SEARCH · Search NASA
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.
A new algorithm, called the CDF-Inversion Algorithm, is described. Using it, one can efficiently learn a probability distribution over a finite set to a specified accuracy and confidence. The algorithm can be extended to learn joint distributions over a vector space. Some implementation results are described.
While load balancing in distributed-memory computing has been well-studied, we present an innovative approach to this problem: a unified, reduced-order model that combines three key components to describe “work” in a distributed system: computation, communication, and memory. Our model enables an optimizer to explore complex tradeoffs in task placement, such as augmented parallelism, at the expense of data replication increasing memory usage. We propose a fully distributed, heuristic-based load balancing optimization algorithm, and demonstrate that it quickly finds close-to-optimal solutions. We formalize the complex optimization problem as a mixed-integer linear program, and compare it to our strategy. Finally, we show that when applied to an electromagnetics code, our approach obtains up to 2.3x speedups for the imbalanced execution.
The Metropolis-Hastings (MH) algorithm is a way to sample a provided target distribution pi(z). It works by repeatedly sampling a separate proposal distribution T(x,x') to generate a random walk {x(t)}. We consider a modification of the MH algorithm in which T is dynamically updated during the walk. The update at time t uses the {x(t' less than t)} to estimate the product distribution that has the least Kullback-Leibler distance to pi. That estimate is the information-theoretically optimal mean-field approximation to pi. We demonstrate through computer experiments that our algorithm produces samples that are superior to those of the conventional MH algorithm.
Double stochastic approximation algorithm for minimizing mean square error in finite expansion of unknown probability distribution functions
This investigation seeks a better understanding of the assorted mechanisms controlling the global distribution of precipitation diurnal variability based on the use of Tropical Rainfall Measuring Mission (TRMM) microwave radiometer and radar data. The horizontal distributions of precipitation's diurnal cycle are derived from eight years of TRMM Microwave Imager (TMI) and Precipitation Radar (PR) measurements involving three TRMM standard rain rate retrieval algorithms -- the resultant distributions analyzed at various spatiotemporal scales. The results reveal the prominent and expected late-evening to early-morning (LE-EM) precipitation maxima over oceans and the counterpart prominent and expected mid- to late-afternoon (MLA) maxima over continents. Moreover, and not generally recognized, the results reveal a widespread distribution of secondary maxima occurring over both oceans and continents -- maxima which generally mirror their counterpart regime's behavior. That is, many ocean regions exhibit clearcut secondary MLA precipitation maxima while many continental regions exhibit just as evident secondary LE-EM maxima. This investigation is the first comprehensive study of these globally prevalent secondary maxima and their widespread nature, a type of study only made possible when the analysis procedure is applied to a high-quality global-scale precipitation dataset. The characteristics of the secondary maxima are mapped and described on global grids using an innovative clock-face format, while a current study to be published at a later date provides physically-based explanations of the seasonal-regional distributions of the secondary maxima. In addition to an "explicit" maxima identification scheme, a "Fourier decomposition" maxima identification scheme is used to examine the amplitude and phase properties of the primary and secondary maxima -- as well as tertiary and quaternary maxima. Accordingly, the advantages, ambiguities, and pitfalls resulting from use of Fourier harmonic analysis are explained.
In the cloud computing environment resources are accessed as services rather than as a product. Monitoring this system for performance is crucial because of typical pay-peruse packages bought by the users for their jobs. With the huge number of machines currently in the cloud system, it is often extremely difficult for system administrators to keep track of all machines using distributed monitoring programs such as Ganglia1 which lacks system health assessment and summarization capabilities. To overcome this problem, we propose a technique for automated anomaly detection using machine performance data in the cloud. Our algorithm is entirely distributed and runs locally on each computing machine on the cloud in order to rank the machines in order of their anomalous behavior for given jobs. There is no need to centralize any of the performance data for the analysis and at the end of the analysis, our algorithm generates error reports, thereby allowing the system administrators to take corrective actions. Experiments performed on real data sets collected for different jobs validate the fact that our algorithm has a low overhead for tracking anomalous machines in a cloud infrastructure.
We evaluated the impact of several newly available sources of meteorological data on mesoscale model forecasts of precipitation produced by the extra-tropical cyclone that struck Florida on February 2, 1998. Precipitation distributions of convective rainfall events were derived from Special Sensor Microwave Imager (SSM/I) and Multi-Channel Passive Microwave Sensor (TMI) microwave radiometric data by means of the Goddard PROFiling (GPROF) algorithm. Continuous lightning distributions were obtained from sferics measurements obtained from a network of VLF radio receivers. Histograms of coincident sferics frequency distributions were matched to those of precipitation to derive bogus convective rainfall rates from the continuously available sferics measurements. SSM/I and TMI microwave data were used to derive Integrated Precipitable Water (IPW) distributions. The TMI also provided sea surface temperatures (SSTS) of the Loop Current and Gulf Stream with improved structural detail. A series of experiments assimilated IPW and latent heating from the bogus convective rainfall for six-hours in the MM5 mesoscale forecast model to produce nine-hour forecasts of all rainfall as well as other weather parameters. Although continuously assimilating latent heating only slightly improved the surface pressure distribution forecast, it significantly improved the precipitation forecasts. Correctly locating convective rainfall was found critical for assimilating latent heating in the forecast model, but measurement of the rainfall intensity proved to be less important. The improved SSTs also had a positive impact on rainfall forecasts for this case. Assimilating bogus rainfall in the model produced nine-hour forecasts of radar reflectivity distributions that agreed well with coincident observations from the TRMM spaceborne precipitation radar, ground based radar and spaceborne microwave measurements.
This project is mainly about the operation optimization of three networked microgrids (MG), including centralized optimization and distributed optimization. The alternating direction method of multipliers (ADMM) algorithm is used for distributed optimization. In the distribution network considered here, there is a Distribution Management system (DMS) as the system coordinator and several networked microgrids. In grid-connected mode, power could be imported or exported at the distribution substation bus according to the utility rate, and the exchanged power at point of common coupling (PCC) of any microgrid has a limitation. In islanded mode, the power imports/exports at the distribution substation are zero. In both grid-connected and islanded mode, the distribution substation is taken as a slack bus with fixed voltage magnitude.
A general-purpose computer program for planning the actions of a spacecraft or other complex system has been augmented by incorporating a subprogram that reasons about uncertainties in such continuous variables as times taken to perform tasks and amounts of resources to be consumed. This subprogram computes parametric probability distributions for time and resource variables on the basis of user-supplied models of actions and resources that they consume. The current system accepts bounded Gaussian distributions over action duration and resource use. The distributions are then combined during planning to determine the net probability distribution of each resource at any time point. In addition to a full combinatoric approach, several approximations for arriving at these combined distributions are available, including maximum-likelihood and pessimistic algorithms. Each such probability distribution can then be integrated to obtain a probability that execution of the plan under consideration would violate any constraints on the resource. The key idea is to use these probabilities of conflict to score potential plans and drive a search toward planning low-risk actions. An output plan provides a balance between the user s specified averseness to risk and other measures of optimality.
A method of incorporating the effects of photographic emulsion grain noise into digital image centering algorithms is presented which improves the accuracy of the derived stellar positions and magnitudes. Theoretical formulae are then derived for the limiting error of the center, and the photometric parameters. For IIIa-J, this error is 0.2-0.3 mu for bright unsaturated images, which agrees quite well with measurements made with the Yale PDS microdensitometer. It is expected that, with further improvements in the positional accuracy of the PDS, it should be possible to reach the emulsion grain noise limit, providing that emulsion shifts or other large scale errors do not dominate. It is also shown that, with appropriate trimming, marginal distribution image centering algorithms can yield an accuracy only slightly poorer than that obtained with two-dimensional distributions.
A new algorithm for an iterative computation of solutions of Laplace's or Poisson's equations in two dimensions, using Green's second identity, is presented. This algorithm converges strongly and geometrically and can be applied to curved, irregular, or moving boundaries with nonlinear and/or discontinuous boundary conditions. It has been implemented in Pascal on a number of micro- and minicomputers and applied to several geometries. Cases with known analytic solutions have been tested. Convergence to within 0.1 percent to 0.01 percent of the theoretical values are obtained in a few minutes on a microcomputer.
The project seeks to develop methods to more efficiently simulate aerospace vehicles. The goals are to reduce model development time, increase accuracy (e.g.,by allowing the integration of multidisciplinary models), facilitate collaboration by geographically- distributed groups of engineers, support uncertainty analysis and optimization, reduce hardware costs, and increase execution speeds. These problems are the subject of considerable contemporary research (e.g., Biedron et al. 1999; Heath and Dick, 2000).
Here, the rapid evolution of modern electric power distribution systems into complex networks of interconnected active devices, distributed generation (DG), and storage poses increasing difficulties for system operators. The large-scale integration of distributed energy resources (DERs) and the rapid exchange of measurement data via communication networks present major opportunities for advancing grid operations but also introduce greater uncertainty, higher data dimensionality, more complex network and device models, and challenging control and optimization problems. Deep reinforcement learning (DRL) algorithms are promising in addressing these challenges. However, they have not been effectively adapted for power systems applications, requiring extensive customization for implementation and evaluation. This has resulted in reproducibility challenges and a steep learning curve for researchers new to applying DRL algorithms to the power systems domain. To bridge these gaps, this tutorial aims to serve as a valuable resource for researchers interested in exploring learning-based algorithms to operate active power distribution networks. Specifically, this work presents a generalized process for translating sequential decision-making problems in power distribution systems into Markov decision process (MDP) formulations, illustrated through concrete grid service examples. Additionally, we introduce a simple environment design strategy to develop and evaluate example DRL algorithms for distribution system applications, complete with an included code repository to guide users through environment construction.
Cooperative behavior algorithms utilizing swarm intelligence are being developed for mobile sensor platforms to inspect jet engines on-wing. Experiments are planned in which several relatively simple autonomous platforms will work together in a coordinated fashion to carry out complex maintenance-type tasks within the constrained working environment modeled on the interior of a turbofan engine. The algorithms will emphasize distribution of the tasks among multiple units; they will be scalable and flexible so that units may be added in the future; and will be designed to operate on an individual unit level to produce the desired global effect. This proof of concept demonstration will validate the algorithms and provide justification for further miniaturization and specialization of the hardware toward the true application of on-wing in situ turbine engine maintenance.
Recap: NASA-Specific Objectives for ICE-POP: Provide real-time observational and NWP data in support of ICE-POP, participate in significant international science effort; GPM (Global Precipitation Measurement) Ground Validation and NASA Weather Program -Direct/physical validation of active/passive satellite-based snowfall retrieval algorithms over coastline and mountains; melting layer interaction with terrain -Physics of snow, coupling to snow water equivalent rate and satellite remote sensor retrieval algorithm assumptions - -Size distributions, types/habit, water equivalent, profiles -NU-WRF (NASA-Unified Weather Research and Forecasting) Model plus Observational analyses: Movement toward “level IV products” leverage intensive and multi-faceted NWP (Numerical Weather Prediction) component -Model precipitation processes (liquid, mixed phase and frozen); Build model testing database for further active/passive remote sensing algorithm development (e.g., satellite data simulators) -"Integrated" validation of products in operational context.
A model of the frequency distributions of the spatial variability in surface temperature is presented. Surface temperature data are obtained from two daytime and two nighttime flights of the Thermal IR Multispectral Scanner (TIMS) over forest land in western Oregon in August, 1985. The temperature values are corrected for atmospheric attenuation and thermal radiation emission with the LOWTRAN-6 algorithm. The temperature distributions were modeled with a two-parameter beta probability density distribution and the fit of the model was evaluated by comparison with the TIMS data set. Use of the model's parameters to identify and classify surface types shows good discrimination among various surfaces for the daytime images, with less distinct discrimination for the nighttime images.
The increasing frequency of extreme events and the integration of distributed energy resources (DERs) into modern grids have elevated the need for resilient and efficient critical load restoration strategies in distribution systems. However, the stochastic nature of renewable DERs, limited energy resource availability and the intricate nonlinearities inherent in complex grid control problem make the problem challenging. Although reinforcement learning (RL) and warm-start RL methods have shown promising results, their performance often falls short in rapidly adapting to new, unseen situations and typically requires exhaustive problem-specific tuning. To address these gaps, we propose a First-Order Meta-based RL (FOM-RL) algorithm within an online framework for adaptive and robust critical load restoration. By harnessing local DERs as the enabling technology, FOM-RL allows the RL agent to swiftly adapt to new unseen scenarios by leveraging previously acquired knowledge of different tasks. Experimental results provide evidence that proposed algorithm learns more efficiently and showcases generalization capabilities across diverse set of operational scenarios. Moreover, a rigorous theoretical analysis yields a tight sublinear regret bound, sensitive to temporal variability, with a task-averaged optimality gap bounded by O(VM+D*/(Tsquare root(M))). These results suggest that optimality improves with task similarity and an increased number of tasks M, reaffirming the efficacy and scalability of the proposed approach in addressing the complexities of critical load restoration in distribution systems.
The basic goal of adaptive-mesh algorithms is to distribute computational resources wisely by increasing the resolution of 'important' regions of the flow and decreasing the resolution of regions that are less important. While this goal is one that is worthwhile, implementing schemes that have this degree of sophistication remains more of an art than a science. In this paper, the basic pieces of adaptive-mesh algorithms are described and some of the possible ways to implement them are discussed and compared. These basic pieces are the data structure to be used, the generation of an initial mesh, the criterion to be used to adapt the mesh to the solution, and the flow-solver algorithm on the resulting mesh. Each of these is discussed, with particular emphasis on methods suitable for the computation of compressible flows.