Masters Theses


"With resources becoming more and more scarse [sic] as well as increasing competition caused by the liberalisation of the energy markets electric load modelling becomes ever more important for proper resource allocation.

This work tries to bridge the gap between long-term modelling done mainly via econometric approaches and short-term modelling in which time series models are more commonplace by focussing [sic] on pure time series modelling [sic] and exploring its limits in the process. Due to various seasonalities present in the data the approach chosen starts with a subdivision of the time axis in different time frames: A model for the yearly horizon based on monthly data, one for the weekly horizon based on daily data and finally a model for the daily horizon based on hourly data is developed. Basis for the case study is data acquired via PJM for American Electric Power's region (AEP) spanning from 1st January 2005 to 31st December 2010.

On the yearly horizon it can be shown that a classical SARIMA-model yields sufficiently good results. On a weekly horizon different approaches had to be experimented with: Inclusion of dummy variables in various settings, a fractionally integrated ARMA-approach to address the possibility of long-term memory as well as a vector-autoregressive approach. A remedy was found in a periodic time series regression establishing an autoregressive model for each day of the week.

Modelling on a daily horizon seasonality could not be accounted for in a classical SARIMA-approach. Thus seasonality in a preparatory step was removed via spline regression. Subsequently a classical ARMA-model was fitted. The pure time series approach at that stage reached its limits: Though a model could be found it didn't account for all effects exhibited in the data and leaving behind white noise. This suggests the inclusion of other explanatory variables in the model"--Abstract, page iii.


Samaranayake, V. A.
Bohner, Martin, 1966-

Committee Member(s)

Gelles, Gregory M.


Mathematics and Statistics

Degree Name

M.S. in Applied Mathematics


Missouri University of Science and Technology

Publication Date

Fall 2011


ix, 96 pages

Note about bibliography

Includes bibliographical references (pages 94-95).


© 2011 Matthias Benjamin Noller, All rights reserved.

Document Type

Thesis - Open Access

File Type




Subject Headings

Electric power systems -- Control
Electric power-plants -- Load -- Forecasting
Time-series analysis -- Mathematical models

Thesis Number

T 9938

Print OCLC #


Electronic OCLC #