火电机组冷端系统真空预测方法与故障诊断OA
Vacuum Prediction Method and Fault Diagnosis of Cold-end System in Thermal Power Units
火电机组冷端系统运行状况的好坏与整个机组的安全性与经济性紧密相关.然而,由于数据采集、传输的过程中往往会出现一些坏点和断点的情况,运行人员对机组运行状况的判断往往存在主观性问题,包括冷端系统在内,火电机组运行状况的诊断及其绩效评价始终是行业技术攻关的难题.针对冷端系统,基于K均值聚类算法得到真空基准值的区间,由斯皮尔曼(Spearman)相关系数得到与凝汽器真空度相关性较强的若干个参数变量,分别建立长短期记忆(long short term memory,LSTM)网络以及基于最小二乘法的高阶多项式拟合模型.结果表明,基于最小二乘法的高阶多项式模型预测精度达98.96％,较LSTM模型(98.09％)具有更优的预测性能.最后,基于建立的局部数学模型构建冷端系统参数应达值预测框架,通过数据纠偏、绩效评估和故障诊断三位一体的应用,实现真空异常的实时预警与系统优化.利用机组历史运行数据构建多项式形式的参数关系模型,该模型兼具计算效率高、预测精度好和工程适用性强等特点,可为火电机组冷端系统的状态监测、运行优化和智能诊断提供有效解决方案.
The operational status of the cold-end system in thermal power units is closely related to the safety and economic efficiency of the entire unit.However,due to frequent occurrences of data collection and transmission anomalies,such as bad points and breakpoints,operational personnel often face challenges in objectively assessing the operational status of the units.The diagnosis of operational status and performance evaluation of thermal power units,including the cold-end system,remains a significant technical challenge in the industry.This study focuses on the cold-end system and determines the interval of the vacuum reference value using the K means clustering algorithm.Based on the Spearman correlation coefficient,several parameter variables with strong correlations to the condenser vacuum are identified.Subsequently,two models are established including a long short-term memory(LSTM)neural network and a high-order polynomial fitting model based on the least squares method.Results indicate that the prediction accuracy of the high-order polynomial model based on the least squares method reaches 98.96％,outperforming the LSTM model(98.09％).Finally,leveraging the developed local mathematical model,a predictive framework for estimating cold-end system parameters is constructed.Through the integration of data correction,performance evaluation and fault diagnosis,real-time early warnings and system optimizations for vacuum anomalies are achieved.Notably,a polynomial-form parameter relationship model is constructed using historical operational data from the unit.This model exhibits high computational efficiency,excellent prediction accuracy and strong engineering applicability,offering an effective solution for condition monitoring,operational optimization and intelligent diagnosis of the cold-end system in thermal power units.
曾丽君;李育昆;李建强;刘若飞;杜康康
国能数智科技开发(北京)有限公司,北京 100011国能数智科技开发(北京)有限公司,北京 100011华北电力大学能源动力与机械工程学院,河北保定 071003||河北省低碳高效发电技术重点实验室,河北保定 071003国能数智科技开发(北京)有限公司,北京 100011华北电力大学能源动力与机械工程学院,河北保定 071003||河北省低碳高效发电技术重点实验室,河北保定 071003
信息技术与安全科学
冷端系统优化最佳真空K均值聚类算法LSTM神经网络最小二乘法故障诊断
cold-end system optimizationoptimal vacuumK means clustering algorithmLSTM neural networkleast square methodfault diagnosis
《广东电力》 2026 (6)
40-49,10
国家自然科学基金项目(52406089)
评论