Control of a nonminimum phase system with varying time-delay using neural networks

This paper deals with the neural control problem of a nonminimum phase discrete-time system with varying time delays. The system considered is a simulated river. The input and output are upstream and downstream flows, respectively. The controller has to propose the necessary upstream flow such that the corresponding downstream flow is as close as possible to a reference trajectory defined in order to satisfy water users. Classical regulation of such a system is performed under the assumption of a fixed time-delay. This assumption causes several problems due to the fact that the time taken by the water to go from the upstream to the downstream of the river is dependent on the average flow in the river. This paper shows that the use of neural networks in place of classical regulators allows to overcome these problems. Since the plant is a nonminimum phase system, the command produced by the neural controller is highly oscillating. Several methods are then compared on the basis of their ability to solve this problem