Optimizing and Modeling of Comprehensive Performance in Control Systems

摘要

In control systems, it is usually short of quantified comprehensive performance index which results in the global optimization and parameter tuning problems. Aimed at these puzzles, in the paper, the comprehensive performance modeling and optimizing problems are studied based on the combined approach EK (Expert Knowledge) &NN (Neural Network) and GA (Genetic Algorithm). Since skillful expert could exactly evaluate the control quality, EK was applied to quantify the comprehensive performance in PID control system. On the basis, the function of comprehensive performance related to adjustable parameters is modeled with NN. When optimizing the NN function of comprehensive performance with GA, the globally optimal parameters were effectively searched. The experimental results reveal the validity of this modeling and optimizing approach. Therefore, with the combined approach EK & NN and GA, the comprehensive performance in control systems can not only be availably modeled and also be globally optimized.