Intelligent predictive control of large-scale solvent-based CO2 capture plant using artificial neural network and particle swarm optimization

Abstract

This paper develops an intelligent predictive controller (IPC) for a large-scale solvent-based post-combustion CO2 capture (PCC) process. An artificial neural network (NN) model is trained to represent the dynamics of the PCC process based on an in-depth behavior investigation of the process under different operating conditions. The resulting NN model can portray the PCC characteristics very well in terms of dynamic trend, response time and steady-state gain. An intelligent predictive controller is thus developed based on the NN model to track the desired CO2 capture level and maintain the given re-boiler temperature, in which the particle swarm optimization (PSO) algorithm is applied to find the best future control sequence for the PCC process. A warm start scheme is proposed in the IPC to improve the quality of initial swarm in the PSO. Dynamic simulations to change CO2 capture level set-point and flue gas flow rate are carried out on the PCC process. The results show that the IPC can adjust CO2 capture level fast and significantly reduce the fluctuations in re-boiler temperature. It is concluded that the proposed IPC is helpful for flexible operation of the solvent-based PCC process.