Simulated annealing-Particle Swarm Optimization (SA-PSO): Particle distribution study and application in Neural Wiener-based NMPC

Good nonlinear optimization plays a vital role in advanced controller such as nonlinear model predictive control (NMPC). Particle swarm optimization (PSO) is one of nonlinear optimization which has a good potential to be implemented in the NMPC. Even though PSO can determine global optimum value, it is less superior in determining a local minimum value. Meanwhile, another optimizer known as simulated annealing (SA), has an opposite capability of PSO in determining the local and global values. Consequently, in this work, the SA and PSO optimizers have been combined to form SA-PSO which expected to cater both local and optimum point searching. From the particle distribution study, the result shows that the SA-PSO has better particle distribution than the original PSO. The proposed SA-PSO optimizer has also successfully applied in NMPC to control temperatures in the MTBE reactive distillation. The set point tracking test show that the NMPC using SA-PSO have good performance with small amount of overshoot, low settling time and small amount of error.