Combining Forecast of Electricity Price with Adaptive Weights Selected by SOM Clustering

DOI

10.2991/isci-15.2015.329How to use a DOI?

Keywords

electricity price forecasting; Combining forecast; SOM; Similar days clustering

Abstract

This paper proposes an efficient method of combining modelling to forecast day-ahead electricity price. The basic idea of this method is to develop a robust technique to select the proper combining weights of individual models according to their historical performance on similar circumstances. Firstly individuals were trained and validated separately, and then self-organising map neural network (SOM) was applied to cluster the circumstance into multiple categories. For day-ahead price forecasting, the circumstances of the coming day were compared with those of past days and then clustered into the same category by SOM. The combining weights of individual models for next-day price forecasting were determined by their performances in same category in the past. The final forecast was a weighted sum of the individual predictions with the time-varying weights obtained above. An experimental study indicated that the proposed method outperforms other price-forecasting techniques with high efficiency more robustness.

Copyright

© 2015, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Yi Feng
PY  - 2015/01
DA  - 2015/01
TI  - Combining Forecast of Electricity Price with Adaptive Weights Selected by SOM Clustering
BT  - Proceedings of the 2015 International Symposium on Computers & Informatics
PB  - Atlantis Press
SP  - 2540
EP  - 2547
SN  - 2352-538X
UR  - https://doi.org/10.2991/isci-15.2015.329
DO  - 10.2991/isci-15.2015.329
ID  - Feng2015/01
ER  -