锅炉再热蒸汽温度具有强非线性和大滞后特性，为解决其软测量中常规单一模型预测精度不足的问题，提出一种基于自适应提升算法(Adaptive Boosting，Adaboost)和极限梯度提升(eXtreme Gradient Boosting，XGBoost)的数据驱动建模方法。利用变分模态分解对数据进行深度解析，通过XGBoost建立预测模型，将其作为弱学习器，经过Adaboost算法的不断迭代，配合误差动态修正(Error Dynamic Correction，EDC)构造出一种再热蒸汽温度动态数据驱动模型。结果表明：模型最终的精度评价指标均方根误差(Root Mean Square Error，RMSE)和平均绝对误差(Mean Absolute Error，MAE)分别为1.733和1.387，与常规的支持向量回归、随机森林及XGBoost模型相比表现更为优异，可以实现再热蒸汽温度的快速准确预测，为后续再热汽温优化控制问题提供有效的参考。

The boiler reheat steam temperature has high nonlinearity and great hysteresis characteristics. In order to improve the prediction precision of conventional single model in soft measurement, a datadriven modeling method based on adaptive boosting (Adaboost) and eXtreme Gradient Boosting (XGBoost) was proposed. The variational mode decomposition (VMD) was utilized to analyze the relevant data deeply, and then, XGBoost was used to build a prediction model as weak learner, based on which a dynamic datadriven model of reheat steam temperature was constructed with error dynamic correction (EDC) after continuous iteration of Adaboost algorithm. The results show that the final root mean square error (RMSE) and mean absolute error (MAE) values as the precision indicators of the model are 1.733 and 1.387 respectively, compared with conventional support vector regression (SVR), random forests (RF) and XGBoost models, indicating an improved performance. Thus, the proposed model can realize a rapid and precise prediction of reheat steam temperature, and provide an effective reference for the subsequent control and optimization of reheat steam temperature.

TP181