A boiler oxygen content and furnace temperature prediction model based on honey badger algorithm optimized neural network

Abstract

Abstract Aiming at the problem that it is difficult to accurately predict the oxygen content and furnace temperature during the combustion process inside the circulating fluidized bed boiler, this paper proposes a prediction model with a mixture of the convolutional neural network(CNN), bi-directional long-short term memory network(biLSTM), and honey badger algorithm(HBA). First, process variables related to furnace temperature and oxygen content were selected as input variables using actual test data from circulating fluidized bed boilers. Secondly, the parsed input variables are analyzed by CNN through an analytical mechanism to extract the data into smaller details and features, which are used as inputs to the biLSTM layer. The BiLSTM layer selectively memorizes the information of the input temporal data and learns the information dependence of the temporal data sequences, which then solves the temporal problem of the training of the irregular trend of the oxygen content and the furnace temperature. Finally, the HBA is utilized to continuously traverse and search the CNN-biLSTM model to find the optimal parameters of the model. The experimental results show: the CNN-biLSTM neural network model mixed with the HBA is able to accurately predict the oxygen content and furnace temperature. The experimental results show that the CNN-biLSTM neural network model with hybrid HBA is able to accurately predict the oxygen content and furnace temperature, and the average MAPE errors for the oxygen content are HBA-CNN-biLSTM (2.92E-03), CNN (7.96E-02), LSTM (5.13E-02), GRU (4.83E-02), RF (4.96E-02), RBF (8.41E-02), SVM (5.71E-02), RNN (5.53E-02), CNN-LSTM (4.79E-02).