Probabilistic approach to handle uncertainties in load forecasting

Damitha K. Ranaweera, George G. Karady, R. G. Farmer

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Scopus citations


This paper presents a novel method to include the uncertainties of weather variables in neural network based load forecasting models. The methodology is based on the probability distribution of uncertain variables. The proposed method consists of traditionally trained neural networks and a set of equations to calculate the mean forecast and the variance of the forecast. This method was tested for daily forecasts for one year. The test results indicated that in addition to the availability of the confidence interval, the new method also provides a more accurate mean forecast than traditional neural networks alone.

Original languageEnglish (US)
Title of host publicationProceedings of the American Power Conference
Editors Anon
PublisherIllinois Inst of Technology
Number of pages6
StatePublished - 1995
EventProceedings of the 57th Annual American Power Conference. Part 1 (of 3) - Chicago, IL, USA
Duration: Apr 18 1995Apr 20 1995


OtherProceedings of the 57th Annual American Power Conference. Part 1 (of 3)
CityChicago, IL, USA

ASJC Scopus subject areas

  • Nuclear Energy and Engineering
  • Electrical and Electronic Engineering
  • Mechanical Engineering


Dive into the research topics of 'Probabilistic approach to handle uncertainties in load forecasting'. Together they form a unique fingerprint.

Cite this