Forecasting day-ahead electricity prices: Utilizing hourly prices

Abstract

Standard-Nutzungsbedingungen:Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden.Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen.Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. This version -May 6, 2013 Terms of use: Documents in AbstractThe daily average price of electricity represents the price of electricity to be delivered over the full next day and serves as a key reference price in the electricity market. It is an aggregate that equals the average of hourly prices for delivery during each of the 24 individual hours. This paper demonstrates that the disaggregated hourly prices contain useful predictive information for the daily average price. Multivariate models for the full panel of hourly prices signicantly outperform univariate models of the daily average price, with reductions in Root Mean Squared Error of up to 16%. Substantial care is required in order to achieve these forecast improvements. Rich multivariate models are needed to exploit the relations between dierent hourly prices, but the risk of overtting must be mitigated by using dimension reduction techniques, shrinkage and forecast combinations.

Day-ahead electricity price forecasting with high-dimensional structures: Univariate vs. multivariate modeling frameworks

“…A white square indicates that forecasts of a model on the X-axis are better for all 24 hours, while a black square that they are not better for a single hour. (2014), Raviv et al (2015), Gaillard et al (2016), Marcjasz et al (2018) and Nowotarski and Weron (2018), among others. In Table 4 we report the MAE values for the better of the two lasso models (this is in row 'Better of the two') and their arithmetic average (row 'Combination').…”

“…Vector AutoRegression (VAR) is the basic modeling structure in the fully multivariate context, for sample EPF applications see Huisman et al (2007), Panagiotelis and Smith (2008), Haldrup et al (2010) and He et al (2015). However, if the number of parameters is very large it may be a good idea to reduce dimensionality of the problem first and consider factor models, as in Garcia-Martos et al (2012), Wu et al (2013), Weron (2015, 2016) and Raviv et al (2015). In one of the few applications in the computational intelligence EPF literature, Yamin et al (2004) use a neural network with 24 nodes in the output layer, hence consider a fully multivariate approach.…”

Comparing the forecasting performances of linear models for electricity prices with high RES penetration

“…More recently, but without accounting for fundamental drivers and looking only at point forecasts, Raviv et al (2015) compared the performances of models for the full panel of 24 hourly prices studying NordPool from 1992 to 2010. Based on univariate AR and multivariate VAR models, they computed forecast combinations and empirically demonstrated that the useful predictive information contained in disaggregated hourly prices improves the forecasts of multivariate models.…”

Stochastic modelling of the price‐volume relationship in electricity markets: Evidence from the Nordic energy exchange

“…It is an aggregate that equals the average of hourly prices for delivery during each of the 24 hours. 12 The use of daily aggregated data in empirical studies has been often motivated by the need for the most parsimonious model relating prices' characteristics such as spikes occurrence and extreme volatility to a limited number of exogenous factors. Relying on daily data simplifies assessing the seasonal components and the influence of many important driving factors.…”