Nonlinear Predictability of Time Series and Its Application for Load Forecast

T. Rauschenbach (Germany)

Keywords

Nonlinear Predictability, Linear Predictability, Load Fore cast, Pattern Based Forecast

Abstract

This contribution shows a specific design of a prediction model for electrical load of a public utility. The model has a forecast horizon of one week and is based on the com putation of the nonlinear predictability of time series. The method does not dependent on an assumed model struc ture. This means a statement about the expected quality of the forecast is possible before model design. Different input variables are tested with the aid of the calculated pre dictability on suitability for the forecast. The best results are obtained with a pattern based approach with additional consideration of the temperature.

Important Links:



Go Back