为了解决表征锅炉受热面表面健康状态的清洁因子在未来时间段内预测时呈现非平稳问题，以省煤器受热面为例，提出一种结合核极限学习机(Kernel Extreme Learning Machine,KELM)和自适应噪声完备集成经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,CEEMDAN)的清洁因子预测方法。首先，通过CEEMDAN分解算法对省煤器表面清洁因子序列进行分解和降低复杂程度，获得各固有模态函数(Intrinsic Mode Function,IMF)；其次，利用皮尔逊相关性分析确定主蒸汽流量、进出口烟温等9个参数为输入，建立核极限学习机模型对清洁因子的各IMF进行预测；最后，将各IMF预测结果相加获得最终预测结果。结果表明：与基本核极限学习机、支持向量机等预测模型相比，本文模型具有较高的预测精度和较优预测时间，可为基于受热面状态开展的锅炉智慧吹灰应用提供参考。

In order to solve the nonstationary problem when the cleaning factor that characterizes the health state of the boiler heating surface is predicted in the future time period,taking the heating surface of the economizer as an example,this paper proposed a cleaning factor prediction method combining the kernel extreme learning machine (KELM) and complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN).Firstly,CEEMDAN algorithm was used to decompose and reduce the complexity of the surface cleaning factor sequence of the economizer,and the intrinsic modal functions (IMF) were obtained; secondly,Pearson correlation analysis was used to determine nine parameters such as main steam flow,inlet and outlet smoke temperatures,etc.as the input,and the KELM model was established to predict each IMF of cleaning factor; finally,the final forecast result was obtained by summing the IMF forecast results.The result shows that compared with several prediction models such as basic KELM and SVM,the model presented in this paper has higher prediction accuracy and better prediction time,which can provide reference for the application of intelligent soot blowing in boiler based on the state of heating surface.

TK223.3