Neural Network Modeling and Predictive Control of Low Oxygen Combustion System in Petrochemical Heating Furnace

The nonlinear and strongly coupled characteristics of the heating furnace system are addressed in this paper through the proposal of a modeling method for the heating furnace combustion system. This method is based on the attention mechanism deep separable convolutional LSTM network and is combined with predictive control to optimize the low-oxygen combustion control process of the petrochemical heating furnace, achieving energy savings and emissions reductions. The predictive control algorithm developed in this study collects data through a sliding time window at each sampling moment to accurately predict the future state of the system. Nonlinear programming is then used to solve the optimal control quantity, reducing the impact of external interference on control decisions and system state. Simulation results demonstrate that the established neural network predictive control (NNPC) algorithm accurately predicts the system state and provides the optimal control quantity at the current moment compared to typical model predictive control algorithms.