基于动态矩阵与特征相似度的AAKR风电机组状态监测  

DYNAMIC MATRIX AND FEATURE SIMILARITY-BASED AUTO ASSOCIATIVE KERNEL REGRESSION FOR CONDITION MONITORING OF WIND TURBINE

作　　者：田雯雯 吕丽霞[2] 刘长良[1,3] 刘帅 Tian Wenwen;Lyu Lixia;Liu Changiang;Liu Shuai(School of Control and Computer Engineering,North China Electric Power University,Beijing 102206,China;Department of Automation,North China Electric Power University,Baoding 071003,China;Baoding Key Laboratory of State Detection and Optimization Regulation for Integrated Energy System,Baoding 071003,China)

机构地区：[1]华北电力大学控制与计算机工程学院,北京102206 [2]华北电力大学自动化系,保定071003 [3]保定市综合能源系统状态检测与优化调控重点实验室,保定071003

出　　处：《太阳能学报》2024年第10期536-543,共8页Acta Energiae Solaris Sinica

基　　金：中央高校基本科研业务费专项资金(2020JG006,2023JG005,2023JC010);河北省高等学校科学技术研究项目(CXY2023001)。

摘　　要：针对传统自组织核回归(AAKR)模型所选记忆矩阵冗余度较高、无法根据在线数据实时更新、计算相似度时未考虑特征参数权值不一的问题,提出一种基于动态矩阵与特征相似度的自组织核回归(DM-FS-AAKR)风电机组状态监测方法。首先基于样本间距离对原始数据集去冗余以降低运算复杂度,形成待选数据集;其次基于k-最近邻算法选取最符合当前运行条件的历史数据构建动态矩阵;为克服相似度计算时不良参数的偏差污染,提出一种特征相似度计算方法为不同参数分配相应权值进一步提高预测精度;最后以河北某风电场SCADA数据为例,对机组故障停机前工况进行验证实验。结果表明,相比于传统AAKR模型,所提算法平均绝对误差降低约15.6%,故障预警时能够提前35天实现预警,具有较高精度和实时性。A dynamic matrix and feature similarity-based Auto Associative Kernel Regression(DM-FS-AAKR)for condition monitoring of wind turbine is proposed to address the problems of high redundancy in the memory matrix selected by the traditional auto associative kernel regression model,inability to update in real-time based on online data,and failure to consider inconsistent feature parameter weights when calculating similarity.Firstly,based on the distance between samples,the original dataset is de-redundant to reduce computational complexity and form a dataset to be selected;Secondly,based on the k-nearest neighbor algorithm,the dynamic matrix is constructed by selecting the historical data that best meets the current operating conditions.To overcome the bias pollution of bad parameters during similarity calculation,a feature similarity calculation method is proposed to assign corresponding weights to different parameters to further improve prediction accuracy.Finally,taking the SCADA data of a wind farm in Hebei as an example,simulation verification experiments were conducted on the operating conditions before the unit malfunctions and shuts down.The results show that compared to the traditional AAKR model,the proposed algorithm reduces the average absolute error by about 15.6%,and the fault warning can be achieved 35 days in advance,with high accuracy and real-time performance.

关 键 词：齿轮箱 风电机组 状态监测 自组织核回归 动态矩阵 特征相似度 

分 类 号：TM614[电气工程—电力系统及自动化]