Uma nova proposta para detecção e correção das distorções nas formas de onda de corrente secundária em TC’s saturados: uma abordagem utilizando RNA’s

Abstract

Protection relays depend on the quality of the output signal from current transformers (CTs) for their correct operation. However, in some conditions, the increase in magnetic flux in the CT core can lead to core saturation, causing many problems in the protection of Electrical Power Systems (EPS). For this reason, this work presents a methodology for detecting and correcting the secondary waveform of a saturated CT with a transformation ratio of 2000:5. The correction algorithm is based on the use of Artificial Neural Networks (ANN), which has as its output the unknown parameters of the fault current coming from the Least Squares Method. A method to detect the instant that saturation occurs is presented and is based on the third difference function technique. The analyzed signals were obtained via simulations from a test system implemented in the ATPDraw software, where different saturation conditions were studied. A database was generated to use in ANN training, which was obtained based on changes in the test system, such as: type of short circuit, fault angle and remaining flux in the CT, with a total of 224 cases analyzed. The methods were implemented in the MATLAB software, in which the results point to a robust algorithm, capable of reconstructing the distorted signal even in cases of high saturation. Therefore, the proposed methodology will contribute to improving EPS protection systems, increasing the reliability of CTs by suppressing the effect of saturation after the occurrence of a fault. Key-words: Current Transformers. Electric Power Systems. Artificial Neural Networks. Saturation. Detection. Correction.