Metodologia para a detecção e correção de outliers em curvas de potência de subestações utilizando técnicas de inteligência artificial

Abstract

One of the main problems of the data acquired by the power utilities is the presence of outliers that affect the database of the measurements throughout the electrical system, damaging the analyzes of the distribution scenario. This work proposes a new module to complement the measurements made by the distributors. A detection technique and three outliers correction techniques were developed, based on fuzzy logic, artificial neural networks and ARIMA. The first technique, with a fuzzy approach, develops an inference system based on the variations of the previous 3 measurements to determine the future variation. In the second algorithm developed using RNA, the outliers were corrected using a prediction model using 10 previous samples. The last correction technique was based on the autoregressive model of the ARIMA type with 96 previous measurements. In order to demonstrate the applicability of the developed methods, a case study was carried out on a real substation in a city in the interior of Paraíba. The three techniques of correction of the outliers presented average relative error less than 5% for all the test scenarios.