Novel Niching Particle Swarm Optimization Algorithms For Complex Optimization Problems

Nowadays, in the real society, a large number of practical problems can be treated as optimization problem. The problem which is needed to be optimized is always complex. The complexity mainly consists of multi-modal, high dimension, multi-objective or dynamic, etc. The traditional evolutionary algorithm is limited by its own mechanism and single structure which results in low convergence precision, sensitivity to initial value, easy falling into local optimum, bad capacity of handling problems of high-dimensional complex problems.Particle swarm optimization algorithm was proposed by Eberhart and Kennedy in1995. It is a typical swarm intelligence optimization algorithm which originated in the fish and bird flock foraging behavior. Because of the outstanding advantages of particle swarm optimization algorithm such as fast search speed, good global search ability and strong robustness, particle swarm optimization algorithm has become the new research spot in the field of computational intelligence and has been applied to many fields.Along with more deep researches about particle swarm optimization algorithm, the algorithm has been successfully applied to solve the static single modal optimization problem. Many optimization problems in the actual production need to be transferred to multi-modal optimization and dynamic optimization problem. These complex optimization problems require optimization algorithm to find the global optimal extreme value point quickly and accurately, to find out all the local optimal extreme value points and track the changes of the global optimal extreme value point in time. This is a new challenge for particle swarm optimization algorithm.This thesis expounds the work from the following three aspects:(1) The development of particle swarm optimization algorithm consists of the linear decreasing inertia weight, full information, one-dimensional search, topology structure, and multi-swarm strategy, etc. In this thesis, the development process and improved versions of particle swarm optimization algorithm are described and analyzed in chapter2.(2) Multi-modal optimization problems are very common in real life such as the path planning, data analysis and prediction of protein structure, etc. This kind of problem not only needs a global optimal extreme value point, but also needs the rest of the local optimum extreme points in some cases. For multi-modal optimization problem, a classical optimization algorithm often falls into local optimal point, so it is difficult to find the global optimal solution, not to mention all local optimal extreme value points. Then, the niche technology which is based on the principle of speciation is introduced to solve the multi-modal optimization problem gradually. The essence of the niche technology is introduced in chapter3and4in detail. And the advantages of niching particle swarm optimization algorithm based on local search was tested and verified.(3) In fact, many problems consist of a lot of dynamic elements. Some variables often vary over time such as the stock market, path planning, logistics allocation, investment allocation, etc. Dynamic optimization problem is the class of problems which are not only intended to obtain the global optimal solution of a problem, but also has the ability to detect changes in the environment timely and track the trajectory of optimal solution which changes over time precisely. Considering that the improved niche particle swarm optimization algorithm discussed above has succeeded in multi-modal function, we did some further research and improvement in chapter5. Some experiments are conducted on dynamic optimization test functions in order to analyze the performance of the algorithm. The results prove the efficiency of the improved algorithm.