Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Learning Algorithm”

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 271 records · Page 15
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Integrating Predictions for Improving Defect Classification Accuracy in NDT-based Assessment of Concrete - 20229

There is an increasing need to create predictive models for defect classification in concrete using the output of non-destructive testing (NDT) techniques. Recent advancement of machine-learning algorithms has offered several techniques for developing classification models for different types of data sets. However, the performance of these algorithms is very uncertain, mainly when applied to small and noisy datasets. For example, when human access is limited (e.g., nuclear facility), robot-based NDT is preferred. But compared to manual tests with humans present on site, the data sets are small, and more noise can exist. Therefore, it is imperative to develop new approaches to ensure a consistently high classification accuracy for inadequate data sets. This study explores the classification performance on NDT dataset using classifiers from different machine-learning algorithms, namely k-Nearest Neighbor (kNN), Decision Tree, Naive Bayes, Logistic Regression, and Support Vector Machine (SVM). The authors further integrated the predictions from these classifiers using proposed methods. The integration strategy combines the output of the classifiers based on two different measures, accuracy, and performance (ACC and PERF), using equations such as sum, average, and square-root-of-sums-of-squares (SRSS). Our results reveal varying classification accuracies across individual classifiers with different misclassifications across the test data set. The integration strategy provided significant improvement in the classification accuracy compared to the individual classifiers. Furthermore, the results indicate minimal variation across the integration methods as compared to the variation across the individual classifiers. To conclude, prediction integration offers a unique approach for combining the output of multiple classifiers to create redundancies with the potential of achieving high classification performance and improved reliability in predictive models for defect detection in concrete. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2014

SU-E-T-48: Automated Quality Assurance for XML Controlled Linacs

Purpose: To automate routine imaging QA procedures so that complying with TG 142 and TG 179 can be efficient and reliable. Methods: Two QA tests for a True Beam Linac were automatized. A Winston Lutz test as described by Lutz et al{sup 1} using the Winston Lutz test kit from BrainLab, Germany and a CBCT Image Quality test as described in TG 179 using the EMMA phantom, Siemens Medical Physics, Germany were performed in our True Beam. For each QA procedure tested, a 3 step paradigm was used. First, the data was automatically acquired using True Beam Developer Mode and XML scripting. Second, the data acquired in the first step was automatically processed using in-home grown Matlab GUIs. Third, Machine Learning algorithms were used to automatically classify the processed data and reports generated. Results: The Winston Luzt test could be performed by an experienced medical physicist in 29.0 ± 8.0 min. The same test, if automated using our paradigm, could be performed in 3.0 ± 0.1 min. In the same lieu, time could be substantially saved for image quality tests. In this case, the amount of time saved will depend on the phantoms used and the initial localization method. Additionally, machine learning algorithms could automatically identify the roots of the problems if any and possibly help reduce machine down time. Conclusion: Modern linear accelerators are equipped with advanced 2D and 3D imaging that are used for patient alignment substantially improving IGRT protocols. However, this extra complexity exponentially increases the number of QA tests needed. Using the new paradigm described above, not only bare minimum but best practice QA programs could be implemented with the same manpower. This work is supported by Varian, Palo Alto, CA.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2017

Bayesian inverse modeling of the atmospheric transport and emissions of a controlled tracer release from a nuclear power plant

Probability distribution functions (PDFs) of model inputs that affect the transport and dispersion of a trace gas released from a coastal California nuclear power plant are quantified using ensemble simulations, machine-learning algorithms, and Bayesian inversion. The PDFs are constrained by observations of tracer concentrations and account for uncertainty in meteorology, transport, diffusion, and emissions. Meteorological uncertainty is calculated using an ensemble of simulations of the Weather Research and Forecasting (WRF) model that samples five categories of model inputs (initialization time, boundary layer physics, land surface model, nudging options, and reanalysis data). The WRF output is used to drive tens of thousands of FLEXPART dispersion simulations that sample a uniform distribution of six emissions inputs. Machine-learning algorithms are trained on the ensemble data and used to quantify the sources of ensemble variability and to infer, via inverse modeling, the values of the 11 model inputs most consistent with tracer measurements. We find a substantial ensemble spread in tracer concentrations (factors of 10 to 10 3 ), most of which is due to changing emissions inputs (about 80%), though the cumulative effects of meteorological variations are not negligible. The performance of the inverse method is verified using synthetic observations generated from arbitrarily selected simulations. When applied to measurements from a controlled tracer release experiment, the inverse method satisfactorily determines the location, start time, duration and amount. In a 2 km × 2 km area of possible locations, the actual location is determined to within 200m. The start time is determined to within 5min out of 2h, and the duration to within 50min out of 4h. Over a range of release amounts of 10 to 1000kg, the estimated amount exceeds the actual amount of 146kg by only 32kg. The inversion also estimates probabilities of different WRF configurations. In conclusion, to best match the tracer observations, the highest-probability cases in WRF are associated with using a late initialization time and specific reanalysis data products.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2019

Bandgap prediction by deep learning in configurationally hybridized graphene and boron nitride

Abstract It is well-known that the atomic-scale and nano-scale configuration of dopants can play a crucial role in determining the electronic properties of materials. However, predicting such effects is challenging due to the large range of atomic configurations that are possible. Here, we present a case study of how deep learning algorithms can enable bandgap prediction in hybridized boron–nitrogen graphene with arbitrary supercell configurations. A material descriptor that enables correlation of structure and bandgap was developed for convolutional neural networks. Bandgaps calculated by ab initio calculations, and corresponding structures, were used as training datasets. The trained networks were then used to predict bandgaps of systems with various configurations. For 4 × 4 and 5 × 5 supercells they accurately predict bandgaps, with a R 2 of >90% and root-mean-square error of ~0.1 eV. The transfer learning was performed by leveraging data generated from small supercells to improve the prediction accuracy for 6 × 6 supercells. This work will pave a route to future investigation of configurationally hybridized graphene and other 2D materials. Moreover, given the ubiquitous existence of configurations in materials, this work may stimulate interest in applying deep learning algorithms for the configurational design of materials across different length scales.

Dong, Yuan↗
DOE OSTI2022

A Wrapper to Use a Machine-Learning-Based Algorithm for Earthquake Monitoring

Seismology is one of the main sciences used to monitor volcanic activity worldwide. Fast, efficient, and accurate seismicity detectors are crucial to assess the activity level of a volcano in near–real time and to issue timely warnings. Traditional real–time seismic processing software uses phase onset pickers followed by a phase association algorithm to declare an event and estimate its location. The pickers typically do not identify whether the detected phase is a P or S arrival, which can have a negative impact on hypocentral location quality and complicates phase association. We implemented the deep–neural–network–based method PhaseNet to identify in real time P and S seismic waves on data from one– and three–component seismometers. We tuned the Earthworm binder_ew associator module to use the phase identification from PhaseNet to detect and locate the events, which we archive in a SeisComP3 database. We assessed the performance of the algorithm by comparing the results with existing catalogs built to monitor seismic and volcanic activity in Mayotte and the Lesser Antilles region. Our algorithm, which we refer to as PhaseWorm, showed promising results in both contexts and clearly outperformed the previous automatic method implemented in Mayotte. As a result, this innovative real–time processing system is now operational for seismicity monitoring in Mayotte and Martinique.

58 GEOSCIENCES↗
DOE OSTI2023

Deep learning for morphological identification of extended radio galaxies using weak labels

Abstract The present work discusses the use of a weakly-supervised deep learning algorithm that reduces the cost of labelling pixel-level masks for complex radio galaxies with multiple components. The algorithm is trained on weak class-level labels of radio galaxies to get class activation maps (CAMs). The CAMs are further refined using an inter-pixel relations network (IRNet) to get instance segmentation masks over radio galaxies and the positions of their infrared hosts. We use data from the Australian Square Kilometre Array Pathfinder (ASKAP) telescope, specifically the Evolutionary Map of the Universe (EMU) Pilot Survey, which covered a sky area of 270 square degrees with an RMS sensitivity of 25–35 $\mu$ Jy beam $^{-1}$ . We demonstrate that weakly-supervised deep learning algorithms can achieve high accuracy in predicting pixel-level information, including masks for the extended radio emission encapsulating all galaxy components and the positions of the infrared host galaxies. We evaluate the performance of our method using mean Average Precision (mAP) across multiple classes at a standard intersection over union (IoU) threshold of 0.5. We show that the model achieves a mAP $_{50}$ of 67.5% and 76.8% for radio masks and infrared host positions, respectively. The network architecture can be found at the following link: https://github.com/Nikhel1/Gal-CAM

Astronomy & Astrophysics↗
DOE OSTI2024

A comparative study on deep learning models for condition monitoring of advanced reactor piping systems

Advanced nuclear reactors offer innovative applications due to their portability, reliability, resiliency, and high capacity factors. To operate them on a wider scale, reducing maintenance life-cycle costs while ensuring their integrity is essential. Autonomous operations in advanced nuclear reactors using augmented Digital Twin (DT) technology can serve as a cost-effective solution by increasing awareness about the system’s health. A key component of nuclear DT frameworks is the condition monitoring of safety systems, such as piping-equipment systems, which involves acquiring and monitoring the plant’s sensor data. Here, this research proposes a condition monitoring methodology utilizing deep learning algorithms, such as multilayer perceptions (MLP) and convolutional neural networks (CNNs), to detect degradation and its severity in nuclear piping-equipment systems. Sensor signals are processed to obtain the power spectral density and the Short-Time Fourier transform, and feature extraction methodologies are proposed to develop degradation-sensitive data repositories. The performance of MLP, one-dimensional (1D) CNN, and 2D CNN within the proposed condition monitoring framework is compared using a finite element model of a 3D piping system subjected to seismic loads as the application case study. Various approaches, such as dropout, k-Fold validation, regularization, and early stopping of training the network, are investigated to avoid overfitting the models to the input sensor data. The predictive capability and computational capacity of the deep learning algorithms are also compared to detect degradation in the Z-pipe system of the Experimental Breeder Reactor II (EBRII). The Z-pipe system is subjected to harmonic excitations that represent normal operating loads, such as pump-induced vibrations. The findings of the study indicate that the proposed artificial intelligence (AI)-driven condition monitoring framework demonstrates superior prediction accuracies with a 2D CNN, whereas the MLP exhibits higher computational efficiency.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2020

Machine Learning-Based PV Reserve Determination Strategy for Frequency Control on the WECC System

Frequency control from photovoltaic (PV) power plants has great potential to address the frequency response challenge of the power system with high penetrations of renewable generation. Using model-based approaches to determine the optimal PV headroom reserve, however, requires significant online computation and is intractable for an interconnection level system. This paper proposes a machine learning based strategy, that is suitable for real-time operation, to determine the optimal PV reserve for frequency control. The proposed machine learning algorithm is trained and tested on 1,987 offline simulations of a 60% renewable penetration Western Electricity Coordinating Council (WECC) system. Furthermore, the proposed reserve determination strategy is applied on a realistic 1-day operation profile of the WECC system and demonstrates a savings of more than 40% PV headroom compared to a conservative approach. It is evident that the proposed strategy can efficiently and effectively determine the optimal PV frequency control reserve for realistic interconnection systems.

frequency control↗
DOE OSTI2017

A Hybrid dasymetric and machine learning approach to high-resolution residential electricity consumption modeling

As urban areas continue to grow and evolve in a world of increasingenvironmental awareness, the need for detailed information regarding residentialenergy consumption patterns has become increasingly important. Though currentmodeling efforts mark significant progress in the effort to better understand thespatial distribution of energy consumption, the majority of techniques are highlydependent on region-specific data sources and often require building- ordwelling-level details that are not publicly available for many regions in the UnitedStates. Furthermore, many existing methods do not account for errors in input datasources and may not accurately reflect inherent uncertainties in model outputs. Wepropose an alternative and more general hybrid approach to high-resolution residentialelectricity consumption modeling by merging a dasymetric model with acomplementary machine learning algorithm. The method’s flexible data requirementand statistical framework ensure that the model both is applicable to a widerange of regions and considers errors in input data sources.

cities↗
DOE OSTI2026

Code for the manuscript "Lagrangian Attention Tensor Networks for Velocity Gradient Statistical Mode

We disclose a python/pytorch implementation of the physics-informed machine learning algorithm described in "Lagrangian Attention Tensor Networks for Velocity Gradient Statistical Modeling", LA-UR-24-30678. Direct numerical simulation (DNS) of ubiquitous turbulence phenomena is computationally infeasible for realistic flows. As a result, reduced modeling for turbulent flows aim to reduce the number of resolved scales while retaining accurate representations of the small-scale physics. The dynamics of the velocity gradient tensor (VGT) is a key ingredient in reduced or subgrid turbulence models. The evolution equation for the VGT involves nonlocal terms, requiring closure modeling. This implementation of the novel methodology of Lagrangian Attention Tensor Networks (LATN), utilizes a structured representation of the history of the VGT to inform a physics-informed machine learning algorithm. This addition of structured memory terms is shown to outperform previous models when trained and evaluated on DNS data.

Livescu, Daniel [LANL]↗
DOE OSTI2021

Interpreting Write Performance of Supercomputer I/O Systems with Regression Models

This work seeks to advance the state of the art in HPC I/O performance analysis and interpretation. In particular, we demonstrate effective techniques to: (1) model output performance in the presence of I/O interference from production loads; (2) build features from write patterns and key parameters of the system architecture and configurations; (3) employ suitable machine learning algorithms to improve model accuracy. We train models with five popular regression algorithms and conduct experiments on two distinct production HPC platforms. We find that the lasso and random forest models predict output performance with high accuracy on both of the target systems. We also explore use of the models to guide adaptation in I/O middleware systems, and show potential for improvements of at least 15% from model-guided adaptation on 70% of samples, and improvements up to 10× on some samples for both of the target systems.

Xie, Bing↗
DOE OSTI2018

An open experimental database for exploring inorganic materials

The use of advanced machine learning algorithms in experimental materials science is limited by the lack of sufficiently large and diverse datasets amenable to data mining. If publicly open, such data resources would also enable materials research by scientists without access to expensive experimental equipment. Here, we report on our progress towards a publicly open High Throughput Experimental Materials (HTEM) Database (htem.nrel.gov). This database currently contains 140,000 sample entries, characterized by structural (100,000), synthetic (80,000), chemical (70,000), and optoelectronic (50,000) properties of inorganic thin film materials, grouped in >4,000 sample entries across >100 materials systems; more than a half of these data are publicly available. This article shows how the HTEM database may enable scientists to explore materials by browsing web-based user interface and an application programming interface. This paper also describes a HTE approach to generating materials data, and discusses the laboratory information management system (LIMS), that underpin HTEM database. Finally, this manuscript illustrates how advanced machine learning algorithms can be adopted to materials science problems using this open data resource.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Automatic Classification of Biological Targets in a Tidal Channel Using a Multibeam Sonar

Multibeam sonars are widely used for environmental monitoring of fauna at marine renewable energy sites. However, they can rapidly accrue vast volumes of data, which poses a challenge for data processing. Here, using data from a deployment in a tidal channel with peak currents of 1–2 m s –1 , we demonstrate the data-reduction benefits of real-time automatic classification of targets detected and tracked in multibeam sonar data. First, we evaluate classification capabilities for three machine learning algorithms: random forests, support vector machines, and k-nearest neighbors. For each algorithm, a hill-climbing search optimizes a set of hand-engineered attributes that describe tracked targets. Here, the random forest algorithm is found to be most effective—in postprocessing, discriminating between biological and nonbiological targets with a recall rate of 0.97 and a precision of 0.60. In addition, 89% of biological targets are correctly classified as either seals, diving birds, fish schools, or small targets. Model dependence on the volume of training data is evaluated. Second, a real-time implementation of the model is shown to distinguish between biological targets and nonbiological targets with nearly the same performance as in postprocessing. From this, we make general recommendations for implementing real-time classification of biological targets in multibeam sonar data and the transferability of trained models.

16 TIDAL AND WAVE POWER↗
DOE OSTI2018

Retrofitting Word Embeddings with the UMLS Metathesaurus for Clinical Information Extraction

Deep learning has surged in popularity and proven to be effective for various artificial intelligence applications including information extraction from cancer pathology reports. Since word representation is a core unit that enables deep learning algorithms to understand words and be able to perform NLP, this representation must include as much information as possible to help these algorithms achieve high classification performance. Therefore, in this work in addition to the distributional information of words in large sized corpora, we use UMLS vocabulary resources to enrich the vector space representation of words with the semantic relations between words. These resources provide many terminologies pertaining to cancer. The refined word embeddings are used with a convolutional neural (CNN) model to extract four data elements from cancer pathology reports; ICD-O-3 tumor topography codes, tumor laterality, behavior, and histological grade. We observed that using UMLS vocabulary resources to enrich word embeddings of CNN models consistently outperformed CNN models without pre-training word embeddings and even with pre-trained word embeddings on a domain specific corpus across all four tasks. The results show marginal improvement on the laterality task, but a significant improvement on the other tasks, especially for the macro-f score. Specifically, the improvements are 3%, 13%, and 15% for tumor site, histological grade, and behavior tasks, respectively. This approach is encouraging to enrich word embeddings with more clinical data resources to be used for information abstraction tasks from clinical pathology reports.

Alawad, Mohammed↗
DOE OSTI2017

Emulating Simulations of Cosmic Dawn for 21 cm Power Spectrum Constraints on Cosmology, Reionization, and X-Ray Heating

Current and upcoming radio interferometric experiments are aiming to make a statistical characterization of the high-redshift 21 cm fluctuation signal spanning the hydrogen reionization and X-ray heating epochs of the universe. However, connecting 21 cm statistics to the underlying physical parameters is complicated by the theoretical challenge of modeling the relevant physics at computational speeds quick enough to enable exploration of the high-dimensional and weakly constrained parameter space. In this work, we use machine learning algorithms to build a fast emulator that can accurately mimic an expensive simulation of the 21 cm signal across a wide parameter space. We embed our emulator within a Markov Chain Monte Carlo framework in order to perform Bayesian parameter constraints over a large number of model parameters, including those that govern the Epoch of Reionization, the Epoch of X-ray Heating, and cosmology. As a worked example, we use our emulator to present an updated parameter constraint forecast for the Hydrogen Epoch of Reionization Array experiment, showing that its characterization of a fiducial 21 cm power spectrum will considerably narrow the allowed parameter space of reionization and heating parameters, and could help strengthen Planck's constraints on σ{sub 8}. We provide both our generalized emulator code and its implementation specifically for 21 cm parameter constraints as publicly available software.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Machine Learning for Improving Surface-Layer-Flux Estimates

Abstract Flows in the atmospheric boundary layer are turbulent, characterized by a large Reynolds number, the existence of a roughness sublayer and the absence of a well-defined viscous layer. Exchanges with the surface are therefore dominated by turbulent fluxes. In numerical models for atmospheric flows, turbulent fluxes must be specified at the surface; however, surface fluxes are not known a priori and therefore must be parametrized. Atmospheric flow models, including global circulation, limited area models, and large-eddy simulation, employ Monin–Obukhov similarity theory (MOST) to parametrize surface fluxes. The MOST approach is a semi-empirical formulation that accounts for atmospheric stability effects through universal stability functions. The stability functions are determined based on limited observations using simple regression as a function of the non-dimensional stability parameter representing a ratio of distance from the surface and the Obukhov length scale (Obukhov in Trudy Inst Theor Geofiz AN SSSR 1:95–115, 1946), $$z/L$$ z / L . However, simple regression cannot capture the relationship between governing parameters and surface-layer structure under the wide range of conditions to which MOST is commonly applied. We therefore develop, train, and test two machine-learning models, an artificial neural network (ANN) and random forest (RF), to estimate surface fluxes of momentum, sensible heat, and moisture based on surface and near-surface observations. To train and test these machine-learning algorithms, we use several years of observations from the Cabauw mast in the Netherlands and from the National Oceanic and Atmospheric Administration’s Field Research Division tower in Idaho. The RF and ANN models outperform MOST. Even when we train the RF and ANN on one set of data and apply them to the second set, they provide more accurate estimates of all of the fluxes compared to MOST. Estimates of sensible heat and moisture fluxes are significantly improved, and model interpretability techniques highlight the logical physical relationships we expect in surface-layer processes.

Meteorology & Atmospheric Sciences↗
DOE OSTI2016

Supply Chain-Based Solution to Prevent Fuel Tax Evasion: Final Report

The primary source of funding for the United States transportation system is derived from motor fuel and other highway use taxes. Loss of revenue attributed to fuel-tax evasion has been assessed to be somewhere between 1 billion and 3 billion per year. Any solution that addresses this problem needs to include not only the tax-collection agencies and auditors, but also the carriers transporting oil products and the carriers customers. This report presents a system developed by the Oak Ridge National Laboratory (ORNL) for the Federal Highway Administration which has the potential to reduce or eliminate many fuel-tax evasion schemes. The solution balances the needs of tax-auditors and those of the fuel-hauling companies and their customers. The system has three main components. The on-board subsystem combined sensors, tracking and communication devices, and software (the on-board Evidential Reasoning System, or obERS) to detect, monitor, and geo-locate the transfer of fuel among different locations. The back office sub-system (boERS) used self-learning algorithms to determine the legitimacy of the fuel loading and offloading (important for tax auditors) and detect potential illicit operations such as fuel theft (important for carriers and their customers, and may justify the deployment costs). The third sub-system, the Fuel Distribution and Auditing System or FDAS, is a centralized database, which together with a user interface allows tax auditors to query the data submitted by the fuel-hauling companies and correlate different parameters to quickly identify any anomalies. Industry partners included Barger Transport of Weber City, Virginia (fleet); Air-Weigh, of Eugene, Oregon (and their wires and harnesses); Liquid Bulk Tank (LBT) of Omaha, Nebraska (three five-compartment trailers); and Innovative Software Engineering (ISE) of Coralville, Iowa(on-board telematics device and back-office system). ORNL conducted a pilot test with the three instrumented vehicles collecting real-world data during an eight-month period (October 2014 to June 2015). The vehicles logged a total of 375,000 miles and transported more than 7.5 million gallons of fuel. The drivers entered information on the telematics device about 77% of the time, with about 15% (10%) of the data showing at least one compartment with loading (offloading) valve activity, but not driver information entry for that compartment. Seven events (or about 0.7% of the trips) were identified by the boERS as having a missing fuel diversion number. The solution developed and tested in the pilot test had federal- and state-level tax auditors as its main audience. Nonetheless, the technology has to be adopted by the fuel carriers and therefore the solution had to address the needs of fuel-hauling companies and their customers (i.e., fuel theft and cocktailing). Sensors in the hatches allowed the obERS recognize when one of those were opened (always a suspicious activity, unless it happens at locations where maintenance is performed on the vehicle). The fuel-theft issue was addressed by a self-learning algorithm deployed on the boERS that continuously processed the data from the field to construct probability distributions of measures such as elapsed time of fuel loading and offloading by driver, vehicle, and compartment, valve actuation sequence, elapsed time between the first two valve actuations (by driver, compartment, and location), and other parameters. Probability thresholds, which can be set up by the carrier, determined how to classify the observed events. The boERS also kept track of valve sequencing at a given location and analyzed these actuations to help identify any suspicious activities. Technical and economic recommendations include: a) simplification of the driver data-entry task; b) reduction of system deployment cost; c) addition of capabilities that make the system more appealing to industry (i.e., identify and avoid fuel misdelivery); and d) incorporation of additional capabilities to the FDAS. Path to commercialization recommendations include: a) identify a company willing to further develop, test, and certify a hardened system with a price point the market will bear; b).transfer ERS to licensee; c) establish a fleet or fleets who want this technology for carrier benefits or trailer manufacturer who wants to offer it as optional technology; and d) identify a location for the deployment of FDAS.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2021

Model-Free Voltage Control of Active Distribution System with PVs Using Surrogate Model-Based Deep Reinforcement Learning

Accurate knowledge of the distribution system topology and parameters is required to achieve good voltage control performance, but this is difficult to obtain in practice. This paper proposes a physical-model-free voltage control method based on a surrogate-model-enabled deep reinforcement learning approach. Specifically, a surrogate model is trained in a supervised manner using the recorded limited number of historical data to learn the relationship between the power injections and voltage fluctuations of each node. Then, the deep reinforcement learning algorithm is applied to learn an optimal control strategy from the experiences obtained by continuous interactions with the surrogate model. The proposed method can achieve physical-model-free control of unbalanced distribution network and inform real-time decisions to deal with fast voltage fluctuations caused by the rapid variation of PV generation. Simulation results on an unbalance IEEE 123-bus system show that the proposed method can achieve similar performance as that of perfect physical-model-based approaches while being advantageous over other traditional methods.

active distribution network↗