Robust Weighted Heterogeneous Feature Ensemble Prediction Model of Temperature in Municipal Solid Waste Incineration Process

Aiming at the challenging problems of the deficient accuracy and generalization ability of the furnace temperature prediction model when the municipal solid waste (MSW) incineration process data has abnormal values and high dimensionality of feature variables, a robust weighted heterogeneous feature ensemble modeling method is proposed to establish the furnace temperature prediction model of the municipal solid waste incineration process. Firstly, the high dimensional feature variables are divided into heterogeneous feature sets according to the incineration process mechanism, and the contribution of each heterogeneous feature set is evaluated by the mutual information and correlation coefficient. Secondly, a robust stochastic configuration network (SCN) with the t mixture distribution is employed to construct base models, and penalty weights of training samples are determined at the same time. Finally, the robust weighted negative correlation learning (NCL) strategy is used to realize the synchronous training of base models. Comparative experiments are carried out using the historical furnace temperature data of a municipal solid waste incineration plant in China. The results show that the furnace temperature prediction model established by the proposed method performs more favourably in accuracy and generalization. © 2024 Science Press. All rights reserved.