A Synchronized Lissajous-Based Method to Detect and Classify Events in Synchro-Waveform Measurements in Power Distribution Networks

Abstract

Waveform measurement units (WMUs) are a new class of smart grid sensors. They capture <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">synchro-waveforms</i> , i.e., time-synchronized high-resolution voltage waveform and current waveform measurements. In this paper, we propose new methods to <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">detect</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">classify</i> power quality events in power distribution systems by using synchro-waveform measurements. The methods are built upon a novel graphical concept, called <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">synchronized Lissajous curve</i> . The proposed event detection and event classification methods work by analyzing the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">shape</i> of the synchronized Lissajous curves during disturbances and events. The impact of challenging factors, such as the angle, the location, and other parameters of the event are discussed. We show that these challenges can be addressed if we treat the synchronized Lissajous curves as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">images</i> , instead of as time series as in the raw synchronized waveform measurements. Hence, we can take advantage of the recent advancements in the field of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">image processing</i> so as to capture the overall characterizing patterns in the shapes of the synchronized Lissajous curves. We develop a Convolutional Neural Network (CNN) method to classify the events, where the input is the synchronized Lissajous images. The effectiveness of the proposed event detection and classification methods is demonstrated through computer simulations, including hardware-in-the-loop simulations, and real-world field data. Multiple case studies verify the performance of the proposed methods. The proposed event detection method can accurately detect events, and identify the start time and the end time of each event. The proposed event classification method can classify power quality events with high accuracy. The proposed detection and classification methods do <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">not</i> require any prior knowledge about the network. They use data from as few as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">only two</i> WMUs.