Equalização adaptativa de transformadores de medição para aplicações em qualidade de energia elétrica

Abstract

This work presents a proposal to implement a blind equalization scheme in instrumentation transformers, aiming to correct harmonic measurements through the application of adaptive filtering techniques and genetic algorithms. Blind adaptive equalization, also known as deconvolution, is a challenging task but has recently emerged as a promising tool for cases where there is no direct access or prior knowledge of the signal of interest in a system. In power systems, the signals captured in measurement transformers (such as voltage and current transformers) are originally sinusoidal but can be distorted by harmonics and white noise. These transformers generally introduce significant distortions, especially in high-order harmonic components. Transformers with a constant response in the frequency band of interest (up to the 50th harmonic, for example) are very expensive. Most methods discussed in the literature for correcting the frequency response of transducers are based on the frequency domain and require laboratory testing, making it impractical for equalizing already installed transducers. Furthermore, the frequency response of these transformers is influenced by factors such as temperature and load impedance, often making the laboratory calibration process inadequate. Therefore, a real-time blind equalization approach is crucial for harmonic measurement. However, many transformer models have zero gain frequency response at specific frequencies, making it challenging to use adaptive FIR (Finite Impulse Response) filters. To overcome this obstacle, a filter with a fixed pole is introduced into the equalization process, and the pole’s location is identified through genetic algorithms (GA). Thus, adaptive algorithms based on FIR filters, such as Least Mean Squares (LMS) or Recursive Least Squares (RLS), can be applied in the equalization process. Additionally, for conventional methods to be employed in equalizing current and voltage meters, signal preprocessing is necessary. This study successfully adapted techniques commonly used in data communication systems equalization, overcoming specific challenges, in order to make their application viable in current and voltage transducers equalization in power distribution and transmission systems, presenting promising results in correcting errors of instrumentation transformers.