基于温度分析的抽水储能发电机故障模式识别

Fault pattern recognition of generator used in pumped storage system based on temperature analysis

为提高抽水蓄能电站系统运行的稳定性和可靠性,提出了一种通过主站发电机瓦温信息构建发电机故障诊断模型的方法。以温度值作为特征量构建了支持向量机故障诊断模型,结合麻雀算法全局寻优和局部寻优自由切换的优势,将故障识别正确率作为迭代更新的目标函数,对支持向量机的惩罚因子以及核函数半径的参数组合进行优化。根据麻雀种群位置优化特点,给出了基于麻雀搜索算法进行SVM参数优化的详细步骤流程,并对采集的水泵发电机实测温度数据进行验证和分析。实测数据表明,所提方法可对下导瓦间隙偏小、冷却器容量不足、透平油老化、杂质混入等故障模式进行识别。与若干现行同类方法进行对比分析可知,基于温度分析的发电机故障诊断方法可为抽水储能发电机运行状态监测及主站电机的安全可靠运行提供有益参考。

To improve the stability and reliability of the pumped storage power plant system operation,a method for constructing a fault diagnosis model of the generator based on the generator tile temperature information at the main station was proposed.A support vector machine(SVM)fault diagnosis model was constructed with temperature value as feature quantities.Taking advantage of the free switching between the global optimization and the local optimization of the Sparrow algorithm,the fault identification accuracy was selected as objective function of interative updating,and the parameter combination of penalty factor and the kernel radius was optimized for support vector machine.According to the iterative characteristics of sparrow population location optimization,the detailed algorithm steps of SVM parameter optimization based on sparrow search algorithmwere also given,and the collected measured temperature data of water pump generator were verified and analyzed.The measured data show that,the proposed method is feasible and effective to accomplish the fault pattern recognition among the lower guide bearing segment insufficient cooler capacity,turbine oil ageing,andmixing of impurities.Compared with several current similar methods,fault diagnosis method of generator based on temperature analysis provides a reference for the operation condition monitoring method of the pumped storage generators.