摘要
为了对风电机组变桨系统的潜在风险进行可靠的动态预测,针对变桨系统部件种类多、系统复杂、故障特征提取困难的问题,文章首先对变桨系统故障点和故障传递过程进行归纳分析,建立故障树;然后将其转化为融合Leaky Noisy-Or节点的动态贝叶斯网络(DBN),保证了模型精度并具备了动态预测能力;最后采用5折交叉验证的方式对模型进行寻优并验证。测试结果表明,该方法在对变桨系统进行风险预测、故障致因分析、风险动态演化过程分析方面准确率较高,可指导变桨系统进行预防性维护,在保证风电机组整体安全方面具有工程应用价值。
In order to make reliable dynamic prediction of the potential risk of pitch system,aiming at the problems of multiple components,complex system and difficult fault feature extraction of pitch system,the fault tree is established through the induction and analysis of its fault point and fault transmission process,and then it is transformed into a dynamic Bayesian network(DBN)integrating Leaky Noisy-or nodes,which ensures the accuracy of the model and has the dynamic prediction ability.The model is optimized and verified by using a 5-fold crossvalidation method.The test results show that this method has high accuracy in risk prediction,fault cause analysis and risk dynamic evolution process analysis of pitch system,and has engineering application value in guiding the preventive maintenance of pitch system and ensuring the overall safety of wind turbine.
作者
冯红岩
朱海娜
邱美艳
冯玉龙
Feng Hongyan;Zhu Haina;Qiu Meiyan;Feng Yulong(Tianjin Sino-German University of Applied Sciences,Tianjin 300350,China;Tianjin Resource Electric Co.,Ltd.,Tianjin 300308,China)
出处
《可再生能源》
CAS
CSCD
北大核心
2024年第4期486-492,共7页
Renewable Energy Resources
基金
天津市科技计划项目技术创新引导专项优秀特派员项目(20YDTPJC01850)。
关键词
变桨系统
动态贝叶斯网络
交叉验证
可靠性评估
pitch system
dynamic Bayesian network
cross validation
reliability evaluation
