Improvement Of Differential Evolution Algorithm And Its Application In Layout Optimized Design For Solar Power Tower Plant

Many problems in science and engineering fields can be abstracted as optimization problems.As the practical systems are becoming more and more complicated,the systems are featured with uncertainty,nonlinearity,and non-differentiable.Developing intelligent optimization algorithms,which are suitable for solving large scale computing,has drawn great attention in academic area.Differential evolution(DE)is a kind of evolutionary algorithm based on population difference.It has been widely used in various fields because of its strong robustness,simpleness and effectiveness since its first introduction in 1995.In this paper,based on the introduction of the basic mechanism and principle of DE algorithm,a improved DE algorithm and framework are proposed according to the advantages and disadvantages of the population structure,the control parameters and the mutation process.Finally,the combined algorithm and framework is applied into the layout optimization for solar power tower plants.The main research contents are as follows:1.In classic mutation operation,the optimum solution has strong leading effect,which will guide the evolution process into local optimum.In addition,the control parameter in basic differential evolution is sensitive.Based on the analysis of these defects,a differential evolution algorithm with dynamic speciation-based mutation is proposed.In this algorithm,a hierarchical population structure is constructed to locate the relatively better solutions in the space.Furthermore,two new mutation operators,one is relatively explorative and the other one is relatively exploitative,are employed simultaneously to balance the exploration and exploitation of the algorithm.Finally,Levy distribution is used to adjust the control parameters dynamically.The improved DE algorithm and several state-of-art DE variants are tested on CEC2014 benchmark and results suggest that the overall performance of our proposal is better than that of the other four DE algorithms.2.The elite regeneration framework is proposed to alleviate the stagnation problem existing in the current DE algorithms.In this framework,Gaussian and Cauchy distribution are used to regenerate elite population.This framework is simple and effective and it is easy to apply it to other DE variants to improve the population diversity and enhance their local searching ability.The experimental results on CEC2014 benchmarks indicate that algorithms combined with this framework perform better than the original algorithms,which validates the effectiveness and university of the framework.3.Finally,this paper studies the layout optimized design for solar power tower plant.Firstly,we build the optical efficiency model of the heliostat field.Secondly,the densest radial staggered layout adopted as the initial layout in our design.The field will be expanded based on this layout,during which the radical spacing between the rows are taken as parameters to be optimized through combining the proposed DE algorithm and framework.The experiment results reveal that the whole optical efficiency of the optimized field is greatly improved,which proves the efficiency of the proposed algorithm and framework again.