Research On Krill Herd Optimization Algorithm

Compared with other intelligent algorithms, Krill Herd algorithm(KH) has a significant advantage in terms of convergence which is easy to implement and has fast convergence when used for solving global optimization problems. The algorithm got common concern for it once proposed by researchers, and it has been applied in the fields of computer, communications, mechanical, neural networks and power systems. However, there also remains a long way for its further improvements and application extension.Based on the analysis of KH principle, the algorithm is studied and verified by simulation test. The main contents of this article as follow:First, the standard KH algorithm is analyzed for deeper understanding of model structure. There are a lot of parameters in KH, and the changes of some parameters have important effects on the performance of the algorithm, such as foraging inertia weight, biggest feeding speed and the step zoom factor. Through a mass of tests, the effects of these three for are found and then the best way for parameter selection is proposed, which can an improve algorithm performance to a certain extent.Second, an improved algorithm of KH with self-adaptive inertia weight(AKH) will be proposed. Theoretical deduction shows that massive invalid iterations in the process of solving global optimization through standard algorithm of KH arise from inertia weight. Based on this, the algorithm is improved with the population particles divided into two categories based on adaption value in the process of algorithm iteration: one with decreased adaption value and the other with increased adaption value. Then, the inertia weight will be adjusted dynamically with that of particles with decreased adaption value reset to zero so as to lessen the adverse impacts of inertia weight and that of particles with increased adaption value remaining the same. Numerical experiment shows that AKH can effectively reduce invalid iterations of algorithm and enhance its convergence accuracy and capacity.