Study on Prediction of Flue Gas Emission Concentration of Solid Waste Incineration Based on RF-LSTM Hybrid Neural Network
The incineration of solid waste produces toxic and harmful flue gas,and predicting the concentration of flue gas emissions can assist in the efficient adjustment of process parameters for solid waste incineration.Autoencoder(AE),Convolutional Neural Networks(CNN),and Long Short Term Memory(LSTM)networks are three common artificial neural networks,while Random Forest(RF)is a highly flexible machine learning algorithm.Based on RF and LSTM networks,a hybrid neural network model is constructed,combined with operational data from a solid waste incineration power plant in Chengdu city,to predict and analyze nitrogen oxide(NOx)concentration.The results show that the root mean squared error and mean absolute error of the RF-LSTM model are reduced by 38.58％and 46.56％respectively compared to the AE-LSTM model,and by 23.77％and 31.96％respectively compared to the CNN-LSTM model;the coefficient of determination of the RF-LSTM model increases by 22.54％compared to the AE-LSTM model and 16.00％compared to the CNN-LSTM model.When interpolating and filling in gaps in the original samples,the RF-LSTM model with a step size of 3 h has the highest prediction accuracy and can effectively predict NOx emission concentrations.
solid waste incinerationflue gas emission concentrationpredictionhybrid neural network model
万方数据
维普
