In accordance with characteristics of V/F inverter and two-motor system, the united mathematic model of two-motor variable frequency speed-regulating system with V/F is given. The system is proved to be invertible and the inverse system can be constructed by a dynamic neural network consisting of a static neural network and an integrator. A linear subsystem of speed and a linear sub-system of tension can be obtained by combining the inverse neural network with the two-motor variable frequency speed-regulating system. Speed and tension control of two-motor variable frequency speed-regulating system can be decoupled by the neural network inverse synchronous control method, and high performance control can be obtained by designing linear close-loop adjustors of speed and tension. Experiments show that the system has good static and dynamic performances and high anti-disturbance of load. Thus a difficult problem of multi-motor synchronous decoupling control is solved.