Improvement And Application In Parameter Estimation Of Bat Algorithm

The bat algorithm（Bat Algorithm,BA）is a swarm intelligent optimization algorithm proposed by Yang Xin-she.Its principle is to simulate the echolocation behavior of bats in nature.The algorithm is simple in structure and robust.BA algorithm has been widely used at home and abroad.Like other heuristic algorithms,bat algorithm still has some disadvantages when solving complex problems,such as population diversity fading in the later iteration period,low convergence accuracy and possibly falling into local optimal solution.In order to solve the problems existing in the bat algorithm,this paper has done the works as follow:（1）A wavelet bat algorithm with cascade chaos and quasi-opposite learning is proposed.The bat algorithm（BA）is improved via cascade chaos mapping,quasi-opposite learning strategy,and wavelet mutation to solve premature problem,population diversity reduction over iterative process,and enhance the search efficiency of BA algorithm.The velocities are updated by adopting cascade chaotic weight with decreasing oscillation to accelerate the convergence rate of the algorithm.Quasi-opposite learning strategy is used to enhance global exploration ability after location update of bats.The current best solution is mutated by wavelet to overcome the problem of premature convergence of the algorithm,which can make the algorithm jump out of the local optima.The proposed CQWMBA algorithm is compared with BA,LBA,PTRBA,TFRWBA algorithms and several swarm intelligent algorithms on 13 benchmark functions of including unimodal and multimodal function,and CEC2014 test sets.The experimental results exhibit that the comprehensive optimization performance of the proposed algorithm is the best.In order to verify the effectiveness of the proposed algorithm in solving practical problems,it is applied to parameter identification of Lorenz chaotic systems.Experiments show the proposed algorithm compared with the contrast algorithm can identify system parameters accurately and effectively.（2）An improved bat algorithm was proposed based on fractional-order strategy and spiral with Levy flight,to improve the search efficiency and avoid falling into local optimal solution.Fractional order strategy with short-term memory characteristics is introduced to update bat position and improve the convergence speed of the algorithm;A new solution is generated locally by the Archimedes spiral with Levy flight strategy,which enhances the local exploitation ability and helps the algorithm jump out of the local optimum;The new nonlinear dynamic mechanism for adjusting loudness and pulse emission rate is to balance the exploration and exploitation abilities of the algorithm.The improved algorithm and BFA,SCA,BA,GCBA are simulated in the CEC2014 benchmark test set.The test results and Friedman statistical analysis show that the search efficiency and the accuracy of the proposed algorithm are improved compared with the contrast algorithm.Finally,the proposed algorithm is used to solve the design problem of mechanical engineering reducer.The experiment results verify the effectiveness of the proposed algorithm compared with PSO-DE,WCA,and APSO.（3）A gradient estimation and bias learning bat algorithm is proposed to verify the parameter identification problem of solar cell models.The S inertial weight is used to balance the exploration and exploitation abilities of the algorithm.According to Newton Lapson’s formula,the updating rule of gradient estimation is used to improve the speed update term of bats and accelerate the convergence performance of the algorithm.The introduction bias learning strategy can strengthen the search efficiency of the algorithm by evaluating fitness values of the current best solution and the bias of the center point,retaining the best,and moving towards individual areas with a large number of small fitness values.The proposed algorithm and SJBA,JCXSBA,GMBA,DE,BFA are simulated in the CEC2014 benchmark test set.The results show that the effectiveness and robustness of the algorithm are greatly improved.In the parameter identification experiment of solar cell model,the proposed algorithm is applied to the parameter identification of single diode and double diode models.The experimental results show that the proposed algorithm has higher accuracy and better stability compared with the comparison algorithm.