Improvement And Application Research Of Firefly Optimization Algorithm

Firefly optimization algorithm simulate the mutual attraction between individual fireflies in nature,thereby achieving the goal of optimization.The algorithm is a population intelligent optimization algorithm,which has the advantages of clear flow,fewer parameters,and easy implementation.At the same time,this algorithm has some defects such as easy to fall into local optimization,individuals prone to oscillations near the peak value,and low resolution.Based on the existing research,this paper conducts in-depth research on firefly optimization algorithm,proposes corresponding improvement strategies for the shortcomings of firefly optimization algorithm,and applies the improved algorithms to engineering design constraint optimization problems and initial parameter optimization problems of BP neural networks.The main work is as follows:(1)Aiming at the problems of slow convergence,low accuracy,and premature convergence of firefly optimization algorithm,a firefly algorithm(AMFA)incorporating covariance matrix adaptive evolution strategy was proposed.First,a new attraction model is used to determine the number of attracted fireflies;Secondly,introducing a search operator into the stepping mechanism improves the chances of fireflies jumping out of local optima while enhancing their global search performance.Finally,the improved firefly optimization algorithm and the covariance matrix adaptive evolution strategy are designed to enhance the diversity and quality of solutions during the iterative process of the algorithm.Test on 9 standard test functions and 8 practical issues in CEC 2011.Experimental results show that the improved algorithm can effectively avoid local optimization problems,and has higher performance than the original algorithm and other common algorithms in terms of solution accuracy and convergence speed.In addition,the algorithm also exhibits good stability.(2)In order to further improve the overall optimization performance and engineering application ability of the Firefly Optimization Algorithm,a Sinusoidal Cosine Based Hybrid Firefly Optimization Algorithm(SCAFA)was proposed.Firstly,a chaotic sequence generated by logical self mapping is used to initialize the individual position of fireflies to improve population diversity;Secondly,the idea of sine cosine algorithm is incorporated into the position update method of the firefly optimization algorithm,and a nonlinear dynamic learning factor is introduced to make the algorithm have better excavation ability and convergence speed,and can balance the local and global exploration ability;Finally,Gaussian mutation is applied to some individuals in the population to enhance the local escape ability of the algorithm.The SCAFA algorithm was tested on 23 standard test functions and compared with other algorithms.The experimental results show that SCAFA algorithm outperforms other algorithms in terms of optimization performance and convergence speed,and has better global search ability and convergence speed,verifying the effectiveness of the improved strategy.(3)In order to verify the practicability of AMFA and SCAFA algorithms,these two algorithms were applied to engineering design constrained optimization problems and BP neural network initialization parameter optimization problems respectively,and the experimental results were analyzed.The experimental results show that AMFA and SCAFA algorithms have good application effects,and can effectively solve engineering design constraint optimization and neural network parameter optimization problems.The application results of AMFA algorithm show that AMFA can quickly find the optimal solution under the premise of satisfying constraints,and has good stability.The application results of SCAFA algorithm show that SCAFA can better handle complex nonlinear constraint problems,and the prediction model optimized by SCAFA algorithm has better generalization ability and prediction accuracy.Experimental results demonstrate the practicality and effectiveness of the proposed algorithm,and provide strong support for further research and application.