Efficient particle swarm optimization method based on RBF proxy model

Abstract

The invention discloses an efficient particle swarm optimization method based on an RBF proxy model. The method comprises the following steps: initializing a population, sampling in a search space to generate NP individuals as an initial population, evaluating the individuals, and adding the individuals into a sample data set; pre-selection: constructing a global agent model, and performing pre-selection by using a particle swarm algorithm as an optimizer; performing local search: constructing a local agent model for neighborhoods of population individuals, and selecting by using a particle swarm algorithm as an optimizer to obtain better individuals of local search; performing updating: using a better individual guide population obtained by local search to perform speed updating and position updating, and selecting a part of individuals to update the population and a sample data set after sorting; judging whether a termination condition is met or not; extracting the guiding information through local search, so the convergence capability of the population is ensured under the limited fitness evaluation times, and the optimization efficiency of the particle swarm algorithm is improved; the method has the characteristic of accelerated population convergence speed.

Description

technical field [0001] The invention belongs to the technical field of agent model-assisted optimization in evolutionary calculation, and in particular relates to an efficient particle swarm optimization method based on an RBF agent model. Background technique [0002] In scientific research and engineering practice, real-valued optimization problems widely exist in the process of design and decision-making. With the continuous improvement of technology and product requirements, in many cases, the actual optimization problem has no explicit analytical formula, and the pros and cons of the solution can only be evaluated through physical experiments and numerical simulations. In addition, the fitness evaluation of complex optimization problems in many engineering practices is expensive, known as Computationally Expensive Problems (CEPs). Therefore, in order to reduce the number of real evaluations, the use of surrogate models with low computational costs has become a hot spot...

AI Technical Summary

Problems solved by technology

However, on the one hand, the existing evolutionary algorithm based on the agent model cannot make full use of the agent model to describe the local characteristics of the problem, participate in the operator update link of EA, and guide the optimization direction of the evolutionary algorithm population, and the search efficiency is low

On the other hand, current algorithms usually use a single agent model as an approximate model, ignoring the errors brought by the agent model construction technology and the influence of the distribution characteristics of the population in the evolutionary algorithm on the agent model.

However, how to make full use of the agent model to guide the evolution process, how to accurately combine the distribution characteristics of the population to construct the agent model and optimize the model management have not yet been fully paid attention to and effectively resolved in the existing SAEA.

Images