| 摘要: |
| 为了对风速进行准确预测，结合分类与回归树(Classification and Regression Tree, CART)、自适应噪声完备集成经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise, CEEMDAN)、回声状态网络与非线性误差修正策略，提出了一种基于回声状态网络(Echo Sate Network, ESN)的混合期风速预测方法。其中，CART用于对原始数据进行重构，得到建模数据集。CEEMDAN用于提取输入特征信息。ESN根据输入特征建立风速预测建模。最后，利用误差修正策略对所得到的模型进行修正。基于国内某风电场的数据实验表明，所提出方法能够准确预测风速，可以指导风场生产，提高生产自动化水平。 |
| 关键词: 风速预测 回声状态网络 CEEMDAN CART 预测值修正 |
| DOI：10.19783/j.cnki.pspc.190923 |
| 投稿时间：2019-07-31修订日期：2019-11-01 |
| 基金项目:国家重点研发计划项目资助(2018YFB1500803) |
|
| Wind speed prediction method based on CEEMDAN and echo state network |
| HAN Hongzhi,TANG Zhenhao |
| (Xinjiang Xinneng Group Co., Ltd.Urumqi Electric Power Construction Debugging Institute, Urumqi 830001, China;School of Automation Engineering, Northeast Electric Power University, Jilin 132012, China) |
| Abstract: |
| In order to predict wind speed accurately, this paper combines CART, CEEMDAN, echo state network and error correction strategy to propose a short-term wind speed prediction method with a multi-processing strategy. CART is applied to reconstruct the original dataset to get the training data. CEEMDAN is employed to extract the feature information. Then, ESN is used to model the wind speed based on the features. Finally, the model is modified by an error correction strategy. The proposed method can predict wind speed accurately, guide the production of a wind farm and improve the automation level of production. This work is supported by National Key Research and Development Program of China (No. 2018YFB1500803). |
| Key words: wind speed prediction echo state network CEEMDAN CART prediction value correction |
