基于数据驱动和ELM的水电机组振动区划分

Data-driven and ELM-based Vibration Region Partition for the Hydroelectric Generating Sets

传统的水电机组振动区划分方法是通过变负荷试验,采用国标规定的限值划分振动区,没有考虑机组型号和运行环境导致的阈值差异。将水电机组状态监测系统的稳定运行数据应用到振动区划分,提出了基于数据驱动和极限学习机(ELM)的水电机组振动区划分模型,筛选能够表征机组稳定性状态的测点,根据全工况稳定性状态样本数据对振动区划分模型进行分类训练,进而实现机组振动区高效、精确划分。实际应用表明,基于数据驱动和ELM的水电机组振动区划分模型,具有快速、高效获取在线状态监测系统有效数据的优点,其振动区划分结果与传统方法相比,覆盖工况区间广,可靠性高。

The traditional method of vibration division for hydropower units adopts the limit value stipulated by national standard through vari⁃able load test,without considering the difference of threshold value caused by unit type and operation environment.Based on the data-driven and limit learning machine(ELM),a model for the vibration division of hydroelectric generating units is proposed,in which the stable opera⁃tion data of hydroelectric generating units are applied to the vibration division,based on the data of steady state under all working condi⁃tions,the model of vibration region division is trained,and then the efficient and accurate division of vibration region is realized.The practi⁃cal application shows that the vibration zoning model based on data-driven and ELM has the advantages of fast and efficient acquisition of ef⁃fective data of on-line condition monitoring system,the wide coverage range and high reliability are of great significance to guide the actual operation of power plants.