Font Size: a A A

The Optimization Design Of Fuzzy Control Systems Based On Multi-population Genetic Algorithms

Posted on:2012-09-03Degree:MasterType:Thesis
Country:ChinaCandidate:H Y ZhangFull Text:PDF
GTID:2178330338955149Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Genetic algorithm (GA) is a kind of convergent algorithm with self-adaptation, elicitation, colony, probability, and iteration on whole range. Optimizing FC using GA's good search characteristics may obtain better control results. In this paper the improved poly-population genetic algorithms are applied to optimize membership functions and control rules.Firstly, this paper concludes and analyzes the general dual-population genetic algorithm for the basic structure and characteristics, based on this proposes an improved dual-population genetic algorithm (CGDPGA): we introduce the strategy of complex-valued encoding and grouping crossover operator by imitating "Tianji horse" among species, in order to enrich information included in single genes and enhance individual diversity; we use immigration method based on the rate of golden mean between populations to speed up the convergence and enhance its optimization ability, so as to avoid falling into the predicament of "early-maturing". Chooses some classic functions to test the improved algorithm, experimental results show that: compared with the general dual-population genetic algorithm, to some extent, the accuracy of the optimal solution and the search for optimal solutions are certain to more advantage in the improved algorithm. Then this algorithm is used to optimize membership functions of FC, scaling factor and quantitative factors for realizing full optimization of FC. Matlab simulation result shows that the proposed method has stronger evolutionary capacity compared with the traditional dual population genetic algorithm, the effect of the control for FC optimized is good.Based on dual-population algorithm, this paper further proposes graded three population genetic algorithm (GTPGA): by imitating the food chain of creature living in natural ecology environment to create an evolutionary chain, the population is divided into sub-populations of three nutrition levels, as follows: low population, sub-optimal population, senior population, three population of different sizes parallel evolve using different kinds of methods from low to high level, and carries out individuals transfer along an evolutionary chain from low-level mode to advanced mode every generation. In addition, the shrinking factor is applied to membership functions, it can reduce encoding size and increase search speed of parameter optimization. Membership functions and fuzzy rule of FC, scaling factor and quantitative factors optimization are optimized by the improved genetic algorithm in order to complete FC optimizing. The comparison experiments shows that FC using this algorithm has small exceeding value, well response characteristic, good stability, and has better performance in disturbance suppressing, strong robustness, high adaptability, the effect is satisfactory. Finally, FC fused this algorithm is applied to an inverted pendulum simulation test and hardware system, the experimental results show that it can stable single inverted pendulum system successfully and high real time control ability.
Keywords/Search Tags:genetic algorithm, dual-population, fuzzy control system, complex-valued encoding, shrinking factor
PDF Full Text Request
Related items