论文部分内容阅读
仿射非线性系统是非线性系统的一种重要形式,近年来出现的状态反馈精确线性化方法对其有较好的控制效果。本文将Hopfield神经网络引入到单输入单输出仿射非线性系统的辨识和控制中。应用具有5个神经元的Hopifeld神经网络对单机无穷大系统进行辨识。对辨识网络进行训练时,采用一种新的学习算法—趋化算法,网络权值进行随机调整。可以有效避免学习中的局部极小值问题,易于计算、构造灵活,适合于动态网络的训练。当网络训练好后,就可以由Hopifeld网络得到状态反馈线性化过程中所需要的李导数和系统状态信息,用状态反馈精确线性化方法将单机无穷大系统转化为简单的线性系统,并设计了一个二次型最优跟踪器对其进行控制。仿真结果表明这种新方法的有效性。
Affine nonlinear systems are an important form of nonlinear systems. In recent years, state feedback precise linearization methods have better control effects on them. In this paper, Hopfield neural network is introduced into the identification and control of affine nonlinear systems with single input and single output. Hopifeld neural network with 5 neurons is used to identify the single machine infinite system. When training identification network, we adopt a new learning algorithm - chemotaxis algorithm and network weight to make random adjustment. It can effectively avoid the problem of local minima in learning, is easy to calculate, has flexible construction and is suitable for the training of dynamic network. When the network is well trained, we can get the Li derivative and the system state information required by the state feedback linearization by the Hopifeld network, and transform the single-machine infinite system into a simple linear system by using the state feedback precise linearization method, and design a Quadratic optimal tracker to control it. Simulation results show the effectiveness of this new method.