The voluminous amount of raw waveform data recorded by triggerless power quality monitors contain conspicuous and inconspicuous disturbance events. Data reduction and detection techniques are needed to efficiently extract useful information hidden in the raw data and identify power quality disturbances. The overall objective of this study is to use step changes in the rms voltage profile as an alternative triggering feature for automatically detecting switching events. The full characterization of the event is based on processing a small portion of the voltage waveform selected around the detected rms voltage step change. A filtering method is proposed to smooth out rapid fluctuations in the rms voltage profile during steady-state operation, while preserving the sharp edges caused by rms voltage step changes. Once the rms voltage profile has been filtered, adaptive limits based on the median absolute deviation are computed for detecting rms voltage step changes. The effectiveness of the proposed technique is evaluated using triggerless voltage waveforms to detect capacitor switching events. The use of the filtered rms voltage profile allows accurate detection of capacitor energizing and de-energizing events, while more than 50% of the detections in the unfiltered profile correspond to false-positives.

Voltage sags and swells are power quality events commonly observed in power systems; however, none of the existing methods allows determining their point-on-wave inception and recovery instants (and consequently, their duration) accurately in all cases. The primary goal of this paper is to determine these instants with little or no delay, by calculating the absolute rms voltage difference between two adjacent sliding windows. The proposed method is based on the assumption that this difference is maximum when the sample under analysis corresponds to either the inception or the recovery instant. This method is valid for both sag and swell events, with or without transients. Evaluation of the proposed method performance for different types of events shows that it is robust and highly accurate in determining the inception and recovery instants. The estimation error for the majority of the events analyzed is either zero or one sample (each sample corresponds to a phase angle difference of 2.81°), while the worst performance is three samples.

Capacitor banks are commonly connected to the power system to enhance its reliability by providing voltage support, improving power factor, and increasing the system capacity. They are usually switched in and out of the system according to the amount of connected load. This paper analyzes the measurement data of a distribution utility with shunt capacitor banks located downstream from the power quality monitor. An algorithm is proposed to identify both energizing and de-energizing capacitor switching events. It is tested in a dataset of 829 files, containing 358 energizing events, 8 de-energizing events, recloser operation during fault, and files with no transient event. Both capacitor energizing and de-energizing identification algorithms were observed to achieve high accuracy ratios, with a small number of misclassifications.

Power quality disturbances commonly observed in power systems have been studied for decades, resulting in numerous algorithms for detecting the events that affect the voltage and/or current waveforms. However, a considerable amount of disturbances is not visually observable in the raw waveforms, especially switching operations. These events must be detected through an alternative feature, such as abrupt variations in the root‐mean‐square (rms) voltage profile. This study examines the methods commonly used for detecting power quality disturbances in the waveform or rms voltage profile domains and identifies their limitations. Afterwards, a novel step change detector is proposed based on a modified median filter and rms voltage gradient values to overcome the deficiencies of the existing methods. The effectiveness of the proposed method is assessed by applying it to both simulated and field data. This assessment shows that the method detects all switching events with no false positives for the datasets under analysis.