A Hybrid Modeling Approach for Batch Process Based on LSTM-RNN

Batch process is characterized by a complex reaction mechanism and has nonlinear and time-varying properties, which makes it more difficult to establish a mathematical model of the process. Therefore, this paper proposes a hybrid model for aniline hydrogenation. Firstly, the input and output variables of the LSTM-RNN are determined, and the model is trained using the production process data to obtain the data-driven reactivity model for aniline hydrogenation. Then, the reactivity is fitted using the trained model based on the reaction temperature, concentration and other information measured in real time during the production process, and the results are passed to the mechanistic model. Finally, the validation of the model was verified by comparing the hybrid model prediction results with the industrial data of aniline hydrogenation.