Marine synchronous generator ARCNN modeling

By the basic analysis and research about chaotic characteristic of Aihara neuron, an Aihara local recurrent chaotic neural network (ARCNN) is established based on Aihara neurons. The ability of global approximation of neural networks for nonlinear mapping is increased by introduced neuron of chaotic characteristic. In marine synchronous generator identification and modeling, power of torque of marine diesel engine and excitation current are chosen as input sample parameters, frequency and terminal voltage and current are chosen as output sample parameters for ARCNN modeling. In ARCNN training, a supervised learning method is used, a dynamic BP learning algorithm is applied with momentum and adaptive learning method for modeling. Then, the marine synchronous generator model is built. The results indicate that the neuron number of the ARCNN hidden layer is few in ARCNN modeling, and the ability of generalization of the ARCNN modeling is well.