水轮机空化现象智能识别的分类模型训练方法  

Classification Model Training Method for Intelligent Recognition of Hydraulic Turbine Cavitation Phenomenon

作　　者：汪刚 王桂虹 骆彦辰 梁权伟 黄曦 吴建平 王智勇 陈梓豪 WANG Gang;WANG Guihong;LUO Yanchen;LIANG Quanwei;HUANG Xi;WU Jianping;WANG Zhiyong;CHEN Zihao(Dongfang Electric Machinery Co.,Ltd.,Dongfang Electric Corporation,Deyang 618000,Sichuan,China)

机构地区：[1]东方电气集团东方电机有限公司,四川德阳618000

出　　处：《水力发电》2023年第7期67-72,77,共7页Water Power

基　　金：国家自然科学基金资助项目(52279088)。

摘　　要：目前国内暂无相关成熟的分类模型训练方法以支持机器自动识别水轮机初生空化现象,针对于此,提出了一种支持向量分类算法(SVCC)用于水轮机空化现象智能识别的分类模型训练,以解决现有技术中分类算法对非线性可分样本数据分类效果不佳的问题。对该分类模型的核函数和超参数选取等环节进行了优化,以更好地适应水轮机空化试验数据的特点。训练好后的分类模型已应用于东方电机有限公司水轮机模型试验台进行水轮机初生空化的识别。实际应用表明,该分类模型能够提高机器对水轮机初生空化现象的识别效率且其最终判别准确率可达80%。At present,there is no relevant and mature classification model training method to support the automatic recognition of nascent cavitation phenomenon of hydraulic turbine in China.In view to this,a support vector calculation(SVCC)algorithm is proposed for the classification model training of intelligent recognition of hydraulic turbine cavitation phenomenon,so as to solve the problem of poor classification effect on nonlinear separable sample data in existing classification algorithm.The kernel function and hyperparameter selection of the classification model are optimized to better adapt to the characteristics of turbine cavitation model test data.The trained classification model has been applied to the turbine model test stand of Dongfang Electric Machine Co.,Ltd.to identify the nascent cavitation of turbine.The actual application shows that the classification model can improve the recognition efficiency of nascent cavitation phenomenon of turbine and its final discrimination accuracy can reach 80%.

关 键 词：水轮机 模型试验 空化现象 分类模型训练方法 最优分类超平面 分类器 升维函数 松弛变量 

分 类 号：TK730[交通运输工程—轮机工程]