Particle Swarm Optimization With Diversity Distribution Parameters And Gravity Learning And Its Application

Optimization, together with modelling and simulation, has been an integrated part of modern industry practice and science research. The main purpose for any design problem is to find a set of good, feasible, ideally the global best, solutions to a given problem, measuring against a set of design criteria such as the minimum costs, high performance, sustainability, recyclability and energy efficiency. Nowadays, design requirements and constraints are more stringent, real-world applications often involve highly complex constraints and uncertainties with often multiple, conflicting objectives. So to find a suitable algorithm with intelligent features for large-scale problems has become the main research objectives. Particle swarm optimization is a new heuristic algorithm based on swarm intelligence, has widely spread in the field of science and engineering because of its characteristics of fast convergence speed, easy to understand and simulation.Considering that the parameters of particle swarm optimization algorithm play an important role in its global searching ability, diversity distribution parameters of particle swarm optimization, which is named as DDPPSO, is proposed. In DDPPSO, particle population and parameter population generated during the initialization respectively. According to the optimization performance of each particle, the weight value of its parameters is determined and then the weighted average values of parameter population is calculated. To achieve diversity distribution parameters, the normal distribution with the weighted average values and adaptive variance is employed to produce the next generation optimal parameter population for the particle swarm. At the same time, good control parameters can get good particles through evolution, which can realize coevolutionary so as to improve the adaptive performance of the algorithm.To improve the global convergence of particle swarm optimization and fully use particle evolution information, a new optimization algorithm is proposed. First of all, the group gravity learning particle, generated by using particle distribution information, is introduced into the velocity updating. The group gravity learning particle, which reflects the individual’s weighted location based on fitness value in the search space, provides distribution information differing from the optimal particle. Secondly, the global cognitive factor, social factor and gravity factor are possessed in the algorithm, and each particle has its inertia weight. Moreover, the individual parameter and the population parameters are optimized through the adaptive feedback regulation utilizing the particle evolution information, so that the algorithm has a good balance between global search and local search. When applied to dynamic optimization of ethanol fermentation process, the results greatly improved compared with the literature values.