Resrarch On Nonlinear System Identification Based On Improved Whale Optimization Algorithm

In order to make up for the theoretical limitations of the Whale Optimization Algorithm(WOA),a series of special improvement strategies(REWOA)are proposed.Numerical test experiments verify the effectiveness and superiority of the strategies,and use REWOA to solve the parameter identification of Hammerstein model.The main work includes:(1)In view of the shortcomings of WOA,an improved WOA algorithm(REWOA)based on two-way operation coordination strategy is proposed.Firstly,different evolution strategies are integrated into the various dimensions of the WOA to optimize the structure and improve search accuracy,and Gaussian distribution can be used to increase the population diversity.Secondly,special enhancements were made to the process of searching for prey,which improved the exploration or exploitation capabilities,and new stepping factor is proposed to improve the ability to escape from the local optimum.Adaptive spiral search can maintain the balance and improve stability.Finally,"last elimination" mechanism is added to enhance the convergence performance.(2)To solve the parameter identification of SISO-Hammerstein model.There are some difficulties of identifying nonlinear systems by traditional method,such as complex calculations and difficult to resolve nonlinearities.REWOA-FLANN identification scheme is proposed to solve these problems.The model nonlinear is constructed with a functional connection neural network(FLANN).MSE is a fitness function,which transforms the parameter identification into parameter optimization problem.By minimizing the linear parameters of the MSE search model,three simulation examples prove the feasibility and effectiveness of the identification scheme.(3)To solve the parameter identification of MIMO-Hammerstein model.Compared with the traditional identification methods,the identification of nonlinear systems under heavy-tail noise could be more attention,which is more closed with the actual production process.The heavy-tail noise interference is difficult to deal with traditional methods.Therefore,REWOA-RBF identification scheme is proposed.The nonlinear link is approximated with radial basis function neural network(RBF),and REWOA algorithm is used to simultaneously perform the model linearization parameter identification and RBFNN parameter training.Simulation experiments confirmed that scheme can accurately identify the parameters of the MIMO-Hammerstein model.