Short-term power load forecasting based on big data

The short-term power load forecasting method had been researched based on the big data. And combined the local weighted linear regression and cloud computing platform, the parallel local weighted linear regression model was established. In order to eliminate the bad data, bad data classification model was built based on the maximum entropy algorithm to ensure the effectiveness of the historical data. The experimental data come from a smart industry park of Gansu province. Experimental results show that the proposed parallel local weighted linear regression model for short-term power load forecasting is feasible; and the average root mean square error is 3. 01% and fully suitable for the requirements of load forecasting, moreover, it can greatly reduce compute time of load forecasting, and improve the prediction accuracy. © 2015 Chin. Soc. for Elec. Eng..