Identification Of Nonlinear Neural Network System Based On Improved Artificial Bee Colony Algorithm

Since2005imitation bee foraging artificial bee colony algorithm (ABC algorithm) many researchers give great efforts to it,expand the field of application. In this thesis, the algorithm used for BP neural network nonlinear system identification, The advantages of bee colony algorithm global search ability for BP neural network training initial weights and thresholds, avoid the random assigning weights and thresholds simple to drop into the department minimum, slow convergence trouble.Firstly the ABC algorithm and LM algorithm, gives ABC-LM algorithm, used to BP neural network nonlinear system identification. Simulation shows that the ABC-LM algorithm is effective nonlinear system identification simulation to four objects.Secondly, the bee colony algorithm exists premature convergence can only search food but shortcomings transformation to it, literature40GA algorithm cross-factor introduced into the ABC algorithm,but the crossover rate fixed immutable and local search ability is not strong. The this thesis GAABC algorithm crossover rate fixed value to change according to the number of population evolutionary adaptive and variation factor, the mutation rate is based on the number of population evolutionary adaptive change, and enhance the local search ability of the bee colony algorithm. The algorithm GAABC and LM algorithm combines training BP neural network and applied to nonlinear system identification. Simulation results show that based on GAABC-LM algorithm neural network identification precision higher than the ABC-LM algorithm.Finally, for further enhance the BP neural network identification accuracy, The Memetic algorithm have all the advantages of GA algorithm to innovative ABC algorithm, for further improve and enhance the local search ability to improve the accuracy of the BP neural network recognition, we introduce the ABC algorithm improvements, in conjunction with the LM algorithm to obtain the initial weights and thresholds of BP Neural Network Identification. With type GAABC algorithm, MMABC algorithm cross factor, the variation factor of the serial working mechanism was changed to work in parallel mechanism. Thereby increasing the innovative capacity of the algorithm, and retains all of ABC algorithm powerful global search capability, increasing the possibility of find the global optimum. Then carried out based on BP neural network algorithm MMABC-LM Nonlinear System Identification, simulation shows that the algorithm has given very good identification results.