Compensation of saturation effects in current transformers using neural networks

Abstract

Magnetic current transformers (CTs) are currently used in electrical devices in order to measure currents. The accuracy of CTs can severely decrease in case of saturation of the magnetic core, which can severely distort the current observed at the secondary coil of the CT. If the current in the primary coil has to be evaluated, to trip a relay for instance, saturation effects must be taken into account. A method using neural networks (NNs) is proposed. First, a large set of current signals encountered in low voltage installations has been built. Saturation has been added with a previously validated CT model. Then, a NN has been trained to invert the saturation effects and to reconstruct the primary current from the distorted one.