基于集成学习的凝汽器真空度预测

Prediction of Condenser Vacuum Degree Based on Ensemble Learning

结合外部气象因素与机组工况数据,建立了基于集成学习的火电厂凝汽器真空度预测模型。首先,运用孤立森林(Isolation Forest,iForest)算法完成了数据的异常值识别和清洗。其次,构建Xgboost模型,按照是否考虑温度、气压差分值两种方式,进行预测模型测试。最后,选取山东省内某机组运行数据对两种方案进行了对比分析,结果表明:考虑外部气温、气压差分的真空度预测模型具备更好的表现。

Combined with meteorological factors and unit operating data,a prediction model of condenser vacuum degree in thermal power plant based on ensemble learning was established.Firstly,the outliers of data were identified and cleaned by using isolation forest(iForest)algorithm.Then,the Xgboost model was constructed,and the prediction model was tested according to whether the difference value of temperature and air pressure was considered.Lastly,the two schemes were compared and analyzed based on the operation data of a unit in Shandong province.The results show that the vacuum prediction model considering the difference of external air temperature and air pressure has better performance.