Ultra Short-Term Power Load Forecasting Based on Similar Day Clustering and Ensemble Empirical Mode Decomposition

With the increasing demand of power industry for load forecasting, improving accuracy forecasting has become increasingly important. In this paper, we propose an ultra short-term method based on similar day clustering and EEMD (Ensemble Empirical Mode Decomposition). detail, K-means algorithm was utilized to divide historical data into different clusters. Through EEMD, each cluster were decomposed several sub-sequences with time scales. The LSTNet (Long- Short-term Time-series Network) adopted as model these sub-sequences. forecast results combined expected result. proposed predicts in next 4 h interval 15 min. experimental show that obtains higher prediction than other comparable models.