Wind Power Ensemble Forecasting

Performance Measures and Ensemble Architectures for Deterministic and Probabilistic Forecasts

Gensler, André

kassel university press, ISBN: 978-3-7376-0636-3, 2019, 214 Pages
(Intelligent Embedded Systems 12)

URN: urn:nbn:de:0002-406378

DOI: 10.19211/KUP9783737606370

Zugl.: Kassel, Univ., Diss. 2018

| Price and available forms -->

Content: This thesis describes performance measures and ensemble architectures for deterministic and probabilistic forecasts using the application example of wind power forecasting and proposes a novel scheme for the situation-dependent aggregation of forecasting models. For performance measures, error scores for deterministic as well as probabilistic forecasts are compared, and their characteristics are shown in detail. For the evaluation of deterministic forecasts, a categorization by basic error measure and normalization technique is introduced that simplifies the process of choosing an appropriate error measure for certain forecasting tasks. Furthermore, a scheme for the common evaluation of different forms of probabilistic forecasts is proposed. Based on the analysis of the error scores, a novel hierarchical aggregation technique for both deterministic and probabilistic forecasting models is proposed that dynamically weights individual forecasts using multiple weighting factors such as weather situation and lead time dependent weighting. In the experimental evaluation it is shown that the forecasting quality of the proposed technique is able to outperform other state of the art forecasting models and ensembles.

Publication is available in following forms:

Full text (pdf-file, printable, Open Access - 6.35 MB)
view PDF