A Class Of Improved Particle Swarm Optimization Algorithm And Application In Parameter Estimation Of Fractional-Order Systems

In recent years,group intelligent optimization algorithms have been rapidly developed due to their excellent ability to deal with black box problems.As a typical swarm intelligence optimization algorithm,particle swarm optimization has been widely concerned by academia and industry.However,traditional particle swarm optimization and most of its variants still have defects in dealing with complex optimization problems and high-dimensional multi-modal optimization problems,and they are easy to fall into local optima.Also,the problem of parameter estimation of fractional order systems has been widely concerned.because there are still many difficulties in using traditional optimization methods,the combination of swarm intelligence optimization algorithm and parameter estimation of fractional order systems is a new breakthrough.In this article,the particle swarm optimization algorithm is improved to deal with high-dimensional complex optimization problems and fractional order system parameter estimation problems.For high-dimensional complex problems,a particle optimization algorithm based on potential information is proposed.Firstly,the potential global optimal position is designed as the third leader,which can improve the overall quality of leaders and mining the hidden information of the current particle swarm fully.Secondly,in order to enhance the exploration ability of particle swarm optimization algorithm,an adaptive mutation strategy based on Lévy flight is designed to make Lévy mutation on particles which have high degree of aggregation.Finally,through two sets of experiments including: comparison with 11 excellent intelligent optimization algorithms on 30 test functions,and comparison with 7 classic particle swarm optimization algorithm variants on CEC2015 sets,the proposed algorithm is verified having great advantages on high-dimensional complex problems.For the optimization of complex optimization problems and the estimation of complex system parameters,an improved surrogate-assisted particle swarm optimization algorithm is proposed.Firstly,in case of blind learning,the particle swarm position update method is modified to consider the individual wisdom of particles.Secondly,the surrogate model is integrated into particle swarm optimization algorithm,which can reduce the computational cost greatly and ensure the effective convergence.In order to further ensure the exploration ability of particle swarm,an adaptive Lévy mutation strategy is proposed to perform Lévy mutation on particles that have not updated for a long time.Finally,we apply this algorithm to the parameter estimation of two fractional order systems.Experiments show that the improved particle swarm optimization algorithm can find the best optimum with faster speed and higher precision.