Probability seminar: Alfred Müller, Universität Siegen
SM3, School of Mathematics
Alfred Müller, Universität Siegen
Probabilistic Forecasting of Medium-Term Electricity Demand: A Comparison of Time Series Models
The uncertainty of customer demand and its relation to the fluctuation of the electricity price is an important risk factor in electricity markets. Therefore there is a need for probabilistic forecasting of medium-term electricity demand for end customers. There already exists comprehensive literature on various load forecasting techniques, but they typically consider the grid load or private households, however. Load forecasting models for companies seem to be rare so far. As consumption patterns of companies vary significantly between different business sectors, model building and calibration depending on the sector seems to be reasonable. In this paper we introduce a whole class of time series models for modeling customer demand. The models vary in their number of parameters for the seasonal patterns, whether or not a dependence on grid load is included, and what kind of distribution is used for the residuals. We use the continuous ranked probability score (CRPS) to compare different time series models. We evaluate the model performance using historical load data of companies from different business sectors. The results reveal that for the yearly seasonality the use of sine and cosine functions typically is better than using dummies for each month. Moreover, hyperbolic distributions often provide a very good fit to the model innovations of the log demand in case of industry customers, whereas normal distributions maybe better in case of customers from the retail and service sector.
(Joint work with Kevin Berk)