基于ICEEMDAN-PE/FE-IGWO-SVR的混合多步分解短期风速预测方法  

Hybrid multi-step decomposition short-term wind speed prediction method based on ICEEMDAN-PE/FE-IGWO-SVR

作　　者：赵鑫 常喜强[1,2,3] 李梦女 ZHAO Xin;CHANG Xiqiang;LI Mengnü(School of Electrical Engineering,Xinjiang University,Urumqi 830047,China;Xinjiang Energy Internet Big Data Laboratory,Urumqi 830002;State Grid Xinjiang Electric Power Co.,Ltd.,Urumqi 830002,China)

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830047 [2]新疆能源互联网大数据实验室,新疆乌鲁木齐830002 [3]国网新疆电力有限公司,新疆乌鲁木齐830002

出　　处：《现代电子技术》2023年第1期124-130,共7页Modern Electronics Technique

基　　金：国家自然科学基金项目(51767023)。

摘　　要：针对风电场季节性风速波动性和时间尺度差异引起的预测滞后性问题,提出一种基于ICEEMDAN-PE/FE-IGWO-SVR的混合多步分解短期风速预测方法。首先,以改进的自适应白噪声完全集合经验模态分解(ICEEMDAN)法对原始时间序列进行一次分解,得到精确的本征模函数(IMF);再以排列熵(PE)和模糊熵(FE)联合判别方法对其进行二次混合分解,进一步削弱风速波动性;然后将分解后的数据代入支持向量机(SVR)进行预测。此外,为了找到更优的SVR参数,文中引入非线性动态更新因子和萤火虫算法的吸引机制对灰狼算法进行改进,并基于改进的灰狼算法对SVR参数寻优,进而对某风电场进行夏季短期风速预测,实验结果证明,与传统预测方法相比,该方法对短时突变型恶劣风况预测精度更高,对提高风电并网灵活性调度具有一定的应用价值。In view of the prediction lag caused by seasonal wind speed fluctuation and time scale difference of wind power plant,a hybrid multi-step decomposition short-term wind speed prediction method based on ICEEMDAN-PE/FE-IGWO-SVR is proposed.The original time series is decomposed once by the improved complete ensemble empirical mode decomposition with adaptive white noise(ICEEMDAN)method to obtain an accurate intrinsic mode function(IMF),and then its secondary hybrid decomposition is performed by the combined discriminant method of permutation entropy(PE)and fuzzy entropy(FE),which further weakens the fluctuation of wind speed.And then,the decomposed data is substituted into the support vector machine(SVM)for prediction.In addition,in order to find out better SVR parameters,the nonlinear dynamic update factor and the attraction mechanism of the firefly algorithm(FA)are introduced to improve the gray wolf optimization(GWO)algorithm.On the basis of the improved GWO(IGWO)algorithm,the SVR parameters are optimized to predict the short-term wind speed of a certain wind power plant in summer.The experimental results show that,in comparison with the traditional prediction methods,the method has higher prediction accuracy for short-term mutation type of harsh wind conditions,and has certain application value for improving the flexible scheduling of wind power grid-connection.

关 键 词：经验模态分解 排列熵 模糊熵 风速预测 混合多步分解 支持向量机 萤火虫算法 改进灰狼算法 

分 类 号：TN919-34[电子电信—通信与信息系统] TM614[电子电信—信息与通信工程]