Mesure de l'intensité du courant électrique et du facteur de puissance par un réseau de capteurs à effet Hall

Abstract

A measurement system, using a wireless Hall Effect Sensor (HES) network was the subject of previous work. HESs have the advantage of a very low cost, and a level of integration that allows a small housing. In addition, their installation does not require the disconnection of the conductor wire to be measured. In return, the measurement gain of this sensor depends on its positioning and its measurements may have crosstalk, since it measures, without discrimination, all the magnetic fields of its environment. This project contributes to advancing the field by improving the measurement capabilities of the previously developed system. Based on the same analog circuit, the sensor is modified to increase the sampling rate to 2KHz and incorporates a communication protocol that synchronizes the information with the central processing unit (CPU). This data synchronization is exploited to add the power factor measurement for each of the circuits measured by the HESs. These enhancements allow the application of most signal desaggregation algorithms, a related research area. The large amount of data exchanged between the sensors and CPU requires the integration of data compression. After analysis, the Discrete Cosine Transform (DCT) is preferred over other methods including Free Lossless Audio Codec (FLAC), Fast Fourier Transform (FFT), and Discrete Wavelet Transform (DWT). The automatic calibration of a HES network, proposed in this project, uses the measurement of a current transformer sensor (CT). The elaborate source separation algorithm (SS) thus dynamically determines the measurement gains, and removes the crosstalk. A band-pass filter, that directly exploits DCT coefficients, reduces the influence of background noise and improves detection of small loads. In the end, the SS and the filter make it possible to reduce the average measurement error of the sensors to 0.2A. The measurement of the phase shift by the HESs has a maximum observed error of 6 degrees, while the CT sensor mesurement has a maximum observed error of 24 degrees. It is recommended to improve the latter in future work, as it influences the overall accuracy of the system. The addition of the normalised least-mean-squares algorithm (NLMS) makes it possible to reduce the difference between the sum of the measurements of the HES sensors and the measurement of the CT sensor at 0.2%. Without the use of NLMS, this gap was 4.7%. However, it increases the average individual sensor measurement error to 0.41A for the SS, filter and NLMS combination.

Date	20 Jul 2018
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Ghyslain Gagnon (Supervisor) & Claude Thibeault (Co-supervisor)