Modelling Seasonalities of HPFCs Using a Parametric Approach

Abstract: Electricity differs from other commodities in that it cannot be stored. This non-storability characteristic results in traditional pricing methods for commodities not being applicable for electricity. An alternative pricing method is therefore needed and the solution is the Hourly Price Forward Curve (HPFC). The HPFC essentially gives the prices as of today for the delivery of electricity at each hour in the future. It is generally constructed in two steps. The first step involves estimating the shape vector, which is a vector of hourly weights reflecting all seasonalities and forward-looking information in the electricity spot price. The second step calibrates the shape vector to futures products in order to make it arbitrage free. In this thesis, we have exclusively studied the first step of the HPFC construction. Specifically, we have modelled the month-to-hour ratios of the shape vector, i.e. a vector of hourly weights normalized for every month. This paper aims to explore the possibilities of modelling the month-to-hour ratios using parametric models with external inputs. This is done by implementing regression models using polynomial and Fourier basis functions, which is there after further developed with the addition of a Kalmanfilter and regularization techniques. The study covers the Nord Pool electricity market and it is conducted using data from E.ON and Nord Pool. It is concluded that parametric models with external inputs are well-suited for constructing a shape vector. It is successful in modelling the intra-yearly, intra-weekly and intra-daily seasonalities and shows robustness against extremities. However, difficulties arose in adequately modelling the summer months and the morning levels.