Search NASA⌕ Search

SEARCH · Search NASA

Results for “Network performance forecasting”

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 235 records · Page 13

A Case Study of the Weather Research and Forecasting Model Applied to the Joint Urban 2003 Tracer Field Experiment. Part III: Boundary-Layer Parametrizations

Numerical weather prediction is often used to supply the mean wind and turbulence fields for atmospheric transport and dispersion plume models as they provide dense geographic coverage in comparison to typically sparse monitoring networks. Here, the Weather Research and Forecasting (WRF) model 4.0 was run over the month-long period of the Joint Urban 2003 field campaign conducted in Oklahoma City. We compare three different simulations in their ability to reproduce the observations, each using a different boundary-layer parametrization. Specifically, we examine the Mellor–Yamada–Janjic (MYJ), Yonsei University (YSU), and Mellor–Yamada–Nakanishi–Niino (MYNN) boundary-layer parametrizations. All three predict the wind speed well during the day but overpredict it at night. The MYNN parametrization is better than MYJ at predicting the daytime turbulence in the surface layer, but both underpredict the nocturnal turbulence. Additionally, the MYJ parametrization is best at predicting the reciprocal Obukhov length, while MYNN and YSU both significantly overpredict thermal stability. Reconstructing the reciprocal Obukhov length from other simulated parameters produces more accurate values for both parametrizations. All three models overpredict the boundary-layer height, particularly under convective conditions. The MYJ parametrization overestimates boundary-layer height the most, while YSU and MYNN have comparable performance with MYNN having an advantage in predicting the stable boundary-layer height. Several days were found where the WRF simulations predict significant deviations from the prevailing diurnal pattern in wind direction, which are not found in the observations.

54 ENVIRONMENTAL SCIENCES↗

Deep Learning Image Segmentation for Atmospheric Rivers

Abstract The identification of atmospheric rivers (ARs) is crucial for weather and climate predictions as they are often associated with severe storm systems and extreme precipitation, which can cause large impacts on society. This study presents a deep learning model, termed ARDetect, for image segmentation of ARs using ERA5 data from 1960 to 2020 with labels obtained from the TempestExtremes tracking algorithm. ARDetect is a convolutional neural network (CNN)-based U-Net model, with its structure having been optimized using automatic hyperparameter tuning. Inputs to ARDetect were selected to be the integrated water vapor transport (IVT) and total column water (TCW) fields, as well as the AR mask from TempestExtremes from the previous time step to the one being considered. ARDetect achieved a mean intersection-over-union (mIoU) rate of 89.04% for ARs, indicating its high accuracy in identifying these weather patterns and a superior performance than most deep learning–based models for AR detection. In addition, ARDetect can be executed faster than the TempestExtremes method (seconds vs minutes) for the same period. This provides a significant benefit for online AR detection, especially for high-resolution global models. An ensemble of 10 models, each trained on the same dataset but having different starting weights, was used to further improve on the performance produced by ARDetect, thus demonstrating the importance of model diversity in improving performance. ARDetect provides an effective and fast deep learning–based model for researchers and weather forecasters to better detect and understand ARs, which have significant impacts on weather-related events such as floods and droughts.

Galea, Daniel↗

Spatial–Temporal Patterns of Historical, Near-Term, and Projected Drought in the Conterminous United States

Major droughts in the United States have heavily impacted the hydrologic system, negatively effecting energy and food production. Improved understanding of historical drought is critical for accurate forecasts. Data from global climate models (GCMs), commonly used to assess drought, cannot effectively evaluate local patterns because of their low spatial scale. This research leverages downscaled (~4 km grid spacing) temperature and precipitation estimates from nine GCMs’ data under the business-as-usual scenario (Representative Concentration Pathway 8.5) to examine drought patterns. Drought severity is estimated using the Palmer Drought Severity Index (PDSI) with the Thornthwaite evapotranspiration method. The specific objectives were (1) To reproduce historical (1966–2005) drought and calculate near-term to future (2011–2050) drought patterns over the conterminous USA. (2) To uncover the local variability of spatial drought patterns in California between 2012 and 2018 using a network-based approach. Our estimates of land proportions affected by drought agree with the known historical drought events of the mid-1960s, late 1970s to early 1980s, early 2000s, and between 2012 and 2015. Network analysis showed heterogeneity in spatial drought patterns in California, indicating local variability of drought occurrence. The high spatial scale at which the analysis was performed allowed us to uncover significant local differences in drought patterns. This is critical for highlighting possible weak systems that could inform adaptation strategies such as in the energy and agricultural sectors.

54 ENVIRONMENTAL SCIENCES↗

Atmospheric Inverse Estimates of Methane Emissions from Central California

Methane mixing ratios measured at a tall-tower are compared to model predictions to estimate surface emissions of CH{sub 4} in Central California for October-December 2007 using an inverse technique. Predicted CH{sub 4} mixing ratios are calculated based on spatially resolved a priori CH{sub 4} emissions and simulated atmospheric trajectories. The atmospheric trajectories, along with surface footprints, are computed using the Weather Research and Forecast (WRF) coupled to the Stochastic Time-Inverted Lagrangian Transport (STILT) model. An uncertainty analysis is performed to provide quantitative uncertainties in estimated CH{sub 4} emissions. Three inverse model estimates of CH{sub 4} emissions are reported. First, linear regressions of modeled and measured CH{sub 4} mixing ratios obtain slopes of 0.73 {+-} 0.11 and 1.09 {+-} 0.14 using California specific and Edgar 3.2 emission maps respectively, suggesting that actual CH{sub 4} emissions were about 37 {+-} 21% higher than California specific inventory estimates. Second, a Bayesian 'source' analysis suggests that livestock emissions are 63 {+-} 22% higher than the a priori estimates. Third, a Bayesian 'region' analysis is carried out for CH{sub 4} emissions from 13 sub-regions, which shows that inventory CH{sub 4} emissions from the Central Valley are underestimated and uncertainties in CH{sub 4} emissions are reduced for sub-regions near the tower site, yielding best estimates of flux from those regions consistent with 'source' analysis results. The uncertainty reductions for regions near the tower indicate that a regional network of measurements will be necessary to provide accurate estimates of surface CH{sub 4} emissions for multiple regions.

54 ENVIRONMENTAL SCIENCES↗

Intercomparison of Deep Learning Model Architectures for Atmospheric River Prediction

With a rapid surge in the application of machine learning (ML) for a diverse range of tasks in climate science, the present study addresses a challenge for climate scientists when selecting the optimal ML or deep learning (DL) architecture for a given application. In particular, a DL intercomparison study was performed with a focus on forecasting the position of atmospheric rivers (ARs) on short-range time scales (up to 5-day lead times). AR predictions from multiple DL architectures, including various types of convolutional autoencoders and a vision transformer (ViT), were compared against ECMWF ERA5 reanalysis and hindcasts from a global climate model. DL models with similar trainable parameters were trained on ERA5 reanalysis data and AR positions derived from a thresholding algorithm to ensure a fair comparison among the DL models. Each model’s performance and accuracy in forecasting AR location and key input fields within a 5-day window were assessed using metrics of root-mean-square error, anomaly correlation, and mean intersection over union. The ViT architecture outperformed other autoencoder models in most of the metrics. Incorporating additional meteorological fields only yielded slight improvements in forecasting certain fields at longer lead times. The results also suggest that a smaller number of input time steps or smaller number of autoregressive steps can achieve better prediction skills, while also improving the overall computational efficiency. This research offers valuable insights into the strengths and weaknesses of different DL techniques for AR forecasting, hopefully guiding the development of improved models for forecasting this phenomenon.

54 ENVIRONMENTAL SCIENCES↗

Can artificial intelligence and data-driven machine learning models match or even replace process-driven hydrologic models for streamflow simulation?: A case study of four watersheds with different hydro-climatic regions across the CONUS

With recent developments in computational techniques, Data-driven Machine Learning Models (DMLs) have shown great potential in simulating streamflow and capturing the rainfall-runoff relationship in given watersheds, which are traditionally fulfilled by Process-based Hydrologic Models (PHMs). There are debates on whether the DMLs can outperform and possibly replace the classical PHMs for streamflow simulation and river forecasting, but no clear conclusions have been made. This study aims to investigate whether the newer DMLs have any potential in further improving the simulation accuracy of classical PHMs, and vice versa. To do this, we compared a few popular PHMs and DMLs over four watersheds across the Continental US (CONUS) that are associated with different input, climate, and regional conditions. A total of five hydrologic models were chosen, including (1) two classical lumped models, i.e., the Sacramento Soil Moisture Accounting (SAC-SMA) and Xinanjiang (XAJ); (2) one modern distributed model, termed Coupled Routing and Excess Storage (CREST); (3) and two DMLs including an Artificial Neural Networks (ANN) and a deep learning model, termed Long Short Term Memory (LSTM). Our results demonstrated that the DMLs still significantly biased when using the baseline input scenario with the PHMs. However, the DMLs fed with delayed input scenarios had great potential and can reach high simulation accuracy. The DMLs, especially the ANN, outperformed other employed models under the rainfall-runoff relationship in which rainfall dominantly drives. Furthermore, the DMLs also showed better performance in the high-flow regime, while the PHMs had a better performance for the low-flow regime, implying both PHMs and DMLs have their own merits and are worthy of joint development. In general, our study indicated a great potential of using DMLs to simulate streamflow, but further studies are still needed to verify the transferability and scalability of DMLs in large-scale experiments, such as the Distributed Model Intercomparison Projects 1&2 conducted by National Weather Services but to compare modern DMLs and PHMs.

58 GEOSCIENCES↗

Lanai high-density irradiance sensor network for characterizing solar resource variability of MW-scale PV system.

Sandia National Laboratories (Sandia) and SunPower Corporation (SunPower) have completed design and deployment of an autonomous irradiance monitoring system based on wireless mesh communications and a battery operated data acquisition system. The Lanai High-Density Irradiance Sensor Network is comprised of 24 LI-COR{reg_sign} irradiance sensors (silicon pyranometers) polled by 19 RF Radios. The system was implemented with commercially available hardware and custom developed LabVIEW applications. The network of solar irradiance sensors was installed in January 2010 around the periphery and within the 1.2 MW ac La Ola PV plant on the island of Lanai, Hawaii. Data acquired at 1 second intervals is transmitted over wireless links to be time-stamped and recorded on SunPower data servers at the site for later analysis. The intent is to study power and solar resource data sets to correlate the movement of cloud shadows across the PV array and its effect on power output of the PV plant. The irradiance data sets recorded will be used to study the shape, size and velocity of cloud shadows. This data, along with time-correlated PV array output data, will support the development and validation of a PV performance model that can predict the short-term output characteristics (ramp rates) of PV systems of different sizes and designs. This analysis could also be used by the La Ola system operator to predict power ramp events and support the function of the future battery system. This experience could be used to validate short-term output forecasting methodologies.

14 SOLAR ENERGY↗

Security Enhancement of Network Constraint Grid-Edge Energy Management System

Network constrained grid edge energy management system (EMS) provides economic solution for active and reactive power dispatch of distributed energy resources (DERs) at the grid edge level. Grid edge EMS ensures secure interconnection of a circuit segment to the distribution system by maintaining grid code requirements (e.g. IEEE 1547–2018). Grid edge EMS is dependent on communication to receive load measurement, which brings a risk of unobservable false data injection attacks (FDIAs). To mitigate the risk, this paper proposes a framework to enhance resilient operation of grid edge EMS by detecting the unobservable FDIAs on loads and replacing them with forecasted values. In this work, a two-step detection algorithm is proposed. In first step, conventional residual based algorithm is deployed. Autoencoder (AE) based data driven mechanism is included in second step to detect the presence of unobservable FDIAs. After ensuring the presence of FDIA, its specific location is detected by checking the maximum residue values till the predefined threshold value is reached. Detected false data injected loads are then replaced with forecasted load values following long-short term memory (LSTM) based forecast to ensure resilient performance of grid edge EMS in the presence of attacks. This proposed security enhancement framework for grid edge EMS is evaluated in IEEE 13 bus system with three integrated DERs. Numerical simulation shows the validation of the proposed framework by reducing voltage violation in real operation of grid edge EMS.

cyber attack detection↗

Forecasting high-dimensional spatio-temporal systems from sparse measurements

This paper introduces a new neural network architecture designed to forecast high-dimensional spatio-temporal data using only sparse measurements. The architecture uses a two-stage end-to-end framework that combines neural ordinary differential equations (NODEs) with vision transformers. Initially, our approach models the underlying dynamics of complex systems within a low-dimensional space; and then it reconstructs the corresponding high-dimensional spatial fields. Many traditional methods involve decoding high-dimensional spatial fields before modeling the dynamics, while some other methods use an encoder to transition from high-dimensional observations to a latent space for dynamic modeling. In contrast, our approach directly uses sparse measurements to model the dynamics, bypassing the need for an encoder. This direct approach simplifies the modeling process, reduces computational complexity, and enhances the efficiency and scalability of the method for large datasets. We demonstrate the effectiveness of our framework through applications to various spatio-temporal systems, including fluid flows and global weather patterns. Although sparse measurements have limitations, our experiments reveal that they are sufficient to forecast system dynamics accurately over long time horizons. Our results also indicate that the performance of our proposed method remains robust across different sensor placement strategies, with further improvements as the number of sensors increases. This robustness underscores the flexibility of our architecture, particularly in real-world scenarios where sensor data is often sparse and unevenly distributed.

97 MATHEMATICS AND COMPUTING↗

A Machine‐Learning‐Assisted Stochastic Cloud Population Model as a Parameterization of Cumulus Convection

Abstract A machine‐learning‐assisted stochastic cloud population model is coupled with the Advanced Research Weather Research and Forecasting (WRF) model to represent fluctuations in the cloud‐base mass flux associated with the life cycles and interactions among cumulus convection cells. In this cloud population model, the size distribution and the associated cloud‐base mass flux of the convective cells are related to their previous state and to the change in the total convective area via a transition function. The convective area tendency in turn is assumed to depend on the cloud‐base mass flux that is resolved by the host WRF model. The transition function is represented by a single hidden‐layer neural network trained by the evolution of convective cell size distributions in a 1‐km grid‐spacing WRF simulation run over the Australian Monsoon region. At every grid point of the host model, the cloud population model predicts the cell size and cloud‐base mass flux distributions from which a random sample of cells is fed to an entraining parcel model that calculates precipitation as well as the associated liquid water potential temperature and total moisture tendencies. These tendencies are averaged over the cells and provided to the host model. Several regional simulations are performed over tropical and midlatitude domains to test this as a potential approach to scale‐aware parameterization. It is shown that such an approach could be a new promising path to simulating realistic precipitation statistics and propagation of precipitation associated with the Madden‐Julian Oscillation while maintaining realistic depictions of the diurnal cycle over both land and ocean.

54 ENVIRONMENTAL SCIENCES↗

Computational Needs for the Next Generation Electric Grid Proceedings

The April 2011 DOE workshop, 'Computational Needs for the Next Generation Electric Grid', was the culmination of a year-long process to bring together some of the Nation's leading researchers and experts to identify computational challenges associated with the operation and planning of the electric power system. The attached papers provide a journey into these experts' insights, highlighting a class of mathematical and computational problems relevant for potential power systems research. While each paper defines a specific problem area, there were several recurrent themes. First, the breadth and depth of power system data has expanded tremendously over the past decade. This provides the potential for new control approaches and operator tools that can enhance system efficiencies and improve reliability. However, the large volume of data poses its own challenges, and could benefit from application of advances in computer networking and architecture, as well as data base structures. Second, the computational complexity of the underlying system problems is growing. Transmitting electricity from clean, domestic energy resources in remote regions to urban consumers, for example, requires broader, regional planning over multi-decade time horizons. Yet, it may also mean operational focus on local solutions and shorter timescales, as reactive power and system dynamics (including fast switching and controls) play an increasingly critical role in achieving stability and ultimately reliability. The expected growth in reliance on variable renewable sources of electricity generation places an exclamation point on both of these observations, and highlights the need for new focus in areas such as stochastic optimization to accommodate the increased uncertainty that is occurring in both planning and operations. Application of research advances in algorithms (especially related to optimization techniques and uncertainty quantification) could accelerate power system software tool performance, i.e. speed to solution, and enhance applicability for new and existing real-time operation and control approaches, as well as large-scale planning analysis. Finally, models are becoming increasingly essential for improved decision-making across the electric system, from resource forecasting to adaptive real-time controls to online dynamics analysis. The importance of data is thus reinforced by their inescapable role in validating, high-fidelity models that lead to deeper system understanding. Traditional boundaries (reflecting geographic, institutional, and market differences) are becoming blurred, and thus, it is increasingly important to address these seams in model formulation and utilization to ensure accuracy in the results and achieve predictability necessary for reliable operations. Each paper also embodies the philosophy that our energy challenges require interdisciplinary solutions - drawing on the latest developments in fields such as mathematics, computation, economics, as well as power systems. In this vein, the workshop should be viewed not as the end product, but the beginning of what DOE seeks to establish as a vibrant, on-going dialogue among these various communities. Bridging communication gaps among these communities will yield opportunities for innovation and advancement. The papers and workshop discussion provide the opportunity to learn from experts on the current state-of-the-art on computational approaches for electric power systems, and where one may focus to accelerate progress. It has been extremely valuable to me as I better understand this space, and consider future programmatic activities. I am confident that you too will enjoy the discussion, and certainly learn from the many experts. I would like to thank the authors of the papers for sharing their perspectives, as well as the paper discussants, session recorders, and participants. The meeting would not have been as successful without your commitment and engagement. I also would like to thank Joe Eto and Bob Thomas for their vision and leadership in bringing together such a well-structured and productive forum.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Long-term Prediction of Spent Fuel Storage Canisters - Issues and Approaches - 18112

The considerations involved in aging management of stainless steel canisters used in dry cask storage of spent fuel are discussed in this paper in a probabilistic framework. The two important modes of degradation in the long-term performance of the casks are localized corrosion and stress corrosion cracking initiating from the external surface. These degradation modes are determined by the evolution of the electrochemical behavior and the strain rate of the stainless steel. Laboratory data suggest that Type 304L/316L stainless steels should suffer from localized corrosion and SCC under salt deposits. However, field inspection of the canisters has not indicated extensive pitting or stress corrosion cracking. This discrepancy can be understood in terms of the joint probabilities of various factors that must come together to cause these failure modes. A Bayesian network approach is suggested as the framework for modeling the probability of failure of canisters and long-term management of aging. The gaps in knowledge to conduct a Bayesian network analysis are discussed. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

OrbNet Denali: A machine learning potential for biological and organic chemistry with semi-empirical cost and DFT accuracy

We present OrbNet Denali, a machine learning model for an electronic structure that is designed as a drop-in replacement for groundstate density functional theory (DFT) energy calculations. The model is a message-passing graph neural network that uses symmetryadapted atomic orbital features from a low-cost quantum calculation to predict the energy of a molecule. OrbNet Denali is trained on a vast dataset of 2.3 × 10 6 DFT calculations on molecules and geometries. This dataset covers the most common elements in biochemistry and organic chemistry (H, Li, B, C, N, O, F, Na, Mg, Si, P, S, Cl, K, Ca, Br, and I) and charged molecules. OrbNet Denali is demonstrated on several well-established benchmark datasets, and we find that it provides accuracy that is on par with modern DFT methods while offering a speedup of up to three orders of magnitude. For the GMTKN55 benchmark set, OrbNet Denali achieves WTMAD-1 and WTMAD-2 scores of 7.19 and 9.84, on par with modern DFT functionals. For several GMTKN55 subsets, which contain chemical problems that are not present in the training set, OrbNet Denali produces a mean absolute error comparable to those of DFT methods. For the Hutchison conformer benchmark set, OrbNet Denali has a median correlation coefficient of R 2 = 0.90 compared to the reference DLPNO-CCSD(T) calculation and R 2 = 0.97 compared to the method used to generate the training data (ωB97X-D3/def2- TZVP), exceeding the performance of any other method with a similar cost. Similarly, the model reaches chemical accuracy for noncovalent interactions in the S66x10 dataset. For torsional profiles, OrbNet Denali reproduces the torsion profiles of ωB97X-D3/def2-TZVP with an average mean absolute error of 0.12 kcal/mol for the potential energy surfaces of the diverse fragments in the TorsionNet500 dataset.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

The EGS Collab Project: Status and Accomplishments

The EGS Collab project, supported by the US Department of Energy, is addressing challenges in implementing enhanced geothermal systems (EGS). This includes improving understanding of the stimulation of crystalline rock to create appropriate flow pathways, and the ability to effectively simulate both the stimulation and the flow and transport processes in the resulting fracture network. The project is performing intensively monitored rock stimulation and flow tests at the 10-m scale in an underground research laboratory. Data and observations from the field test are compared to simulations to understand processes and to build confidence in numerical modeling of the processes. In Experiment 1, we examined hydraulic fracturing an underground test bed at the Sanford Underground Research Facility (SURF) in Lead, South Dakota, at a depth of approximately 1.5 km. We drilled eight sub-horizontal boreholes in a well-characterized phyllite. Six of the boreholes were instrumented with many sensor types to allow careful monitoring of stimulation events and flow tests, and the other two boreholes were used for water injection and production. We performed a number of stimulations and flow tests in the testbed. Our monitoring systems allowed detailed observations and collection of numerous data sets of processes occurring during stimulation and during dynamic flow tests. Long-term ambient temperature and chilled water flow tests were performed in addition to many tracer tests to examine system behavior. Data were rapidly analyzed, allowing adaptive control of the tests. Numerical simulation was used to answer key experimental design questions, to forecast fracture propagation trajectories and extents, and to analyze and evaluate results. Many simulations were performed in near-real-time in conjunction with the field experiments, with more detailed process study simulations performed on a longer timeframe. Experiment 2 will examine hydraulic shearing in a test bed being built at the SURF at a depth of about 1.25 km in amphibolite under a different set of stress and fracture conditions than Experiment 1. Five sets of fracture orientations were considered in design, and three orientations seem to be consistently observed.

EGS Collab↗

Observation of CH4 and other Non-CO2 Green House Gas Emissions from California

In 2006, California passed the landmark assembly bill AB-32 to reduce California's emissions of greenhouse gases (GHGs) that contribute to global climate change. AB-32 commits California to reduce total GHG emissions to 1990 levels by 2020, a reduction of 25 percent from current levels. To verify that GHG emission reductions are actually taking place, it will be necessary to measure emissions. We describe atmospheric inverse model estimates of GHG emissions obtained from the California Greenhouse Gas Emissions Measurement (CALGEM) project. In collaboration with NOAA, we are measuring the dominant long-lived GHGs at two tall-towers in central California. Here, we present estimates of CH{sub 4} emissions obtained by statistical comparison of measured and predicted atmospheric mixing ratios. The predicted mixing ratios are calculated using spatially resolved a priori CH{sub 4} emissions and surface footprints, that provide a proportional relationship between the surface emissions and the mixing ratio signal at tower locations. The footprints are computed using the Weather Research and Forecast (WRF) coupled to the Stochastic Time-Inverted Lagrangian Transport (STILT) model. Integral to the inverse estimates, we perform a quantitative analysis of errors in atmospheric transport and other factors to provide quantitative uncertainties in estimated emissions. Regressions of modeled and measured mixing ratios suggest that total CH{sub 4} emissions are within 25% of the inventory estimates. A Bayesian source sector analysis obtains posterior scaling factors for CH{sub 4} emissions, indicating that emissions from several of the sources (e.g., landfills, natural gas use, petroleum production, crops, and wetlands) are roughly consistent with inventory estimates, but livestock emissions are significantly higher than the inventory. A Bayesian 'region' analysis is used to identify spatial variations in CH{sub 4} emissions from 13 sub-regions within California. Although, only regions near the tower are significantly constrained by the tower measurements, CH{sub 4} emissions from the south Central Valley appear to be underestimated in a manner consistent with the under-prediction of livestock emissions. Finally, we describe a pseudo-experiment using predicted CH{sub 4} signals to explore the uncertainty reductions that might be obtained if additional measurements were made by a future network of tall-tower stations spread over California. These results show that it should be possible to provide high-accuracy estimates of surface CH{sub 4} emissions for multiple regions as a means to verify future emissions reductions.

54 ENVIRONMENTAL SCIENCES↗

A low-complexity non-intrusive approach to predict the energy demand of buildings over short-term horizons

Reliable, non-intrusive, short-term (of up to 12 hours ahead) prediction of a building's energy demand is a critical component of intelligent energy management applications. A number of such approaches have been proposed over time, utilizing various statistical and, more recently, machine learning techniques, such as decision trees, neural networks and support vector machines. Importantly, all of these works barely outperform simple seasonal auto-regressive integrated moving average models, while their complexity is significantly higher. Here, we propose a novel low-complexity non-intrusive approach that improves the predictive accuracy of the state-of-the-art by up to ~10%. The backbone of our approach is a K-nearest neighbours search method, that exploits the demand pattern of the most similar historical days, and incorporates appropriate time-series pre-processing and easing. In the context of this work, we evaluate our approach against state-of-the-art methods and provide insights on their performance.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Neural-based time series forecasting of loss of coolant accidents in nuclear power plants

During the last few years, deep learning in neural networks has demonstrated impressive successes in the areas of computer vision, speech and image recognition, text generation, and many others. However, sensitive engineering areas such as nuclear engineering benefited less from these efficient techniques. In this work, deep learning expert systems are utilized to model and predict time series progression of a design-basis nuclear accident, featuring a loss of coolant accident. Two major findings are accomplished in this work. First, the ability to train expert systems with high accuracy, which could help nuclear power plant operators to figure out plant responses during the accident. Second, building fast, efficient, and accurate deep models to simulate nuclear phenomena, which could be valuable to nuclear computational science. In this work, large amount of time series data is obtained from simulation tools by simulating different conditions of the base-case/nominal accident scenario. Four critical outputs/responses are monitored during the accident (e.g. temperature, pressure, break flow rate, water level). Two approaches are adopted in this work. The first approach is to use feedforward deep neural networks (DNN) to fit all time steps and outputs in a single model. The second approach is to use long short-term memory (LSTM) to fit all time steps together for each reactor response separately. Both DNN and LSTM demonstrate very good performance in predicting the test and base-case scenarios, with accuracy as low as 92% and as high as 99%, where these test scenarios are unknown to the expert systems and are not included in the model training. In addition, both approaches demonstrate a significant reduction in computational costs, as the deep expert system is able to accurately predict the accident 100,000 times faster than the original simulation tool. Given sufficient data, the methodology adopted in this study demonstrates that DNN/LSTM expert systems can be used as a decision support system to model advanced time series phenomena within nuclear power plants with high accuracy and negligible computational costs.

42 ENGINEERING↗

Karhunen–Loève deep learning method for surrogate modeling and approximate Bayesian parameter estimation

We evaluate the performance of the Karhunen-Loève Deep Neural Network (KL-DNN) framework for surrogate modeling and approximate Bayesian parameter estimation in partial differential equation models. In the surrogate model, the Karhunen-Loève (KL) expansions are used for the dimensionality reduction of the number of unknown parameters and variables, and a deep neural network is employed to relate the reduced space of parameters to that of the state variables. The KL-DNN surrogate model is used to formulate a maximum-a-posteriori-like least-squares problem, which is randomized to draw samples of the posterior distribution of the parameters. We test the proposed framework for a hypothetical unconfined aquifer via comparison with the forward MODFLOW and inverse PEST++ iterative ensemble smoother (IES) solutions as well as the state-of-the-art Fourier neural operator (FNO) and deep operator networks (DeepONets) operator learning surrogate models. Our results show that the KL-DNN surrogate model outperforms FNO and DeepONet for forward predictions. For solving inverse problems, the randomized algorithm provides the same or more accurate Bayesian predictions of the parameters than IES as evidenced by the higher log-predictive probability of both the estimated parameter field and the forecast hydraulic head. The posterior mean obtained from the randomized algorithm is closer to the reference parameter field than that obtained with FNO as the maximum a posteriori estimate.

Approximate Bayesian inference↗