Search NASA⌕ Search

SEARCH · Search NASA

Results for “Real Time 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 505 records · Page 28

SU-D-213-03: Towards An Optimized 3D Scintillation Dosimetry Tool for Quality Assurance of Dynamic Radiotherapy Techniques

Purpose: The purpose of this work is to simulate a multi-focus plenoptic camera used as the measuring device in a real-time three-dimensional scintillation dosimeter. Simulating and optimizing this realistic optical system will bridge the technological gap between concept validation and a clinically viable tool that can provide highly efficient, accurate and precise measurements for dynamic radiotherapy techniques. Methods: The experimental prototype, previously developed for proof of concept purposes, uses an off-the-shelf multi-focus plenoptic camera. With an array of interleaved microlenses of different focal lengths, this camera records spatial and angular information of light emitted by a plastic scintillator volume. The three distinct microlens focal lengths were determined experimentally for use as baseline parameters by measuring image-to-object magnification for different distances in object space. A simulated plenoptic system was implemented using the non-sequential ray tracing software Zemax: this tool allows complete simulation of multiple optical paths by modeling interactions at interfaces such as scatter, diffraction, reflection and refraction. The active sensor was modeled based on the camera manufacturer specifications by a 2048×2048, 5 µm-pixel pitch sensor. Planar light sources, simulating the plastic scintillator volume, were employed for ray tracing simulations. Results: The microlens focal lengths were determined to be 384, 327 and 290 µm. A realistic multi-focus plenoptic system, with independently defined and optimizable specifications, was fully simulated. A f/2.9 and 54 mm-focal length Double Gauss objective was modeled as the system’s main lens. A three-focal length hexagonal microlens array of 250-µm thickness was designed, acting as an image-relay system between the main lens and sensor. Conclusion: Simulation of a fully modeled multi-focus plenoptic camera enables the decoupled optimization of the main lens and microlens specifications. This work leads the way to improving the 3D dosimeter’s achievable resolution, efficiency and build for providing a quality assurance tool fully meeting clinical needs. M.R. is financially supported by a Master’s Canada Graduate Scholarship from the NSERC. This research is also supported by the NSERC Industrial Research Chair in Optical Design.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Dose reconstruction for real-time patient-specific dose estimation in CT

Purpose: Many recent computed tomography (CT) dose reduction approaches belong to one of three categories: statistical reconstruction algorithms, efficient x-ray detectors, and optimized CT acquisition schemes with precise control over the x-ray distribution. The latter category could greatly benefit from fast and accurate methods for dose estimation, which would enable real-time patient-specific protocol optimization. Methods: The authors present a new method for volumetrically reconstructing absorbed dose on a per-voxel basis, directly from the actual CT images. The authors’ specific implementation combines a distance-driven pencil-beam approach to model the first-order x-ray interactions with a set of Gaussian convolution kernels to model the higher-order x-ray interactions. The authors performed a number of 3D simulation experiments comparing the proposed method to a Monte Carlo based ground truth. Results: The authors’ results indicate that the proposed approach offers a good trade-off between accuracy and computational efficiency. The images show a good qualitative correspondence to Monte Carlo estimates. Preliminary quantitative results show errors below 10%, except in bone regions, where the authors see a bigger model mismatch. The computational complexity is similar to that of a low-resolution filtered-backprojection algorithm. Conclusions: The authors present a method for analytic dose reconstruction in CT, similar to the techniques used in radiation therapy planning with megavoltage energies. Future work will include refinements of the proposed method to improve the accuracy as well as a more extensive validation study. The proposed method is not intended to replace methods that track individual x-ray photons, but the authors expect that it may prove useful in applications where real-time patient-specific dose estimation is required.

60 APPLIED LIFE SCIENCES↗

Power Minimization techniques for Networked Data Centers.

Our objective is to develop a mathematical model to optimize energy consumption at multiple levels in networked data centers, and develop abstract algorithms to optimize not only individual servers, but also coordinate the energy consumption of clusters of servers within a data center and across geographically distributed data centers to minimize the overall energy cost and consumption of brown energy of an enterprise. In this project, we have formulated a variety of optimization models, some stochastic others deterministic, and have obtained a variety of qualitative results on the structural properties, robustness, and scalability of the optimal policies. We have also systematically derived from these models decentralized algorithms to optimize energy efficiency, analyzed their optimality and stability properties. Finally, we have conducted preliminary numerical simulations to illustrate the behavior of these algorithms. We draw the following conclusion. First, there is a substantial opportunity to minimize both the amount and the cost of electricity consumption in a network of datacenters, by exploiting the fact that traffic load, electricity cost, and availability of renewable generation fluctuate over time and across geographical locations. Judiciously matching these stochastic processes can optimize the tradeoff between brown energy consumption, electricity cost, and response time. Second, given the stochastic nature of these three processes, real-time dynamic feedback should form the core of any optimization strategy. The key is to develop decentralized algorithms that can be implemented at different parts of the network as simple, local algorithms that coordinate through asynchronous message passing.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Three-dimensional NDE of VHTR core components via simulation-based testing. Final report

A next generation, simulation-driven-and-enabled testing platform is developed for the 3D detection and characterization of defects and damage in nuclear graphite and composite structures in Very High Temperature Reactors (VHTRs). The proposed work addresses the critical need for the development of high-fidelity Non-Destructive Examination (NDE) technologies for as-manufactured and replaceable in-service VHTR components. Centered around the novel use of elastic (sonic and ultrasonic) waves, this project deploys a robust, non-iterative inverse solution for the 3D defect reconstruction together with a non-contact, laser-based approach to the measurement of experimental waveforms in VHTR core components. In particular, this research (1) deploys three-dimensional Scanning Laser Doppler Vibrometry (3D SLDV) as a means to accurately and remotely measure 3D displacement waveforms over the accessible surface of a VHTR core component excited by mechanical vibratory source; (2) implements a powerful new inverse technique, based on the concept of Topological Sensitivity (TS), for non-iterative elastic waveform tomography of internal defects - that permits robust 3D detection, reconstruction and characterization of discrete damage (e.g. holes and fractures) in nuclear graphite from limited-aperture NDE measurements; (3) implements state-of-the art computational (finite element) model that caters for accurately simulating elastic wave propagation in 3D blocks of nuclear graphite; (4) integrates the SLDV testing methodology with the TS imaging algorithm into a non-contact, high-fidelity NDE platform for the 3D reconstruction and characterization of defects and damage in VHTR core components; and (5) applies the proposed methodology to VHTR core component samples (both two- and three-dimensional) with a priori induced, discrete damage in the form of holes and fractures. Overall, the newly established SLDV-TS testing platform represents a next-generation NDE tool that surpasses all existing techniques for the 3D ultrasonic imaging of material damage from non-contact, limited-aperture waveform measurements. Outlook. The next stage in the development of this technology includes items such as (a) non-contact generation of mechanical vibrations in VHTR components via thermal expansion created by high-intensity laser; (b) development and incorporation of Synthetic Aperture Focusing Technique (SAFT) for elevating the accuracy of 3D imaging in highly noisy environments with minimal accessible surface; (c) further analytical and computational developments to facilitate the reconstruction of diffuse damage (e.g. microcracks) in nuclear graphite as they lead to the dispersion of elastic waves, (d) concept of model updating for accurate tracking of the evolution of material damage via periodic inspections; (d) adoption of the Bayesian framework to obtain information on the certainty of obtained images; and (e) optimization of the computational scheme toward real-time, model-based imaging of damage in VHTR core components.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

ML based control systems for nuclear physics experiments

The Experimental Physics Software and Computing Infrastructure (EPSCI) group at Jefferson Lab is leading the use of machine learning (ML) to enhance control systems in nuclear physics experiments. Collaborating closely with domain experts and data scientists, we have developed an ML-based control system that uses a Gaussian process to dynamically adjust the high voltage of the GlueX Central Drift Chamber. This results in stable detector performance by adapting to environmental changes, thereby reducing the offline calibration effort. Furthermore, we are developing ML-driven systems for optimizing the polarization of photon beams and polarized cryotargets. These systems will maintain the optimal microwave frequency in cryogenic targets and make real-time adjustments to diamond radiators for polarized photon sources, tasks traditionally handled by human operators. By automating these functions, we aim to optimize the polarization, reduce downtime, and minimize human error. This talk will highlight the development of reliable ML-based control systems and the policies to ensure they are both effective and trustworthy.

Jeske, Torri↗

Quantum annealing for combinatorial optimization: a benchmarking study

Quantum annealing (QA) has the potential to significantly improve solution quality and reduce time complexity in solving combinatorial optimization problems compared to classical optimization methods. However, due to the limited number of qubits and their connectivity, the QA hardware did not show such an advantage over classical methods in past benchmarking studies. Recent advancements in QA with more than 5000 qubits, enhanced qubit connectivity, and the hybrid architecture promise to realize the quantum advantage. Here, we use a quantum annealer with state-of-the-art techniques and benchmark its performance against classical solvers. To compare their performance, we solve over 50 optimization problem instances represented by large and dense Hamiltonian matrices using quantum and classical solvers. The results demonstrate that a state-of-the-art quantum solver has higher accuracy (~0.013%) and a significantly faster problem-solving time (~6561×) than the best classical solver. Our results highlight the advantages of leveraging QA over classical counterparts, particularly in hybrid configurations, for achieving high accuracy and substantially reduced problem solving time in large-scale real-world optimization problems.

97 MATHEMATICS AND COMPUTING↗

Laboratory Evaluation of Federated, Hierarchical Controls for Distribution Power System Management: Preprint

The connection of more loads and distributed energy resources (DERs) to the distribution power system brings both challenges and opportunities to system operators. There are opportunities to aggregate flexible loads and DERs to provide transmission grid services, but the coordinated actions of DERs being managed by independent, third-party DER aggregators to support transmission system operations can present challenges. We developed a federated DER management architecture and control framework that aims to manage heterogeneous DERs to deliver reliable transmission grid services while respecting distribution system constraints. The controls include stochastic day-ahead optimization, model predictive control, and a simple real-time management scheme. We present simulation results obtained from a realistic laboratory test bed of federated controls managing DERs within a substation service area to make the substation net power follow the optimal net power determined by the day-ahead optimization based on cost and limiting reverse power flow.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Energy-efficient multimodal mobility networks in transportation digital twins: Strategies and optimization

The study proposes a comprehensive Transportation Mobility (TransitMo) framework covering conceptual design, model formulation, optimization, simulation, and impact analysis of the transportation mobility system. TransitMo is composed of a transportation digital twin developed in Simulation of Urban MObility (SUMO) and an Intelligent Traffic Management and Control Center (ITMCC) that identifies the best ways to improve the movement of people within urban areas using various modes of transportation. This study encompasses advanced modeling techniques, algorithms, and strategic testing to optimize energy efficiency and mobility in a multimodal shared mobility network. TransitMo’s practical applications are exemplified through a city-scaled simulation network in Chattanooga, TN, employing demographic data to analyze historical traffic patterns and forecast future demands. Central to this methodology are three models: the User Preference Model (UP), the Energy Consumption Model (EC), and the System Optimization Model (SO). These models work in concert to iteratively devise the optimal travel incentives and minimize the total system cost in a real-time manner. In conclusion, test results verified that the proposed adaptive incentive program and optimized bus scheduling can improve network performance by increasing public transit ridership.

42 ENGINEERING↗

Development of Digital Twin-Informed Predictive Maintenance for Critical Components in Advanced Reactors

Small modular reactors (SMRs) and microreactors, along with other advanced reactor (AR) technologies, are key to the future of nuclear energy. For these systems to achieve low operating costs, high reliability, and flexibility across applications, their operation and maintenance must be optimized. Digital twin (DT) technology is one of the technologies that enables real-time (or faster than real-time) monitoring and prognosis of critical components which are vital for operational efficiency, low costs, and enhanced safety of ARs, accelerating their deployment. DT technology provides dynamic virtual representation of physical assets by integrating real-time data, physics-based models, and advanced analytics, which is critical to optimizing the performance of the entire energy system throughout the life cycle. DTs empower engineers and operators to virtually explore different scenarios, configurations, and control strategies, allowing for the identification of optimal solutions that maximize reactor efficiency, safety, and economics.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗

Multicolor multifocal 3D microscopy using in-situ optimization of a spatial light modulator

Abstract Multifocal microscopy enables high-speed three-dimensional (3D) volume imaging by using a multifocal grating in the emission path. This grating is typically designed to afford a uniform illumination of multifocal subimages for a single emission wavelength. Using the same grating for multicolor imaging results in non-uniform subimage intensities in emission wavelengths for which the grating is not designed. This has restricted multifocal microscopy applications for samples having multicolored fluorophores. In this paper, we present a multicolor multifocal microscope implementation which uses a Spatial Light Modulator (SLM) as a single multifocal grating to realize near-uniform multifocal subimage intensities across multiple wavelength emission bands. Using real-time control of an in-situ-optimized SLM implemented as a multifocal grating, we demonstrate multicolor multifocal 3D imaging over three emission bands by imaging multicolored particles as well as Escherichia coli ( E. coli ) interacting with human liver cancer cells, at $$\sim 2.5$$ ∼ 2.5 multicolor 3D volumes per second acquisition speed. Our multicolor multifocal method is adaptable across SLM hardware, emission wavelength band locations and number of emission bands, making it particularly suited for researchers investigating fast processes occurring across a volume where multiple species are involved.

47 OTHER INSTRUMENTATION↗

Regulation of Dynamical Systems to Optimal Solutions of Semidefinite Programs: Algorithms and Applications to AC Optimal Power Flow

This paper considers a collection of networked nonlinear dynamical systems, and addresses the synthesis of feedback controllers that seek optimal operating points corresponding to the solution of pertinent network-wide optimization problems. Particular emphasis is placed on the solution of semidefinite programs (SDPs). The design of the feedback controller is grounded on a dual e-subgradient approach, with the dual iterates utilized to dynamically update the dynamical-system reference signals. Global convergence is guaranteed for diminishing stepsize rules, even when the reference inputs are updated at a faster rate than the dynamical-system settling time. The application of the proposed framework to the control of power-electronic inverters in AC distribution systems is discussed. The objective is to bridge the time-scale separation between real-time inverter control and network-wide optimization. Optimization objectives assume the form of SDP relaxations of prototypical AC optimal power flow problems.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Automated ICRF heating surrogate modeling via machine learning

This work introduces automated machine learning workflows that address critical bottlenecks in surrogate model development for Ion Cyclotron Range of Frequencies (ICRF) heating applications. The automated framework includes data analysis tools that transform raw datasets into actionable insights in seconds, replacing weeks of manual exploratory effort and ensuring consistent, reproducible dataset characterization. By integrating advanced hyperparameter optimization (HPO) methods including Bayesian optimization via BoTorch and Tree-structured Parzen Estimators (TPE), the framework significantly reduces model development time from weeks to hours, decreasing computational cost and required expertise, while enabling high-accuracy surrogate models. Compared to traditional hyperparameter scanning (HPS) techniques such as methodical, randomized, and grid searches, HPO methods achieve superior convergence and predictive performance, even when compared to already well-tuned reference models. On NSTX High Harmonic Fast Wave (HHFW) heating datasets, both Random Forest Regressor (RFR) and neural network surrogates demonstrate improved accuracy, achieving R 2 values beyond 0.97 and 0.98, respectively. The results show that while HPO gains are modest for robust architectures like RFR, they become essential for more sensitive models such as neural networks, highlighting the trade-offs across optimization strategies. Through automated workflows that eliminate manual hyperparameter tuning and require minimal ML expertise, this work enables widespread adoption of high-fidelity surrogate models across the fusion community for real-time plasma control, uncertainty quantification, rapid experimental scenario development, and integrated system optimization.

Sanchez-Villar, Alvaro [Princeton Plasma Physics L↗

Real time intelligent process control system for thin film solar cell manufacturing

This project addresses the problem of lower solar conversion efficiency and waste in the typical solar cell manufacturing process. The work from the proposed development will lead toward developing a system which should be able to increase solar panel conversion efficiency by an additional 12-15% resulting in lower cost panels, increased solar technology adoption, reduced carbon emissions and reduced dependency on foreign oil. All solar cell manufacturing processes today suffer from manufacturing inefficiencies that currently lead to lower product quality and lower conversion efficiency, increased product cost and greater material and energy consumption. This results in slower solar energy adoption and extends the time solar cells will reach grid parity with traditional energy sources. The thin film solar panel manufacturers struggle on a daily basis with the problem of thin film thickness non-uniformity and other parameters variances over the deposited substrates, which significantly degrade their manufacturing yield and quality. Optical monitoring of the thin films during the process of the film deposition is widely perceived as a necessary step towards resolving the non-uniformity and non-homogeneity problem. In order to enable the development of an optical control system for solar cell manufacturing, a new type of low cost optical sensor is needed, able to acquire local information about the panel under deposition and measure its local characteristics, including the light scattering in very close proximity to the surface of the film. This information cannot be obtained by monitoring from outside the deposition chamber (as traditional monitoring systems do) due to the significant signal attenuation and loss of its scattering component before the reflected beam reaches the detector. In addition, it would be too costly to install traditional external in-situ monitoring systems to perform any real-time monitoring over large solar panels, since it would require significant equipment refurbishing needed for installation of multiple separate ellipsometric systems, and development of customized software to control all of them simultaneously. The proposed optical monitoring system comprises AccuStrata’s fiber optics sensors installed inside the thin film deposition equipment, a hardware module of different components (beyond the scope of this project) and our software program with iterative predicting capability able to control material bandgap and surface roughness as films are deposited. Our miniature fiber optics monitoring sensors are installed inside the vacuum chamber compartments in very close proximity where the independent layers are deposited (an option patented by us in 2003). The optical monitoring system measures two of the most important parameters of the photovoltaic thin films during deposition on a moving solar panel - material bandgap and surface roughness. In this program each sensor array consists of two fiber optics sensors monitoring two independent areas of the panel under deposition. Based on the monitored parameters and their change in time and from position to position on the panel, the system is able to provide to the equipment operator immediate information about the thin films as they are deposited. This DoE Supply Chain program is considered the first step towards the development of intelligent optical control system capable of dynamically adjusting the manufacturing process “on-the-fly” in order to achieve better performance. The proposed system will improve the thin film solar cell manufacturing by improving the quality of the individual solar cells and will allow for the manufacturing of more consistent and uniform products resulting in higher solar conversion efficiency and manufacturing yield. It will have a significant impact on the multibillion-dollar thin film solar market. We estimate that the financial impact of these improvements if adopted by only 10% of the industry ($7.7 Billion) would result in about $1.5 Billion in savings by 2015 (at the assumed 20% improvement). This can be achieved by optimizing the manufactured product and process in real time without changing the manufacturing technology.

14 SOLAR ENERGY↗

Low-Temperature Plasma-Based Metrology of Lithium-Ion Battery Electrode Materials (CRADA Final Report)

As part of the Cyclotron Road program, SirenOpt Inc. evaluated its low-temperature plasma-based metrology sensor prototype for measuring multiple critical properties of lithium-ion battery electrode materials in parallel and in real-time. Cost-effective, minimal-waste manufacturing of high-performance battery electrode materials will be vital for achieving society’s net-zero carbon emission goals. Because existing electrode metrology sensors cannot operate within most sections of manufacturing lines, manufacturers often complete hundreds of processing steps before they can test their products and detect problems. When manufacturers perform these offline tests, they typically only test a small portion of the manufactured products. Current electrode manufacturing thus often yields many low-quality products, or off-spec products that must be thrown away all together. For example, at least 6% of the total lithium-ion battery manufacturing cost (i.e., over $250 million/year for the average gigafactory) is devoted to processing defective electrodes that are not scrapped until performance tests are failed during late-stage quality control checks. Electrode variability also leads manufacturers to build extra cells into battery packs to reduce the risk of poor performance. For example, many electric vehicle (EV) manufacturers include up to 10% more cells than needed, which substantially increases the cost and weight of the final EV product. The SirenOpt sensor can potentially enable early detection of poorly manufactured electrodes and allow them to be removed earlier from manufacturing lines, which can save battery manufacturers (hundreds of) millions of dollars per year. The sensor can further be used to improve product quality by accelerating R&D and process optimization, improving quality control, and enabling real-time process control. Overall, a real-time, in-situ metrology strategy can create unprecedented opportunities for implementation of smart manufacturing practices and advanced quality and process control solutions to realize higher battery electrode throughput and performance.

25 ENERGY STORAGE↗

Enhanced Performance Assessment System (EPAS) for carbon sequestration.

Carbon capture and sequestration (CCS) is an option to mitigate impacts of atmospheric carbon emission. Numerous factors are important in determining the overall effectiveness of long-term geologic storage of carbon, including leakage rates, volume of storage available, and system costs. Recent efforts have been made to apply an existing probabilistic performance assessment (PA) methodology developed for deep nuclear waste geologic repositories to evaluate the effectiveness of subsurface carbon storage (Viswanathan et al., 2008; Stauffer et al., 2009). However, to address the most pressing management, regulatory, and scientific concerns with subsurface carbon storage (CS), the existing PA methodology and tools must be enhanced and upgraded. For example, in the evaluation of a nuclear waste repository, a PA model is essentially a forward model that samples input parameters and runs multiple realizations to estimate future consequences and determine important parameters driving the system performance. In the CS evaluation, however, a PA model must be able to run both forward and inverse calculations to support optimization of CO{sub 2} injection and real-time site monitoring as an integral part of the system design and operation. The monitoring data must be continually fused into the PA model through model inversion and parameter estimation. Model calculations will in turn guide the design of optimal monitoring and carbon-injection strategies (e.g., in terms of monitoring techniques, locations, and time intervals). Under the support of Laboratory-Directed Research & Development (LDRD), a late-start LDRD project was initiated in June of Fiscal Year 2010 to explore the concept of an enhanced performance assessment system (EPAS) for carbon sequestration and storage. In spite of the tight time constraints, significant progress has been made on the project: (1) Following the general PA methodology, a preliminary Feature, Event, and Process (FEP) analysis was performed for a hypothetical CS system. Through this FEP analysis, relevant scenarios for CO{sub 2} release were defined. (2) A prototype of EPAS was developed by wrapping an existing multi-phase, multi-component reservoir simulator (TOUGH2) with an uncertainty quantification and optimization code (DAKOTA). (3) For demonstration, a probabilistic PA analysis was successfully performed for a hypothetical CS system based on an existing project in a brine-bearing sandstone. The work lays the foundation for the development of a new generation of PA tools for effective management of CS activities. At a top-level, the work supports energy security and climate change/adaptation by furthering the capability to effectively manage proposed carbon capture and sequestration activities (both research and development as well as operational), and it greatly enhances the technical capability to address this national problem. The next phase of the work will include (1) full capability demonstration of the EPAS, especially for data fusion, carbon storage system optimization, and process optimization of CO{sub 2} injection, and (2) application of the EPAS to actual carbon storage systems.

54 ENVIRONMENTAL SCIENCES↗

Sci—Thur PM: Planning and Delivery — 06: Real-Time Interactive Treatment Planning

Purpose: To describe and evaluate a novel system for generalized Real-Time Interactive Planning (RTIP) applied to head and neck (H and N) VMAT. Methods: The clinician interactively manipulates dose distributions using DVHs, isodoses, or rate of dose fall-off, which may be subjected to user-defined constraints. Dose is calculated using a fast Achievable Dose Estimate (ADE) algorithm, which simulates the limits of what can be achieved during treatment. After each manipulation contributing fluence elements are modified and the dose distribution updates in effectively real-time. For H and N VMAT planning, structure sets for 11 patients were imported into RTIP. Each dose distribution was interactively modified to minimize OAR dose while constraining target DVHs. The resulting RTIP DVHs were transferred to the Eclipse™ VMAT optimizer, and conventional VMAT optimization was performed. Results: Dose calculation and update times for the ADE algorithm ranged from 2.4 to 22.6 milliseconds, thus facilitating effectively real-time manipulation of dose distributions. For each of the 11 H and N VMAT cases, the RTIP process took ∼2–10 minutes. All RTIP plans exhibited acceptable PTV coverage, mean dose, and max dose. 10 of 11 RTIP plans achieved substantially improved sparing of one or more OARs without compromising dose to targets or other OARs. Importantly, 10 of the 11 RTIP plans required only one or two post-RTIP optimizations. Conclusions: RTIP is a novel system for manipulating and updating achievable dose distributions in real-time. H and N VMAT plans generated using RTIP demonstrate improved OAR sparing and planning efficiency. Disclosures: One author has a commercial interest in the presented materials.

60 APPLIED LIFE SCIENCES↗

Scalable Prediction of Energy Consumption using Incremental Time Series Clustering

Time series datasets are a canonical form of high velocity Big Data, and often generated by pervasive sensors, such as found in smart infrastructure. Performing predictive analytics on time series data can be computationally complex, and requires approximation techniques. In this paper, we motivate this problem using a real application from the smart grid domain. We propose an incremental clustering technique, along with a novel affinity score for determining cluster similarity, which help reduce the prediction error for cumulative time series within a cluster. We evaluate this technique, along with optimizations, using real datasets from smart meters, totaling ~700,000 data points, and show the efficacy of our techniques in improving the prediction error of time series data within polynomial time.

Simmhan, Yogesh↗

Deep reinforcement learning control for co-optimizing energy consumption, thermal comfort, and indoor air quality in an office building

With the recent demand for decarbonization and energy efficiency, advanced HVAC control using Deep Reinforcement Learning (DRL) becomes a promising solution. Due to its flexible structures, DRL has been successful in energy reduction for many HVAC systems. However, only a few researches applied DRL agents to manage the entire central HVAC system and control multiple components in both the water loop and the air loop, owing to its complex system structures. Moreover, those researches have not extended their applications by incorporating the indoor air quality, especially both CO2 and PM2.5concentrations, on top of energy saving and thermal comfort, as achieving those objectives simultaneously can cause multiple control conflicts. What's more, DRL agents are usually trained on the simulation environment before deployment, so another challenge is to develop an accurate but relatively simple simulator. Therefore, we propose a DRL algorithm for a central HVAC system to co-optimize energy consumption, thermal comfort, indoor CO2 level, and indoor PM2.5 level in an office building. To train the controller, we also developed a hybrid simulator that decoupled the complex system into multiple simulation models, which are calibrated separately using laboratory test data. The hybrid simulator combined the dynamics of the HVAC system, the building envelope, as well as moisture, CO2, and particulate matter transfer. Three control algorithms (rule-based, MPC, and DRL) are developed, and their performances are evaluated on the hybrid simulator environment with a realistic scenario (i.e., with stochastic noises). The test results showed that, the DRL controller can save 21.4 % of energy compared to a rule-based controller, and has improved thermal comfort, reduced indoor CO2 concentration. The MPC controller showed an 18.6 % energy saving compared to the DRL controller, mainly due to savings from comfort and indoor air quality boundary violations caused by unmeasured disturbances, and it also highlights computational challenges in real-time control due to non-linear optimization. Finally, we provide the practical considerations for designing and implementing the DRL and MPC controllers based on their respective pros and cons.

Guo, Fangzhou↗