Dynamic Optimization Method Based On Intelligent Computing And Its Application

Chemical processes strictly are the dynamic processes, where the state variables vary with the time and the position. Dynamic processes are often described by the dynamic model of a group of differential equations. Dynamic optimization is to make a performance index optimal by controlling operational variables in dynamic model. For complex dynamic optimization problems, it is difficult to obtain analytic solution. The general method, including steepest descent method, conjugate gradient method and dynamic programming, is to search piecewise functions as the approximation with numerical methods on the base of Bellman’s principle of optimality or the Hamiltonian function. In recent years, the novel intelligent algorithms based on bionics become more and more popular for solving dynamic optimization problems. The intelligent algorithms are applicable especially for the cases that the gradients are not available. This thesis aims at solving dynamic optimization problems using control vector parameterization (CVP) method based on intelligent evolutionary algorithm. The main contributions can be summarized as follows:1. For solving dynamic optimization problems with single control variable and no constraint, a hybrid genetic algorithm (HGA) is proposed. Genetic algorithm (GA) has been proved to be a feasible method when the gradient is difficult to calculate. Its advantage is that the control profiles at all time stages are optimized simultaneously. However, in the later period of evolution, GA has a very slow convergence speed. Simplex method (SM) is used to perform the local search in the neighborhoods of the optimal solution. By using SM, the ideal searching direction of global optimal solution could be found as soon as possible and the convergence speed of the algorithm is improved.In HGA, some improvements like employing adaptive crossover and Gaussian mutation operators are also presented. The efficiency of the proposed algorithm is demonstrated with solving several dynamic optimization problems. An approach that combines HGA and CVP is also proposed to solve the dynamic optimization problems of chemical processes using numerical methods. In the new CVP method, control variables are approximated with polynomials based on state variables and time in the entire time interval. The iterative method, which reduces redundant expense and improves computing efficiency, is used with HGA to reduce the width of the search region. Results demonstrated the feasibility and robustness of the proposed methods. 2. For solving dynamic optimization more accuratly, the piecewise linear function parameterization method is proposed, including equal time interval distribution and changeable time interval distribution. A novel hybrid evolutionary algorithm (HEA) by combining GA and particle swarm optimization (PSO) is proposed. Based on the characteristics of dynamic optimization problems, the concept of "search region reduction" is integrated into the HEA to improve the convergence rate. The results of the case studies demonstrate the feasibility and efficiency of the proposed methods. In order to fairly evaluate their advantages, a careful and critical comparison with several other direct approaches is provided. The results indicate that the proposed approach presents the best compromise between robustness and efficiency.3. For solving dynamic optimization problems with multi control variables and with constraints, the methods for dealing with multi variables and constraints are proposed. Multi control variables dynamic optimization problems and constrained dynamic optimization problems are more difficult than the simple dynamic optimization problems. A new method that embeds the information of infeasible chromosomes into the evaluation function is introduced in this study to solve dynamic optimization problems with or without constraint. The results indicate that the proposed approaches can effectively solve the dynamic optimization problems with multi control variables and with constraints.4. Focusing on ethylene oxide (EO) hydration reactor industrial equipment, the reaction mechanism model is established. Based on the principle of material balance, energy balance and kinetics of the reactions of ethylene oxide with water, partial least squares regression (PLSR) is used in the model to establish a corresponding relationship between the reaction rate constant and the reaction temperature. With kinetic parameters correction by using field data, the results are more tallies with the actual operation. According to the established model, influences of water/EO molar ratio and inlet temperature on product quality, outlet temperature and energy consumption are analyzed. The results show that the model can preferably reflect the performance of EO hydration reactor and have certain guidance functions to the further advanced control strategies. At last, the dynamic optimization algorithm is used to slove the temperature distribution problem of the EO hydration reactor. The optimal concentration distribution of the product and the optimal temperature distribution of the reactor are obtained.