摘要
风力发电机组运行过程中出现故障的样本相对较少,而正常运行的样本数量通常较多,导致训练模型时样本分布不平衡。为此,提出了基于边缘AI计算架构的风力发电机组传动链故障预警算法。归一化处理样本数据,基于卷积神经网络,提取原始输入数据的关键特征;采用支持向量机算法完成故障预警机制的建立,引入非线性核函数实行最优计算,实现故障预警数据属性分类;基于边缘AI计算架构,根据风力发电机组传动链数据特点,实现风力发电机组传动链故障预警。实验结果表明,所提方法可以预警出齿轮箱、主轴承出现的故障,主动预警准确率高于80%,故障预警曲线与实际曲线拟合度更高,故障预警效果较好。
There are relatively few samples that failure in the operation of wind turbine,while the number of samples in normal operation is usually more,which leads to the unbalanced sample distribution when training the model.This paper,the fault warning algorithm of wind turbine based on edge AI was proposed.Based on the sample data,the key features of the original input data;the support vector machine algorithm was used to complete the establishment of the fault warning mechanism,the nonlinear kernel function was introduced to implement the optimal calculation and realize the fault warning data attribute classification;based on the edge AI computing architecture,according to the characteristics of the wind turbine transmission chain.The experimental results show that the proposed method can warn the failure of the gearbox and the main bearing,the active warning accuracy is higher than 80%,the fault warning curve and the actual curve fit higher,and the fault warning effect is better.
作者
朱振军
曾佳佳
邓睿
刘旭东
ZHU Zhenjun;ZENG Jiajia;DENG Rui;LIU Xudong(Wuling Power Corporation LTD.,Changsha 410029,China)
出处
《微电机》
2024年第6期51-54,68,共5页
Micromotors
关键词
边缘AI计算架构
风力发电机组
传动链
故障预警
edge AI computing architecture
wind turbine
transmission chain
fault warning
