汕头大学学报（自然科学版）

针对凝汽器变工况特性传统获取方法存在不能实时反映凝汽器真实运行情况的问题，在机组历史运行数据的基础上，采用XGBoost算法，选取主蒸汽流量、循环水流量、循环水入口温度、凝汽器清洁系数为特征参数建立凝汽器的压力预测模型并通过控制变量法详细分析各特征参数变化对凝汽器压力的影响，实现凝汽器的变工况特性研究.经某机组的实例验证，结果表明：通过历史运行数据建立的凝汽器压力预测模型，其预测平均绝对百分比误差仅为0.59%，具有较好的预测精度，可用于凝汽器的变工况运行特性分析并为冷端系统的运行优化提供参考.

In view of the problem that the traditional acquisition method of condenser off design condition characteristics can not reflect the real operation of the condenser in real time, the xgboost algorithm based on the historical operation data of the unit is used to select the main steam flow, circulating water flow, circulating water inlet temperature and condenser cleaning coefficient as the characteristics parameter to establish the pressure prediction model of the condenser. The influence of each characteristic parameter change on the condenser pressure is analyzed in detail by the control variable method, so as to realize the research on the off design characteristics of the condenser. The example of a unit shows that the average absolute percentage error of the condenser pressure prediction model based on the historical operation data is only 0.59%, which has good prediction accuracy. It can be used to analyze the off design operation characteristics of the condenser and provide references for the operation optimization of the cold end system.