Prediction of NO_x emissions from co-disposal of municipal solid waste and sludge using a GA-LSTM neural network

Accurately predicting NO_x emissions is crucial for effectively controlling pollution during the incineration of municipal solid waste (MSW). This study focuses on the application of genetic algorithm (GA) and long short-term memory (LSTM) neural networks in modeling the relationship between operating parameters and NO_x emissions for an 850 t/d MSW incinerator. After data cleaning, principal component analysis (PCA) was used to eliminate correlations among input variables and GA was applied to optimize the hyperparameters of the LSTM model which was compiled with the Adam optimizer. Lastly, a NO_x emission trend prediction model with practical engineering value was proposed, specifically considering the co-incineration of sludge and waste. The model was thoroughly validated using both actual operational data from the waste incineration process and numerical simulation results. Analysis on prediction performance indicates that even the GA-LSTM model maintains a strong capability for predicting NO_x emissions for MSW incinerator, even when handling large amounts of high-dimensional data.