Application of neural network as voltmeter and regulator of electric voltage

摘要

In the paper, we propose using the neural network in the electric voltage monitoring system and in the voltage control system. Our investigations are based on the mathematical modelling idea. In the first case, as a `teacher' for the neural network, the mathematical semi-field model of the toroidal saturated choke is used while in the second case, the neural network is learned using signals of the classical regulator. The mathematical model of the choke is described by partial differential equations of the quasistationary electromagnetic field and ordinary differential electromagnetic circuit equations (semi-field model). From a numerical method point of view, semi-field model gives better accuracy and stability of solution in opposite to the electric circuit model. Results of computation are given. These results confirm that using neural networks in two considered systems gives some advantages, which can not be obtained by classical methods.