论文部分内容阅读
提出了一种新的两层反馈型神经网络模型.该网络采用正弦基函数作为权值,神经元激活函数为线性函数,连接形式为两层反馈型结构.研究并定义了该反馈型神经网络的能量函数,分析了网络运行的稳定性问题,并证明了在Liapunov意义下网络运行的稳定性.网络运行过程中,其权值不做调整(但随时间按正弦规律变化),网络状态不断地转换.随着网络状态变化其能量不断减小,最终在达到稳定时能量到达极小点.由于该反馈型神经网络权值为正弦函数,特别适合于周期信号的自适应逼近和检测,为实际中周期性信号检测与处理提供了一种新的、有效的网络模型和方法.作为应用实例把该网络应用于电力系统中电压凹陷特征量实时检测,仿真结果表明,网络用于信号检测不仅有很高的静态精度,而且有非常好的动态响应特性.
This paper proposes a new two-layer feedback neural network model which uses the sine basis function as the weight, the neuron activation function as a linear function and the connection form as a two-layer feedback structure. The feedback neural network , The stability of network operation is analyzed, and the stability of network operation in the sense of Liapunov is proved. During network operation, its weight is not adjusted (but changes sinusoidally with time), and the network status keeps constant As the energy of the network decreases continuously, the energy reaches a minimum point when it reaches a steady state. Since the weight of the feedback neural network is a sine function, it is especially suitable for the adaptive approximation and detection of periodic signals. In practice, the periodic signal detection and processing provides a new and effective network model and method.As an application example, the network is applied to real-time detection of voltage sag features in power system. The simulation results show that the network is not only used for signal detection High static accuracy and very good dynamic response.