Reducing Power Flow Simulation Cost Using Universal Function Approximators

• Author / Creator

  Bardwell, Michael

• By integrating universal function approximators into existing simulation software, it is possible to reduce the cost of repeating simulations and thereby increase research output. In this thesis, support vector regression, random forest and artificial neural networks are deployed as universal function approximators. It is shown in an applied non-linear power flow problem that each model can achieve a maximum absolute error below 2 percent, and a root mean squared error below 0.2 percent. For selecting the number of hidden layer neurons in a single hidden layer artificial neural network, a method known as extrema equivalence is trialled. The extrema equivalence algorithm successfully identifies the approximately most sparse hidden layer size that produces near-perfect R2 scores for smooth, continuous functions. Lastly, a generic file management software is proposed that can be implemented into simulation programs to save users time when re-simulating the same models with different inputs.

• Subjects / Keywords

  • extrema equivalence
  • power flow
  • regression

• Graduation date

  Fall 2019

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-e210-st50

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.