School of Economics and Management, North China Electric Power University, Beijing, Chinia

---

Abstract—In this paper, an efficient combined modeling based on FHNN similar-day clustering to forecast shortterm power load is proposed. As the performance of individual models varies under different circumstances, the combination weights of forecast model should change with the circumstances. Here we classify historical power load into three parts including training set, validation set and test set model. Four methods, including Autoregressive Moving Average (ARMA), Generalized Autogressive Conditional Heteroscedasticity (GRACH), Artificial Neural Network (ANN) and Support Vector Machine (SVM), are selected as candidate models. For short load forecasting, the circumstance of the coming day is compared with those of past days and then clustered into the same category by Fuzzy Hopfield neural network (FHNN). The combining weights are obtained according to mean absolute percentage errors of different models. Then the combined forecasting model with ARMA-GRACH-ANN-SVM weighted by average with the weights obtained from FHNN clustering is got. A case study shows that the proposed combined model outperforms other forecast methods.

Index Terms—Short-term power load, combined forecasting, ARMA-GRACH-ANN-SVM, FHNN, similar days clustering.

[PDF]

Cite: Dongxiao Niu, Yanan Wei, "An Improved Short-Term Power Load Combined Forecasting With ARMA-GRACH-ANN- SVM Based on FHNN Similar-Day Clustering," Journal of Software vol. 8, no. 3, pp. 716-723, 2013.