摘要
提出了一种改进的动态过程神经网络模型辨识方法,该方法在传统的模型输出与样本输出误差平方和性能指标基础上,添加了相邻采样周期模型输出变化量与样本输出变化量之差的平方和项,作为模型辨识性能指标的一部分,并给出了相应的模型辨识算法。以单元机组过热汽温为对象使用改进的方法进行神经网络模型辨识研究,仿真结果表明与传统神经网络模型辨识方法相比,在相同的辨识精度条件下,该方法可提高所建模型的数据拟合能力和泛化能力,有效提高模型的质量。
In this paper,an improved dynamic process neural network model identification method is proposed.On the basis of performance index of the squared sum of the differences between traditional model output and sample output,increasing by squared sum of the differences between the output variation of model in the adjacent sampling period and the sample which is part of new performance index,and the corresponding model identification algorithm is given.Researching on neural network model identification based on the object of unit's superheated steam temperature through improved method.
出处
《工业控制计算机》
2019年第12期45-46,共2页
Industrial Control Computer
关键词
神经网络
动态模型
模型辨识
neural network
dynamic model
model identification