Abstract ：In this paper, the design and the prototype implementation of a Power Quality (PQ) disturbance detector and compressor are described. This instrument, named Power System Smart Waveform Recorder (PSSWR), is able to acquire the samples of the Electric Power System (EPS) signals and process them in order to detect, compress and store the disturbances waveforms into an SD card, from which it can be reconstructed and analyzed offline with a suitable computer application. The prototype uses, among other devices, Field Programmable Gate Array (FPGA) and ARM platforms to work with the EPS signals in a smart way. The PSSWR is able to detect and record either the waveform of the well-known PQ dis- turbance as well the new PQ disturbances not yet observed in EPS, thanks to the Novelty Detection con- cept. This characteristic is important in the new context of smart grids where hidden disturbances can be detected by the methodology.

© 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Nowadays there is a growing concern about Power Quality (PQ) problems. These problems are due to the massive use of non-linear loads and electronic-based equipment in residences, commercial centers, and industrial plants, and the proliferation of distributed generation in the Electric Power System (EPS). Therefore, the mon- itoring of EPS in real-time, along with off-line analysis using both centralized and decentralized schemes, has grown in importance [1].

In several applications the continuous data acquisition and stor- age are necessary. This is sustained by the fact that the post- processing of this data can uncover information not previously observed, allowing system enhancement, troubleshoot, algorithm optimization, among others. For example, high frequency distur- bances like transient oscillations or switching processes are only visible in the full waveform, the signature hidden in raw data can be used to predict the breaking of cables, etc. Although aggregation is useful for data reduction and comparison, deep data analysis should be enabled considering full observation possibilities   [2].

However, continuous raw data recordings of electrical signal is not a simple task due the large amount of data to be recorded    and

later transferred. Besides, few commercial types of equipment are currently available aiming at recording continuous raw data at high sampling rate [3]. Most of the conventional recorders are application oriented and are used only for acquiring either a short-term of failure signal or disturbance signal [4,5]. A thorough survey of the leading manufacturers of PQ analyzers was made. Nine of the top brands were examined, totaling 27 devices. From the investigation of their manuals and data sheets, it was observed that all of them are able to record the PQ parameters (data) for a long period of time. Depending on the aggregation time, some equipment are able to record over than a year of PQ parameters. Nevertheless, only two of them are able to save waveform record- ings for a long period. The one described in [6] was used in this paper for a comparison purpose, because it is able to recording about 1 year of gapless waveform using a compression technique. The system described in this paper, named Power System Smart Waveform Recorder (PSSWR), is able to reconstruct the entire waveform signal, acquired at a high sample rate. However, a con- tinuous raw data recording is not employed. Instead, only samples around the detected electrical disturbance are packed, compressed and recorded for offline reconstruction of the whole signal. Apart those samples, an information about the frequency estimation of each cycle of the signal is also recorded. These informations are used to produce a smooth signal between two disturbances, recov- ering the entire electrical signal at any time, either when the  fre-

quency is time varying.

To further reduce the amount of data to be stored or transmit- ted, a compression algorithm is added to the system. The compres- sion consists in eliminating the redundant information present in the signal. In general it is done trough a transform. The Discrete Wavelet Transform (DWT) [7–9] has been extensively used to this purpose due to its ability to sparse the   signal.