MA Li xin, YIN Jing jing, ZHENG Xiao dong, LUAN Jian

 

(School of Optical Electrical and Computer Engineering, University of Shanghai for

Science and Technology, Shanghai 200093, China)

 

 

Abstract: Aiming at solving the problems that the research objects are complex, the load randomness is strong and the short term power load forecasting has low forecasting accuracy and long computation time in smart electricity consumption environment, a new power load forecasting method that based on extreme learning machine (ELM) and Adaboost algorithm was proposed. The Adaboost algorithm was extended to this method. Firstly, the historical data were processed, the sample weights were initialized and the ELM network was trained repeatedly to predict output data. The forecasting process of ELM was simple and fast. The weight of weak predictors was identified through the Adaboost algorithm adjusting the weight of test samples. Finally the weak predictors were assembled together to constitute a strong predictor. The power load data of one city were used for simulating and comparing. The results indicate that this method not only has high prediction precision and good generalization performance, but also has a certain theoretical significance and good application prospect.

 

Key words: load forecasting; extreme learning machine; adaboost algorithm; strong predictor; neural network