Research On Evolutionary Modeling And Control For Water Supply System In Controlled Watercurtain Cooling Process

The modeling and optimizing control for the water supply system, a key subsystem of controlled watercurtain cooling system in Steel Plate Plant, Shougang Group, are investigated in this thesis. The water supply system is a nonlinear, tight coupled multivariable system which is difficult to accurately describe with traditional mathematical model. Due to the fact that fuzzy control, compared to conventional controls, is more suitable for those controlled plants that are difficult to model analytically and exhibit coupling features as well as nonlinear characteristics, the fuzzy logic methodology is employed to solve the modeling and optimizing control problems of water supply system. To this end, the author took part in the on-site test of the controlled watercurtain cooling and obtained the measured input-output data for the water supply system. Then, the fuzzy model of water supply system is built based on compositional rule of inference using test data. Simulations show the fuzzy model built can approximate the dynamic response process of the objective system quite well, and this model is then optimized using Genetic Algorithms. Simulations results demonstrate the optimized model has a higher approximation to the objective system than does the previous one. Next, fuzzy controller for the supply system is designed as follows: first, the structural decoupling is performed to formally decompose the 2-input 3-output controller as three 2-input single-output ones, which greatly simplifies the design of the fuzzy controller. Second, rudimentary control rules are identified and then improved by trial-and-error adjustment. Control simulations prove that the designed control system has quick response and high steady-state accuracy. But, for the reason that the design of a fuzzy controller is to a great extent dependent on rule of thumb, it is hard to guarantee the fuzzy controller to be the optimal. So, at last, Genetic Algorithms are capitalized on to optimize the membership functions of the input and output linguistic variables of the fuzzy controller so as to better its performance. Simulations results indicate that the optimized controller based on the evolutionary principle of GAs outperforms the conventional one greatly in overshoot, settling time and mean square error.