NOX Concentration Prediction in Cement Denitrification Process Based on EEMD-MImRMR-BASBP

NOx concentration is an important indicator of the response to ammonia dosage and nitrogen emissions, its accurate prediction allows for efficient rational optimal control dosage. Due large external noise, time lag non-linearity cement denitrification process, it difficult derive mathematical models. Therefore, a new machine learning model, namely EEMD-MImRMR-BASBP, developed. Firstly, Ensemble Empirical Mode Decomposition (EEMD) median-averaged filtering used process data remove noise. In order handle lags, non-smoothness among variables, mutual information (MI) based on entropy principle proposed calculate non-linear system; furthermore, according feature variable selection method Max-Relevance Min-Redundancy (mRMR), factors with strong influence are selected as input variables model in combination results mechanism analysis. Then, EEMD-MImRMR-BASBP predict NOX constructed, which initialization parameters Back Propagation Neural Network (BP) searched by Beetle Antennae Search (BAS) effectively overcome parameter problem traditional BP Finally, was applied real plant Jiang xi Fu ping compared classical BP-based BASBP root means square error (RMSE) mean absolute (MAE) two production lines only 0.2927, 0.3513 0.1795, 0.2383, have better performance current model.