Prediction of hour of coincident daily peak load
Document Type
Conference Proceeding
Publication Date
8-8-2019
Department
Department of Computer Science
Abstract
To manage the power system well, it is important to keep the balance between the electricity supply and its demand. For many utilities, consumers need to pay additional money for the electricity during the peak time on days with the highest demands - coincident peak pricing. Therefore, if the consumers are informed the daily load peak a day or hours in advance, such surcharges may be avoided. The accurate forecasting of load peak time not only helps to provide a reliable electricity supply, but also be useful to reduce the cost of electricity for consumers. In this paper, we study the historical data and then build the classification models for winter and summer to predict the time of daily peaks 24 hours ahead. Several classification algorithms, including Naïve Bayes, SVM, Random Forest, AdaBoost, CNN, LSTM, Stacked Autoencoder, are applied to solve this problem. Finally, the performance of these methods have been examined and compared with LSTM having the best overall accuracy, precision, and recall.
Publication Title
2019 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT)
Recommended Citation
Liu, J.,
&
Brown, L. E.
(2019).
Prediction of hour of coincident daily peak load.
2019 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT).
http://doi.org/10.1109/ISGT.2019.8791587
Retrieved from: https://digitalcommons.mtu.edu/michigantech-p/909
Publisher's Statement
Copyright © 2019, IEEE. Publisher’s version of record: https://doi.org/10.1109/ISGT.2019.8791587