Risk Analysis of Power Bidding Market Based on Extreme Value Theory

This paper proposes three econometric AR-EGARCH models, whose error is governed by normal, students' t and GED distribution respectively, to nine o'clock spot price serial of Nordpool power market from 2001 to 2005 year. In addition, POT method of extreme value theory is adopted to model the tails of the return distribution, and the parameters of GPD distribution are estimated, then VaR and CVaR are measured under 1% and 5% confidence level. In the end, factual return serial consisted of 175 datas in 2006 year is adopted to test the proposed models, the statistical test shows that CVaR, not VaR, is a good tool to measure risk, and the hypothesis of AR-EGARCH models' error are students' t and GED distribution provides appropriate estimate, which one is better estimate factual return lies on confidence level.