The Influence of Different Factors on Machine Learning Based Current Transformer Algorithm

The continuous construction of smart grid has put forward new requirements on the efficiency and accuracy of current measurement in all parts of the grid. In a previous study, a neural network-based algorithm for current transformers was proposed, which was applied to a circular magnetic field sensor array based on giant magnetoresistance effect (GMR). The essence of this algorithm is an inverse computation from magnetic field to current magnitude. The feasibility of this algorithm is verified under simplified conditions in previous study. In this article, we study the influence of the number of sensors, the inclination angle between current wire and the sensors plane, the wire's eccentricity level, and the environmental noise on the accuracy of this algorithm, and then determine the feasibility of this algorithm in practical applications.