Contactless voltage and current estimation using signal processing and machine learning

Abstract

This thesis describes a contactless sensor developed to estimate the line currents and line-to-line voltages of a multi-phase cable in the presence of significant external disturbances. The current estimates are derived from an array of point magnetic-field measurements processed by a linear least-square-error estimator. The gains in the estimator are chosen using a probabilistic model of measurement errors created by external magnetic field sources. Test bed validation of the estimates demonstrates estimation errors below 1% even in the presence of nearby cables carrying comparable currents, metal plates that could support eddy currents, and large magnetizable cores. The voltage estimates are derived using actively-guarded electrodes that capacitively couple to the cable conductors. Knowing the coupling capacitance, test bed validation of the estimates again demonstrates estimation errors below 1% even in the presence of nearby cables carrying comparable voltages, and metal plates. A method involving capacitively coupling signals onto the cables is also proposed and demonstrated to determine the coupling capacitance without operator intervention.