Estimating a Levy Multifactor Market Model for Electricity Futures Market by Using ICA

In this paper, we develop an Electricity Market Model (EMM) for electricity futures and forwards contracts. The dynamic of these derivatives is modelled as multi-factor market model where the idea is to match the observed volatility term structure and correlation surface among different electricity futures deliveries. Empirical analysis shows that the distributions of electricity forward log-returns are non-normal. A Levy multi-factor model for electricity futures contracts with non-overlapping delivery periods, and in the particular case of normal inverse Gaussian (NIG), is proposed in order to capture the heavy tails that are not described by the normal distribution. The large deviation from normality of electricity futures price returns produces an unbiased volatility estimation by Principal Component Analysis (PCA). For this reason, we consider a method to decompose the correlation/covariance matrix that can handle leptokurtic data: the Independent Component Analysis (ICA). Having identified the K independent components that affect the forward term structure, we look at the time series of these components for jumps and fit a Levy-type model for each principal component. Finally we implement the model and provide some numerical examples using data from the EEX and Powernext electricity market