A novel virtual instrument for power quality surveillance based in higher-order statistics and case-based reasoning

This paper presents the performance results of a Virtual Instrument (VI) based in Case-Based Reasoning (CBR), conceived to online monitor the power-quality. The PC-based instrument receives data through a DAQ board and a differential probe, while maintaining economy by eliminating the extra network construction and hardware. Being flexible, presents an user-friendly interface and a large data storage capacity, since it uses the hard disk. The computational guts of the instrument are based in third and fourth-order statistics (along with the variance), which enhance detection capability and reject noise influence. A time-domain sliding window sweeps the register under test and offers a time-variation pattern which reflects the deviation of the statistical estimator with respect to the steady state. This three-valued time-series comprises variance, skewness and kurtosis evolution, and constitutes a triple input to the innovative CBR module, which in turn is capable of distinguishing electrical anomalies among five categories (the sixth is reserved to the healthy signal): non-50 Hz, 50-Hz-asymmetrical, 50-symmetrical non-sinusoidal, swell and sag. Online surveillance tests developed over the local electrical network show acceptable accuracy (96%).