摘要
针对变速变桨风速发电机组如何在低风速时最大限度地捕获风能,提出了一种基于风速预测的改进爬山法最大功率追踪策略。首先搭建由模糊粗糙集(FRS)和神经网络预测(LSTM)两部分组成的预测模型,利用模糊粗糙集对噪声的敏感性,对风机多传感器采集的自然特性时间序列参数进行分析,通过属性约简,将输入信息的空间维数简化,确定神经网络的输入参数,作为后者LSTM神经网络预测模型部分的输入。利用LSTM在时间深度上有效避免梯度传播消失的特性,通过训练学习,抽取逼近隐含的输入输出的非线性关系,捕获时间序列风速上各个信息的关联度和时间延展度,得到风速的提前一步预测。然后依靠预测的风速信息,从搜索方向确定,搜索区间优化,避免最大功率点附近频繁波动3个方面改进了传统爬山法,避免传统方法的不足。通过GH bladed软件实验仿真表明:提出的最大功率追踪控制策略能够有效避免了风速变化情况下错误的搜索方向,提高追踪速度,明显减少风机在MPP点处的振荡,有效提高了风能捕获效率。
In order to capture the maximum wind energy at low wind speed,an improved hill climbing method based on wind speed prediction is proposed.Firstly,a prediction model consisting of two parts,namely fuzzy rough set(FRS)and neural network prediction(LSTM),is constructed.Making use of the sensitivity of the fuzzy rough set to the noises and analysing the natural time series parameters collected by the wind turbine sensor,their space dimensions of input information are simplified by attribute reduction.Through the LSTM neural network,the prediction of the subsequent wind speed can be realized.With the predicted wind speed information to determine the search direction and optimize the search interval,the hill climbing method is then improved.The simulation results show that the new hill climbing method can quickly judge the search direction and improve the tracking speed,and avoid the fluctuations near the MPP point as well.
作者
姚万业
贾昭鑫
黄璞
YAO Wanye;JIA Zhaoxin;HUANG Pu(School of Control and Computer Engineering,North China Electric Power University,Baoding 071003,China)
出处
《电力科学与工程》
2018年第2期44-49,共6页
Electric Power Science and Engineering
基金
中央高校基本科研业务费专项资金资助项目(2014MS138)
