other sequence and structural features. Results Tested on three homodimer test datasets (Homo_std dataset, DeepHomo dataset and CASP-CAPRI dataset), the precision of DRCon for top L/5 interchain contact predictions (L: length of monomer in a homodimer) is 43.46%, 47.10% and 33.50% respectively at 6 Å contact threshold, which is substantially better than DeepHomo and DNCON2_inter and similar to Glinter. Moreover, our experiments demonstrate that using predicted tertiary structure or intrachain contacts of monomers in the unbound state as input, DRCon still performs well, even though its accuracy is lower than using true tertiary structures in the bound state are used as input. Finally, our case study shows that good interchain contact predictions can be used to build high-accuracy quaternary structure models of homodimers. Availability and implementation The source code of DRCon is available at https://github.com/jianlin-cheng/DRCon. The datasets are available at https://zenodo.org/record/5998532#.YgF70vXMKsB. Supplementary information Supplementary data are available at Bioinformatics online.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Accurate flux predictions using tissue-specific gene expression in plant metabolic modeling

The accurate prediction of complex phenotypes such as metabolic fluxes in living systems is a grand challenge for systems biology and central to efficiently identifying biotechnological interventions that can address pressing industrial needs. The application of gene expression data to improve the accuracy of metabolic flux predictions using mechanistic modeling methods such as flux balance analysis (FBA) has not been previously demonstrated in multi-tissue systems, despite their biotechnological importance. We hypothesized that a method for generating metabolic flux predictions informed by relative expression levels between tissues would improve prediction accuracy. Relative gene expression levels derived from multiple transcriptomic and proteomic datasets were integrated into FBA predictions of a multi-tissue, diel model of Arabidopsis thaliana’s central metabolism. This integration dramatically improved the agreement of flux predictions with experimentally based flux maps from 13 C metabolic flux analysis compared with a standard parsimonious FBA approach. Disagreement between FBA predictions and MFA flux maps was measured using weighted averaged percent error values, and for parsimonious FBA this was 169%–180% for high light conditions and 94%–103% for low light conditions, depending on the gene expression dataset used. This fell to 10%-13% and 9%-11% upon incorporating expression data into the modeling process, which also substantially altered the predicted carbon and energy economy of the plant.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Genomic prediction of hybrid performance for agronomic traits in sorghum

Hybrid breeding in sorghum [Sorghum bicolor (L.) Moench] utilizes the cytoplasmic-nuclear male sterility (CMS) system for seed production and subsequently harnesses heterosis. Since the cost of developing and evaluating inbred and hybrid lines in the CMS system is costly and time-consuming, genomic prediction of parental lines and hybrids is based on genetic data genotype. We generated 602 hybrids by crossing two female (A) lines with 301 diverse and elite male (R) lines from the sorghum association panel and collected phenotypic data for agronomic traits over two years. We genotyped the inbred parents using whole genome resequencing and used 2,687,342 high quality (minor allele frequency > 2%) single nucleotide polymorphisms for genomic prediction. For grain yield, the experimental hybrids exhibited an average mid-parent heterosis of 40%. Genomic best linear unbiased prediction (GBLUP) for hybrid performance yielded an average prediction accuracy of 0.76–0.93 under the prediction scenario where both parental lines in validation sets were included in the training sets (T2). However, when only female tester was shared between training and validation sets (T1F), prediction accuracies declined by 12–90%, with plant height showing the greatest decline. Mean accuracies for predicting the general combining ability of male parents ranged from 0.33 to 0.62 for all traits. Our results showed hybrid performance for agronomic traits can be predicted with high accuracy, and optimizing genomic relationship is essential for optimal training population design for genomic selection in sorghum breeding.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2017

Stochastic Short-term High-resolution Prediction of Solar Irradiance and Photovoltaic Power Output

The increased penetration of solar photovoltaic (PV) energy sources into electric grids has increased the need for accurate modeling and prediction of solar irradiance and power production. Existing modeling and prediction techniques focus on long-term low-resolution prediction over minutes to years. This paper examines the stochastic modeling and short-term high-resolution prediction of solar irradiance and PV power output. We propose a stochastic state-space model to characterize the behaviors of solar irradiance and PV power output. This prediction model is suitable for the development of optimal power controllers for PV sources. A filter-based expectation-maximization and Kalman filtering mechanism is employed to estimate the parameters and states in the state-space model. The mechanism results in a finite dimensional filter which only uses the first and second order statistics. The structure of the scheme contributes to a direct prediction of the solar irradiance and PV power output without any linearization process or simplifying assumptions of the signal’s model. This enables the system to accurately predict small as well as large fluctuations of the solar signals. The mechanism is recursive allowing the solar irradiance and PV power to be predicted online from measurements. The mechanism is tested using solar irradiance and PV power measurement data collected locally in our laboratory.

Olama, Mohammed M.↗
DOE OSTI2014

Predictability of the terrestrial carbon cycle

Abstract Terrestrial ecosystems sequester roughly 30% of anthropogenic carbon emission. However this estimate has not been directly deduced from studies of terrestrial ecosystems themselves, but inferred from atmospheric and oceanic data. This raises a question: to what extent is the terrestrial carbon cycle intrinsically predictable? In this paper, we investigated fundamental properties of the terrestrial carbon cycle, examined its intrinsic predictability, and proposed a suite of future research directions to improve empirical understanding and model predictive ability. Specifically, we isolated endogenous internal processes of the terrestrial carbon cycle from exogenous forcing variables. The internal processes share five fundamental properties (i.e., compartmentalization, carbon input through photosynthesis, partitioning among pools, donor pool‐dominant transfers, and the first‐order decay) among all types of ecosystems on the Earth. The five properties together result in an emergent constraint on predictability of various carbon cycle components in response to five classes of exogenous forcing. Future observational and experimental research should be focused on those less predictive components while modeling research needs to improve model predictive ability for those highly predictive components. We argue that an understanding of predictability should provide guidance on future observational, experimental and modeling research.

Luo, Yiqi↗
DOE OSTI2020

Traffic Prediction for Merging Coordination Control in Mixed Traffic Scenarios

Connected and autonomous vehicles (CAVs) have the potential to bring in safety, mobility, and energy benefits to transportation. The control decisions of CAVs are usually determined for a look-ahead horizon based on previewed traffic information. This requires an effective prediction of future traffic conditions and its integration with the CAV control framework. However, the short-term traffic prediction using information from connectivity is a challenging research topic, especially for mixed traffic scenarios. This work focuses on the development of a traffic prediction framework for a merging coordination controller. The previously developed merging controller coordinates the merging sequence and travel speed of CAVs to maximize the energy efficiency and overall mobility. In mixed traffic scenarios, the controller receives information regarding the position of all the vehicles traveling inside a control zone and controls the desired speed of all CAVs. The controller has no control on the human-driven vehicles. The merging controller does not have direct information or an explicit prediction on the behaviors of human-driven vehicles. Aiming to improve the performance of the merging controller in various mixed traffic conditions, a traffic prediction algorithm is developed and evaluated in this work. The performance of this traffic prediction algorithm is investigated for various penetration rates of connectivity for a single-lane secondary road merging to a single-lane primary road. The results show that compared to a constant speed assumption of human-driven vehicles, the proposed traffic prediction algorithm is able to reduce the prediction error of the arrival time of human-driven vehicles at the merging zone by more than 50%.

Shao, Yunli↗
DOE OSTI2014

SU-C-BRF-07: A Pattern Fusion Algorithm for Multi-Step Ahead Prediction of Surrogate Motion

Purpose: To assure that tumor motion is within the radiation field during high-dose and high-precision radiosurgery, real-time imaging and surrogate monitoring are employed. These methods are useful in providing real-time tumor/surrogate motion but no future information is available. In order to anticipate future tumor/surrogate motion and track target location precisely, an algorithm is developed and investigated for estimating surrogate motion multiple-steps ahead. Methods: The study utilized a one-dimensional surrogate motion signal divided into three components: (a) training component containing the primary data including the first frame to the beginning of the input subsequence; (b) input subsequence component of the surrogate signal used as input to the prediction algorithm: (c) output subsequence component is the remaining signal used as the known output of the prediction algorithm for validation. The prediction algorithm consists of three major steps: (1) extracting subsequences from training component which best-match the input subsequence according to given criterion; (2) calculating weighting factors from these best-matched subsequence; (3) collecting the proceeding parts of the subsequences and combining them together with assigned weighting factors to form output. The prediction algorithm was examined for several patients, and its performance is assessed based on the correlation between prediction and known output. Results: Respiratory motion data was collected for 20 patients using the RPM system. The output subsequence is the last 50 samples (∼2 seconds) of a surrogate signal, and the input subsequence was 100 (∼3 seconds) frames prior to the output subsequence. Based on the analysis of correlation coefficient between predicted and known output subsequence, the average correlation is 0.9644±0.0394 and 0.9789±0.0239 for equal-weighting and relative-weighting strategies, respectively. Conclusion: Preliminary results indicate that the prediction algorithm is effective in estimating surrogate motion multiple-steps in advance. Relative-weighting method shows better prediction accuracy than equal-weighting method. More parameters of this algorithm are under investigation.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

SU-F-E-09: Respiratory Signal Prediction Based On Multi-Layer Perceptron Neural Network Using Adjustable Training Samples

Purpose: Dynamic tracking of moving organs, such as lung and liver tumors, under radiation therapy requires prediction of organ motions prior to delivery. The shift of moving organ may change a lot due to huge transform of respiration at different periods. This study aims to reduce the influence of that changes using adjustable training signals and multi-layer perceptron neural network (ASMLP). Methods: Respiratory signals obtained using a Real-time Position Management(RPM) device were used for this study. The ASMLP uses two multi-layer perceptron neural networks(MLPs) to infer respiration position alternately and the training sample will be updated with time. Firstly, a Savitzky-Golay finite impulse response smoothing filter was established to smooth the respiratory signal. Secondly, two same MLPs were developed to estimate respiratory position from its previous positions separately. Weights and thresholds were updated to minimize network errors according to Leverberg-Marquart optimization algorithm through backward propagation method. Finally, MLP 1 was used to predict 120∼150s respiration position using 0∼120s training signals. At the same time, MLP 2 was trained using 30∼150s training signals. Then MLP is used to predict 150∼180s training signals according to 30∼150s training signals. The respiration position is predicted as this way until it was finished. Results: In this experiment, the two methods were used to predict 2.5 minute respiratory signals. For predicting 1s ahead of response time, correlation coefficient was improved from 0.8250(MLP method) to 0.8856(ASMLP method). Besides, a 30% improvement of mean absolute error between MLP(0.1798 on average) and ASMLP(0.1267 on average) was achieved. For predicting 2s ahead of response time, correlation coefficient was improved from 0.61415 to 0.7098.Mean absolute error of MLP method(0.3111 on average) was reduced by 35% using ASMLP method(0.2020 on average). Conclusion: The preliminary results demonstrate that the ASMLP respiratory prediction method is more accurate than MLP method and can improve the respiration forecast accuracy.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

SU-F-J-218: Predicting Radiation-Induced Xerostomia by Dosimetrically Accounting for Daily Setup Uncertainty During Head and Neck IMRT

Purpose: To determine if the accumulated parotid dosimetry using planning CT to daily CBCT deformation and dose re-calculation can predict for radiation-induced xerostomia. Methods: To track and dosimetrically account for the effects of anatomical changes on the parotid glands, we propagated physicians’ contours from planning CT to daily CBCT using a deformable registration with iterative CBCT intensity correction. A surface mesh for each OAR was created with the deformation applied to the mesh to obtain the deformed parotid volumes. Daily dose was computed on the deformed CT and accumulated to the last fraction. For both the accumulated and the planned parotid dosimetry, we tested the prediction power of different dosimetric parameters including D90, D50, D10, mean, standard deviation, min/max dose to the combined parotids and patient age to severe xerostomia (NCI-CTCAE grade≥2 at 6 mo follow-up). We also tested the dosimetry to parotid sub-volumes. Three classification algorithms, random tree, support vector machine, and logistic regression were tested to predict severe xerostomia using a leave-one-out validation approach. Results: We tested our prediction model on 35 HN IMRT cases. Parameters from the accumulated dosimetry model demonstrated an 89% accuracy for predicting severe xerostomia. Compared to the planning dosimetry, the accumulated dose consistently demonstrated higher prediction power with all three classification algorithms, including 11%, 5% and 30% higher accuracy, sensitivity and specificity, respectively. Geometric division of the combined parotid glands into superior-inferior regions demonstrated ∼5% increased accuracy than the whole volume. The most influential ranked features include age, mean accumulated dose of the submandibular glands and the accumulated D90 of the superior parotid glands. Conclusion: We demonstrated that the accumulated parotid dosimetry using CT-CBCT registration and dose re-calculation more accurately predicts for severe xerostomia and that the superior portion of the parotid glands may be particularly important in predicting for severe xerostomia. This work was supported in part by NIH/NCI under grant R42CA137886 and in part by Toshiba big data research project funds.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2023

Machine Learning for Daily Forecasts of Arctic Sea Ice Motion: An Attribution Assessment of Model Predictive Skill

Physics-based simulations of Arctic sea ice are highly complex, involving transport between different phases, length scales, and time scales. Resultantly, numerical simulations of sea ice dynamics have a high computational cost and model uncertainty. We employ data-driven machine learning (ML) to make predictions of sea ice motion. The ML models are built to predict present-day sea ice velocity given present-day wind velocity and previous-day sea ice concentration and velocity. Models are trained using reanalysis winds and satellite-derived sea ice properties. We compare the predictions of three different models: persistence (PS), linear regression (LR), and a convolutional neural network (CNN). We quantify the spatiotemporal variability of the correlation between observations and the statistical model predictions. Additionally, we analyze model performance in comparison to variability in properties related to ice motion (wind velocity, ice velocity, ice concentration, distance from coast, bathymetric depth) to understand the processes related to decreases in model performance. Results indicate that a CNN makes skillful predictions of daily sea ice velocity with a correlation up to 0.81 between predicted and observed sea ice velocity, while the LR and PS implementations exhibit correlations of 0.78 and 0.69, respectively. The correlation varies spatially and seasonally: lower values occur in shallow coastal regions and during times of minimum sea ice extent. LR parameter analysis indicates that wind velocity plays the largest role in predicting sea ice velocity on 1-day time scales, particularly in the central Arctic. Regions where wind velocity has the largest LR parameter are regions where the CNN has higher predictive skill than the LR.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Evidential Deep Learning: Enhancing Predictive Uncertainty Estimation for Earth System Science Applications

Abstract Robust quantification of predictive uncertainty is a critical addition needed for machine learning applied to weather and climate problems to improve the understanding of what is driving prediction sensitivity. Ensembles of machine learning models provide predictive uncertainty estimates in a conceptually simple way but require multiple models for training and prediction, increasing computational cost and latency. Parametric deep learning can estimate uncertainty with one model by predicting the parameters of a probability distribution but does not account for epistemic uncertainty. Evidential deep learning, a technique that extends parametric deep learning to higher-order distributions, can account for both aleatoric and epistemic uncertainties with one model. This study compares the uncertainty derived from evidential neural networks to that obtained from ensembles. Through applications of the classification of winter precipitation type and regression of surface-layer fluxes, we show evidential deep learning models attaining predictive accuracy rivaling standard methods while robustly quantifying both sources of uncertainty. We evaluate the uncertainty in terms of how well the predictions are calibrated and how well the uncertainty correlates with prediction error. Analyses of uncertainty in the context of the inputs reveal sensitivities to underlying meteorological processes, facilitating interpretation of the models. The conceptual simplicity, interpretability, and computational efficiency of evidential neural networks make them highly extensible, offering a promising approach for reliable and practical uncertainty quantification in Earth system science modeling. To encourage broader adoption of evidential deep learning, we have developed a new Python package, Machine Integration and Learning for Earth Systems (MILES) group Generalized Uncertainty for Earth System Science (GUESS) (MILES-GUESS) ( https://github.com/ai2es/miles-guess ), that enables users to train and evaluate both evidential and ensemble deep learning. Significance Statement This study demonstrates a new technique, evidential deep learning, for robust and computationally efficient uncertainty quantification in modeling the Earth system. The method integrates probabilistic principles into deep neural networks, enabling the estimation of both aleatoric uncertainty from noisy data and epistemic uncertainty from model limitations using a single model. Our analyses reveal how decomposing these uncertainties provides valuable insights into reliability, accuracy, and model shortcomings. We show that the approach can rival standard methods in classification and regression tasks within atmospheric science while offering practical advantages such as computational efficiency. With further advances, evidential networks have the potential to enhance risk assessment and decision-making across meteorology by improving uncertainty quantification, a longstanding challenge. This work establishes a strong foundation and motivation for the broader adoption of evidential learning, where properly quantifying uncertainties is critical yet lacking.

Schreck, John S.↗
DOE OSTI2020

Two-Stream Multi-Channel Convolutional Neural Network for Multi-Lane Traffic Speed Prediction Considering Traffic Volume Impact

Traffic speed prediction is a critically important component of intelligent transportation systems. Recently, with the rapid development of deep learning and transportation data science, a growing body of new traffic speed prediction models have been designed that achieved high accuracy and large-scale prediction. However, existing studies have two major limitations. First, they predict aggregated traffic speed rather than lane-level traffic speed; second, most studies ignore the impact of other traffic flow parameters in speed prediction. To address these issues, the authors propose a two-stream multi-channel convolutional neural network (TM-CNN) model for multi-lane traffic speed prediction considering traffic volume impact. In this model, the authors first introduce a new data conversion method that converts raw traffic speed data and volume data into spatial–temporal multi-channel matrices. Then the authors carefully design a two-stream deep neural network to effectively learn the features and correlations between individual lanes, in the spatial–temporal dimensions, and between speed and volume. Accordingly, a new loss function that considers the volume impact in speed prediction is developed. A case study using 1-year data validates the TM-CNN model and demonstrates its superiority. This paper contributes to two research areas: (1) traffic speed prediction, and (2) multi-lane traffic flow study.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Accelerated coupled Monte Carlo-Thermal hydraulic calculations using a hybrid GTF-diffusion-based prediction block: first results

Accurate predictions of spatial power and temperature distributions require the coupling of a neutron transport solver with a thermal-hydraulic (TH) feedback. Nowadays, Monte Carlo (MC) codes are widely coupled to TH solvers, typically via a Picard iteration (PI) method, due to the higher fidelity that such frameworks can produce. To speed up a PI, a prediction step can produce an improved initial guess for a source distribution and feed it to the MC code. Recent work investigated a prediction step that uses generalized transfer functions (GTFs) to predict the macroscopic cross sections' variations following a perturbation in TH properties, such as coolant density. The previous method also relied on first order perturbation (FOP) theory to predict perturbed power profiles, rather than using an expensive MC iterate. The implemented FOP method relied on generating a fission matrix from which the forward and adjoint Eigenmodes were extracted and later used to for power calculations. The generation of the fission matrix can introduce a significant computational overhead, therefore undermining the performance of the proposed hybrid technique when applied to high-dimensional problems, e.g., full core calculations. This work attempts to improve the GTF-FOP prediction step by replacing the FOP solver with a nodal diffusion solver, thus eliminating the need to calculate a fission matrix. The GTF-diffusion step was tested for various moderator density perturbations. In each case, the predicted power distribution showed good agreement with the reference case. The latter is attributed to the generally good prediction of most spatially distributed macroscopic cross sections, except the transport cross section, which will become the focus of future work. (authors)

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2015

Total Dissolved Gas Prediction and Optimization in RIVERWARE

Management and operation of dams within the Columbia River Basin (CRB) provides the region with irrigation, hydropower production, flood control, navigation, and fish passage. These various system-wide demands can require unique dam operations that may result in both voluntary and involuntary spill, thereby increasing tailrace levels of total dissolved gas (TDG) which can be fatal to fish. Appropriately managing TDG levels within the context of the systematic demands requires a predictive framework robust enough to capture the operationally related effects on TDG levels. Development of the TDG predictive methodology herein attempts to capture the different modes of hydro operation, thereby making it a viable tool to be used in conjunction with a real-time scheduling model such as RiverWare. The end result of the effort will allow hydro operators to minimize system-wide TDG while meeting hydropower operational targets and constraints. The physical parameters such as spill and hydropower flow proportions, accompanied by the characteristics of the dam such as plant head levels and tailrace depths, are used to develop the empirically-based prediction model. In the broader study, two different models are developed a simplified and comprehensive model. The latter model incorporates more specific bubble physics parameters for the prediction of tailrace TDG levels. The former model is presented herein and utilizes an empirically based approach to predict downstream TDG levels based on local saturation depth, spillway and powerhouse flow proportions, and entrainment effects. Representative data collected from each of the hydro projects is used to calibrate and validate model performance and the accuracy of predicted TDG uptake. ORNL, in conjunction with IIHR - Hydroscience & Engineering, The University of Iowa, carried out model adjustments to adequately capture TDG levels with respect to each plant while maintaining a generalized model configuration. Validation results indicate excellent model performance with coefficient of determination values exceeding 92% for all sites. This approach enables model extension to an increasingly wider array of hydropower plants, i.e., with the proper data input, TDG uptake can be calculated independent of actual physical component design. The TDG model is used as a module in the systematic optimization framework of RiverWare, a river and reservoir modeling tool used by federal agencies, public utility districts, and other dam owners and operators to forecast, schedule, and manage hydropower assets. The integration and testing of the TDG module within RiverWare, led by University of Colorado's Center for Advanced Decision Support for Water and Environmental Systems (CADSWES), will allow users to generate optimum system schedules based on the minimization of TDG. Optimization analysis and added value will be quantified as system wide reductions in TDG achieved while meeting existing hydropower constraints. Future work includes the development of a method to predict downstream reservoir forebay TDG levels as a function of upstream reservoir tailrace TDG values based on river hydrodynamics, hydro operations, and reservoir characteristics. Once implemented, a holistic model that predicts both TDG uptake and transport will give hydropower operators valuable insight into how system-wide environmental effects can be mitigated while simultaneously balancing stakeholder interests.

13 HYDRO ENERGY↗
DOE OSTI2019

Computational Bayesian Framework for Quantification and Reduction of Predictive Uncertainty in Subsurface Environmental Modeling

Subsurface environmental systems are open and complex, in which intricate hydrologic, microbiologic, and geochemical processes occur and interact at multiple scales. Understanding and predicting system responses to natural forces (e.g., climate changes) and human activities (e.g., contaminant remediation and CO 2 sequestration) is indispensable for managing water resources, cleaning subsurface contamination at many sites of the Department of Energy (DOE), and providing expert analysis to inform policy-making on long-term stewardship of nuclear waste disposal and CO 2 storage sites. Groundwater reactive transport modeling is necessary for understanding and predicting the complex subsurface environmental system. However, as complex systems are never completely predictable, model predictions of the subsurface system are inherently uncertain, and uncertainty is one of the greatest obstacles in simulating the subsurface environment. Quantifying predictive uncertainty of the complex system is critical to test model predictive power, support decision-making, and guide experimental design and data collection for uncertainty reduction. This project tackled two major obstacles to efficient and effective quantification and reduction of predictive uncertainty in simulating the complex subsurface environment. One obstacle is that existing methods of uncertainty quantification are too fragmented and incomplete for understanding and predicting the complex subsurface environment as a whole. The other obstacle is that existing methods of uncertainty quantification are computationally demanding and may not be applicable to groundwater reactive transport modeling that is always computationally expensive. The above two obstacles to uncertainty quantification directly affect the design of data collection schemes for uncertainty reduction. In this project, we developed a comprehensive Bayesian framework to remove the first obstacle. The second obstacle was tackled by using the sparse grid methods, which were originally developed and analyzed for numerical integration in the field of applied mathematics. We integrated the sparse grid methods into the Bayesian framework seamlessly, which led to a sparse-grid-based method of Bayesian uncertainty analysis. The developed computational Bayesian framework has been applied to a number of subsurface environmental problems through collaborating with scientists at the Pacific Northwest National Laboratory and Oak Ridge National Laboratory. The applications demonstrated that the computational Bayesian framework is general and compatible with any models and numerical codes of subsurface modeling. It is thus can be applied to gain insights for subsurface environmental modeling. This project helped the PI achieve his career goal to advance the state-of-the-art computational methods in uncertainty quantification for subsurface environmental modeling. This project supported two doctoral students, two master students, and one post-doc. The two doctoral students, after their graduation, worked as post-docs at the Pacific Northwest National Laboratory and the Oak Ridge National Laboratory. The project produced a total of 29 journal articles (four under review), one book chapter, and a number of conference proceedings and abstracts.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Application of Modified Meshgraphnets for Subsurface Prediction during CO2 Sequestration

In the face of the increasingly dire consequences of anthropogenic climate change, capturing and storing carbon dioxide is paramount. However, several impediments exist to the safe and effective subsurface storage of CO2, such as cost of transport, identification of suitable sites for subsurface storage, and assessment of long-term risk from storage in subsurface aquifers. Accurate subsurface modeling is necessary to ensure that CO2 storage is both safe and effective. Still, such modeling has traditionally required either substantial time and computational power (numerical simulation) or a substantial amount of pre-existing data for training (machine learning models). Additionally, these models lack flexibility in dealing with both changes in discretization of the input data and generalizability beyond the data on which they are trained. In order to address these issues, this research applies graph neural networks (GNNs) to predict subsurface saturation and pressure during CO₂ injection in a model of the Illinois Basin-Decatur Project (IBDP). GNNs provide a flexible, intuitive method for representing and manipulating complex unstructured data, which is often found in many practical domain problems such as fluid flow and subsurface characterization. These unstructured grids are easily represented in GNNs by representing spatially-localized features such as permeability, porosity, saturation, and pressure as nodes in a graph and relationships between these properties as edges connecting these nodes. This research applies a specific GNN model called MeshGraphNets (MGN) to model the change in CO2 saturation and pressure over a 50-month time period (36 months of injection, 14 months post-injection). The MGN model leverages a message passing process that allows the network to learn both the spatial and temporal dynamics of this system simultaneously. Additionally, training on a limited dataset (64 realizations, 20 time points each) resulted in a high degree of accuracy in saturation prediction both within the same timeframe as the training (20 months, 0.039 average RMSE) and when projecting out to the end of injection (36 months, 0.053 average RMSE). Temporal predictions such as those generated by MGNs and other similar models are prone to accumulated error over time; in order to address this, a multi-step rollout (MSR) training process was applied to calculate training loss. This method mimics the forward prediction during inference by “rolling out” multiple time points in a single training step using the previous prediction as input to the MGN model. By calculating the loss several time steps forward from the current prediction, the model is forced to find a more stable state over time. Application of MSR to the MGN model resulted in an average 15% reduction in inference error over time during forward prediction. This study showcases the immense potential of GNNs as a game-changing methodology for predicting pressure and saturation evolution in CCS projects, ultimately paving the way for more sustainable and effective carbon storage solutions. Presentation prepared for the 2024 AiChE Annual Meeting, October 27 to November 1 2024, San Diego, CA.

Holcomb, Paul↗
DOE OSTI2021

Hardware-Based Emulator with Deep Learning Model for Building Energy Control and Prediction Based on Occupancy Sensors’ Data

Heating, ventilation, and air conditioning (HVAC) is the largest source of residential energy consumption. Occupancy sensors’ data can be used for HVAC control since it indicates the number of people in the building. HVAC and sensors form a typical cyber-physical system (CPS). In this paper, we aim to build a hardware-based emulation platform to study the occupancy data’s features, which can be further extracted by using machine learning models. In particular, we propose two hardware-based emulators to investigate the use of wired/wireless communication interfaces for occupancy sensor-based building CPS control, and the use of deep learning to predict the building energy consumption with the sensor data. We hypothesize is that the building energy consumption may be predicted by using the occupancy data collected by the sensors, and question what type of prediction model should be used to accurately predict the energy load. Another hypothesis is that an in-lab hardware/software platform could be built to emulate the occupancy sensing process. The machine learning algorithms can then be used to analyze the energy load based on the sensing data. To test the emulator, the occupancy data from the sensors is used to predict energy consumption. The synchronization scheme between sensors and the HVAC server will be discussed. We have built two hardware/software emulation platforms to investigate the sensor/HVAC integration strategies, and used an enhanced deep learning model—which has sequence-to-sequence long short-term memory (Seq2Seq LSTM)—with an attention model to predict the building energy consumption with the preservation of the intrinsic patterns. Because the long-range temporal dependencies are captured, the Seq2Seq models may provide a higher accuracy by using LSTM architectures with encoder and decoder. Meanwhile, LSTMs can capture the temporal and spatial patterns of time series data. The attention model can highlight the most relevant input information in the energy prediction by allocating the attention weights. The communication overhead between the sensors and the HVAC control server can also be alleviated via the attention mechanism, which can automatically ignore the irrelevant information and amplify the relevant information during CNN training. Our experiments and performance analysis show that, compared with the traditional LSTM neural network, the performance of the proposed method has a 30% higher prediction accuracy.

Ye, Zhijing↗
DOE OSTI2020

Predictions of the glass transition temperature and viscosity of organic aerosols from volatility distributions

Abstract. Volatility and viscosity are important properties of organic aerosols (OA), affecting aerosol processes such as formation, evolution, and partitioning of OA. Volatility distributions of ambient OA particles have often been measured, while viscosity measurements are scarce. We have previously developed a method to estimate the glass transition temperature (Tg) of an organic compound containing carbon, hydrogen, and oxygen. Based on analysis of over 2400 organic compounds including oxygenated organic compounds, as well as nitrogen- and sulfur-containing organic compounds, we extend this method to include nitrogen- and sulfur-containing compounds based on elemental composition. In addition, parameterizations are developed to predict Tg as a function of volatility and the atomic oxygen-to-carbon ratio based on a negative correlation between Tg and volatility. This prediction method of Tg is applied to ambient observations of volatility distributions at 11 field sites. The predicted Tg values of OA under dry conditions vary mainly from 290 to 339 K and the predicted viscosities are consistent with the results of ambient particle-phase-state measurements in the southeastern US and the Amazonian rain forest. Reducing the uncertainties in measured volatility distributions would improve predictions of viscosity, especially at low relative humidity. We also predict the Tg of OA components identified via positive matrix factorization of aerosol mass spectrometer (AMS) data. The predicted viscosity of oxidized OA is consistent with previously reported viscosity of secondary organic aerosols (SOA) derived from α-pinene, toluene, isoprene epoxydiol (IEPOX), and diesel fuel. Comparison of the predicted viscosity based on the observed volatility distributions with the viscosity simulated by a chemical transport model implies that missing low volatility compounds in a global model can lead to underestimation of OA viscosity at some sites. The relation between volatility and viscosity can be applied in the molecular corridor or volatility basis set approaches to improve OA simulations in chemical transport models by consideration of effects of particle viscosity in OA formation and evolution.

54 ENVIRONMENTAL SCIENCES↗