[Zurück]


Vorträge und Posterpräsentationen (mit Tagungsband-Eintrag):

S. Grosswindhager, A. Voigt, M. Kozek:
"Online Short-Term Forecast of System Heat Load in District Heating Networks";
Vortrag: 31st International Symposium on Forecasting, Prag; 26.06.2011 - 29.06.2011; in: "ISF 2011 - Prague PROCEEDINGS", (2011), 1997-41.



Kurzfassung englisch:
This paper presents an on-line short term forecasting approach for system heat load in district heating networks (DHN) using the popular Seasonal Autoregressive Integrated Moving Average (SARIMA) models in state space representation. The system heat load itself is a non-stationary random process composed of the individual consumer
heat demands plus heat losses from pipes. Short term load forecasting is essential for effective operational production planning. It was found that the recurring pattern of the process based on half-hourly data are well described by a SARIMA(2,1,1)(0,1,1)48 model. The adequacy of the model was confirmed by standard regression diagnostics. Furthermore, the identified SARIMA model was incorporated into the state space framework where classical Kalman
Recursion allows convenient calculation of on-line forecasting values. Moreover, exogenous effects such as weather
effects are explicitly accounted for by decomposition of the original time series into an outdoor temperature dependent
part and a social component part, where the latter was again modeled as SARIMA process. The relationship between
system heat load and outdoor temperature may appropriately be expressed by a piece-wise linear function. Finally,
the performance of the proposed model is validated on real data by calculating the mean absolute percentage error
(MAPE) value for 48-steps-ahead (24h) estimates. The on-line performance for the basic and the temperature adapted
model was assessed by computing rolling 24-steps ahead MAPE values for approx. 20 days of real data. In this
work, the Kalman procedure is presented as an elegant approach for prediction of SARIMA processes in state space
representation. Specifically, it is shown that the proposed methods are suitable for on-line short term forecasting of
system heat load in district heating networks.

Schlagworte:
Heat Load Forecast, SARIMA, State Space Models, Kalman Filter, District Heating Networks


Elektronische Version der Publikation:
http://publik.tuwien.ac.at/files/PubDat_202018.pdf


Erstellt aus der Publikationsdatenbank der Technischen Universitšt Wien.