Research And Application Of Interval Swarm Optimization Algorithm

Interval optimization algorithm uses interval variables instead of point variables, which can involve all of the available response function curves in the acceptable range. In practical control, we can choose one of them to achieve the control function. Interval algorithm can realize the global optimization, so it’s suitable to solve the non-linear or uncertainty problems.The paper improved the existing interval swarm algorithm and made a study of multi-objective interval swarm optimization based on non-dominated sorting method. Meanwhile, according to the characteristics of glutamic acid fermentation process, the paper modeled it using dynamic neural network. Based on it, the optimization of the glutamic acid fermentation process’s single objective and multi-objectives had been carried out. The paper includes the following aspects:Firstly, the paper made a summary about the interval algorithm and classification about the domestic and international’s study on interval algorithm. The basic concept and the basic ideas about the interval algorithm had been introduced in detail in this paper. Meanwhile, the traditional interval optimization algorithm-interval bisection had been introduced and examples had been simulated in the paper.Secondly, the paper improved the single target interval swarm algorithm respectively on the interval size comparison method, interval swarm iterative formula and the convergence of decision variables. The simulation examples demonstrated the effectiveness of the algorithm.Thirdly, the paper introduced the NSGA-Ⅱ paradigm to the interval swarm optimization of the multi-objective problems. The dominated relationship of the objective functions’ solution had been ranked though the reliability, the advantages had been selected by the crowding distance of the evolution individuals and the new generation had been predicted by the swarm algorithm. After that, the new algorithm had been introduced to solve the problems of the constrained nonlinear multi-objective instances and the results demonstrated the algorithm’s effectiveness.At last, according to the characteristic of the glutamic acid fermentation process, the paper established a interval dynamic neural network. The fermentation process is divided into two stages:the single objective optimization before the 20th hour and the multi-objectives optimization between the 20th and 30th hours. The single objective optimization was carried out respectively by the rolling method and the global method and the multi-objectives optimization were carried out only by the global method. The simulation results proved that the improved algorithm can effectively meet the needs of the fermentation process optimization.The paper also presented the next improving direction based on the above research results.