Research On Improvements And Applications Of Chicken Swarm Optimization Algorithm

Chicken Swarm Optimization(CSO)is an intelligent optimization algorithm proposed in2014,which simulates the hierarchical order and the foraging behavior of chicken swarm.Chicken swarm are divided into several groups,each consisting of one cock and many hens and chicks.In a certain hierarchical order,chickens of different orders compete with one another and follow different rules of locomotion.The chicken swarm optimization algorithm has the advantages of clear structure,being easy to understand and strong searching ability,so it has been widely concerned by scholars at home and abroad.With the in-depth exploration of the chicken swarm optimization algorithm,scholars have found that the algorithm also has problems such as slow convergence speed,low solution accuracy and being easy to fall into local optima,which seriously limit the development of chicken swarm optimization algorithm.In this paper,aiming at the existing problems of the chicken swarm optimization algorithm,a series of improvements over the algorithm are made and the improved chicken swarm optimization algorithms are applied to function optimization and model parameter estimation.The main research work is as follows:(1)In order to improve the performance of CSO,in Chapter two,we improve the updating mode of CSO and propose an adaptive dynamic learning chicken swarm optimization algorithm(ADLCSO).Firstly,we add a nonlinear decline learning factor and a reverse foraging mechanism to the rooster position updating process to enhance the ability of the population to escape local extrea.Secondly,in the process of hen updating,we propose a new population similarity index,and use the index to adjust the position of each hen adaptively so as to inhibit the attenuation of population diversity and improve the solution precision of the algorithm.Through the simulation experiment,the superiority of ADLCSO is verified.(2)In order to further optimize the performance of CSO,in Chapter three,we simplify the structure of the chicken swarm optimization algorithm,and propose an adaptive simplified chicken swarm optimization algorithm based on inverted S-shaped inertial weight(ASCSO-S).Firstly,from the perspective of improving the structure of the chicken swarm algorithm,we remove the chicks from the chicken swarm and propose a simplified chicken swarm optimization algorithm(SCSO).Secondly,the inverted S-shaped inertia weight is introduced into the SCSO to dynamically adjust the moving step size of individuals.Finally,when the algorithm falls into the local optimum,we introduce the adaptive updating method in the updating process of hens to enhance the exploration ability of the algorithm.Through the simulation experiment,the results show that ASCSO-S is superior to other contrast algorithms in terms of convergence speed,solution precision and solution stability.(3)In view of the fact that ADLCSO and ASCSO-S are improved from different perspectives,we compare the performance of the two improved algorithms in Chapter four.Firstly,ADLCSO and ASCSO-S are tested on the test functions,and the experimental results show that both algorithms have high accuracy and stability,and ASCSO-S has faster convergence speed than ADLCSO.Secondly,these two improved algorithms are applied to the parameter estimation of Richards model.Through experiments,it is verified that both ADLCSO and ASCSO-S remarkably outperform the other three algorithmsl,and ASCSO-S is superior to ADLCSO in the two evaluation indexes of determination coefficient and mean absolute error.(4)In order to improve the simulation and prediction accuracy of the traditional GM(1,1)model,in Chapter five,we propose an improved GM(1,1)model based on chicken swarm algorithm(CSO-GM(1,1)).The initial condition of the model is improved by introducing disturbance factor,and an optimization model based on the principle of new information priority is designed,and then the parameters of the model are optimized by using the chicken swarm algorithm.Through experiments,the CSO-GM(1,1)has a wider application range and higher accuracy of simulation and prediction and in order to verify the effectiveness of ADLCSO and ASCSO-S in the parameter estimation of GM(1,1)model,we use these two algorithms instead of CSO to estimate the parameters of the improved model in this paper.The experimental results show that the two improved algorithms proposed in this paper have more advantages in the parameter estimation of GM(1,1) model.