Efficient and accurate fault diagnosis of the coal pulverizing system is of paramount importance for guaranteeing safe and reliable operation of thermal power plant, but it is a non-trivial task due to varying severities of a fault and wide range of operation of thermal power generator units. This paper presented a novel diagnosis solution considering different fault severity degrees based on combination of Least-Square support vector machines(LS-SVM)and optimal zoom factor search. The variation rules of feature variables of the same fault with different severities are used in the zoom factor search process. The symptom calculation is jointly used for on-line fault detect, aiming to achieve a timely and reliable performance of fault diagnosis. The numerical results obtained from a realistic 660-MW power plant verify the effectiveness of proposed solution for real-time diagnosis of a multi-degree fault, in terms of diagnosis correctness and efficiency.%针对火电厂磨煤机故障诊断问题,通过采用系统运行特性的动态数学建模的方法来逼近真实磨煤机系统,并利用真实运行数据对模型参数进行辨识.结合典型制粉系统故障类型,研究其不同故障程度下特征参数的变化规律来完善故障知识库,从而有利于对磨煤机运行故障的快速和精确诊断.同时,根据不同故障严重程度的特征分析,针对每类故障的故障样本数据进行离线训练,提出了一种通过故障征兆的计算和缩放因子搜索实现故障的在线辨识方法,获取更为快速和可靠的故障诊断结果.最终,结合山西河曲电厂660-MW 机组的双进双出BBD3854型磨煤机验证了将所提出的故障诊断方法的有效性.
展开▼
