The Research Of Optimization Method Of Reference Sky Classification Based On Ant Colony Algorithm

In recent years, energy problems become more and more tension. It has greatsignificance to full use of nature light and save the energy. The sky luminancedistribution is a very important factor for the use of natural light. In this paper, we willuse the research result about sky luminance distribution model at home and abroad,base on the observatory data, departure from the meteorological parameters of the skyluminance distribution and use ant colony algorithm to research the classificationmethod of the sky,then obtain the sky luminance distribution model. Finally, providea theoretical basis for lighting design.Ant colony algorithm is a bionic swarm optimization algorithm which havesuperior distributed solving features. In recent years, the algorithm has achieved greatsuccess in solving discrete combinatorial optimization problems and also attractedmuch attention to domestic and foreign researchers. But its discrete nature limited itsapplication in solving the continuous optimization problems. Because solving the skyluminance distribution model parameters is also an continuous space optimizationproblem. So how to use ant colony algorithm to solve the continuous optimizationproblems is a focus of the paper.Ant colony algorithm is also a probabilistic selection algorithm. Ants move isclosely related with the pheromone distribution model.So in order to design an efficientalgorithm, it is necessary to design a suitable pheromone distribution model first. Inthis paper, we propose two continuous ant colony optimization algorithms, the firstone is based on mesh and named DACO, another is based on normal distribution andnamed GACO. DACO has borrowed the idea of basic ant colony algorithm to optimizethe discrete problems, each dimension of the solution space is divided into grids, onevery dimension, the pheromone distributes on the grid points discrete. Ants departurefrom the starting point, depending on the content of the pheromone which distributes onthe grid and choice a grid for every dimension. After N times choices, finally all antsreach the end point. After several classic functions’ test, the experiment results showthat DACO is suitable for solving lower dimensional and relatively simple optimizationproblems. While GACO is quite different from DACO which the pheromone distributescontinuously and follow a normal distribution at each dimension. The ants statetransition rules of GACO is according to sampling of pheromone distribution function. The update of pheromone depends on updating the location of the optimal ant and thewidth value of the pheromone distribution function. In addition, in order to improve thesolving ability of GACO, we use pattern search strategy. The experiments show thatGACO has the ability to solving high-dimensional complex problems.Finally, through the simulation experiment,use the observatory data as theexperimental data and use GACO to analyze and optimize the impact of variouselements of the model of sky brightness. The experiments show that GACO canoptimize this problem very well.In a word, in this paper we study the continuous ant colony algorithms in-depth.What’s more two continuous ant colony algorithms had been proposed. One of thealgorithm GACO was used to solve practical problem. Finally, the conclusions aresummarized.