Wind power forecast based on improved Long Short Term Memory network. (15th December 2019)

• Record Type: Journal Article
• Authors: Han, Li; Jing, Huitian; Zhang, Rongchang; Gao, Zhiyu
• Is Part Of: Energy. Volume 189(2019)
• Publication Date: 2019-12-15
• Subjects: Wind power forecast -- Improved LSTM -- Random memory suppression -- VMD; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2019.116300
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: In order to improve the forecast accuracy of wind power, an Improved Long Short Term Memory (ILSTM) network structure is proposed. Firstly, Variational Mode Decomposition (VMD) method is adopted to decompose wind power signal to the long-term component, the fluctuation component and the random component, which are used as the input of forecast model. Then a parameter was defined and added to the memory cell to suppress the random component to long term memory of neural network. To provide a pass for the current random component, the output gate was modified accordingly. Compared with the traditional Long Short Term Memory (LSTM), the improved LSTM can reduce the impact of random component on the patterns in long term memory cells, while maintain the current random component in the short term memory of network. As a result, the learning for the real patterns of wind power is strengthened, avoiding over-fitting and achieve a better generalized forecast model. Finally, the performance of the forecast method proposed in this paper is tested by using the wind power data from the Belgian ELIA website. Highlights: Variational Mode Decomposition method is improved to decompose wind power signal. In improved LSTM neural network, a parameter was added to the memory cells. ILSTM can reduce the impact of random component on the long-term memory cells.