The identification of dynamic system based on memory RBF neural network

The memory RBF (MRBF) neural network (NN) is obtained by introducing memory neuron into the RBF neural network, and it functions a recurrent NN. The MRBF NN can identify dynamic system without feedbacking the past input and output, because the present output is related with the past input. Thus, the NN can identify the system of which the order or delay is unknown. In this paper, the learning algorithm is given and the correlative theory is proved. The simulation of dynamic system identification shows that method is valid, and can provide great potential for self-adaptive control.