A High-Sensitivity Magnetic Protection Method for Reactors Based on Statistical Testing of Linear Regression Residuals

Abstract

Slight inter-turn short circuits in reactors produce only weak changes in terminal electrical quantities, resulting in insufficient sensitivity of conventional protection methods. Magnetic differential protection can sensitively reflect magnetic field changes during internal faults, but its sensitivity is affected by factors such as the installation accuracy of magnetic sensors in engineering applications. To further improve the performance of magnetic protection, a reactor magnetic protection method based on linear regression residual testing is proposed. First, the internal magnetic field signals of the reactor are acquired through magneto-optical measurement sensors, and the least squares method is used to perform linear regression modeling on the magnetic field data. Subsequently, based on the statistical characteristics of the regression residuals, a t-test is employed to determine the significant deviation of the residuals, thereby achieving sensitive identification of slight inter-turn short circuits in the reactor. Finally, the effectiveness of the proposed method under different fault conditions is evaluated through finite-element simulation. The simulation results show that the proposed method can detect a one-turn inter-turn short circuit under the modeled conditions, while no maloperation is observed for the investigated sensor position and angle deviations. It can quickly and reliably distinguish between internal and external faults of the reactor, providing a new perspective for reactor condition monitoring and protection.