Hostname: page-component-cd9895bd7-gvvz8 Total loading time: 0 Render date: 2024-12-27T21:53:04.166Z Has data issue: false hasContentIssue false
  • English
  • Français

A LOGNORMAL MODEL FOR DEMAND FORECASTING IN THE NATIONAL ELECTRICITY MARKET

Part of: General and miscellaneous specific topics

Published online by Cambridge University Press:  15 February 2016

J. MAISANO Open the ORCID record for J. MAISANO [Opens in a new window] ,
A. RADCHIK Open the ORCID record for A. RADCHIK [Opens in a new window]  and
T. LING
J. MAISANO*
Affiliation:
Trading Technology Australia Pty. Ltd., Australia email joe@tta.com.au University of Technology Sydney, NSW, Australia email Timothy.Ling-1@uts.edu.au
A. RADCHIK
Affiliation:
Green Trading Systems Pty. Ltd., Australia email aradchik@aksong.com University of Technology Sydney, NSW, Australia email Timothy.Ling-1@uts.edu.au
T. LING
Affiliation:
University of Technology Sydney, NSW, Australia email Timothy.Ling-1@uts.edu.au
*
✉joe@tta.com.au
Rights & Permissions [Opens in a new window]

Abstract

Core share and HTML view are not available for this content. However, as you have access to this content, a full PDF is available via the ‘Save PDF’ action button.

Many electricity market participants have a requirement to calculate the probabilistic risk measures, such as earnings at risk (EaR) and value at risk (VaR), for compliance reporting purposes. This requirement is currently hindered by the lack of analytical representations for forecasts of demand (load) and price curves; this motivates numerical simulation and models that need extensive calibration. In this paper, we derive an analytical representation of a state demand forecast which is the aggregated usage of all electricity consumers in a particular region (such as New South Wales or Victoria). We have used two probabilistic benchmarks from the Australian energy market operator as input, which are expressed as forecasted probability of exceedance.

Due to a number of considerations, including asymmetry of these quantiles with respect to the median, we have selected a series of truncated lognormal distributions with two parameters. The procedure of finding these parameters has been reduced to solving (for every half-hour) a single nonlinear equation. As a result, the two-year half-hourly forecast (expected curve) and demand volatility are found by explicit integration with the set of derived distributions. We have also tested an alternative method based on simplifying assumptions; using a nontruncated lognormal distribution, we found that under the test conditions this method produces an identical forward load and volatility curve.

Keywords

electricitydemand forecastingnational electricity market

MSC classification

Primary: 00A69: General applied mathematics
Type
Research Article
Information
The ANZIAM Journal , Volume 57 , Issue 3 , January 2016 , pp. 369 - 383
Copyright
© 2016 Australian Mathematical Society 

References

Abramowitz, M. and Stegun, I., Handbook of mathematical functions with formulas, graphs, and mathematical tables (Dover Publications, New York, 1972); http://people.math.sfu.ca/∼cbm/aands/intro.htm.Google Scholar
AEMO, “Short term PASA process description”, Technical Report, AEMO Electricity Market Performance, ver. 006 (2012); http://www.aemo.com.au/Electricity/Market-Operations/Dispatch/STPASA-Process-Description.Google Scholar
Witkovsky, V., “Computing the distribution of a linear combination of inverted gamma variables”, Kybernetika 37 (2001) 79–90; http://www.researchgate.net/publication/236323890_Computing_the_distribution_of_a_linear_combination_of_inverted_gamma_variables.Google Scholar