An Improved Bat Algorithm Based On Local Search And Its Application

In 2010,Professor Yang in Cambridge University had established the field of bat algorithm(BA),which is based on echolocation theory of bats.The bat algorithm searches the optimal location of the bats by changing the volume of sound,emissivity of the pulse and frequency,which has advantages of relatively less parameters,faster speed of finding optimal places and easier to realize.Bat algorithm is an effective method,which is not only novel but also reliable to search optimal solutions from an overall perspective.However,this algorithm also exists some drawbacks when refers to the mechanism of its own iteration,such as low speed of convergence in latter period,inaccuracy of convergence and easy to fall into minimal point of the local.Therefore,the basic bat algorithm still to be of high values to research.As for unconstrained optimization problems,this paper uses Gaussian disturbance,steepest descent method and Lévy flight respectively to modify the basic bat algorithm,raising three new approaches to make improvements.As for the fact of basic bat algorithm that doesn't implement overstepping treatment on bats' speed,and disturbs insufficiently in the local search,causing the drawbacks of local optimal solutions,this paper puts up with the new algorithm based on velocity overstepping treatment and improvements of Gaussian disturbance in order to ensure the algorithm could have the ability of global searching.And as for the disadvantages of basic bat algorithm while could not find the best bats' location during local search compared with the global stage,this paper puts forward advanced bat algorithm on the basis of velocity overstepping treatment and steepest descent method,which decreases the uncertainty of part searching.Lastly,considering the insufficiency of basic bat algorithm when directly steps into local search stage after global ones,the paper raises newly established algorithm referring to inertia weight and Lévy flight,to forbid the results are led astray in local optimal solutions.Among the three improved bat algorithms mentioned above,the first and third are to enhance their global search ability,and the second is to improve their local search utility.The results show that these methods have higher accuracy than basic one through quantitative optimization tests without any constraints,and have smaller standard deviation.As for constrained optimization problems,referring to inertia weight and Lévy flight,the paper introduces a new way to transfer the problem with boundary constraints—Augmented Lagrange multiplier method;and proposes the improved algorithm base on this.By applying quantitative optimization tests with constraints,the results shows that the improved algorithm has better accuracy and smaller standard deviation,compared with other optimization methods.