A design method
based on Bayesian-Gaussian neural network(BGNN) is proposed for the
inverse modeling problem of nonlinear system. The training
procedures of the BGNN model include two steps. The
structure parameters of the BGNN model are optimized by using the swarm
intelligence optimization algorithm. Furthermore, the trained BGNN
model combining with the history data is applied to the on-line
prediction of the nonlinear system inverse model. The inverse model
simulation experiments are conducted on a nonlinear system of
hydraulic turbine, and the results show that the inverse model design
method based on BGNN has the easy topology determination and
effective identification accuracy, and it is fit for nonlinear system
inverse model design.