skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Energy prediction using spatiotemporal pattern networks

Abstract

This paper presents a novel data-driven technique based on the spatiotemporal pattern network (STPN) for energy/power prediction for complex dynamical systems. Built on symbolic dynamical filtering, the STPN framework is used to capture not only the individual system characteristics but also the pair-wise causal dependencies among different sub-systems. To quantify causal dependencies, a mutual information based metric is presented and an energy prediction approach is subsequently proposed based on the STPN framework. To validate the proposed scheme, two case studies are presented, one involving wind turbine power prediction (supply side energy) using the Western Wind Integration data set generated by the National Renewable Energy Laboratory (NREL) for identifying spatiotemporal characteristics, and the other, residential electric energy disaggregation (demand side energy) using the Building America 2010 data set from NREL for exploring temporal features. In the energy disaggregation context, convex programming techniques beyond the STPN framework are developed and applied to achieve improved disaggregation performance.

Authors:
; ; ; ; ORCiD logo
Publication Date:
Research Org.:
National Renewable Energy Lab. (NREL), Golden, CO (United States)
Sponsoring Org.:
National Science Foundation (NSF)
OSTI Identifier:
1399848
Report Number(s):
NREL/JA-5500-70286
Journal ID: ISSN 0306-2619
DOE Contract Number:
AC36-08GO28308
Resource Type:
Journal Article
Resource Relation:
Journal Name: Applied Energy; Journal Volume: 206; Journal Issue: C
Country of Publication:
United States
Language:
English
Subject:
32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATION; 42 ENGINEERING; wind power; symbolic dynamical filtering; spatiotemporal pattern network; probabilistic finite state automata; NILM

Citation Formats

Jiang, Zhanhong, Liu, Chao, Akintayo, Adedotun, Henze, Gregor P., and Sarkar, Soumik. Energy prediction using spatiotemporal pattern networks. United States: N. p., 2017. Web. doi:10.1016/j.apenergy.2017.08.225.
Jiang, Zhanhong, Liu, Chao, Akintayo, Adedotun, Henze, Gregor P., & Sarkar, Soumik. Energy prediction using spatiotemporal pattern networks. United States. doi:10.1016/j.apenergy.2017.08.225.
Jiang, Zhanhong, Liu, Chao, Akintayo, Adedotun, Henze, Gregor P., and Sarkar, Soumik. 2017. "Energy prediction using spatiotemporal pattern networks". United States. doi:10.1016/j.apenergy.2017.08.225.
@article{osti_1399848,
title = {Energy prediction using spatiotemporal pattern networks},
author = {Jiang, Zhanhong and Liu, Chao and Akintayo, Adedotun and Henze, Gregor P. and Sarkar, Soumik},
abstractNote = {This paper presents a novel data-driven technique based on the spatiotemporal pattern network (STPN) for energy/power prediction for complex dynamical systems. Built on symbolic dynamical filtering, the STPN framework is used to capture not only the individual system characteristics but also the pair-wise causal dependencies among different sub-systems. To quantify causal dependencies, a mutual information based metric is presented and an energy prediction approach is subsequently proposed based on the STPN framework. To validate the proposed scheme, two case studies are presented, one involving wind turbine power prediction (supply side energy) using the Western Wind Integration data set generated by the National Renewable Energy Laboratory (NREL) for identifying spatiotemporal characteristics, and the other, residential electric energy disaggregation (demand side energy) using the Building America 2010 data set from NREL for exploring temporal features. In the energy disaggregation context, convex programming techniques beyond the STPN framework are developed and applied to achieve improved disaggregation performance.},
doi = {10.1016/j.apenergy.2017.08.225},
journal = {Applied Energy},
number = C,
volume = 206,
place = {United States},
year = 2017,
month =
}
  • Following several successful applications of feedforward neural networks (NNs) to the building energy prediction problem a more difficult problem has been addressed recently: namely, the prediction of building energy consumption well into the future without knowledge of immediately past energy consumption. This paper will report results on a recent study of six months of hourly data recorded at the Zachry Engineering Center (ZEC) in College Station, TX. Also reported are results on finding the R and C values for buildings from networks trained on building data.
  • Identification of vessel failures provides operators and technical support center personnel with important information to manage severe accidents in a nuclear power plant. It may be very difficult, however, for operators to identify a reactor vessel failure simply by watching temporal trends of some parameters because they have not experienced severe accidents. Therefore, the authors propose a methodology on the identification of pressurized water reactor (PWR) vessel failure for severe accident management using spatiotemporal neural network (STN). STN can deal directly with the spatial and temporal aspects of input signals and can well identify a time-varying problem. Target patterns ofmore » seven parameter signals were generated for training the network from the modular accident in nuclear power plants. They integrated MAAP code with STN in on-line system to mimic real accident situation in nuclear power plants. Using new pattern of signals that had never been used for training, the identification capability of STN was tested in a real-time manner. At the tests, STN developed in this study demonstrated acceptable performance in identifying the occurrence of a vessel failure. It is found that STN techniques can be extended to the identification of other key events such as onset of core uncovery, coremelt initiation, containment failure, etc.« less
  • The asymmetric attractor neural networks designed by the Monte Carlo- (MC-) adaptation rule are shown to be promising candidates for pattern recognition. In such a neural network with relatively low symmetry, when the members of a set of template patterns are stored as fixed-point attractors, their attraction basins are shown to be isolated islands embedded in a ''chaotic sea.'' The sizes of these islands can be controlled by a single parameter. We show that these properties can be used for effective pattern recognition and rejection. In our method, the pattern to be identified is attracted to a template pattern ormore » a chaotic attractor. If the difference between the pattern to be identified and the template pattern is smaller than a predescribed threshold, the pattern is attracted to the template pattern automatically and thus is identified as belonging to this template pattern. Otherwise, it wanders in a chaotic attractor for ever and thus is rejected as an unknown pattern. The maximum sizes of these islands allowed by this kind of neural networks are determined by a modified MC-adaptation rule which are shown to be able to dramatically enlarge the sizes of the islands. We illustrate the use of our method for pattern recognition and rejection with an example of recognizing a set of Chinese characters.« less
  • Power system generators may experience self excitation overvoltages due to certain contingencies. In an attempt to derive a new self-excitation (SE) preventive scheme, a prediction system based on pattern recognition techniques is described in this paper. Several design approaches are explained. A hyperplane discriminant is used to define the predictor surface. An algorithm is developed to reduce the number of telemetered channels required and hence the complexity of the prediction system structure. Moreover, a fast corrective algorithm designed to provide security improvement action is explained. The studied system is the Manitoba Hydro northern ac collector system where SE is themore » operating problem of concern. The results obtained prove the effectiveness of the proposed prediction scheme.« less
  • Jominy hardness profiles of steels were predicted from chemical composition and austenitizing temperature using an artificial neural network. The neural network was trained using some 4,000 examples, covering a wide range of steel compositions. The performance of the neural network is examined as a function of the network architecture, the number of alloying elements, and the number of data sets used for training. A well-trained network predicts the Jominy hardness profile with an average error of about 2 HRC. Special attention was devoted to the effect of boron on hardenability. A network trained using data only from boron steels producedmore » results similar to those of a network trained using all data available. The accuracy of the predictions of the model is compared with that of an analytical model for hardenability and with that of a partial least-squares model using the same set of data.« less