Search NASA⌕ Search

SEARCH · Search NASA

Results for “distributed 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 145 records · Page 8
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Deep Reinforcement Learning Based Volt-VAR Optimization in Smart Distribution Systems

This paper develops a model-free volt-VAR optimization (VVO) algorithm via multi-agent deep reinforcement learning (DRL) in unbalanced distribution systems. This method is novel since we cast the VVO problem in distribution networks to an intelligent deep Q-network (DQN) framework, which avoids solving a specific optimization model directly when facing time-varying operating conditions in the systems. We consider statuses/ratios of switchable capacitors, voltage regulators, and smart inverters installed at distributed generators as the action variables of the agents. A delicately designed reward function guides these agents to interact with the distribution system, in the direction of reinforcing voltage regulation and power loss reduction simultaneously. The forward-backward sweep method for radial three-phase distribution systems provides accurate power flow results within a few iterations to the DRL environment. The proposed method realizes the dual goals for VVO. We test this algorithm on the unbalanced IEEE 13-bus and 123-bus systems. Numerical simulations validate the excellent performance of this method in voltage regulation and power loss reduction.

41 EE - Solar Energy Technologies Office (EE-4S)↗
DOE OSTI2019

Solving a system of linear equations: From centralized to distributed methods

For a wide range of control engineering applications, the problem of solving a system of linear equations is often encountered and has been well studied. Traditionally, this problem has been mainly solved in a centralized manner. However, for applications related to large-scale complex networked systems, centralized algorithms are often subjected to some practical issues due to limited computational power and communication bandwidth. As a promising and viable alternative, distributed algorithms can effectively address the issues associated with centralized algorithms by solving the problem efficiently in a multi-agent setting that accords with the distributed nature of networked systems. Distributed algorithms decompose the entire problem into many sub-problems that are solved by individual agents in a cooperative manner. In this survey paper, we provide a detailed overview of the state of the art relevant to distributed algorithms for solving a system of linear equations. We will first review basic distributed algorithms including both discrete-time and continuous-time algorithms. Then we will discuss the extended algorithms to achieve communication efficiency. Furthermore, we will also introduce distributed algorithms to obtain the minimum-norm solution for a system of linear equations with multiple solutions, as well as the least-squares solutions when there is no solution. Finally, the relationship of distributed algorithms for solving a system of linear equations to the existing distributed optimization algorithms is discussed.

Wang, Peng↗
DOE OSTI2022

Impact of cyber attacks on distributed compressive sensing based state estimation in power distribution grids

Modern power distribution grids suffer from multiple vulnerabilities due to the tight integration between the physical system and the cyber infrastructure. Sophisticated and malicious cyber attacks continue to adversely impact the grid operation leading to performance degradation, service interruption, and grid failure. State estimation plays an essential role in grid monitoring and advancing cyber-attack situational awareness. In this regard, this paper first proposes a distributed compressive sensing (CS) state estimation approach for an unobservable distribution grid. Further, the proposed distributed CS approach divides the distribution grid into sub-areas to perform local state estimation. Then an alternating direction method of multipliers (ADMM) based iterative information exchange among neighboring areas is employed to complete the estimation process. In this estimation process, the impact of loss of measurement data, false data injection (FDI), replay, and neighborhood cyber-attacks is analyzed. Extensive simulations are performed on the IEEE 37-bus and IEEE 123-bus standard networks to demonstrate the algorithm’s robustness to the aforementioned cyber-attacks. A quantitative analysis of computational complexity and simulation time of the distributed CS based approach is also presented.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2025

Estimating energy consumption and GHG emissions in the U.S. food supply chain for net-zero

This work provides a database of the U.S. food system’s energy consumption and GHG emissions at the national and state levels by food supply chain (FSC) stage, fuel type, and food commodity. We estimate that the U.S. FSC consumed a total 4660 TBTU (4900 PJ) of site energy, 7130 TBTU (7500 PJ) of primary energy, and generated 970 MMT of GHG emissions in 2016. Among all the stages, on-farm production is the largest energy consumer (31% primary energy) and GHG emissions contributor (70%), largely due to raising animals. Optimizing distribution can reduce the stage’s energy consumption and GHG emissions and increase products’ shelf-life. Reducing food loss and waste is another good option, as it decreases the amount of food necessary to grow, thus impacting the overall FSC. The database can help stakeholders identify stage- and region-specific strategies and measures to curtail the environmental footprint of the U.S. food system.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2023

Networked Microgrids for Improved Resilient Operation: A Case Study in Adjuntas Puerto Rico

This paper presents a case study of a network microgrid orchestrator designed to allow coordinated operation of microgrids. Through Control Hardware in the Loop (CHIL), the proposed microgrid orchestrator was validated using, as case study two community-owned microgrids in Adjuntas, Puerto Rico. These two microgrids in Adjuntas were designed to provide affordable and reliable access to electricity to 14 businesses located in the town square. In its current design, these two microgrids operate independently from each other. This paper studies how networking them would increase their resiliency metrics while operating as an island. A distributed optimization is implemented in the microgrid orchestrator due to its superior scalability compared to centralized approach as well as to maintain data privacy. Two operational modes are considered in this study to showcase the advantages of networking microgrids: a) normal island, and b) degraded island. CHIL simulations are performed to validate the proposed microgrid orchestrator. CHIL results show meaningful resiliency improvements obtained by networking microgrids during contingencies such as loss of PV generation.

Ferrari Maglia, Max↗
DOE OSTI2020

A Robust Hierarchical Dispatch Scheme for Active Distribution Networks Considering Home Thermal Flexibility

Distribution networks are changing from passive absorbers of electric energy to active distribution networks (ADNs) capable of operating and participating in electricity markets. In the context of residential microgrids, which is a type of ADNs, aggregated home heating, ventilation and air-conditioning (HVAC) loads present a key opportunity to drive operational and economic objectives, facilitate high renewable energy penetration, and enhance both system resiliency and flexibility. A robust, hierarchical dispatch scheme is developed and presented in this paper, which connects an upper level multi-phase distribution optimal power flow (DOPF) to a lower level model predictive control (MPC)-based HVAC fleet controller. The approach is tested and verified on a modified IEEE 13 bus system in an intraday market application. The results demonstrate that the proposed hierarchical dispatch scheme is able to drive both economic and operational objectives for the ADN operator.

Rooks, Cody D.↗
DOE OSTI2021

The effects of allelochemical transfer on the dynamics of hosts, parasitoids, and competing hyperparasitoids

Allelochemicals produced by plants may be ingested by herbivorous insects and transferred to higher trophic levels with potentially deleterious effects. We develop a system of differential equations to investigate the effect of the transfer of allelochemicals, such as nicotine, on the population dynamics of a system of hosts, parasitoids, and two competing hyperparasitoids that attack different life stages of the parasitoids. We find both somewhat deleterious effects of nicotine on the larvae-attacking hyperparasitoids and increased attack rates for the pupae-attacking hyperparasitoids can promote coexistence. We also use an evolutionary game-theoretic approach to determine the optimal distribution of hyperparasitoid attacks among nicotine-producing and nicotine-free plants. With strong deleterious effects of nicotine and increased attack rates for the pupae-attacking hyperparasitoid, we find both species attack parasitoids on the nicotine-free plant but only pupae-attacking hyperparasitoids attack parasitoids on the nicotine-producing plant.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Interpreting Ionic Conductivity for Polymer Electrolyte Fuel Cell Catalyst Layers with Electrochemical Impedance Spectroscopy and Transmission Line Modeling

A cathode catalyst layer containing optimally distributed ionomer is critical to reduce the platinum loading and increase its utilization in polymer electrolyte fuel cells. Here, electrochemical impedance spectroscopy (EIS) was used to measure effective ionic conductivity of pseudo catalyst layers (PCLs) at a relative humidity (RH) range of 50%-120%. These results are compared to previous work using the hydrogen pump (HP) method. EIS effective ionic conductivity results reported here are higher than those from the HP because in the HP set-up ionic pathways must be effectively connected through the PCL to be counted, whereas in the EIS measurement, ionomer segments that are in contact with the membrane but are not effectively connected all the way through the PCL can be detected. Double layer capacitances and effective ionic conductivities of Pt/C catalyst layers with various supports and ionomer to carbon (I/C) ratios were studied. High surface area carbon support resulted in a lower effective ionic conductivity compared to the graphitized carbon support due to worse ionomer dispersion. Effective ionic conductivities of Pt/C layers were compared to that of PCLs. On average, effective ionic conductivities of Pt/C layers were higher than PCLs because of possible carbon agglomeratio n within the PCLs.

25 ENERGY STORAGE↗
DOE OSTI2020

Wattile: Probabilistic Deep Learning-based Forecasting of Building Energy Consumption [SWR-20-94]

Accurate energy forecasting is becoming critical due to many reasons: i ) optimal distributed energy resources operations and dispatch, ii) fault detection and diagnostics, and iii) meeting operational energy efficiency targets. Wattile uses deep learning (DL) for the building's short-term load forecasting application. Two specific types of neural networks called, Long Short Term Memory (LSTM) and Sequence-to-Sequence (S2S) models are used to make predictions. Forecasting models are trained using online historical weather and occupancy indicator data streams from the Intelligent Campus Program's data acquisition systems at the National Renewable Energy Laboratory (NREL) for main meters and sub-meters of multiple building types. These models use probabilistic methods to provide quantile-based forecasts in addition to nominal conditional median predictions of electricity consumption.

Frank, Stephen↗
DOE OSTI2023

Multi-injection investigation of a high-volatility diesel in advanced compression ignition combustion for NO x control

Traditional selective catalytic reduction aftertreatment technologies used to reduce [Formula: see text] are very limited at exhaust temperatures below [Formula: see text]. Therefore, under these low engine load conditions, having effective in-cylinder control of [Formula: see text] emissions is important. Previous work by the authors explored the effect of fuel physical properties on the ability to control [Formula: see text] in-cylinder. That work was limited to one direct injection near top dead center. Modern diesel high-pressure fuel systems have the capability of five or more injections in one engine cycle. A higher-volatility diesel fuel and high amounts of exhaust gas recirculation to delay ignition could provide an opportunity for reduction in engine-out [Formula: see text] through an increased level of fuel premixing. By appropriately timing multiple short injections, a more optimal distribution of fuel in-cylinder may be achieved, which could reduce [Formula: see text] while maintaining an efficient combustion phasing. A computational fluid dynamics model previously validated against experimental data was used to explore several injection strategies with increased levels of fuel premixing to assess the potential trade-offs between [Formula: see text] and CO/unburned hydrocarbon (UHC) emissions and thus reduce reliance on the aftertreatment system for [Formula: see text] control. The results show that the devised injection strategies resulted in an increased level of fuel premixing. However, none of the attempted injection strategies resulted in significant [Formula: see text] reductions, and all strategies showed a significant increase in CO and UHC emissions.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2022

Fed-DeepONet: Stochastic Gradient-Based Federated Training of Deep Operator Networks

The Deep Operator Network (DeepONet) framework is a different class of neural network architecture that one trains to learn nonlinear operators, i.e., mappings between infinite-dimensional spaces. Traditionally, DeepONets are trained using a centralized strategy that requires transferring the training data to a centralized location. Such a strategy, however, limits our ability to secure data privacy or use high-performance distributed/parallel computing platforms. To alleviate such limitations, in this paper, we study the federated training of DeepONets for the first time. That is, we develop a framework, which we refer to as Fed-DeepONet, that allows multiple clients to train DeepONets collaboratively under the coordination of a centralized server. To achieve Fed-DeepONets, we propose an efficient stochastic gradient-based algorithm that enables the distributed optimization of the DeepONet parameters by averaging first-order estimates of the DeepONet loss gradient. Then, to accelerate the training convergence of Fed-DeepONets, we propose a moment-enhanced (i.e., adaptive) stochastic gradient-based strategy. Finally, we verify the performance of Fed-DeepONet by learning, for different configurations of the number of clients and fractions of available clients, (i) the solution operator of a gravity pendulum and (ii) the dynamic response of a parametric library of pendulums.

Moya, Christian↗
DOE OSTI2019

Getting the Most Out of Your Crystals: Data Collection at the New High-Flux, Microfocus MX Beamlines at NSLS-II

Advances in synchrotron technology are changing the landscape of macromolecular crystallography. The two recently opened beamlines at NSLS-II—AMX and FMX—deliver high-flux microfocus beams that open new possibilities for crystallographic data collection. They are equipped with state-of-the-art experimental stations and automation to allow data collection on previously intractable crystals. Optimized data collection strategies allow users to tailor crystal positioning to optimally distribute the X-ray dose over its volume. Vector data collection allows the user to define a linear trajectory along a well diffracting volume of the crystal and perform rotational data collection while moving along the vector. This is particularly well suited to long, thin crystals. We describe vector data collection of three proteins—Akt1, PI3Kα, and CDP-Chase—to demonstrate its application and utility. For smaller crystals, we describe two methods for multicrystal data collection in a single loop, either manually selecting multiple centers (using H108A-PHM as an example), or “raster-collect”, a more automated approach for a larger number of crystals (using CDP-Chase as an example).

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

Designing for Zero Energy and Zero Carbon on a Multi-Building Scale Using URBANopt: Preprint

Groundbreaking efforts are necessary to mitigate contributors increasing impacts of climate change. In parallel to inventing pioneering clean energy technologies it is even more fundamental to rethink designing energy systems within a singular facility and collectively to function as a district. Facilities should not be continuously passive by just consuming; there is a need to shift to perform more dynamically. Designing for zero energy and zero carbon on a multi-building scale can uncover opportunities for building energy efficiency, decarbonization, demand flexibility, and resiliency that are not accessible at an individual building scale. This approach can be challenging without innovative tools to evaluate the multitude of possibilities. As an investigated result, we highlight the use of a campus-scale energy modeling platform - URBANopt™ - for the expansion of the National Renewable Energy Laboratory's (NREL's) South Table Mountain campus in Golden, Colorado. Programmatic growth included the design of three new all-electric, zero-energy, and zero-carbon, mixed used buildings (a combination of research laboratories and office space). This investigation is critical to NREL reaching net-zero emissions for its operational footprint, which will occur in phases over the next decade. Leveraging URBANopt's capabilities, we evaluate 1) high-performance building energy efficiency and decarbonization measures, 2) 4th generation district heating and cooling (4th GDHC) systems, 3) optimized onsite generation and energy storage assets that meet zero-energy and zero-carbon targets at minimum life-cycle costs, and 4) cost-optimal distributed energy technology mixes, dispatch strategies, and associated capacities that increase resiliency to grid outages. This work demonstrates the use and capabilities of URBANopt through a real-world case study on a multi-building scale.

community energy model↗
DOE OSTI2022

A Deep Reinforcement Learning-based Reserve Optimization in Active Distribution Systems for Tertiary Frequency Regulation

Federal Energy Regulatory Commission (FERC)Orders 841 and 2222 have recommended that distributed energy resources (DERs) should participate in energy and reserve markets; therefore, a mechanism needs to be developed to facilitate DERs’ participation at the distribution level. Although the available reserve from a single distribution system may not be sufficient for tertiary frequency regulation, stacked and coordinated contributions from several distribution systems can enable them participate in tertiary frequency regulation at scale. This paper proposes a deep reinforcement learning (DRL)-based approach for optimization of requested aggregated reserves by system operators among the clusters of DERs. The co-optimization of cost of reserve, distribution network loss, and voltage regulation of the feeders are considered while optimizing the reserves among participating DERs. The proposed framework adopts deep deterministic policy gradient (DDPG), which is an algorithm based on an actor-critic method. The effectiveness of the proposed method for allocating reserves among DERs is demonstrated through case studies on a modified IEEE 34-node distribution system.

deep reinforcement learning, distributed energy re↗
DOE OSTI2022

Towards Optimal and Executable Distribution Grid Restoration Planning With a Fine-Grained Power-Communication Interdependency Model

Distribution service restoration (DSR) under natural disasters is always a critical and challenging problem for utility companies. An effective solution must not ignore the power-communication interdependency as various systems are getting increasingly connected in the Smart Grid era. In this paper, we propose a two-layer distribution system model with both power and communication components. Based on this model, we formulate the restoration process as a routing problem that schedules the path and action sequence of utility crews that involves repairing damaged components, closing power switches, and enabling communication paths between the control center and remote field devices. Further, we develop a simulation-based method to quantitatively evaluate the restoration process with public reference models of large-scale power systems. The experimental results show that our method improves the total restored energy up to 57.6% and reduces the recovery time up to 63% by considering the power-communication interdependency.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2025

Decentralized Distributed Proximal Policy Optimization (DD-PPO) for High Performance Computing Scheduling on Multi-User Systems

Resource allocation in High Performance Computing (HPC) environments presents a complex and multifaceted challenge for job scheduling algorithms. Beyond the efficient allocation of system resources, schedulers must account for and optimize multiple performance metrics, including job wait time and system throughput. Traditional heuristic-based scheduling algorithms increasingly struggle and lack the efficiency needed to meet the demands and address the complexity and scale of modern HPC systems. Consequently, recent research efforts have focused on leveraging advancements in Artificial Intelligence (AI) and Deep Learning (DL), particularly Reinforcement Learning (RL), to develop more adaptable and intelligent scheduling strategies. Previous RL-based scheduling approaches have explored a range of algorithms, from Deep Q-Networks (DQN) to Proximal Policy Optimization (PPO), and more recently, hybrid methods that integrate Graph Neural Networks (GNNs) with RL techniques. However, a common limitation across these methods is their reliance on relatively small datasets, with few methods being evaluated using large-scale, multi-million-job trace datasets representative of real-world HPC workloads. Moreover, existing RL schedulers face scalability issues due to centralized policy updates, which hinder training efficiency and performance when applied to large datasets. This study introduces a novel RL-based scheduler utilizing Decentralized Distributed Proximal Policy Optimization (DD-PPO) algorithm, which supports large-scale distributed training across multiple workers without requiring parameter synchronization at every step. By eliminating reliance on centralized updates to a shared policy, the DD-PPO scheduler enhances scalability, training efficiency, and sample utilization. Experimental validation using a large real-world dataset containing over 11.5 million job traces collected from petascale HPC systems over six years assesses the influence of dataset scale on training effectiveness and compares DD-PPO performance to traditional and advanced scheduling approaches. The experimental results demonstrate improved scheduling performance in comparison to both heuristic-based schedulers and existing RL-based scheduling algorithms.

AI↗