A hybrid approach to price forecasting incorporating exogenous variables for a day ahead electricity Market

Abstract

Predicting Location Based Market Price (LBMP) in real time is an important and complex task that involves a great accuracy to model extreme spikes. Occurrence of spikes in real time LBMP is due to uncertainty in load demand as well as a significant variation in surrounding temperature throughout the day. These fluctuations cause extreme spikes in market clearing price due to transmission outages, shortage of supply and consumption of costly generation reserves. Therefore, the market participants are forced to purchase the electricity at very high price but the regulation set on the price cap by market authorities results into an economic loss to retailers and market participants. Accordingly, accurate forecasting of price is an important exercise to hedge against volatility. This paper proposes a hybrid technique comprising singular spectrum analysis and neural network incorporating temperature and load data for forecasting in day ahead electricity market. The results obtained with this hybrid technique are found to be more accurate when compared to existing methods described in literature.