Affiliations: [a] School of Electrical Engineering, Shandong University,
Jinan, Shandong, China | [b] School of Electrical and Electronic Engineering, North
China Electric Power University, Beijing, China
Correspondence:
[*]
Corresponding author: Haoxi Cong, School of Electrical
Engineering, Shandong University, Jinan 250061, Shandong, China. E-mail:
lqmeee@ncepu.edu.cn
Abstract: J-A model is widely used in hysteresis modeling as well as
performance simulation of the magnetic materials. To achieve preferable
adequacy, a modified J-A model is given, however, accurate solution of the
parameters is very important for the modified J-A model, especially for
over-determined nonlinear equations. In response to solving the over-determined
nonlinear equations, this paper first turned the problem of the over-determined
nonlinear equations into solving the minimum value of a multivariate function
by means of the least square method. While the multivariate function is of high
nonlinearity (the function is not continuous and the matrix of the partial
derivatives is singular), solution methods using derivative calculation were
abandoned, and the direct seeking methods with no derivative calculations
(simplex algorithm) were involved to solve the problem. In the end the solution
was validated with the use of the genetic algorithm and the simulated annealing algorithm.
