证明满足一定条件的单输入单输出系统的α阶逆系统（与原系统复合起来可构成α阶积分或时延系统）存在且唯一，阐明静态多层网络加若干积分器或时延因子可以逼近这样的α阶逆系统，对神经网络α阶逆系统方法用于线性、非线性控制的可行性作了探讨。 It is proved that the α-order inverse system of a single-input / single-output system that satisfies certain conditions (which can be combined with the original system to form an α-order integral or delay system) exists and is unique. It shows that the static multilayer network plus some integrators or delay factors can approximate Such an α-order inverse system, the neural network α-order inverse system method for linear and nonlinear control of the feasibility are discussed.