Load forecasting using elastic gradient descent

• Main Authors: Hong, Yuan, Xia, Changhao, Zhang, Shixiang, Wu, Lin, Yuan, Chao, Huang, Ying, Wang, Xuxu, Zhu, Haifeng
• Format: Conference Proceeding
• Language: English
• Subjects: Accuracy; Computer simulation; Conferences; Convergence; Descent; elastic gradient descent method; error back propagation artificial neural network; Forecasting; Load forecasting; Load modeling; Mathematical models; Neural networks; Predictive models; Principal component analysis; time sequence; Training
• ISSN: 2157-9555; 2157-9563
• DOI: 10.1109/ICNC.2013.6817979

The article describes in detail the theoretical basis of the elastic gradient descent method which combines the principal component analysis (PCA) and the time sequence method. In the short-term forecasting instance, the elastic gradient descent neural networks which combines the PCA and the time sequence method was used. The result verifies the effectiveness and feasibility of the introducing the PCA and the time sequence method in processing network optimization. The simulation result shows that this method has good prediction accuracy and convergence speed. In the long-term forecasting instance, the elastic gradient descent method which combines PCA method was used for that forecasting. The result indicated the superiority of the introducing the principal component analysis method in processing large amounts of data. As used herein, the model has good ductility and also lots of factors can be considered in. The prediction accuracy and generalization is good. And it will have a further application prospect in the actual forecast.