Title: Power load forecasting using data mining and knowledge discovery technology

Authors: Yongli Wang, Dongxiao Niu, Ling Ji

Addresses: School of Economics and Management, North China Electric Power University, 2 Beinong Road, Changping District, Beijing 102206, China. ' School of Economics and Management, North China Electric Power University, 2 Beinong Road, Changping District, Beijing 102206, China. ' School of Economics and Management, North China Electric Power University, 2 Beinong Road, Changping District, Beijing 102206, China

Abstract: Considering the importance of the peak load to the dispatching and management of the electric system, the error of peak load is proposed in this paper as criteria to evaluate the effect of the forecasting model. This paper proposes a systemic framework that attempts to use data mining and knowledge discovery (DMKD) to pretreat the data. And a new model is proposed which combines artificial neural networks with data mining and knowledge discovery for electric load forecasting. With DMKD technology, the system not only could mine the historical daily loading which had the same meteorological category as the forecasting day to compose data sequence with highly similar meteorological features, but also could eliminate the redundant influential factors. Then an artificial neural network is constructed to predict according to its characteristics. Using this new model, it could eliminate the redundant information, accelerate the training speed of neural network and improve the stability of the convergence. Compared with single BP neural network, this new method can achieve greater forecasting accuracy.

Keywords: knowledge discovery; data mining; power systems; artificial neural networks; ANNs; power load forecasting; China; peak load; electric load forecasting; forecasting accuracy.

DOI: 10.1504/IJIIDS.2011.042531

International Journal of Intelligent Information and Database Systems, 2011 Vol.5 No.5, pp.452 - 467

Received: 26 Jul 2010
Accepted: 31 Oct 2010
Published online: 21 Oct 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article