Rescarch On P Systems Based Optimiza-tion Algorithms And Applications

P systems (also known as membrane computing) are presented originally as a distributed and parallel computational model that is abstracted from the structure and functioning of living cells, tissues or organs. Based on the fruits of existing membrane computing, inspired by the cells’structure and other behaviors of life, this dissertation discusses series of methodologies and their applications in addressing the process modeling and parameter estimation problems which possess great nonlinearity and deception. The contributions and innovations of this dissertation are as follows:(1) A P systems based optimization algorithm with nested membrane structure is proposed. In this algorithm, the nested membrane and new rules such as autophagy rule, adaptive mutation rule and partial migration rule are designed to help algorithm to adjust its population distribution with procedure, avoid premature and to improve the algorithm’s global convergence performance. Studies on some typical unconstrained benchmark functions indicate that the proposed algorithm possess great reliability, fast convergence speed and searching accuracy. The algorithm is also used to solve the parameter estimation problems of proton exchange membrane fuel cell modeling, and the results show that those models established by the proposed method can reflect the nonlinear external dynamic property of the practical systems.(2) A P systems based hybrid optimization algorithm with expansion and contraction mechanism (PHOA) is proposed. In PHOA, the expansion of the search space makes the proposed algorithm realize exploration and the dynamic contraction of search space makes it realize exploitation. The real-coded quantum update rules make full use of the current best solution and guide the rest mutating toward the potential global optimum, which accelerate the algorithm’s convergence speed. Studies on some benchmark functions show that the proposed scheme outperforms DNA based genetic algorithm (DNA-GA) and standard genetic algorithm (SGA) both in search efficiency and accuracy. With the parameters obtained by this algorithm, the discrete transfer function matrix model of FCCU reactor-regenerator in an oil refinery is established. The results validate the effectiveness of the proposed approach.(3) DNA molecular operation based P systems optimization algorithm (DNA-PSOA) is proposed. Combining with the nested membrane structure and the traditional communication rule and selection rule, the proposed DNA rearrangement and recombination rules are designed to improve the distribution of object set and strengthened the algorithm’s abilities of anti-premature. The dynamic complementary mutation mechanism and variable gain in site-directed mutagenesis rule make this algorithm realize dynamic shift from exploration and exploitation in the whole processes. Numerical experiments validate the effectiveness of the proposed algorithm. Then the algorithm is applied to design the llR digital filters that possess multimodal error surface property. The results and comparison with others show that the proposed method can obtain better filters with close-to-ideal amplitude responses.(4) A box-density-distribution-strategy based multi-objective P systems optimization algorithm (Box-MPSOA) is proposed. In the framework of P systems, the dynamic membrane structure is adopted to changing in the number of targets in multi-objective optimization problems. The rules such as selecting, communication, crossover and adaptive mutation used in the Box-MPSOA make the algorithm converge faster. Based on Pareto domination, box-density-distributing-strategy redistribute the individuals which are located in the high-density box, so that the distribution of final Pareto solution set will be more uniform on the nondominated solution set front. Benchmark multi-objective optimization problems are tested to validate the proposed algorithm. Simulation results show that the final non-dominated solution set obtained by Box-MPSOA reaches or is close to the true Pareto front with good diversity. It can be concluded that the proposed algorithm is capable to solve the multi-objective optimization problems efficiently.