EM Algorithm for Electricity Pool Price Prediction and Errors-in-variables Process Identification

• Author / Creator

  Wu, Ouyang

• In this thesis, under the EM algorithm framework, a multiple model approach is developed towards electricity price prediction, and the identification problem for errors-in-variables (EIV) systems is studied. Alberta's electricity price, which shows high volatility and erratic nature, is considered as an example of a nonlinear process. A Markov regime-switching model is applied to predict the price using the local models through the investigations of characteristics of the pool price sequence. The expectation maximization (EM) algorithm is applied to solve the maximum likelihood (ML) estimation problem for model parameters, and several initialization methods are proposed to generate the initial values for the EM algorithm. The validations are presented to verify the proposed approach, which demonstrate an improvement on the existing price prediction for the range of high electricity prices. A dynamic system that has both input and output measurement errors is considered as an errors-in-variables (EIV) system. Employment of traditional identification strategies for EIV systems will result in biased estimates and inaccurate estimation of system parameters. EIV approaches such as the subspace EIV method has been proposed, but the subspace approach does not possess the optimality such as ML estimation. However, the direct application of ML approach for EIV model parameter estimation can lead to intractable solutions. In this work, we assume a dynamic model for noise-free input and propose to solve the ML problem using the EM algorithm. To identify industrial nonlinear EIV processes that operate along an operating trajectory, a linear parameter varying (LPV) EIV model is proposed to approximate the global models. The EM algorithm is used to solve the ML estimation for LPV EIV model parameters. Various numerical simulations and pilot-scale experiments are used to demonstrate the effectiveness of the proposed approach.

• Subjects / Keywords

  • EM algorithm
  • Errors-in-variables
  • Pool price prediction
  • Multiple model

• Graduation date

  Spring 2016

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/R36T0H190

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.