船舶大功率发电机混沌神经网络建模

MARINE HIGH POWER GENERATOR MODELING BASED ON CHAOTIC NEURAL NETWORK

在分析和研究了Aihara神经元混沌特性的基础上,建立了基于Aihara混沌神经元的Elman局部递归混沌神经网络(CNN),神经元引入混沌特性后增强了神经网络对非线性映射的全局逼近能力。在船舶大功率同步发电机建模中,以船用柴油机输出转矩功率和发电机输入励磁电流作为CNN建模与辨识的输入参数;以发电机的输出频率、发电机端电压和输出电流作为CNN建模与辨识的输出参数;采用有导师学习方式,运用基于BP的动态训练方法,最终完成了船舶大功率发电机的动态建模。与其它的ANN建模相比较,用CNN建立的模型的隐层神经元数量少,系统的泛化能力强。

On the foundation of analysis and research of chaotic characteristic of Aihara neuron, an Elman local recurrent chaotic neural network （CNN） was established based on Aihara neuron. The neuron introduce of chaotic characteristic boosted up ability of global approximation of neural network for nonlinear mapping. In marine high power synchronization generator modeling, output torque power of marine diesel engine and input excitation current are selected as input parameters of the CNN for modeling and identification, output of frequency and terminal voltage and current are selected as output parameters of CNN for modeling and identification. A form of supervised learning was used. A dynamic BP algorithm was applied for the CNN training. Finally, marine high power synchronization generator model was built. Compare to other ANN modeling, the neuron number of the CNN hidden layer is few, and the ability of generalization of the CNN system is well.