Search NASA⌕ Search

SEARCH · Search NASA

Results for “optimization algorithm”

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 1,531 records · Page 85

Online Stochastic Optimization of Networked Distributed Energy Resources

Our work investigates distributed control and incentive mechanisms to coordinate distributed energy resources (DERs) with both continuous and discrete decision variables as well as device dynamics in distribution grids. We formulate a multi-period social welfare maximization problem, and based on its convex relaxation propose a distributed stochastic dual gradient algorithm for managing DERs. Moreover, we extend it to an online realtime setting with time-varying operating conditions, asynchronous updates by devices, and feedback being leveraged to account for nonlinear power flows as well as reduce communication overhead. The resulting algorithm provides a general online stochastic optimization algorithm for coordinating networked DERs with discrete power setpoints and dynamics to meet operational and economic objectives and constraints. We characterize the convergence of the algorithm analytically and evaluate its performance numerically.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Optimization and Evaluation of Energy Savings for Connected and Autonomous Off-Road Vehicles

Off-road vehicles, such as wheel loaders, excavators, and harvesters, are extensively utilized across a wide range of industries, including construction, agriculture, and mining. These machines have become indispensable in supporting the day-to-day operational needs of a nation, playing a critical role in various sectors' infrastructure and productivity. However, despite their utility, off-road vehicles are significant consumers of fossil fuels, resulting in substantial emissions that contribute to environmental degradation. This highlights the pressing need for research and technological advancements aimed at improving their energy efficiency and reducing their carbon footprint. There are, however, two primary challenges that must be addressed to achieve these goals. First, off-road vehicles typically perform both driving and working tasks simultaneously, which introduces a high level of complexity into their overall dynamic systems. Analysis the interactions between these functions is challenging. Second, research into off-road vehicles is inherently interdisciplinary, demanding expertise across several domains such as fluid power systems, vehicle dynamics, control theory, optimization techniques, and real-world implementation. Recognizing these challenges, we proposed the project titled "Optimization and Evaluation of Energy Savings for Connected and Autonomous Off-Road Vehicles" as a comprehensive solution to enhance fuel efficiency while simultaneously improving productivity. This project specifically focuses on autonomous off-road vehicles, with particular attention to wheel loaders, and seeks to develop novel methods to optimize energy consumption without sacrificing operational performance. The project integrates real-time control algorithms, vehicle dynamics modeling, and co-optimization of powertrain system and vehicle system to achieve these goals. Our optimization strategy dynamically co-optimizes critical parameters at both the powertrain and vehicle levels, including vehicle speed, working tool movements, powertrain dynamics, and engine operations in real-time. To streamline this optimization process, we developed a vehicle model that captures the key dynamics while significantly enhancing computational efficiency. This allows the system to intelligently minimize fuel consumption, all while maintaining or even improving productivity through real-time calculations during various off-road operations. To validate the effectiveness of this energy optimization method, we introduced a state-of-the-art Hardware-in-the-Loop (HIL) testbed. This reconfigurable testbed seamlessly integrates the actual engine with virtual models of the wheel loader's subsystems, allowing for accurate emulation of real-world operational loads and environments. By simulating these conditions, the HIL testbed enables us to evaluate the wheel loader’s performance under diverse working scenarios, ensuring the developed solution is applicable in real-world operations. This testbed proved to be instrumental in validating the optimization algorithms and demonstrating the system's practical effectiveness. During the evaluation and testing phase, we employed the HIL testbed to rigorously assess the energy savings and productivity improvements generated by the optimized system. The results were highly encouraging, revealing that the automated wheel loader achieved over 30% fuel savings compared to traditional, human-operated cycles, with comparable or even enhanced levels of productivity. The insights gained from this HIL-based testing provided critical validation of our approach and highlighted the potential for deploying these optimized autonomous technologies in real-world off-road vehicles.

33 ADVANCED PROPULSION SYSTEMS↗

A Hybrid Bulk Algorithm to Predict Turbulent Fluxes over Dry and Wet Bare Soils

Measurements made in the Columbia River basin (Oregon) in an area of irregular terrain during the second Wind Forecast Improvement Project (WFIP2) field campaign are used to develop an optimized hybrid bulk algorithm to predict the surface turbulent fluxes from readily measured or modeled quantities over dry and wet bare or lightly vegetated soil surfaces. The hybrid (synthetic) algorithm combines (i) an aerodynamic method for turbulent flow, which is based on the transfer coefficients (drag coefficient and Stanton number), roughness lengths, and Monin–Obukhov similarity; and (ii) a modified Priestley–Taylor (P-T) algorithm with physically based ecophysiological constraints, which is essentially based on the surface energy budget (SEB) equation. Soil heat flux in the latter case was estimated from measurements of soil temperature and soil moisture. In the framework of the hybrid algorithm, bulk estimates of the momentum flux and the sensible heat flux are derived from a traditional aerodynamic approach, whereas the latent heat flux (or moisture flux) is evaluated from a modified P-T model. Direct measurements of the surface fluxes (turbulent and radiative) and other ancillary atmospheric/soil parameters made during WFIP2 for different soil conditions (dry and wet) are used to optimize and tune the hybrid bulk algorithm. The bulk flux estimates are validated against the measured eddy-covariance fluxes. We also discuss the SEB closure over dry and wet surfaces at various time scales based on the modeled and measured fluxes. Although this bulk flux algorithm is optimized for the data collected during the WFIP2, a hybrid approach can be used for similar flux-tower sites and field campaigns.

54 ENVIRONMENTAL SCIENCES↗

Synchronization-Avoiding Graph Algorithms

Because they were developed for optimal sequential complexity, classical graph algorithms as found in textbooks have strictly-defined orders of operations. Enforcing a prescribed order of operations, or even an approximate order, in a distributed memory setting requires significant amounts of synchronization, which in turn can severely limit scalability. As a result, new algorithms are typically required to achieve scalable performance, even for solving well-known graph problems. Yet, even in these cases, parallel graph algorithms are written according to parallel programming models that evolved for, e.g., scientific computing, and that still have inherent, and scalability-limiting, amounts of synchronization. In this paper we present a new approach to parallel graph algorithms: synchronization-avoiding algorithms. To eliminate synchronization and its associated overhead, synchronization-avoiding algorithms perform work in an unordered and fully asynchronous fashion in such a way that the result is constantly refined toward its final state. “Wasted” work is minimized by locally prioritizing tasks using problem-dependent task utility metrics. We classify algorithms for graph applications into two broad categories: algorithms with monotonic updates (which evince global synchronization) and algorithms with nonmonotonic updates (which evince vertex-centric synchronization). We apply our approach to both classes and develop novel, synchronization-avoiding algorithms for solving exemplar problems: SSSP and connected components for the former, graph coloring for the latter. We demonstrate that eliminating synchronization in conjunction with effective scheduling policies and optimizations in the runtime results in improved scalability for both classes of algorithms.

Firoz, Jesun S.↗

SU-E-T-446: Group-Sparsity Based Angle Generation Method for Beam Angle Optimization

Purpose: This work is to develop the effective algorithm for beam angle optimization (BAO), with the emphasis on enabling further improvement from existing treatment-dependent templates based on clinical knowledge and experience. Methods: The proposed BAO algorithm utilizes a priori beam angle templates as the initial guess, and iteratively generates angular updates for this initial set, namely angle generation method, with improved dose conformality that is quantitatively measured by the objective function. That is, during each iteration, we select “the test angle” in the initial set, and use group-sparsity based fluence map optimization to identify “the candidate angle” for updating “the test angle”, for which all the angles in the initial set except “the test angle”, namely “the fixed set”, are set free, i.e., with no group-sparsity penalty, and the rest of angles including “the test angle” during this iteration are in “the working set”. And then “the candidate angle” is selected with the smallest objective function value from the angles in “the working set” with locally maximal group sparsity, and replaces “the test angle” if “the fixed set” with “the candidate angle” has a smaller objective function value by solving the standard fluence map optimization (with no group-sparsity regularization). Similarly other angles in the initial set are in turn selected as “the test angle” for angular updates and this chain of updates is iterated until no further new angular update is identified for a full loop. Results: The tests using the MGH public prostate dataset demonstrated the effectiveness of the proposed BAO algorithm. For example, the optimized angular set from the proposed BAO algorithm was better the MGH template. Conclusion: A new BAO algorithm is proposed based on the angle generation method via group sparsity, with improved dose conformality from the given template. Hao Gao was partially supported by the NSFC (#11405105), the 973 Program (#2015CB856000) and the Shanghai Pujiang Talent Program (#14PJ1404500)

60 APPLIED LIFE SCIENCES↗

NONPARAMETRIC BAYESIAN ESTIMATION OF PERIODIC LIGHT CURVES

Many astronomical phenomena exhibit patterns that have periodic behavior. An important step when analyzing data from such processes is the problem of identifying the period: estimating the period of a periodic function based on noisy observations made at irregularly spaced time points. This problem is still a difficult challenge despite extensive study in different disciplines. This paper makes several contributions toward solving this problem. First, we present a nonparametric Bayesian model for period finding, based on Gaussian Processes (GPs), that does not make assumptions on the shape of the periodic function. As our experiments demonstrate, the new model leads to significantly better results in period estimation especially when the light curve does not exhibit sinusoidal shape. Second, we develop a new algorithm for parameter optimization for GP which is useful when the likelihood function is very sensitive to the parameters with numerous local minima, as in the case of period estimation. The algorithm combines gradient optimization with grid search and incorporates several mechanisms to overcome the high computational complexity of GP. Third, we develop a novel approach for using domain knowledge, in the form of a probabilistic generative model, and incorporate it into the period estimation algorithm. Experimental results validate our approach showing significant improvement over existing methods.

79 ASTRONOMY AND ASTROPHYSICS↗

Methodology for Formulating Diesel Surrogate Fuels with Accurate Compositional, Ignition-Quality, and Volatility Characteristics

In this study, a novel approach was developed to formulate surrogate fuels having characteristics that are representative of diesel fuels produced from real-world refinery streams. Because diesel fuels typically consist of hundreds of compounds, it is difficult to conclusively determine the effects of fuel composition on combustion properties. Surrogate fuels, being simpler representations of these practical fuels, are of interest because they can provide a better understanding of fundamental fuel-composition and property effects on combustion and emissions-formation processes in internal-combustion engines. In addition, the application of surrogate fuels in numerical simulations with accurate vaporization, mixing, and combustion models could revolutionize future engine designs by enabling computational optimization for evolving real fuels. Dependable computational design would not only improve engine function, it would do so at significant cost savings relative to current optimization strategies that rely on physical testing of hardware prototypes. The approach in this study utilized the stateof- the-art techniques of 13C and 1H nuclear magnetic resonance spectroscopy and the advanced distillation curve to characterize fuel composition and volatility, respectively. The ignition quality was quantified by the derived cetane number. Two wellcharacterized, ultra-low-sulfur #2 diesel reference fuels produced from refinery streams were used as target fuels: a 2007 emissions certification fuel and a Coordinating Research Council (CRC) Fuels for Advanced Combustion Engines (FACE) diesel fuel. A surrogate was created for each target fuel by blending eight pure compounds. The known carbon bond types within the pure compounds, as well as models for the ignition qualities and volatilities of their mixtures, were used in a multiproperty regression algorithm to determine optimal surrogate formulations. The predicted and measured surrogate-fuel properties were quantitatively compared to the measured target-fuel properties, and good agreement was found. This paper is dedicated to the memory of our friend and colleague Jim Franz. Funding for this research was provided by the U.S. Department of Energy (U.S. DOE) Office of Vehicle Technologies, and by the Coordinating Research Council (CRC) and the companies that employ the CRC members. The study was conducted under the auspices of CRC. The authors thank U.S. DOE program manager Kevin Stork for supporting the participation of the U.S. national laboratories in this study.

Mueller, Charles J↗

Joint Optimization of Well Completions and Controls for CO 2 Enhanced Oil Recovery and Storage

CO 2 storage through CO 2 enhanced oil recovery (EOR) is considered as one of the technologies to help promote larger scale deployment of CO 2 storage because of associated economic benefits through oil recovery, 45Q tax credits and the utilization of existing infrastructure. The objective of this study is to demonstrate how optimal reservoir management and operation strategies (including well completions and controls) can be used to optimize both CO 2 storage and oil recovery. The optimization problem was focused on jointly estimating the well completions (i.e., fraction of injection/production well perforations in each reservoir layer) and CO 2 injection/oil production controls that maximize the net present value (NPV) in a CO 2 EOR and storage operation. We utilized the newly developed StoSAG algorithm, one of the most efficient optimization algorithms in the reservoir management community, to solve the optimization problem. The performance of joint optimization approach was compared with the performance of well control only optimization approach. In addition, the performance of co-optimization of CO 2 storage and oil recovery approach was compared with the performances of maximization of only CO 2 storage and maximization of only oil recovery approaches. The optimization results showed that a joint optimization of well completions and well controls can achieve an 8.84% higher final NPV than the one obtained from the optimization of only well controls. It was observed that the NPV incremental for joint optimization is mainly due to the fact that the optimal well completions and controls approach results in efficient CO 2 storage and oil production from different reservoir layers depending on the differences in individual layer properties. Comparison of co-optimization (i.e., maximization of NPV) and maximization of only CO 2 storage or only oil recovery showed that the co-optimization and maximization of only oil recovery result in significantly higher final NPV than that obtained through maximization of only CO 2 storage approach while maximization of only CO 2 storage can achieve significantly higher CO 2 storage in the reservoir compared to the other two scenarios. The similar results for co-optimization and maximization of oil production are obtained because of the difference in oil revenue compared to CO 2 storage tax credit. To the best of our knowledge, this is the first study in oil/gas industry and CO 2 storage community to perform joint optimization of well completions and well controls in the fields. We expect that the proposed optimization framework will be a useful and efficient tool for field engineers to optimally manage CO 2 EOR projects to maximize revenue through oil recovery as well as CO 2 storage by taking advantage of the new 45Q tax law.

Artificial intelligence↗

Dosimetric Comparison of Manual and Beam Angle Optimization of Gantry Angles in IMRT

Dosimetric comparison of manual beam angle selection (MBS) and beam angle optimization (BAO) for IMRT plans is investigated retrospectively for 15 head and neck and prostate patients. The head and neck and prostate had planning target volumes (PTVs) ranging between 96.0 and 319.9 cm{sup 3} and 153.6 and 321.3 cm{sup 3}, whereas OAR ranged between 8.3 and 47.8 cm{sup 3} and 68.3 and 469.2 cm{sup 3}, respectively. In MBS, a standard coplanar 7-9 fields equally spaced gantry angles were used. In BAO, the selection of gantry angle was optimized by the algorithm for the same number of beams. The optimization and dose-volume constraints were kept the same for both techniques. Treatment planning was performed on the Eclipse treatment planning system. Our results showed that the dose-volume histogram for PTV are nearly identical in both techniques but BAO provided superior sparing of the organs at risk compared with the MBS. Also, MBS produced statistically significant higher monitor units (MU) and segments than the BAO; 13.1 {+-} 6.6% (p = 0.012) and 10.4 {+-} 13.6% (p = 0.140), and 14.6 {+-} 5.6% (p = 1.003E-5) and 12.6 {+-} 7.4% (p = 0.76E-3) for head and neck and prostate cases, respectively. The reduction in MU translates into the reduction in total body and integral dose. It is concluded that BAO provides advantage over MBS for most intenisty-modulated radiation therapy cases.

61 RADIATION PROTECTION AND DOSIMETRY↗

RLMolLM: Reinforcement Learning-Enhanced Language Model Framework for Inverse Molecular Design

Inverse molecular design faces significant challenges due to vast chemical space and complex property requirements. While language models show promise for molecular generation, they struggle with validity, multi-property optimization, and structural constraints. This work presents RLMolLM, a reinforcement learning framework combining Proximal Policy Optimization (PPO) with genetic algorithms to address these limitations. Our approach optimizes multiple user-specified properties including quantitative estimates of drug-likeness (QED), synthetic accessibility (SA), and ADMET (absorption, distribution, metabolism, excretion, and toxicity) endpoints without requiring complete model retraining, while maintaining capability for scaffold-constrained generation where specific substructures must be preserved. We outperform state-of-the-art methods for molecular optimization, achieving best QED scores across GDB13, Moses, and Zinc datasets with up to 31% improvement over previous methods while maintaining excellent validity, uniqueness, and novelty metrics. For simultaneous multi-property optimization, our framework achieves substantial improvements in ADMET properties including 4.5-fold reduction in hERG toxicity and enhanced Caco-2 permeability compared to Moses dataset. Under structural constraints, the framework significantly improves molecular validity while preserving scaffolds and effectively optimizing properties. In conclusion, this versatile solution advances pharmaceutical and materials molecular design through effective integration of reinforcement learning and genetic algorithms with multi-property optimization and scaffold preservation.

Genetic algorithms↗

CATS: The Hubble Constant from Standardized TRGB and Type Ia Supernova Measurements

The tip of the red giant branch (TRGB) provides a luminous standard candle for constructing distance ladders to measure the Hubble constant. In practice, its measurements via edge-detection response (EDR) are complicated by the apparent fuzziness of the tip and the multipeak landscape of the EDR. Previously, we optimized an unsupervised algorithm, Comparative Analysis of TRGBs, to minimize the variance among multiple halo fields per host without relying on individualized choices, achieving state-of-the-art ~<0.05 mag distance measures for optimal data. Here we apply this algorithm to an expanded sample of SN Ia hosts to standardize these to multiple fields in the geometric anchor, NGC 4258. In concert with the Pantheon+ SN Ia sample, this analysis produces a (baseline) result of H 0 = 73.22 ± 2.06 km s -1 Mpc -1 . The largest difference in H 0 between this and similar studies employing the TRGB derives from corrections for SN survey differences and local flows used in the most recent SN Ia compilations that were absent in earlier studies. The SN-related differences total ~2.0 km s -1 Mpc -1 . A smaller share, ~1.4 km s -1 Mpc -1 , results from the inhomogeneity of the TRGB calibration across the distance ladder. We employ a grid of 108 variants around the optimal TRGB algorithm and find that the median of the variants is 72.94 ± 1.98 km s -1 Mpc -1 with an additional uncertainty due to algorithm choices of 0.83 km s -1 Mpc -1 . None of these TRGB variants result in an H 0 of less than 71.6 km s -1 Mpc -1 .

79 ASTRONOMY AND ASTROPHYSICS↗

First-order convex feasibility algorithms for x-ray CT

Purpose: Iterative image reconstruction (IIR) algorithms in computed tomography (CT) are based on algorithms for solving a particular optimization problem. Design of the IIR algorithm, therefore, is aided by knowledge of the solution to the optimization problem on which it is based. Often times, however, it is impractical to achieve accurate solution to the optimization of interest, which complicates design of IIR algorithms. This issue is particularly acute for CT with a limited angular-range scan, which leads to poorly conditioned system matrices and difficult to solve optimization problems. In this paper, we develop IIR algorithms which solve a certain type of optimization called convex feasibility. The convex feasibility approach can provide alternatives to unconstrained optimization approaches and at the same time allow for rapidly convergent algorithms for their solution-thereby facilitating the IIR algorithm design process. Methods: An accelerated version of the Chambolle-Pock (CP) algorithm is adapted to various convex feasibility problems of potential interest to IIR in CT. One of the proposed problems is seen to be equivalent to least-squares minimization, and two other problems provide alternatives to penalized, least-squares minimization. Results: The accelerated CP algorithms are demonstrated on a simulation of circular fan-beam CT with a limited scanning arc of 144 Degree-Sign . The CP algorithms are seen in the empirical results to converge to the solution of their respective convex feasibility problems. Conclusions: Formulation of convex feasibility problems can provide a useful alternative to unconstrained optimization when designing IIR algorithms for CT. The approach is amenable to recent methods for accelerating first-order algorithms which may be particularly useful for CT with limited angular-range scanning. The present paper demonstrates the methodology, and future work will illustrate its utility in actual CT application.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Developing a Drilling Optimization System for Improved Overall Rate of Penetration in Geothermal Wells

Geothermal energy is renewable, reliable and environmentally friendly source of energy. The major cost in the development of geothermal wells is the actual drilling of the wells. The main objective of this paper is to introduce a new real-time drilling optimization system designed for granite formation to reduce the overall drilling cost. In this study, a drilling optimization system is verified using drilling data from Utah-Forge well 58-32. The drilling optimization system used the Utah-Forge well 58-32 data to achieve real-time unconfined compressive strength (UCS). Based on the UCS value from the previous feet, the system simulates the ROP for the next drilling feet. The drilling optimization system utilizes the Differential Evolution Algorithm (DEA), which is a metaheuristic method to search the space of solution, to find the best operating parameters (i.e. WOB and RPM) for the next drilling foot. The optimization algorithm takes a maximum cutter temperature into account as a constraint and avoids the accelerated wear. The developed drilling optimization system improves ROP responses and reduces the drilling cost of geothermal wells. The simulated ROP results from the system show a good agreement with the ROP from Utah-Forge well 58-32 drilling data. The drilling time before and after optimization for both intervals were presented.

15 GEOTHERMAL ENERGY↗

Analytic Characterization of the Wake Behind In-Stream Hydrokinetic Turbines

Analytical algorithms developed and optimized for quantifying the wake behind in-stream hydrokinetic turbines are presented. These algorithms are based on wake expressions originally developed for wind turbines. Unlike previous related studies, the optimization of empirical coefficients contained in these algorithms is conducted using centerline velocity data from multiple published experimental studies of the wake velocities behind in-stream hydrokinetic turbine models or porous disks and not using computational fluid dynamics. Empirical coefficients are first individually optimized based on each set of experimental data, and then empirically based coefficient expressions are created using all of the data sets collectively, such that they are functions of ambient turbulence intensity. This expands the applicability of the created algorithms to cover the expected range of operating conditions for in-stream hydrokinetic turbines. Wind turbine wake model expressions are also modified to characterize the dependence of wake velocities on radial location from the centerline of in-stream hydrokinetic turbines. Thus, expressions with empirically optimized coefficients for calculating wake velocities behind in-stream hydrokinetic turbines are described in terms of both centerline and radial positions. Wake predictions made using the Larsen model for radial dependence are shown to diverge from experimental measurements near the wake radius defined by the Jensen model, suggesting that this is a good indication of the cutoff point beyond which numerical estimations no longer apply. Results suggest that using a combined Larsen/Ainslie approach or combined Jensen/Ainslie approach for characterizing wake have similar mean errors to using only a Larsen approach.

16 TIDAL AND WAVE POWER↗

SU-E-T-21: A Novel Sampling Algorithm to Reduce Intensity-Modulated Radiation Therapy (IMRT) Optimization Time

Purpose: The IMRT optimization problem requires substantial computer time to find optimal dose distributions because of the large number of variables and constraints. Voxel sampling reduces the number of constraints and accelerates the optimization process, but usually deteriorates the quality of the dose distributions to the organs. We propose a novel sampling algorithm that accelerates the IMRT optimization process without significantly deteriorating the quality of the dose distribution. Methods: We included all boundary voxels, as well as a sampled fraction of interior voxels of organs in the optimization. We selected a fraction of interior voxels using a clustering algorithm, that creates clusters of voxels that have similar influence matrix signatures. A few voxels are selected from each cluster based on the pre-set sampling rate. Results: We ran sampling and no-sampling IMRT plans for de-identified head and neck treatment plans. Testing with the different sampling rates, we found that including 10% of inner voxels produced the good dose distributions. For this optimal sampling rate, the algorithm accelerated IMRT optimization by a factor of 2–3 times with a negligible loss of accuracy that was, on average, 0.3% for common dosimetric planning criteria. Conclusion: We demonstrated that a sampling could be developed that reduces optimization time by more than a factor of 2, without significantly degrading the dose quality.

60 APPLIED LIFE SCIENCES↗

Efficient block preconditioned eigensolvers for linear response time-dependent density functional theory

We present two efficient iterative algorithms for solving the linear response eigen- value problem arising from the time dependent density functional theory. Although the matrix to be diagonalized is nonsymmetric, it has a special structure that can be exploited to save both memory and floating point operations. In particular, the nonsymmetric eigenvalue problem can be transformed into a product eigenvalue problem that is self-adjoint with respect to a K-inner product. This product eigenvalue problem can be solved efficiently by a modified Davidson algorithm and a modified locally optimal block preconditioned conjugate gradient (LOBPCG) algorithm that make use of the K-inner product. The solution of the product eigenvalue problem yields one component of the eigenvector associated with the original eigenvalue problem. However, the other component of the eigenvector can be easily recovered in a postprocessing procedure. Therefore, the algorithms we present here are more efficient than existing algorithms that try to approximate both components of the eigenvectors simultaneously. The efficiency of the new algorithms is demonstrated by numerical examples.

97 MATHEMATICS AND COMPUTING↗

Optimizing Discharge Capacity of Li-O 2 Batteries by Design of Air-Electrode Porous Structure: Multifidelity Modeling and Optimization

We develop a new mathematical framework to study the optimal design of air electrode microstructures for lithium-oxygen (Li-O2) batteries. It can eectively reduce the number of expensive experiments for testing dierent air-electrodes, thereby minimizing the cost in the design of Li-O2 batteries. The design parameters to characterize an air-electrode microstructure include the porosity, surface-to-volume ratio, and parameters associated with the pore-size distribution. A surrogate model (also known as response surface) for discharge capacity is rst constructed as a function of these design parameters. The surrogate model is accurate and easy to evaluate such that an optimization can be performed based on it. In particular, a Gaussian process regression method, co-kriging, is employed due to its accuracy and eciency in predicting high-dimensional responses from a combination of multidelity data. Specically, a small amount of data from high-delity simulations are combined with a large number of data obtained from computationally ecient low-delity simulations. The high-delity simulation is based on a multiscale modeling approach that couples the microscale (pore-scale) and macroscale (device-scale) models. Whereas, the low-delity simulation is based on an empirical macroscale model. The constructed response surface provides quantitative understanding and prediction about how air electrode microstructures aect the discharge performance of Li-O2 batteries. The succeeding sensitivity analysis via Sobol indices and optimization via genetic algorithm ultimately oer a reliable guidance on the optimal design of air electrode microstructures. The proposed mathematical framework can be generalized to investigate other new energy storage techniques and materials.

25 ENERGY STORAGE↗

ZEUS: An Efficient GPU Optimization Method Integrating PSO, BFGS, and Automatic Differentiation

We introduce a novel, efficient computational method, ZEUS, for numerical optimization, and provide an open-source implementation. It has four key ingredients: (1) particle swarm optimization (PSO), (2) the use of the Broyden-Fletcher-Goldfarb-Shanno (BFGS) method, (3) automatic differentiation (AD), and (4) GPUs. Our approach addresses the computational challenges inherent in high-dimensional, non-convex optimization problems. In the first phase of the algorithm, we get a potentially good set of starting points using PSO. Thereafter, we run BFGS independently in parallel from these starting points. BFGS is one of the best-performing algorithms for numerical optimization. However, it requires the gradient of the function being optimized. ZEUS integrates automatic differentiation into BFGS thus avoiding the need for the user to calculate derivatives explicitly. The use of GPUs allows ZEUS to speed up the calculations substantially. We carry out systematic studies to explore the trade-offs between the number of PSO iterations taken, starting points, and BFGS iteration depth. We show that a handful of iterations of PSO can improve global convergence when combined with BFGS. We also present performance studies using common test functions. The source code can be found at https://github.com/fnal-numerics/global-optimizer-gpu.

Soos, Dominik [Old Dominion U.]↗