An improved method of HDP for optimal control in wiped film molecular distillation systems

Abstract

For the wiped film molecular distillation system which has the characteristics of multiple parameter, large inertia, large time delay, nonlinearity and others, adjustment it's parameters mainly bases on human experience. In order to stabilize the production process, Extreme Learning Machine network is used to model the molecular distillation system and proposes heuristic dynamic programming algorithm based on extreme learning machine. For the purpose of verifying the effectiveness of the algorithm, the algorithm is used to control the wiped film molecular distillation system and the optimizing control results show that the heuristic dynamic programming algorithm has good control effect and improves the stability of the molecular distillation process. At the same time, the convergence rate of Heuristic Dynamic Programming based on the Extreme Learning Machine is faster than the Heuristic Dynamic Programming based on the BP network by analyzing experiment result.