| Genetic algorithm (GAs) is a new software algorithm, a self-organize, self-adaptive and self-optimized strategy, based on the combination of theory of evolution and computer science. Lately, progress and achievements have been made in theory and application of the field, thus making it possible for its application in such fields as computer science, automation, combination optimization, image and signal procession, artificial biology, management science and social science.Systematic discussion and research have been made on reality during applying GAs. A idea of poly-intergrowth was proposed in this study. Key points to help improve the efficiency of the algorithm have been pointed out after analyzing basic algorithm. With the results, we can make improvement in different aspects put forward new strategy, construct improved operator and methods. The study and improvement have been done in order of briefness, cooperation and poly-intergrowth. Finally, it can be used with Neural Network (NN) to recognise algae and forecast water pollution. It has following six main characteristics or achievements.1. By analyzing, comparing and applying the evolution mechanism of GAs, problems and their reason have been pointed out to sketch out the map of six main factors and three dynamic equilibriums. These are key points and have significance in improving GAs efficiency and provide an idea in future research.2. Several improved operators have been put forward based on basic algorithm research under a simple and useful idea starting from rise individual multiplicity, the first important factor of GAs. Of all the operators,α°changable crossover operator is based on gene dislocation to solve balky process often happened in evolution. Mix crossover operator is made to meet different diversity requirements in each phase. In the idea of one-to-many results produced, self-adaptive and conflict mutation operator has been designed by realizing the role of environment in mutation mixed with momentum principle.3. To solve the problem of more-than-one-objective optimization, research on cooperation GAs has been made from the aspect of improving both accessibility and diversity of solution, which is the second important factor of GAs. On the one hand, niche method base on crowding mechanism instead of regular method is used to exert its superiority in eliminating inferior population controlling selecting pressure and keeping global search ability. On the other hand, some new operators are been constructed such as adapt potential well crossover operator, symbiosis balance crossover operator(SYGA), transient self-adaptive operator (TGA) and so on. They all take on ability of priority, adjustment intensity of randomness and self-adaptability in crossover process, so that accurate search in small range and global optimized value in large range are guaranteed.4. Overall efficiency-improving method of GAs, which is a poly-intergrowth genetic algorithm, has been studied to solve complicated optimization problem. Differential selecting method in low dimension, a new niche algorithm, is more complete than any existing crowding mechanism and simpler than sharing mechanism. So this method is more realizable and reliable. A new trial is made in GAs restructuring, the configuration form on poly-intergrowth genetic algorithm (PGA) and ethic genetic algorithm (EGA). In the view of poly-factor and global optimization, the two form and their algorithms make every step of GAs improving by combination of ploy conflict, mix, cooperate and complement between different operators and groups. This helps to form a better algorithm in the whole area. Moreover, poly-intergrowth method has an original advantage in design way. It is a changeable frame method and applicable to different situation with corresponding structure, thus improving convenience and adaptability of GAs application.5. Combined Modeling method on NN and GAs is studied in this paper. A grade optimization algorithm on NN is put forward, together with optimized evolution steps based on improved genetic operators. Some improved GAs in this paper, such as TGA, EGA and PGA are successfully used in NN's optimization. To solve the problem of data lack in Intelligence Recognition, micrograph pretreatment method of sea red tide and blue-green alga in lakes is researched. A quick noise removing algorithm is proposed to provide essential data source for NN recognition. By compare experiment of numerical value and algae, better improved GANN for algae recognition and forecasting is basically selected.6. Tests are made on these improved GA and GANN. Three better improved GANN are selected after testing their efficiency through typical algae grow state forecasting, engineering requirement in the real world. Amount, space and time characteristics of more ten algae are obtained by these GANN, thus completing growth data void of some dominance algae. And the result of the three GANN in algae shoot-up growth experiment is satisfactory.
|
PDF Full Text Request