【摘要】 神经网络凭借其对非线性的处理能力被广泛应用于实际系统的黑箱建模,但在理论上可以任意逼近模型的神经网络在实际应用中的能力是有限的,对于复杂动态特性的实际系统基于神经网络的模型在逼近效果和泛化能力上都存在不足。提出了基于神经网络的混合模型建模方法,建立的模型由通过传统方法建立的基本系统和由神经网络建立的逼近实际系统和基本模型之间差值的不确定部分组成,用此方法建模大大提高了模型的精度和对不同输入的泛化能力,通过对多个系统的建模仿真结果验证其可行性。更多还原

【Abstract】 Relying on its non-linear handling ability, the neural network is widely applied to the actual system black box modeling, but theoretically its practical application ability is limited. Regarding the complex dynamic characteristics of actual system, its generalization and approaching ability is insufficient. This paper proposed a mixed model modeling method based on the neural network, composed of the normal model through the traditional method and the neural network model for approaching the difference between the actual system and normal model to improve the model’s approaching and generalization ability. The feasibility is verified through emulation.更多还原