Improvement And Application Of Harris Hawks Optimization Algorithm

Harris hawks optimization algorithm(HHO)is a swarm intelligence algorithm that simulates the predation and hunting behavior of hawks.It has disadvantages such as low calculation accuracy,easy to fall into local optimum,and difficult to balance exploration and exploitation.This paper proposes two improved HHO algorithms,and applies them to engineering constraint optimization and data clustering.Two kinds of grey prediction models based on improved HHO algorithms are established for power prediction and COVID-19 cumulative population prediction,which expands the application field of HHO algorithm.Specific research contents are as follows:(1)Proposed a harris hawks optimization algorithm based on random unscented Sigma point mutation(IHHO).Use quasi-opposite and quasi-reflection learning strategies to update population individuals according to probability,which improves the diversity of the population;uses logarithmic nonlinear convergence factor balance algorithm exploration and development;carries out random unscented Sigma point mutation for the optimal individual to avoid the algorithm falls into the local optimum to effectively improve the calculation accuracy.The 30 standard test functions of CEC2014 and 15 typical test functions of CEC2005 verify the numerical validity of IHHO,and 3 engineering constraint optimization problems verify the practicability of the IHHO algorithm.(2)An elite fractional derivative mutation harris hawks optimization algorithm(FHHO)is proposed.The memory and storage of Grunwald-Letnikov(GL)fractional derivatives are integrated into the HHO algorithm to mutate the elite group to improve the accuracy and convergence speed of the algorithm;use three randomly selected individuals to design a new exploratory formula,enhance the global exploration ability of the algorithm.Combining the FHHO algorithm with the density peak clustering algorithm(DPC),the FHHO-DPC algorithm is proposed.The 29 test functions of CEC2017 verify the effectiveness of the FHHO algorithm,and the clustering results of 7 synthetic data sets and 5 real data sets verify the superiority of the FHHO-DPC algorithm.(3)A discrete fractional time delayed grey model with triangular residual correction(TDFTDGM)based on IHHO algorithm is established,which expands the application range of the IHHO algorithm.Firstly,the relationship between the continuous form and the discrete form is analyzed,and a discrete fractional time delayed grey model(DFTDGM)is established;secondly,use the triangular residual model to correct the error to obtain the TDFTDGM model;finally,the IHHO algorithm is used to optimize the selection The optimal parameter value in the TDFTDGM model improves the efficiency of model parameter setting.The TDFTDGM model was used to predict nuclear energy consumption in china and power consumption in shaanxi province,which verified the higher accuracy of the built model.(4)A conformable fractional non-homogeneous grey bernoulli model CFNGBM(1,1,b,c)based on IHHO algorithm is established.By using the non-equal weight property of the conformable fractional order accumulation operator and the advantages of the non-homogeneous exponential model in parameter setting,the linear correction term is introduced into the grey bernoulli model,and the CFNGBM(1,1,b,c)model based on the conformable fractional order accumulation is established.IHHO algorithm is used to optimize the selection of model parameters to improve the prediction accuracy of the model.The CFNGBM(1,1,b,c)model was used to predict the cumulative number of COVID-19 in the united states,india and russia,and the high prediction accuracy of the established model was verified.