摘要
高海拔、低温作业下的风电机组常伴有机翼结冰现象。针对风机数据纬度高,传统模型无法挖掘数据间时序关系、收敛速度慢、预测精度低等问题,提出一种基于维度融合优化与长短期记忆(LSTM)网络的结冰检测模型。结合特征消减算法筛选建模特征,通过主成分分析(PCA)降低数据耦合性并引入改进的麻雀搜索算法(ISSA)建立长短期记忆网络结冰检测模型。实验验证,维度融合与改进麻雀搜索算法优化的结冰检测模型判决准确率得到较好的改善,平均具有99.85%的判决准确率。
Wing icing is often associated with wind turbines operating at high altitude and low temperature.Aiming at the problems of high latitude fan data,the traditional model is unable to mine the time-series relation between data,slow convergence speed and low prediction accuracy,an icing detection model based on dimension fusion optimization and long short-term memory(LSTM)network is proposed.Combining feature reduction algorithm to screen modeling features,principal component analysis(PCA)is used to reduce data coupling and an improved sparrow search algorithm(SSA)is introduced to establish the icing detection model of long short-term memory network.Experimental results show that the accuracy of ice detection model based on dimension fusion and improved sparrow search algorithm is improved,with an average of 99.85%accuracy.
作者
聂福印
李强
黄秋凤
黄玲琳
NIE Fuyin;LI Qiang;HUANG Qiufeng;HUANG Linglin(School of Information Engineering,Southwest University of Science and Technology,Mianyang 621010,China)
出处
《传感器与微系统》
CSCD
北大核心
2022年第6期118-121,共4页
Transducer and Microsystem Technologies
基金
四川省科技计划资助项目(2019JDTD0019)
国家重点研发计划资助项目(2019YFB1705100)。
关键词
结冰检测
特征消减
主成分分析
麻雀搜索算法
长短期记忆网络
icing detection
feature reduction
principal component analysis(PCA)
sparrow search algorithm(SSA)
long short-term memory(LSTM)network
