Detector de novidades de sinais de qualidade de energia baseado em transformada de stockwell utilizando um processador soft core embarcado em plataforma FPGA

Abstract

With the increasing implementation of Smart Grids (SGs), where the presence of non-linear loads and new energy generation sources become increasingly prominent, there is significant potential for the occurrence of unknown disturbances. In the midst of this scenario, the exponential volume of data from smart meters requires efficient management to preserve relevant information and promote effective data storage. In this context, Novelty Detection (ND) techniques emerge as a promising solution. The Stockwell Transform (ST) stands out as a time-frequency distribution that demonstrates remarkable ability in detecting novelties associated with changes in signal stationarity, applicable to both voltage and current signals in the power grid. This work describes, therefore, the use of ST for novelty detection in Power Quality (PQ) signals. To enable implementation on FPGA platforms, the adoption of a soft-core processor is proposed to optimize available hardware resources. Furthermore, a voice selection strategy is proposed to reduce the complexity and execution time of the algorithm in real-time implementations while maintaining detection capability. The proposed ND technique was implemented in a functional prototype based on FPGA, using both synthesized and real signals recorded in the field, thus proving the feasibility of equipment development. Through the voice selection method, it was possible to reduce the processing time of a 66.6ms duration frame from 109ms to 8.9ms, at a rate of 10MHz, enabling real-time implementation of the detector.