Modeling of German Electricity Load for Pricing of Forward Contracts

Nowadays, a large part of the traded volume in power markets represents energy to be consumed or produced in the future as a forward or future product. Hence, the need for precise forecasting of load demand and prices arises for all market players. This project proposes an hourly load model for Germany with focus on long term forecast using tools provided from the econometrics theory, such as regression analysis, and time series analysis with stochastic calculus. The main purpose of this model is for applications in forward markets, such as fundamental modeling of spot prices, pricing of forward contracts and the formulation of the hourly price forward curve (HPFC). The proposed model captures the deterministic component of the load, as well as its stochastic variations. Decomposition of daily load level and hourly load was made in order to achieve more accurate results. Then, each part was modeled separately and aggregated in the end. A linear regression approach was used for capturing the deterministic patterns of the load such as yearly,weekly seasonality and intraday patterns, taking into account holidays, weather and economic trends. An autoregressive model (AR) was used for the modeling of the stochastic variations. Parameter calibration was made once in a 5 year period (2006-2010) and once in a 4 year period (2006-2009). Then, the the model was evaluated and validated out of sample for 2011 and 2010-2011 respectively. The results presented in Chapter 5 were promising and led us to produce a future forecast for the period 2012-2014.