Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “surrogate approximator”

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 253 records · Page 14
All sourcesNASA NTRSDOE OSTI
NASA NTRS2008

Aerosol and Cloud Interaction Observed From High Spectral Resolution Lidar Data

Recent studies utilizing satellite retrievals have shown a strong correlation between aerosol optical depth (AOD) and cloud cover. However, these retrievals from passive sensors are subject to many limitations, including cloud adjacency (or 3D) effects, possible cloud contamination, uncertainty in the AOD retrieval. Some of these limitations do not exist in High Spectral Resolution Lidar (HSRL) observations; for instance, HSRL observations are not a ected by cloud adjacency effects, are less prone to cloud contamination, and offer accurate aerosol property measurements (backscatter coefficient, extinction coefficient, lidar ratio, backscatter Angstrom exponent,and aerosol optical depth) at a neospatial resolution (less than 100 m) in the vicinity of clouds. Hence, the HSRL provides an important dataset for studying aerosol and cloud interaction. In this study, we statistically analyze aircraft-based HSRL profiles according to their distance from the nearest cloud, assuring that all profile comparisons are subject to the same large-scale meteorological conditions. Our results indicate that AODs from HSRL are about 17% higher in the proximity of clouds (approximately 100 m) than far away from clouds (4.5 km), which is much smaller than the reported cloud 3D effect on AOD retrievals. The backscatter and extinction coefficients also systematically increase in the vicinity of clouds, which can be explained by aerosol swelling in the high relative humidity (RH) environment and/or aerosol growth through in cloud processing (albeit not conclusively). On the other hand, we do not observe a systematic trend in lidar ratio; we hypothesize that this is caused by the opposite effects of aerosol swelling and aerosol in-cloud processing on the lidar ratio. Finally, the observed backscatter Angstrom exponent (BAE) does not show a consistent trend because of the complicated relationship between BAE and RH. We demonstrate that BAE should not be used as a surrogate for Angstrom exponent, especially at high RH.

Su, Wenying↗
DOE OSTI2023

Probabilistic partition of unity networks for high–dimensional regression problems

We explore the probabilistic partition of unity network (PPOU-Net) model in the context of high-dimensional regression problems and propose a general framework focusing on adaptive dimensionality reduction. With the proposed framework, the target function is approximated by a mixture of experts model on a low-dimensional manifold, where each cluster is associated with a fixed-degree polynomial. We present a training strategy that leverages the expectation maximization (EM) algorithm. During the training, we alternate between (i) applying gradient descent to update the DNN coefficients; and (ii) using closed-form formulae derived from the EM algorithm to update the mixture of experts model parameters. Under the probabilistic formulation, step (ii) admits the form of embarrassingly paralleliazable weighted least-squares solves. The PPOU-Nets consistently outperform the baseline fully-connected neural networks of comparable sizes in numerical experiments of various data dimensions. Here, we also explore the proposed model in applications of quantum computing, where the PPOU-Nets act as surrogate models for cost landscapes associated with variational quantum circuits.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

A General Framework for Bounding Approximate Dynamic Programming Schemes

For years, there has been interest in approximation methods for solving dynamic programming problems, because of the inherent complexity in computing optimal solutions characterized by Bellman’s principle of optimality. A wide range of approximate dynamic programming (ADP) methods now exists. It is of great interest to guarantee that the performance of an ADP scheme be at least some known fraction, say ß , of optimal. This letter introduces a general approach to bounding the performance of ADP methods, in this sense, in the stochastic setting. The approach is based on new results for bounding greedy solutions in string optimization problems, where one has to choose a string (ordered set) of actions to maximize an objective function. This bounding technique is inspired by submodularity theory, but submodularity is not required for establishing bounds. Instead, the bounding is based on quantifying certain notions of curvature of string functions; the smaller the curvatures the better the bound. The key insight is that any ADP scheme is a greedy scheme for some surrogate string objective function that coincides in its optimal solution and value with those of the original optimal control problem. The ADP scheme then yields to the bounding technique mentioned above, and the curvatures of the surrogate objective determine the value ß of the bound. The surrogate objective and its curvatures depend on the specific ADP.

discrete event systems↗
NASA NTRS2015

Nitric Oxide PLIF Visualization of Simulated Fuel-Air Mixing in a Dual-Mode Scramjet

Nitric oxide (NO) planar induced laser fluorescence (PLIF) measurements have been performed in a small scale scramjet combustor at the University of Virginia Aerospace Research Laboratory at nominal simulated Mach 5 flight. A mixture of NO and N2 was injected at the upstream end of the inlet isolator as a surrogate for ethylene fuel, and the mixing of this fuel simulant was studied with and without a shock train. The shock train was produced by an air throttle, which simulated the blockage effects of combustion downstream of the cavity flame holder. NO PLIF signal was imaged in a plane orthogonal to the freestream at the leading edge of the cavity. Instantaneous planar images were recorded and analyzed to identify the most uniform cases, which were achieved by varying the location of the fuel injection and shock train. This method was used to screen different possible fueling configurations to provide optimized test conditions for follow-on combustion measurements using ethylene fuel. A theoretical study of the selected NO rotational transitions was performed to obtain a LIF signal that is linear with NO mole fraction and approximately independent of pressure and temperature.

Cantu, Luca M. L.↗
NASA NTRSDate not supplied

Uncertainty Quantification using Deep Ensembles for Decision Making in Cyber-Physical-Human Systems

In this paper and its companion, Differential Equation Approximation Using Gradient-Boosted Quantile Regression, Robison et al., we examine an approach to quantifying model uncertainty with the aim of increasing the trustworthiness of computational models in human-machine interactions. In Differential Equation Approximation Using Gradient-Boosted Quantile Regression, we focus on gradient-boosted decision trees, while in this one, we give more details about deep ensembles. Uncertainty quantification is crucial for building trustworthy autonomous decision-making agents in human-machine teams. There are two types of uncertainties: aleatoric and epistemic. The former is related to the inherent stochasticity (noise) of the process, whereas the latter is associated with the lack of knowledge or representation capability of models, such as neural networks. By lack of knowledge, we mean the model’s inability to accurately predict outputs for all possible inputs. The aleatory uncertainty can be estimated fairly easily with, for example, filters, whereas epistemic uncertainty is challenging to compute. This paper uses deep ensembles to quantify both aleatory and epistemic uncertainty. It can act as an uncertainty-aware surrogate transition model for decision-making frameworks. "Uncertainty-aware" means that the surrogate transition model should make predictions along with confidence in those predictions. In the context of decision-making, the transition models are ordinary differential equations (ODEs). Since ODEs can be simulated to make one-step or multi-step predictions, a good surrogate model for them should perform reasonably well in both modes. In a multi-step approach, the trajectory sampling method TS∞ was used to propagate uncertainty over multiple steps. The cartpole dynamical system was selected to demonstrate the ability of deep ensembles as good surrogate transition models for decision-making frameworks. The deep ensembles modeled the dynamics of cartpole ODEs and made uncertainty-aware predictions in single-step and multi-step transition modes.

CPH systems↗
DOE OSTI2022

Towards multi-fidelity deep learning of wind turbine wakes

We report engineering wake models that accurately predict wake in a computationally efficient manner are very important for tasks such as layout optimization and control of wind farms. In this paper, we explore an application of deep learning (DL) to learn the wake model from hierarchies of physics-based approaches ranging from analytical models to an approximate form of the Reynolds-averaged Navier-Stokes equations. We first illustrate the application of principal component analysis to obtain a lower-dimensional representation that allows a computationally tractable training and deployment of DL models. Then, the DL model is trained to learn the mapping from input parameter space to the principal components, which are then used to reconstruct the three-dimensional flow field. Additionally, we investigate a composite framework consisting of two neural networks to learn the correlation between low- and high-fidelity data with Gauss and curl models treated as proxies for low- and high-fidelity models, respectively. The prediction from both DL models matches well with the high-fidelity data with a maximum relative percentage error for the kinetic energy flux of <1%. This work opens up possibilities for data-efficient construction of surrogate models for wake prediction that can be used to study the influence of wind speed and yaw angles on wind farm power production.

17 WIND ENERGY↗
DOE OSTI2025

Advancements on Multi-Fidelity Random Fourier Neural Networks: Application to Hurricane Modeling for Wind Energy

Multi-fidelity approaches are emerging as effective strategies in computational science to handle otherwise intractable tasks like Uncertainty Quantification (UQ), training of Machine Learning (ML) models, and optimization, for expensive high-fidelity applications in which the amount of available simulations or data is limited. The main idea is simple: large datasets generated for low-fidelity approximations of the problem at hand are fused with a much sparser dataset for the target (high-fidelity) system. In this paper, we build on our recent success in designing random Fourier Neural Networks (rFNNs) [1] to target problems arising in wind energy applications and in particular problems of interest for hurricane modeling. In this context, data for the high-fidelity models are limited and lower fidelity alternatives are needed. In this work, we introduce a novel multi-fidelity training approach for our rFNNs and demonstrate its use on a simple verification problem and on a hurricane modeling problem in which high-fidelity data are generated via Large-Eddy Simulations (LES), while low-fidelity data are given by a mesoscale model. Initial results demonstrate how the multi-fidelity training approach can improve the quality of the resulting surrogate.

Fourier Neural Networks↗
NASA NTRS1996

Effect of space flight on cytokine production and other immunologic parameters of rhesus monkeys

During a recent flight of a Russian satellite (Cosmos #2229), initial experiments examining the effects of space flight on immunologic responses of rhesus monkeys were performed to gain insight into the effect of space flight on resistance to infection. Experiments were performed on tissue samples taken from the monkeys before and immediately after flight. Additional samples were obtained approximately 1 month after flight for a postflight restraint study. Two types of experiments were carried out throughout this study. The first experiment determined the ability of leukocytes to produce interleukin-1 and to express interleukin-2 receptors. The second experiment examined the responsiveness of rhesus bone marrow cells to recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF). Human reagents that cross-reacted with monkey tissue were utilized for the bulk of the studies. Results from both studies indicated that there were changes in immunologic function attributable to space flight. Interleukin-1 production and the expression of interleukin-2 receptors was decreased after space flight. Bone marrow cells from flight monkeys showed a significant decrease in their response to GM-CSF compared with the response of bone marrow cells from nonflight control monkeys. These results suggest that the rhesus monkey may be a useful surrogate for humans in future studies that examine the effect of space flight on immune response, particularly when conditions do not readily permit human study.

unmanned↗
DOE OSTI2021

Identifying Entangled Physics Relationships Through Sparse Matrix Decomposition to Inform Plasma Fusion Design

We report a sustainable burn platform through inertial confinement fusion (ICF) has been an ongoing challenge for over 50 years. Mitigating engineering limitations and improving the current design involves an understanding of the complex coupling of physical processes. While sophisticated simulation codes are used to model ICF implosions, these tools contain necessary numerical approximation but miss physical processes that limit predictive capability. Identification of relationships between controllable design inputs to ICF experiments and measurable outcomes (e.g., neutron yield, neutron velocity, areal density) from performed experiments can help guide the future design of experiments and development of simulation codes, to potentially improve the accuracy of the computational models used to simulate ICF experiments. We use sparse matrix decomposition methods to identify clusters of a few related design variables. Sparse principal component analysis (SPCA) identifies groupings that are related to the physical origin of the variables (laser, hohlraum, and capsule). A variable importance analysis finds that in addition to variables highly correlated with neutron yield, such as picket power and laser energy, variables that represent a dramatic change of the ICF design, such as number of pulse steps, are also very important. The obtained sparse components are then used to train a random forest (RF) regression surrogate for predicting total yield. The RF performance on the training and testing data compares with the performance of the RF trained using all the design variables considered. This work is intended to inform design changes in future ICF experiments by augmenting the expert intuition and simulation results.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2022

Learning high-dimensional parametric maps via reduced basis adaptive residual networks

We propose a scalable framework for the learning of high-dimensional parametric maps via adaptively constructed residual network (ResNet) maps between reduced bases of the inputs and outputs. When just few training data are available, it is beneficial to have a compact parametrization in order to ameliorate the ill-posedness of the neural network training problem. By linearly restricting high-dimensional maps to informed reduced bases of the inputs, one can compress high-dimensional maps in a constructive way that can be used to detect appropriate basis ranks, equipped with rigorous error estimates. A scalable neural network learning framework is thus to learn the nonlinear compressed reduced basis mapping. Unlike the reduced basis construction, however, neural network constructions are not guaranteed to reduce errors by adding representation power, making it difficult to achieve good practical performance. Inspired by recent approximation theory that connects ResNets to sequential minimizing flows, we present an adaptive ResNet construction algorithm. This algorithm allows for depth-wise enrichment of the neural network approximation, in a manner that can achieve good practical performance by first training a shallow network and then adapting. We prove universal approximation of the associated neural network class for $L^2_v$ functions on compact sets. Our overall framework allows for constructive means to detect appropriate breadth and depth, and related compact parametrizations of neural networks, significantly reducing the need for architectural hyperparameter tuning. Numerical experiments for parametric PDE problems and a 3D CFD wing design optimization parametric map demonstrate that the proposed methodology can achieve remarkably high accuracy for limited training data, and outperformed other neural network strategies we compared against.

42 ENGINEERING↗
DOE OSTI2023

A study of explosive-induced fracture in polymethyl methacrylate (PMMA)

The fracture response of geologic materials is of interest for applications, including geothermal energy harnessing and containment of underground explosions. To better understand the explosively induced fracture response of geomaterials, polymethyl methacrylate (PMMA) was used as a transparent rock surrogate to allow imaging of internal shock propagation and fracture growth processes. Experiments were conducted using high-speed shadowgraphy and photon Doppler velocimetry (PDV), which were compared to numerical simulations. Experiments measured fractures produced in 304.8mm × 304.8mm × 304.8mm PMMA cubes with two simultaneously initiated detonators. The cubes were subjected to varying amounts and directions of externally applied uniaxial stresses, including no stress, 2 MPa stress, and 20 MPa stress. The fracture radius as a function of time was extracted from the high-speed videos. Post-test images of the PMMA cubes aided in the determination of three-dimensional effects not directly imaged by the cameras. The surface velocity history and the shock response captured in PDV and the high-speed videos were compared to the simulated explosive-induced shock response. The simulation results indicate that the shock drives the fracture for the first 20 μs corresponding to a fracture radius of approximately 15 mm in the experiments. The gas-driven fracture extent was estimated analytically using an equilibrium stress distribution calculated after the shock wave propagation through the sample. Reduction in the gas pressure due to the leakage of the explosive products through the crack as a function of time was accounted for. In conclusion, the estimated fracture lengths were in agreement with the experimentally observed fracture lengths.

15 GEOTHERMAL ENERGY↗
NASA NTRS2010

Advancements in Dual-Pump Broadband CARS for Supersonic Combustion Measurements

Space- and time-resolved measurements of temperature and species mole fractions of nitrogen, oxygen, and hydrogen were obtained with a dual-pump coherent anti-Stokes Raman spectroscopy (CARS) system in hydrogen-fueled supersonic combustion free jet flows. These measurements were taken to provide time-resolved fluid properties of turbulent supersonic combustion for use in the creation and verification of computational fluid dynamic (CFD) models. CFD models of turbulent supersonic combustion flow currently facilitate the design of air-breathing supersonic combustion ramjet (scramjet) engines. Measurements were made in supersonic axi-symmetric free jets of two scales. First, the measurement system was tested in a laboratory environment using a laboratory-scale burner (approx.10 mm at nozzle exit). The flow structures of the laboratory-burner were too small to be resolved with the CARS measurements volume, but the composition and temperature of the jet allowed the performance of the system to be evaluated. Subsequently, the system was tested in a burner that was approximately 6 times larger, whose length scales are better resolved by the CARS measurement volume. During both these measurements, weaknesses of the CARS system, such as sensitivity to vibrations and beam steering and inability to measure temperature or species concentrations in hydrogen fuel injection regions were indentified. Solutions were then implemented in improved CARS systems. One of these improved systems is a dual-pump broadband CARS technique called, Width Increased Dual-pump Enhanced CARS (WIDECARS). The two lowest rotational energy levels of hydrogen detectable by WIDECARS are H2 S(3) and H2 S(4). The detection of these lines gives the system the capability to measure temperature and species concentrations in regions of the flow containing pure hydrogen fuel at room temperature. WIDECARS is also designed for measurements of all the major species (except water) in supersonic combustion flows fueled with hydrogen and hydrogen/ethylene mixtures (N2, O2, H2, C2H4, CO, and CO2). This instrument can characterize supersonic combustion fueled with surrogate fuel mixtures of hydrogen and ethylene. This information can lead to a better understanding of the chemistry and performance of supersonic combustion fueled with cracked jet propulsion (JP)-type fuel.

Tedder, Sarah Augusta Umberger↗
DOE OSTI2025

Low‐dimensional manifold learning for uncertainty quantification in complex multi‐scale stochastic systems

Broadly speaking, the goals of the project are to develop techniques to use manifold learning to develop reduced‐order and surrogate models for "hyper‐reduction" of very high‐dimensional complex multi‐scale systems. This is being achieved by employing a newly proposed form of manifold projection and learning that leverages recent advancements in computational geometry and data‐driven modeling. In particular, we are applying a manifold projection technique to project the solutions of very high‐dimensional systems onto the so‐called Grassmannmanifold, a Reimannian manifold comprised of orthonormal matrices. We then apply data‐driven machine learning techniques to classify the solutions on the manifold (e.g. clustering techniques) according to their proximity on the manifold and leverage a further nonlinear dimension reduction to organize the structured data on the manifold. Finally, we are developing novel techniques that enable us to directly interpolate the hyper‐reduced data such that we can predict the solution of the complex, high‐ dimensional system without need to call the full expensive computational model. Given their adherence to the underlying structure of the solution of the physical system, it is expected that these approximate solutions will be sufficiently constrained so as to (approximately) adhere to physical principles.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Accelerating Hamiltonian Monte Carlo for Bayesian inference in neural networks and neural operators

Hamiltonian Monte Carlo (HMC) is a powerful and accurate method to sample from the posterior distribution in Bayesian inference. However, HMC techniques are computationally demanding for Bayesian neural networks due to the high dimensionality of the network’s parameter space and the non-convexity of their posterior distributions. Therefore, various approximation techniques, such as variational inference (VI) or stochastic gradient MCMC, are often employed to infer the posterior distribution of the network parameters. Such approximations introduce inaccuracies in the inferred distributions, resulting in unreliable uncertainty estimates. In this work, we propose a hybrid approach that combines inexpensive VI and accurate HMC methods to efficiently and accurately quantify uncertainties in neural networks and neural operators. The proposed approach leverages an initial VI training on the full network. We examine the influence of individual parameters on the prediction uncertainty, which shows that a large proportion of the parameters do not contribute substantially to uncertainty in the network predictions. This information is then used to significantly reduce the dimension of the parameter space, and HMC is performed only for the subset of network parameters that strongly influence prediction uncertainties. This yields a framework for accelerating the full batch HMC for posterior inference in neural networks. We demonstrate the efficiency and accuracy of the proposed framework on deep neural networks and operator networks, showing that inference can be performed for large networks with tens to hundreds of thousands of parameters. Finally, we show that this method can effectively learn surrogates for complex physical systems by modeling the operator that maps from upstream conditions to wall-pressure data on a cone in hypersonic flow.

Bayesian inference↗
DOE OSTI2025

Improving Trustworthiness of Data-Driven Power Grid Contingency Analysis With Bayesian Residual Graph Neural Networks

The evolving energy landscape requires novel tools to efficiently perform contingency analysis and reliability assessment of power grids, potentially in real-time. The high computational cost of traditional power flow solvers limits their applicability in practice. Machine learning (ML) surrogates such as deep neural networks (NNs) accelerate power flow solvers computations, enabling high-order contingency analysis and real-time decision-making by learning highly nonlinear functions and integrating grid topology via graph architectures. However, (graph) NNs lack predictive power away from training data and do not provide predictive confidence estimates. Here, we present a Bayesian residual graph NN that integrates knowledge from low-fidelity data via residual training and embeds granular quantification of uncertainties, improving trustworthiness critical for high-consequence decision-making. Applying Bayesian concepts to NNs is challenging due to the high-dimensionality of both the parameter space, complicating derivation of a meaningful prior, and the output space in large grid systems, requiring enhanced techniques to assess the predicted high-dimensional uncertainties. Our contributions include: (1) Deriving a prior for fully connected and graph NNs that leverages low-fidelity data to guide mean predictions and appropriately control prior predictive uncertainty. (2) Integrating this prior within an ensembling with anchoring scheme for efficient approximate posterior inference. (3) Deriving enhanced metrics to assess accuracy of both the mean and uncertainty predictions in high dimensions, appropriately accounting for correlations propagated through graph layers. The resulting Bayesian residual graph NN is tested on a contingency analysis task for 14-bus and 118-bus grids.

24 - POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2021

Predicting Hugoniot equation of state in erythritol with ab initio and reactive molecular dynamics

Erythritol has been proposed as an inert surrogate for developing theoretical and computational models to study aging in energetic materials. In this work, we present a comparison of mechanical and shock properties of erythritol computed using the ReaxFF reactive force field and from ab initio calculations employing density functional theory (DFT). We screened eight different ReaxFF parameterizations, of which the CHO parameters developed for hydrocarbon oxidation provide the most accurate predictions of mechanical properties and the crystal structure of erythritol. Further validation of the applicability of this ReaxFF parameterization for modeling erythritol is demonstrated by comparing predictions of the elastic constants, crystal structure, vibrational density of states, and Hugoniot curves against DFT calculations. The ReaxFF predictions are in close agreement with the DFT simulations for the elastic constants and shock Hugoniot when the crystal is loaded along its c axis but show as much as 30% disagreement in the elastic constants in the ab plane and 12% difference in shock pressures when shocked along the a or b crystal axes. Last, we compare thermomechanical properties predicted from classical molecular dynamics with those calculated using the quasi-harmonic approximation and show that quantum mechanical effects produce large discrepancies in the computed values of heat capacity and thermal expansion coefficients compared with classical assumptions. Combining classical molecular dynamics predictions of mechanical behavior with phonon-based calculations of thermal behaviors, we show that predicted shock-induced temperatures for pressures up to 6.5 GPa do not exceed the pressure-dependent melting point of erythritol.

Hu, Jing↗
NASA NTRS2006

Solar CIV Vacuum-Ultraviolet Fabry-Perot Interferometers

Aims: A tunable, high spectral resolution, high effective finesse, vacuum ultraviolet (VUV) Fabry-Perot interferometer (PPI) is designed for obtaining narrow-passband images, magnetograms, and Dopplergrams of the transition region emission line of CIV (155 nm). Methods: The integral part of the CIV narrow passband filter package (with a 2-10 pm FWHM) consists of a multiple etalon system composed of a tunable interferometer that provides high-spectral resolution and a static low-spectral resolution interferometer that allows a large effective free spectral range. The prefilter for the interferometers is provided by a set of four mirrors with dielectric high-reflective coatings. A tunable interferometer, a VUV piezoelectric-control etalon, has undergone testing using the surrogate F2 eximer laser line at 157 nm for the CIV line. We present the results of the tests with a description of the overall concept for a complete narrow-band CIV spectral filter. The static interferometer of the filter is envisioned as being hudt using a set of fixed MgF2 plates. The four-mirror prefilter is designed to have dielectric multilayer n-stacks employing the design concept used in the Ultraviolet Imager of NASA's Polar Spacecraft. A dual etalon system allows the effective free spectral range to be commensurate with the prefilter profile. With an additional etalon, a triple etalon system would allow a spectrographic resolution of 2 pm. The basic strategy has been to combine the expertise of spaceflight etalon manufacturing with VUV coating technology to build a VUV FPI which combines the best attributes of imagers and spectrographs into a single compact instrument. Results. Spectro-polarimetry observations of the transition region CIV emission can be performed to increase the understanding of the magnetic forces, mass motion, evolution, and energy release within the solar atmosphere at the base of the corona where most of the magnetic field is approximately force-free. The 2D imaging of the full vector magnetic field at the height of maximum magnetic influence (minimum plasma beta) can be accomplished, albeit difficult, by measuring the Zeeman splitting of the CIV resonance pair. Designs of multiple VUV FPIs can be developed for integration into future orbiting solar observatories to obtain rapid cadence, spectral imaging of the transition region.

Gary, G. Allen↗
DOE OSTI2020

Optimal carbon storage reservoir management through deep reinforcement learning

Model-based optimization plays a central role in energy system design and management. The complexity and high-dimensionality of many process-level models, especially those used for geosystem energy exploration and utilization, often lead to formidable computational costs when the dimension of decision space is also large. This work adopts elements of recently advanced deep learning techniques to solve a sequential decision-making problem in applied geosystem management. Specifically, a deep reinforcement learning framework was formed for optimal multiperiod planning, in which a deep Q-learning network (DQN) agent was trained to maximize rewards by learning from high-dimensional inputs and from exploitation of its past experiences. To expedite computation, deep multitask learning was used to approximate high-dimensional, multistate transition functions. Both DQN and deep multitask learning are pattern based. As a demonstration, the framework was applied to optimal carbon sequestration reservoir planning using two different types of management strategies: monitoring only and brine extraction. Both strategies are designed to mitigate potential risks due to pressure buildup. Results show that the DQN agent can identify the optimal policies to maximize the reward for given risk and cost constraints. Finally, experiments also show that knowledge the agent gained from interacting with one environment is largely preserved when deploying the same agent in other similar environments.

15 GEOTHERMAL ENERGY↗