The objective of this paper is to propose a reference model neural control of a system, which change its internal structure from a linear system of first order to a linear system of second order, applying for this task a recurrent neural network. Two schemes of reference model neural control, for the above mentioned system, are presented. One characteristic feature of the neural network used, is that a feedback weight restriction is applied, which preserved its stability during the learning The first control scheme uses one neural network for identification of the variable structure system; the second control scheme uses two neural networks so to separate the identification of each subsystem.