Aiming at the impact of inaccurate seasonal division and dynamic changes of meteorological conditions on short-term power load prediction

a kind of short-term prediction algorithm of power load based on load clustering and trend adaptive segmentation is proposed. The algorithm firstly clusters the historical data to achieve seasonal division of load curve. And then

the trend of each season load variation is extracted and the load curve is subject to adaptive division. Finally

the multiple segmentation modeling and prediction are achieved by way of improving ORELM model. The experimental results show that

compared withsuchexistedpredictionmethodsassupportvectorregressionandextremelearningmachine

thealgorithm in this paper obtains achieves better prediction accuracy and robustness to different seasons and meteorological conditions.