自组织网络及其在发电机模型研究中的应用

Self-Organizing Network and Its Application to Turbogenerator Modelling

本文将一种能够自行优化、调整网络结构和参数的自组织网络用于发电机组模型的研究中。网络结构的组织是通过对样本数据正交化处理并运用误差减小比指标进行的。对不断从系统获得的新的样本数据，本文运用分批送推的处理方式，通过修正累加矩阵，对网络参数不断进行调节。本文绘出的应用实例表明了这一方法的有效性。

A new kind of network named self-organizing neural netwotk is proposed in this paper. The network can organize its structure and weights automatically by orthogonalizing the sample data and producing the hidded layer processing elements using the index called error redution ratio. The network will be modified recursively when the new sample date is collected batch after batch so that the network can follow the change of the plant where the sample data is collected . The network has been successfully employed in modelling turbogenerator which is considered as a nonlinear multi-inpot multi-output system. The simulation examples illustrate the effectiveness of the estabilished NARX model of the system.