期刊
RENEWABLE & SUSTAINABLE ENERGY REVIEWS
卷 60, 期 -, 页码 960-981出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.rser.2016.01.114
关键词
Multi-step wind speed forecast; Validation cuckoo search; EEMD; Lazy learning; Robustness
资金
- National Natural Science Foundation of China [71171102]
Wind energy, which is clean, inexhaustible and free, has been used to mitigate the crisis of conventional resource depletion. However, wind power is difficult to implement on a large scale because the volatility of wind hinders the prediction of steady and accurate wind power or speed values, especially for multi-step-ahead and long horizon cases. Multi-step-ahead prediction of wind speed is challenging and can be realized by the Weather Research and Forecasting Model (WRF). However, a large error in wind speed will occur due to inaccurate predictions at the beginning of the synoptic process in WRF. Multi-step wind speed predictions using statistical and machine learning methods have rarely been studied because greater numbers of forecasting steps correspond to lower accuracy. In this study, a detailed review of wind speed forecasting is presented, including the application of wind energy, time horizons for wind speed prediction and wind speed forecasting methods. This paper presents eight strategies for realizing multi-step wind speed forecasting with machine-learning methods and compares 48 different hybrid models based on these eight strategies. The results show good consistency among the different wind farms, with COMB-DIRMO models generally having a higher prediction accuracy than the other strategies. Thus, this paper introduced three methods of combining these COMB-DIRMO models, an analysis of their performance improvements over the original models and a comparison among them. Valid experimental simulations demonstrate that ALL-DDVC, one combination of the COMB-DIRMO models, is a practical, effective and robust model for multi-step-ahead wind speed forecasting. (C) 2016 Elsevier Ltd. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据