Search NASA⌕ Search

SEARCH · Search NASA

Results for “Dynamic optimization”

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 523 records · Page 29

How good are learning-based control v.s. model-based control for load shifting? Investigations on a single zone building energy system

Both model predictive control (MPC) and deep reinforcement learning control (DRL) have been presented as a way to approximate the true optimality of a dynamic programming problem, and these two have shown significant operational cost saving potentials for building energy systems. Furthermore, there is still a lack of in-depth quantitative studies on their approximation levels to the true optimality, especially in the building energy domain. To fill in the gap, this paper provides a numerical framework that enables the evaluation of the optimality levels of different controllers for building energy systems. This framework is then used to comprehensively compare the optimal control performance of both MPC and DRL controllers with given computation budgets for a single zone fan coil unit system. Note the optimality is estimated based on a user-specific selection of trade-off weights among energy costs, thermal comfort and control slew rates. Compared with the best optimality we can find through expensive optimization simulations, the best DRL agent can maximally approximate the optimality by 96.54%, which outperforms the best MPC whose optimality level is 90.11%. However, due to the stochasticity, the DRL agent is only expected to approximate the optimality by 90.42%, which is almost equivalent to the best MPC. Except for Proximal Policy Optimization (PPO), all DRL agents can have a better approximation to the optimality than the best MPC, and are expected to have better approximation than the MPC with a prediction horizon of 32 steps (15 min per step). In terms of reducing energy cost and thermal discomfort, MPC can outperform the rule-based control (RBC) by 18.47%–25.44%. DRL can be expected to outperform RBC by 18.95%–25.65% ,and the best DRL control policy can outperform RBC by 20.29%–29.72%. Although the comparison of the optimality level is performed in a perfect setting, e.g., MPC assumes perfect models, and DRL assumes a perfect offline training process and online deployment process, this can shed insight on their capabilities of approximating to the original dynamic programming problem.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Bayesian sparse learning with preconditioned stochastic gradient MCMC and its applications

Deep neural networks have been successfully employed in an extensive variety of research areas, including solving partial differential equations. Despite its significant success, there are some challenges in effectively training DNN, such as avoiding overfitting in over-parameterized DNNs and accelerating the optimization in DNNs with pathological curvature. Here, we propose a Bayesian type sparse deep learning algorithm. The algorithm utilizes a set of spike-and-slab priors for the parameters in the deep neural network. The hierarchical Bayesian mixture will be trained using an adaptive empirical method. That is, one will alternatively sample from the posterior using preconditioned stochastic gradient Langevin Dynamics (PSGLD), and optimize the latent variables via stochastic approximation. The sparsity of the network is achieved while optimizing the hyperparameters with adaptive searching and penalizing. A popular SG-MCMC approach is Stochastic gradient Langevin dynamics (SGLD). However, considering the complex geometry in the model parameter space in nonconvex learning, updating parameters using a universal step size in each component as in SGLD may cause slow mixing. To address this issue, we apply a computationally manageable preconditioner in the updating rule, which provides a step-size parameter to adapt to local geometric properties. Moreover, by smoothly optimizing the hyperparameter in the preconditioning matrix, our proposed algorithm ensures a decreasing bias, which is introduced by ignoring the correction term in the preconditioned SGLD. According to the existing theoretical framework, we show that the proposed algorithm can asymptotically converge to the correct distribution with a controllable bias under mild conditions. Numerical tests are performed on both synthetic regression problems and learning solutions of elliptic PDE, which demonstrate the accuracy and efficiency of the present work.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Fast, accurate, and robust automatic marker detection for motion correction based on oblique kV or MV projection image pairs

Purpose: A robust and accurate method that allows the automatic detection of fiducial markers in MV and kV projection image pairs is proposed. The method allows to automatically correct for inter or intrafraction motion. Methods: Intratreatment MV projection images are acquired during each of five treatment beams of prostate cancer patients with four implanted fiducial markers. The projection images are first preprocessed using a series of marker enhancing filters. 2D candidate marker locations are generated for each of the filtered projection images and 3D candidate marker locations are reconstructed by pairing candidates in subsequent projection images. The correct marker positions are retrieved in 3D by the minimization of a cost function that combines 2D image intensity and 3D geometric or shape information for the entire marker configuration simultaneously. This optimization problem is solved using dynamic programming such that the globally optimal configuration for all markers is always found. Translational interfraction and intrafraction prostate motion and the required patient repositioning is assessed from the position of the centroid of the detected markers in different MV image pairs. The method was validated on a phantom using CT as ground-truth and on clinical data sets of 16 patients using manual marker annotations as ground-truth. Results: The entire setup was confirmed to be accurate to around 1 mm by the phantom measurements. The reproducibility of the manual marker selection was less than 3.5 pixels in the MV images. In patient images, markers were correctly identified in at least 99% of the cases for anterior projection images and 96% of the cases for oblique projection images. The average marker detection accuracy was 1.4{+-}1.8 pixels in the projection images. The centroid of all four reconstructed marker positions in 3D was positioned within 2 mm of the ground-truth position in 99.73% of all cases. Detecting four markers in a pair of MV images takes a little less than a second where most time is spent on the image preprocessing. Conclusions: The authors have developed a method to automatically detect multiple markers in a pair of projection images that is robust, accurate, and sufficiently fast for clinical use. It can be used for kV, MV, or mixed image pairs and can cope with limited motion between the projection images.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Extending The Realm of Optimization for Complex Systems: Uncertainty, Competition, And Dynamics

Research reported addressed these topics: the development of analytical and algorithmic tools for distributed computation of Nash equilibria; synchronization in mean-field oscillator games, with an emphasis on learning and efficiency analysis; questions that combine learning and computation; questions including stochastic and mean-field games; modeling and control in the context of power markets.

97 MATHEMATICS AND COMPUTING↗

QMMM 2023: A program for combined quantum mechanical and molecular mechanical modeling and simulations

Combined quantum mechanical and molecular mechanical (QM/MM) methods play an important role in multiscale modeling and simulations. QMMM 2023 is a general-purpose program for single-point calculations, geometry optimizations, transition-state optimizations, and molecular dynamics (MD) at the QM/MM level. It calls a QM package and an MM package to perform the required single-level calculations and combines them into a QM/MM energy by a variety of schemes. QMMM 2023 supports GAMESS-US, Gaussian, and ORCA as QM packages and Tinker as the MM package. Four types of treatments are available for embedding the QM subsystem in the MM environment: mechanical embedding with gas-phase calculations of the QM region, electronic embedding that allows polarization of the QM region by the MM environment, polarizable embedding for mutual polarization of the QM and MM regions, and flexible embedding for both mutual polarization and partial charge transfer between the QM and MM regions. Boundaries between QM and MM regions that pass through covalent bonds can be treated by several methods, including the redistributed charge (RC) scheme, redistributed charge and dipole (RCD) scheme, balanced-RC scheme, balanced-RCD scheme, screened charge scheme that takes account of charge penetration effects, and smeared charge scheme that delocalizes the MM charges near the QM–MM boundary. Geometry optimization can be done using the optimizer implemented in QMMM 2023 or the Berny optimizer in Gaussian through external calls to Gaussian. Molecular dynamics simulations can be performed at the pure-MM level, pure-QM level, fixed-partitioning QM/MM level, and adaptive-partitioning QM/MM level. As a result, the adaptive-partitioning treatments permit on-the-fly relocation of the QM–MM boundary by dynamically reclassifying atoms or groups into the QM or MM subsystems.

97 MATHEMATICS AND COMPUTING↗

Online Optimization as a Feedback Controller: Stability and Tracking

This paper develops and analyzes feedback-based online optimization methods to regulate the output of a linear time-invariant (LTI) dynamical system to the optimal solution of a time-varying convex optimization problem. The design of the algorithm is based on continuous-time primal-dual dynamics, properly modified to incorporate feedback from the LTI dynamical system, applied to a proximal augmented Lagrangian function. The resultant closed-loop algorithm tracks the solution of the time-varying optimization problem without requiring knowledge of (time-varying) disturbances in the dynamical system. The analysis leverages integral quadratic constraints to provide linear matrix inequality (LMI) conditions that guarantee global exponential stability and bounded tracking error. Analytical results show that, under a sufficient time-scale separation between the dynamics of the LTI dynamical system and the algorithm, the LMI conditions can be always satisfied. The paper further proposes a modified algorithm that can track an approximate solution trajectory of the constrained optimization problem under less restrictive assumptions. As an illustrative example, the proposed algorithms are showcased for power transmission systems, to compress the time scales between secondary and tertiary control, and allow to simultaneously power re-balancing and tracking of DC optimal power flow points.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Optimizing Distributed Photovoltaic System Set Points Under Uncertainty

We explore the potential for using dynamic programming (DP) and approximate dynamic programming (ADP) techniques to optimize the set points of distributed photovoltaic (DGPV) inverters under uncertainty about future DGPV deployment. We consider a case where a large (=1MW) system is installed first, and growth in deployment of small rooftop systems is anticipated, but uncertain. We find that for a real feeder (EPRI's J1 feeder), a significant reduction in the number and severity of voltage violations can be expected when DP or ADP is used compared to selecting set points based on the current conditions (the traditional myopic approach). Additionally, we find that using a simple ADP algorithm, sampled backward induction, is more than twice as fast as DP with similar outcomes.

14 SOLAR ENERGY↗

Robust Dynamics and Control of a Partially Observed Markov Chain

In a seminal paper, Martin Clark (Communications Systems and Random Process Theory, Darlington, 1977, pp. 721-734, 1978) showed how the filtered dynamics giving the optimal estimate of a Markov chain observed in Gaussian noise can be expressed using an ordinary differential equation. These results offer substantial benefits in filtering and in control, often simplifying the analysis and an in some settings providing numerical benefits, see, for example Malcolm et al. (J. Appl. Math. Stoch. Anal., 2007, to appear).Clark's method uses a gauge transformation and, in effect, solves the Wonham-Zakai equation using variation of constants. In this article, we consider the optimal control of a partially observed Markov chain. This problem is discussed in Elliott et al. (Hidden Markov Models Estimation and Control, Applications of Mathematics Series, vol. 29, 1995). The innovation in our results is that the robust dynamics of Clark are used to compute forward in time dynamics for a simplified adjoint process. A stochastic minimum principle is established.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Reverse and Forward Engineering of Local Voltage Control in Distribution Networks

The increasing penetration of renewable and distributed energy resources in distribution networks calls for real-time and distributed voltage control. As such, in this paper we investigate local Volt/VAR control with a general class of control functions, and show that the power system dynamics with non-incremental local voltage control can be seen as a distributed algorithm for solving a well-defined optimization problem (reverse engineering). The reverse engineering further reveals a fundamental limitation of the non-incremental voltage control: the convergence condition is restrictive and prevents better voltage regulation at equilibrium. This motivates us to design two incremental local voltage control schemes based on the subgradient and pseudo-gradient algorithms respectively for solving the same optimization problem (forward engineering). The new control schemes decouple the dynamical property from the equilibrium property, and have much less restrictive convergence conditions. This work presents another step towards developing a new foundation—network dynamics as optimization algorithms—for distributed real-time control and optimization of future power networks.

42 ENGINEERING↗

Dynamical fingerprints for probing individual relaxation processes in biomolecular dynamics with simulations and kinetic experiments

There is a gap between kinetic experiment and simulation in their views of the dynamics of complex biomolecular systems. Whereas experiments typically reveal only a few readily discernible exponential relaxations, simulations often indicate complex multistate behavior. Here, a theoretical framework is presented that reconciles these two approaches. The central concept is dynamical fingerprints which contain peaks at the time scales of the dynamical processes involved with amplitudes determined by the experimental observable. Fingerprints can be generated from both experimental and simulation data, and their comparison by matching peaks permits assignment of structural changes present in the simulation to experimentally observed relaxation processes. The approach is applied here to a test case interpreting single molecule fluorescence correlation spectroscopy experiments on a set of fluorescent peptides with molecular dynamics simulations. The peptides exhibit complex kinetics shown to be consistent with the apparent simplicity of the experimental data. Moreover, the fingerprint approach can be used to design new experiments with site-specific labels that optimally probe specific dynamical processes in the molecule under investigation.

60 APPLIED LIFE SCIENCES↗

Modeling Water Vapor Transport at Liquid/Membrane Interfaces for Applications in Liquid Desiccant Air Conditioners (CRADA CRD-17-679 Final Report)

Conventional compressor-based cooling has undergone incremental changes over the past 100 years. Further improvements in efficiency require evermore complex systems, especially under humid conditions. Meso-porous membranes offer unique opportunities for efficient humidity control in buildings using an absorbent desiccant solution, but membranes are not designed or optimized for this purpose. This project will use molecular dynamic simulations to determine optimal membrane properties for these air conditioning applications, focusing on the membrane properties at the membrane/liquid/air interface. Optimal membrane designs will enable smaller, more durable, and less expensive designs of these membrane air conditioning systems.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Diagnostics for Linac Optimization With Machine Learning

The Fermilab Linac delivers 400 MeV H- beam to the rest of the accelerator chain. Providing stable intensity, energy, and emittance is key since it directly affects downstream machines. To counter fluctuations of Linac output due to various effects to be described below we are working on implementing dynamic longitudinal parameter optimization based on Machine Learning (ML). As inputs for the ML model, signals from beam diagnostics have to be well understood and reliable. In this paper we discuss the status and plans for ML-based optimization as well as preliminary results of diagnostics studies.

43 PARTICLE ACCELERATORS↗

Towards optimal extraction of cosmological information from nonlinear data

One of the main unsolved problems of cosmology is how to maximize the extraction of information from nonlinear data. If the data are nonlinear the usual approach is to employ a sequence of statistics (N-point statistics, counting statistics of clusters, density peaks or voids etc.), along with the corresponding covariance matrices. However, this approach is computationally prohibitive and has not been shown to be exhaustive in terms of information content. Here we instead develop a hierarchical Bayesian approach, expanding the likelihood around the maximum posterior of linear modes, which we solve for using optimization methods. By integrating out the modes using perturbative expansion of the likelihood we construct an initial power spectrum estimator, which for a fixed forward model contains all the cosmological information if the initial modes are gaussian distributed. We develop a method to construct the window and covariance matrix such that the estimator is explicitly unbiased and nearly optimal. We then generalize the method to include the forward model parameters, including cosmological and nuisance parameters, and primordial non-gaussianity. We apply the method in the simplified context of nonlinear structure formation, using either simplified 2-LPT dynamics or N-body simulations as the nonlinear mapping between linear and nonlinear density, and 2-LPT dynamics in the optimization steps used to reconstruct the initial density modes. We demonstrate that the method gives an unbiased estimator of the initial power spectrum, providing among other a near optimal reconstruction of linear baryonic acoustic oscillations.

79 ASTRONOMY AND ASTROPHYSICS↗

Optimizing Data Locality for Fork/Join Programs Using Constrained Work Stealing

We present an approach to improving data locality across different phases of fork/join programs scheduled using work stealing. The approach consists of: (1) user-specified and automated approaches to constructing a steal tree, the schedule of steal operations and (2) constrained work stealing algorithms that constrain the actions of the scheduler to mirror a given steal tree. These are combined to construct work stealing schedules that maximize data locality across computation phases while ensuring load balance within each phase. These algorithms are also used to demonstrate dynamic coarsening, an optimization to improve spatial locality and sequential overheads by combining many finer-grained tasks into coarser tasks while ensuring sufficient concurrency for locality-optimized load balance. Implementation and evaluation in Cilk demonstrate performance improvements of up to 2.5x on 80 cores. We also demonstrate that dynamic coarsening can combine the performance benefits of coarse task specification with the adaptability of finer tasks.

cilk↗

Development of scalable design optimization parameters for bi-component protective systems

Bi-component protection-systems need to withstand a variety of extreme loading conditions such as high-velocity projectile impact to shock and blast mitigation. The large number of parameters at each stage of bi-component protection system design process results in a near-infinite number of possible variables for optimization. The aim of this project is to identify the scaling laws linking material and geometric scales of the problem to the final impact performance, further including the effects of wave propagation/interaction and inter-component bonding. Specifically, the impact dynamics for different optimal projectile/target density ratios and relative component mass fractions are quantified for bonded ceramic/titanium alloy systems. The impact behavior such as wave mechanics and damage propagation within the target system is characterized and quantified via digital image correlation or via x-ray phase-contrast imaging.

36 MATERIALS SCIENCE↗

Preliminary study of auto-differentiation algorithm in beam dynamics with stochastic process

Modern particle accelerator optimization requires sophisticated computational methods to address the inherently stochastic nature of beam dynamics. This research develops a framework applying AD to SDEs that specifically addresses beam dynamics challenges in particle accelerators, focusing on accurately modeling and optimizing beam behavior in regimes dominated by stochastic processes. By incorporating key physical phenomena such as synchrotron radiation, wakefield effects, and quantum excitation, the framework aims to provide auto differentiation on the figure of merit of the phase space evolution and beam dynamics. The methodology will enable effective optimization method in a dynamic system with stochastic process.

Accelerator Physics↗

Origin of core-to-core x-ray emission spectroscopy sensitivity to structural dynamics

Recently, coherent structural dynamics in the excited state of an iron photosensitizer was observed through oscillations in the intensity of Kα x-ray emission spectroscopy (XES). Understanding the origin of the unexpected sensitivity of core-to-core transitions to structural dynamics is important for further development of femtosecond time-resolved XES methods and, we believe, generally necessary for interpretation of XES signals from highly non-equilibrium structures that are ubiquitous in photophysics and photochemistry. Here, we use multiconfigurational wavefunction calculations combined with atomic theory to analyze the emission process in detail. The sensitivity of core-to-core transitions to structural dynamics is due to a shift of the minimum energy metal–ligand bond distance between 1s and 2p core-hole states. A key effect is the additional contraction of the non-bonding 3s and 3p orbitals in 1s core-hole states, which decreases electron– electron repulsion and increases overlap in the metal–ligand bonds. The effect is believed to be general and especially pronounced for systems with strong bonds. The important role of 3s and 3p orbitals is consistent with the analysis of radial charge and spin densities and can be connected to the negative chemical shift observed for many transition metal complexes. The XES sensitivity to structural dynamics can be optimized by tuning the emission energy spectrometer, with oscillations up to ±4% of the maximum intensity for the current system. The theoretical predictions can be used to design experiments that separate electronic and nuclear degrees of freedom in ultrafast excited state dynamics.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Optimizing a Thomson scattering diagnostic for fast dynamics and high background

The Madison Symmetric Torus (MST) presents challenging conditions for Thomson scattering (TS) measurements. The MST plasmas are reversed-field pinches (RFPs) with electron density n{sub e}<3x10{sup 13} cm{sup -3}, typically 1x10{sup 13} cm{sup -3}. The TS system was designed to measure from 10 eV to 2 keV; however, six polychromators were upgraded from four to eight spectral channels to resolve to 10 keV. There is no diverter or vertical field, so wall interaction results in high background light both from ion and neutral bremsstrahlungs and from line radiation. Also during standard plasmas, the RFP exhibits regular reconnection sawteeth events during which the plasma current, density, and temperature profiles are flattened. These events are of interest both due to the reconnection physics and to their contribution to the MST equilibrium and confinement. These events occur over 100 {mu}S and exhibit large changes in background light and fast changes in temperature. During improved confinement plasmas, there are no sawteeth; the background is low but the temperature can be over an order of magnitude higher. Data analysis of the system has been developed to accommodate both the large dynamic range of the temperature, the fast dynamics, and the fast changing, high amplitude background. Special attention has been paid to the sources of error, in particular, the contribution of the background. A response-function method reduces the measured uncertainty by a factor of 2. Numerical techniques have been developed which are extremely robust. Two methods are used, a conventional {chi}{sup 2} minimization using a Levenberg-Marquardt algorithm coupled with Monte Carlo modeling for the error bar and a Bayesian statistics method. The Bayesian method computes the probability distribution for the measured photons and electron temperature and this information can be used to ensemble data and will allow future integrated data analysis efforts.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗