Research On Membrane Computing Optimization Methods For Control System Design

Combining the standard ideas of membrane computing with the fruits of evolutionary computing , the membrane computing optimization algorithms and the design of control systme are researched. Some test functions are used to test the proposed algorithms, and the simulations experiments are alse made. The main contents of this thesis are as follows:(1) Based on the framework of the standard membrane computing , and inspired by the structure of DNA molecules, a membrane computing optimization algorithm(dsDAN-MC) with double stranded DNA structure is presented. The algorithm adopts the membrane structure with two subsystems, its objects are encoded by double strings, and the evolutionary rules used in the algorithm are rewritting rule, splicing rule, communication rule and inversion rule of RNA computing. In order to compare with other methods, eight benchmark functions are adopted to form the test environment. The experimental results demonstrate the effectiveness of the proposed method, especially in the advantages of searching ability and convergence speed .(2) The dsDAN-MC algorithm is applied for tuning parameters of PID controllers, and compared with Z-N and SGA methods. The results illuminate the advantages of the proposed algorithm. The dsDNA-MC algorithm is also used to get the optimal parameters of the fuzzy-neuron controller.(3) Combining the framework of membrane computing with the fruits of the evolutionary multi-objective optimization algorithm, the membrane computing based multi-objective optimization algorithm(MCMO) is proposed. The algorithm adopts the membrane structure, which divides the population into several subpopulations. The communication rule is applied betweent different subpopulations. In terms of maintaining diversity of the population and protecting elitist, the proposed algorithm adopts the strategy of NSGA-Ⅱ. Comparisons of MCMO with NSGA and VEGA with three benchmark functions show that the optimal solutions of MCMO are the most closed to the true pareto frontier and distribute well. The proposed algorithm is adopted to solve the dynamic multi-objective optimization problem of the PID controller for a time-varying plant, and satisfactory performance is reached.