Research and simulation of Volterra series kernel identification on wiener model

Non-linear system identification is a hot point problem in the field of electrical information and automatically control. Among many analysis methods, Volterra series model has rigorous theory basic, which can approximate in any precision, and reconstruct non-linear system, so, it has been applied into many non-linear system identification and equipment fault diagnosis. But, with the improvement of memory length and identification degree, Volterra high degree kernel will become complex and hard to identification, and cause reduction of identification precision, even to failure. This paper uses a kind of reproducing kernel Hilbert space method, transfer calculating Volterra series kernel into reproducing kernel or Hilbert space, which can reduce operation amount largely, and can calculate infinite degree Volterra series kernel coefficient in theory. Through simulation to some non-linear system, this solution has higher identification precision, and second amplitude frequency response error is less than 5dB, phase frequency response is less than 3 degree, which proving algorithm´s accuracy.