基于自适应多目标模糊聚类的多模型软测量

Multi-Model Soft Sensing Modeling Based on Adaptive Multi-Objective Fuzzy Clustering

针对复杂非线性系统单模型软测量存在建模精度低、模型泛化能力差的问题,提出一种采用自适应多目标模糊聚类的多模型高斯过程回归(GPR)软测量建模方法。首先使用自适应多目标聚类方法自动确定聚类个数并得到最优数据子集,避免了聚类个数不易人为给定的问题;然后对各数据子集分别建立GPR子模型,最后采用子模型加权融合方法得到最终的预测结果。使用火电厂历史运行数据建立烟气含氧量软测量模型验证该方法,仿真结果表明,该方法可以提高软测量模型精度,提升模型泛化能力。

Aiming at the problem of low modeling accuracy and generalization ability of single soft sensor modeling for complex nonlinear systems,we proposed a multi-model Gauss Process Regression soft sensor modeling method based on adaptive multi-objective fuzzy clustering in the paper.Firstly,the adaptive multi-objective clustering method was used to automatically determine the number of clusters and get the optimal data subset,which avoids the problem that the number of clusters is not easy to be given.Then,GPR sub-models were established for each data subset,and the final prediction results were obtained with sub-model weighted fusion method.The method was validated by using the historical operation data of thermal power plants to establish a soft sensor model for oxygen content in flue gas.The simulation results show that the method can improve the accuracy and the generalization ability of the soft sensor model.