(197g) Extended SBX-RCGA Neural Network-Based Multi-Model Predictive Control for Wastewater Neutralization Process

Wastewater neutralization process is considered to be a challenging control problem due to its inherent time varying nonlinearity and variable time delay characteristics. Although many control algorithms have been used for pH neutralization process, most of them employ only one controller, which uses only acid or base to control such a highly nonlinear and time-varying process. In this work a multiple-controller algorithm which includes, an acid, a base, and a ?transient controller? is proposed to control the pH neutralization process more effectively and efficiently. A modified simulated binary crossover (SBX) method was incorporated a in real code genetic algorithm (RCGA) to find the hidden layer network weights of a Neural Network efficiently, and with high speed; this proposed finds the near-optimal solution. A Micro-Genetic Algorithm was applied to locally fine-tune the solution to reach the real optimum. Three nonlinear MPC controllers were then designed to deal with the three different operating regions partitioned by fuzzy c-means clustering. Simulation results demonstrate the effectiveness of the proposed method. The following figures illustrate some of the results.