Deep learning for power system data analysis

Elena Mocanu, Phong H. Nguyen, Madelaeine Gibescu

Research output: Chapter in Book/Report/Conference proceedingChapterAcademicpeer-review

3 Citations (Scopus)

Abstract

Unprecedented high volumes of data are available in the smart grid context, facilitated by the growth of home energy management systems and advanced metering infrastructure. In order to automatically extract knowledge from, and take advantage of this useful information to improve grid operation, recently developed machine learning techniques can be used, in both supervised and unsupervised ways. The proposed chapter will focus on deep learning methods and will be structured as follows: Firstly, as a starting point with respect to the state of the art, the most known deep learning concepts, such as deep belief networks and high-order restricted Boltzmann machine (i.e., conditional restricted Boltzmann machine, factored conditional restricted Boltzmann machine, four-way conditional restricted Boltzmann machine), are presented. Both, their theoretical advantages and limitations are discussed, such as computational requirements, convergence, and stability. Consequently, two applications for building energy prediction using supervised and unsupervised deep learning methods will be presented. The chapter concludes with a glimpse into the future trends highlighting some open questions as well as new possible applications, which are expected to bring benefits toward better planning and operation of the smart grid, by helping customers to adopt energy conserving behaviors and their transition from a passive to an active role.
Original languageEnglish
Title of host publicationBig Data Application in Power Systems
EditorsReza Arghandeh, Yuxun Zhou
PublisherElsevier
Chapter7
Pages125-158
Number of pages34
ISBN (Electronic)978-0-12-811969-3
ISBN (Print)978-0-12-811968-6
DOIs
Publication statusPublished - 27 Nov 2017
Externally publishedYes

Keywords

  • Deep Learning
  • Energy prediction
  • Reinforcement learning
  • Supervised learning
  • Transfer learning
  • Unsupervised learning
  • Time series

Fingerprint Dive into the research topics of 'Deep learning for power system data analysis'. Together they form a unique fingerprint.

  • Cite this

    Mocanu, E., Nguyen, P. H., & Gibescu, M. (2017). Deep learning for power system data analysis. In R. Arghandeh, & Y. Zhou (Eds.), Big Data Application in Power Systems (pp. 125-158). Elsevier. https://doi.org/10.1016/B978-0-12-811968-6.00007-3