当前火电厂中的汽动给水泵普遍采用单预警模型,有误报、漏报的风险,而给水泵的故障会给汽水系统带来严重的影响。对此提出了一种双重故障预警的健康度评价模型。首先利用斯皮尔曼(Spearman)相关性分析筛选相关的特征参数作为输入,并通过双树复小波包变换(DTCWPT)对给水泵的振动数据进行降噪处理,提升输入数据的质量。其次,对iTransformer模型加以改进,建立了iTransformer-KAN故障预警模型,对给水泵驱动端轴承温度进行预测。最后,基于朗肯循环确定了给水泵的核心故障监测参数,并基于时间动态规整(DTW)映射出期望给水流量与实际出口流量的偏离度。在计算出预测残差值和实时偏离度值的基础上,基于核密度估计法(KDE)设定阈值,实现故障预警的功能；同时,为方便运行人员做出决策,构建健康指数指标,直观地展现出给水泵的运行状态和机组状态等级的变化。实验结果表明,所提出的方法在实际数据上的表现优异,预警效果优于其他对比模型,能够有效评估给水泵的运行状态,保障机组的安全运行。

The current steam driven feed water pump in thermal power plants generally adopts a single warning model, which carries the risk of false alarms and missed alarms, and the failure of the feed water pump can have a serious impact on the steam water system. A dual fault warning health evaluation model has been proposed for this purpose. Firstly, Spearman correlation analysis is used to screen relevant feature parameters as inputs, and the vibration data of the feedwater pump is denoised through the Double Tree Complex Wavelet Packet Transform (DTCWPT) to improve the quality of the input data. Secondly, the iTransformer model was improved and an iTransformer KAN fault warning model was established to predict the temperature of the drive end bearing of the feedwater pump. Finally, the core fault monitoring parameters of the feedwater pump were determined based on the Rankine cycle, and the deviation between the expected feedwater flow rate and the actual outlet flow rate was mapped based on Time Dynamic Warping (DTW). Based on the calculation of predicted residual values and real-time deviation values, a threshold is set using kernel density estimation (KDE) to achieve fault warning function; At the same time, in order to facilitate decision-making by operators, a health index indicator is constructed to visually display the changes in the operating status of the feedwater pump and the level of unit status. The experimental results show that the proposed method performs well on actual data, with better warning effects than other comparative models. It can effectively evaluate the operating status of the feedwater pump and ensure the safe operation of the unit.

TK221

1.中国华能集团有限公司2022年度科技项目(HNKJ22-HF22) 2.上海市“科技创新行动计划”地方院校能力建设专项项目(19020500700)