抄録
Model parameter identification is essential in the area of modeling of chemical processes. These problems become very difficult to solve due to the fact that the contours are basically multi-optima and even rugged in the solution area. In this work, a novel information guided genetic algorithm is developed to solve these model identification problems. The information theory derived by Shannon is implemented to reduce the mutation steps and hence increase the efficiency of this algorithm. The total number of function evaluations to reach the optimum is drastically reduced, and hence solution of complex problems, such as distillation column model identification is made possible.
Several benchmark problems are solved and compared to well-known global optimization approaches. The parameter identification problems of a linear system and a computation intensive distillation system are solved to show the superiority of this approach. Note that this distillation problem is for the first time solved using parallel processing algorithm in this work. This approach should be interesting for many researchers and industries that implement complex chemical engineering systems.
