Authors: Amitava Bandyopadhayay; Soumya Roy; Dipojjwal Ghosh
Addresses: Indian Statistical Institute, SQC and OR Unit, Kolkata, 203 Barrackpore Trunk Road, Kolkata-700 108, India ' Department of Management Studies, Indian Institute of Science, Bangalore-560012, India ' Analytics and Information Management, Wipro Technologies, Block DM, Sector-5, Saltlake, Kolkata-700091, India
Abstract: The electricity market is being deregulated all over the world. Deregulation has brought in a variety of trading systems such as day-ahead trading and has also introduced high volatility of electricity prices. The large variability of price increases the risk for the market participants and forces the business houses to look for a forecasting accuracy of about ± 3%. This paper provides a method to predict next-day electricity prices using dynamic regression methodology where the price was regressed on selected demand, as well as supply side variables available in the public domain, and the error has been modelled using ARIMA/SARIMA models. The results were found to be very encouraging with MAPE lying in the range of ± 3.5% in most cases. In order to reduce the complexity associated with developing many models, a clustering methodology was used to group the different hours of the day so as to reduce the number of forecasting models to be fitted. Agglomerative hierarchical clustering with single linkage was used and models for representative hours had the required level of accuracy.
Keywords: deregulated markets; day-ahead prices; price forecasting; multiple linear regression; MLR; ARIMA; dynamic regression; electricity prices; modelling; clustering.
International Journal of Business Excellence, 2013 Vol.6 No.5, pp.584 - 604
Published online: 14 Aug 2013 *Full-text access for editors Access for subscribers Purchase this article Comment on this article