NO X Concentration Prediction at Secondary Combustion Chamber Outlet in Small-Scale Domestic Waste Incineration System

Small-scale municipal solid waste incineration systems (distributed waste treatment technology) have great potential. Yet, secondary combustion is highly sensitive to pyrolysis furnace outlet parameter fluctuations, causing significant NOX fluctuations at the secondary combustion chamber outlet─complicating monitoring, impairing stability, limiting emission performance, and hindering application. To address these challenges, this study proposes a soft sensing method using a Least Squares Support Vector Machine (MIC-ISSA-LSSVM) optimized with an Improved Sparrow Search Algorithm for predicting NOX concentration. Key variables influencing NOX concentration were identified through maximum mutual information (MIC) dimensionality reduction and Spearman correlation coefficient screening and were subsequently used to train the ISSA-LSSVM model. Compared with the ISSA-LSSVM model, the proposed MIC-ISSA-LSSVM model achieved an R2 improvement of 1.84%, a RMSE reduction of 17.72%, and a mean absolute error (MAE) reduction of 17.89%. Compared with the MIC-SVM, MIC-LSSVM, MIC-BP, MIC-RBF, MIC–CNN, and MIC-LSTM, the MIC-ISSA-LSSVM model achieved R2 improvements of 0.46%, 0.44%, 13.97%, 8.39%, 8.90%, and 6.10%, respectively; RMSE reductions of 5.51%, 5.36%, 51.37%, 42.64%, 43.52%, and 36.93%, respectively; and MAE reductions of 9.00%, 4.05%, 51.84%, 40.37%,42.90%, and 35.00%, respectively. In summary, the MIC-ISSA-LSSVM model showed strong performance in predicting secondary combustion chamber outlet parameters, offering valuable reference for its design and potential application to other large/medium incinerators and related equipment.