Research On Improved Harris Hawk Algorithm And Optimized Hybrid Kernel Extreme Learning Machin

Kernel Extreme Learning Machine(KELM)is the introduction of kernel function into the Extreme Learning Machine(ELM),which has more stable generalization performance.However,due to the existence of kernel function,it is difficult to determine the optimal penalty coefficient and kernel parameters.The Harris Hawk algorithm(HHO),as a new type of swarm intelligence optimization algorithm,has good performance in finding the optimum,which can compensate for the drawback that the parameters of the kernel extreme learning machine need to be set by human.This paper improve the Harris Hawk algorithm to optimize the kernel parameters and construct a photovoltaic power generation prediction model,The main work is as follows.(1)In order to address the shortcomings of the Harris Hawk algorithm,which is prone to fall into local optimum,poor search accuracy and slow convergence speed,four strategies are proposed to improve the convergence accuracy and convergence speed of the original algorithm.Firstly,the Sine chaos mapping and the elite backward learning strategy are used to increase the population diversity and preserve the good initial position of the Harris Hawk population,which improves the convergence speed and the global search ability of the Harris Hawk algorithm;Secondly,in order to balance the transition between the global search phase and the local search phase of the algorithm,a nonlinear energy decreasing strategy is proposed,which enables the algorithm to keep the escape energy at the beginning of the iterative period at a large value.thus improving the global search ability of the algorithm and preventing premature maturity.Finally,the accuracy of the algorithm is improved by applying the cauchy variation to the population optimal individuals so that the algorithm can jump out of the local optimal solution and improve the local search ability.Further simulation analysis by three types of test functions shows that the improved Harris Hawk algorithm(SPHHO)has higher performance than PSO,GWO,BOA,SSA,HHO and CHHO algorithms.The mean value of each algorithm was also subjected to Wilcoxon rank sum test,indicating that SPHHO is significantly different from other algorithms.(2)Construction of a SPHHO-MKELM model and example analysis.Firstly,in order to address the shortcomings of the single kernel extreme learning machine model,which cannot have both high learning ability and high generalization ability,this paper constructs a multiple kernel extreme learning machine model(MKELM)based on the combination of gaussian radial basis kernel function and polynomial kernel function.Secondly,SPHHO is used to optimize the penalty coefficient,kernel parameters and combined weights of MKELM,which overcomes the shortcomings of the parameter selection efficiency of the multiple kernel extreme learning machine and the difficulty of effectively selecting the global optimal parameters.Therefore,the MKELM model has strong learning ability and good generalization ability.Finally,based on the data,the model proposed in this paper(SPHHO-MKELM)is compared with MKELM and ELM.It shows that the SPHHO-MKELM prediction model has high accuracy and reliability.