基于CEEMDAN—DBN的风向预测研究

Study on Wind Direction Prediction based on CEEMDAN--DBN

高精度的风向预测有利于提高发电量、制定偏航控制方法和维持风电机组的稳定运行。由于风向具有随机性和不确定性,文章提出一种基于CEEMDAN和深度置信网络(DBN)组合的短期风向预测模型。首先,采用随机森林算法对原始风向数据进行分类,同时再利用拉伊达准则对其进行异常值的处理;然后采用CEEMDAN进一步剔除选定输入序列中所含有的无用的输入信息,提取建模所需特征信息;最后利用深度置信网络进行建模,结合中国某风电场的数据完成风向预测。实验结果表明,文章所用方法相比于BP、LSTM、ELM算法提高了预测精度,对偏航研究提供了支持。

High-precision wind direction prediction is conducive to improving power generation,formulating yaw control methods and maintaining the stable operation of wind turbine.Due to the randomness and uncertainty of wind direction,this study proposes a short-term wind direction prediction model based on the combination of CEEMDAN and deep confidence network(DBN).First,the random forest algorithm is used to classify the original wind direction data,and the outliers are processed by the Raida criterion;then CEEMDAN is used to further eliminate the useless input information in the selected input sequence and extract the characteristic information required for modeling;finally,the deep confidence network is used to model the wind direction prediction based on the data of a wind farm in China.Experimental results show that the method used improves the prediction accuracy compared with BP,LSTM and ELM algorithms and supports the yaw research.