Please use this identifier to cite or link to this item: http://hdl.handle.net/10419/234065
Authors: 
Vogler, Arne
Ziel, Florian
Year of Publication: 
2019
Series/Report no.: 
HEMF Working Paper No. 11/2019
Abstract: 
In this paper we present an evaluation framework for predictions of binary events in probabilistic electricity price forecasting. It employs the MSE-equivalent QPS together with the DM test and allows for further insights about deficiencies of the considered models. Additionally, techniques from the field of classification are considered, which extend our framework and are particularly suited for the evaluation of predictions of rare events. We consider binary events with direct applicability to a generator's daily decision making such as profitability of a pumped-hydro storage plant and evaluate the respective forecasts statistically. We show that the task of forecast evaluation can be simplified from assessing a multivariate distribution over prices to assessing a univariate distribution over a binary outcome, fully characterized by a single probability.
Subjects: 
Probabilistic Forecasting
Binary Predictions
Classification
Electricity Price Forecasting
JEL: 
C53
C38
Q47
Document Type: 
Working Paper

Files in This Item:
File
Size
2.68 MB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.