基于SCADA和支持向量回归的风机状态监测  

Wind Turbine Condition Monitoring Based on SCADA and Support Vector Regression

作　　者：梁涛 钱思琦 姜文 龚思远 LIANG Tao;QIAN Si-qi;JIANG Wen;GONG Si-yuan(College of Artificial Intelligence and Data Science,Hebei University of Technology,Tianjin 300130,China;Hebei Construction Investment Energy Investment Co.Ltd,Shijiazhuang 050000,China)

机构地区：[1]河北工业大学人工智能与数据科学学院,天津300130 [2]河北建投能源投资股份有限公司,河北石家庄050000

出　　处：《控制工程》2020年第8期1317-1323,共7页Control Engineering of China

基　　金：河北省科技计划项目(19210108D);石家庄科技局重点研发项目(181060481A)。

摘　　要：为充分利用集控中心风机(Supervisory Control and Data Acquisition,SCADA)系统采集的数据,采用智能化的机器学习算法,挖掘集控中心海量数据,提出基于机组运行状态特征参量数据挖掘和支持向量回归算法(Support Vactor Regression,SVR)结合的机组状态监测模型。该模型采用基于灰色关联度算法构建风电机组特征参量,然后建立SVR数据模型,模型以机组功率、叶轮转速、桨距角为输出向量,特征参量为模型的输入向量,采用遗传算法结合交叉验证方法对SVR模型参数寻优,并对距离阈值进行分析。最后,将模型应用于某实际风场,验证了该模型的可行性和有效性。To make full use of the data of SCADA systems in a wind farm centralized control center,an intelligent machine learning algorithm is adopted for mining the massive amounts of data from the centralized control center.A state monitoring model is set based on the combination the SVR(support vector regression)algorithm and the characteristic parameter data mining of the unit operation state.In this model,the characteristic parameters of wind turbine are constructed based on grey correlation degree algorithm,and then the SVR model is established for the data model.The turbine power,impeller speed,and pitch angle are used as the output vectors,and the characteristic parameters are the input vectors of the model.Genetic algorithm combined with cross validation method is used to optimize the parameters of SVR model,and the distance threshold is analyzed.Finally,the model is applied to an actual wind farm,and the feasibility and effectiveness of the model are verified.

关 键 词：SCADA 风机 灰色关联度 SVR 状态监测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]