NO_x emission prediction for gas turbines based on CNN-BAS-BiGRU

To address the difficulty of accurately monitoring NO_x emissions from gas turbines, this paper proposes a NO_x emission prediction model based on a convolutional neural network and a bidirectional gated recurrent unit network optimized using the beetle antennae algorithm (CNN-BAS-BiGRU). First, partial least squares (PLS) analysis is employed to calculate the correlation characteristics among variables, thereby reducing the influence of redundant information on model accuracy. Next, a convolutional neural network (CNN) is used to extract multilevel spatial features from time-series data. The extracted features are then input into a bidirectional gated recurrent unit (BiGRU) network to further capture the underlying temporal dependencies of the data. To address the difficulty of selecting optimal BiGRU parameters, the beetle antennae search (BAS) algorithm is adopted for parameter optimization. Finally, a NO_x emission prediction model is established using data from the UCI Machine Learning Repository. Comparative experiments with conventional machine learning and deep learning models demonstrate that the proposed method achieves superior prediction performance, with prediction accuracy improved by more than 15% .