Ultra-short-term wind power interval prediction based on hybrid temporal inception convolutional network model

被引：17

作者：

Han, Yuchao ^{[1
]}

Tong, Xiangqian ^{[1
]}

Shi, Shuyan ^{[1
]}

Li, Feng ^{[2
]}

Deng, Yaping ^{[1
]}

机构：

[1] Xian Univ Technol, 58 Yanxiang Rd, Xian 710061, Shaanxi Provinc, Peoples R China

[2] Power Res Inst State Grid Ningxia Power Co, 288 Changcheng Rd, Yinchuan 750011, Ningxia Provinc, Peoples R China

来源：

ELECTRIC POWER SYSTEMS RESEARCH | 2023年 / 217卷

关键词：

Ultra-short-term wind power prediction; Variational mode decomposition; Temporal convolution network; Deep learning; Hybrid prediction model; SPEED; DECOMPOSITION; ALGORITHM;

D O I：

10.1016/j.epsr.2023.109159

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

Wind speed interval prediction is essential for wind power production. The possible variation interval of wind speed cannot be reflected in deterministic prediction, which is an excellent opportunity to build up interval prediction. This paper proposed a hybrid interval prediction model based on temporal inception convolutional network (TICN) and variational mode decomposition (VMD). Firstly, Pearson correlation is used to determine the decompose level of variational mode decomposition, and the simplified sub-series can be used to extend the trend of wind speed. Then referring to the concept of inception network, the proposed interval prediction model is built up with multiple temporal convolutional networks with different kernels. At last, data sets from real wind farms are used to evaluate the performance, and experiment results show the proposed model outperforms all other benchmark models in multi-step interval predictions. It is indicated that the proposed wind speed interval prediction model has better performance and can be used in practice.

引用

页数：11