Data mining techniques for smart grid load forecasting

Abstract

Smart grids, or intelligent electricity grids that utilize modern IT/communication/control technologies, become a global trend nowadays. Forecasting of future grid load (electricity usage) is an important task to provide intelligence to the smart gird. Accurate forecasting will enable a utility provider to plan the resources and also to take control actions to balance the supply and the demand of electricity. In this paper, our contribution is the proposal of a new data mining scheme to forecast the peak load of a particular consumer entity in the smart grid for a future time unit. We utilize least-squares version of support vector regression with online learning strategy in our approach. Experimental results using two datasets, each containing two sub-datasets, show that our method is able to provide more accurate results than an existing forecasting method which is reported to be one of the best. On Germany dataset, our method can provide 98.4–98.7% of average accuracy whilst the state-of-the-art method by Lv et al. is able to provide only 96.7% of average accuracy. On Abu Dhabi dataset, which is less predictable, our method still can provide 95.7–96.7% of average accuracy whilst the method by Lv et al. provides lightly less average accuracy of 95.3%–96.4%. Our method is also computationally efficient and can potentially be used for large scale load forecasting applications.

Date of Award	2012
Original language	American English
Supervisor	U Zeyar Aung (Supervisor)