An Integrated Affine Jump Diffusion Framework to Manage Power Portfolios in a Deregulated Market
Electricity markets around the world have gone through, or are currently in a deregulation phase. As a result, power companies that formerly enjoyed a monopoly are now facing risks. In order to cover (hedge) these risks, futures markets have emerged, in parallel with the spot price markets. Then, markets of more complex derived products have appeared to better hedge the risk exposures of power suppliers and consumers.
An Affine Jump Diffusion (AJD) framework is presented here to coherently model the dynamics of the spot price of electricity and all the futures contracts. The non-storability of electricity makes it indeed impossible to use it in hedging strategies. Futures contracts, however, are standard financial contracts that can be stored and used in hedging strategies. We thus propose to consider the set of futures contracts as the primary commodities to be modelled and jointly estimate the parameters of the spot and futures prices based on their historical time series.
The estimation is done by Maximum Likelihood, using a Kalman Filter recursive algorithm that has been updated to account for non-Gaussian errors. This procedure has been applied to the German European Energy index (EEX) based in Frankfurt for electricity, to the Brent for Crude oil, and to the NBP for natural gas.
The AJD framework is very powerful because the characteristic function of the underlying stochastic variables can be obtained just by solving a system of complex valued ODEs. We took advantage of this feature and developed a novel approach to estimate expectations of arbitrary functions of random variables that does not require the probability density function of the stochastic variables, but instead, their characteristic function. This approach, relying on the Parseval Identity, provided closed form solutions for options with payoff functions that have an analytical Fourier transform. In particular, European calls, puts and spread options could be computed as well as the value of multi-fuel power plants that can be viewed as an option to exchange the most economic fuel of the moment against electricity.
A numerical procedure has also been developed for options with payoff functions that do not have an analytical Fourier transform. This numerical approach is indeed using a Fast Fourier Transform of the payoff function, and can be used in Dynamic Programming algorithms to price contracts with endogenous exercise strategies.
Finally, it is showed that the (mathematical) partial derivatives of these contracts, often referred to as the Greeks, could also be computed at low cost. This allows to build hedging strategies to shape the risk profile of a given producer, or consumer.
School:Université catholique de Louvain
Source Type:Master's Thesis
Keywords:derivatives pricing hedging stochastic model estimation
Date of Publication:01/24/2003