DETECCIÓN E IDENTIFICACIÓN DE EVENTOS DE LA CALIDAD DE LA ENERGÍA ELÉCTRICA UTILIZANDO LA TRANSFORMADA WAVELET DISCRETA Y REDES NEURONALES ARTIFICIALES
DETECTION AND IDENTIFICATION OF POWER QUALITY EVENTS BASED ON DISCRETE WAVELET TRANSFORM (DWT) AND NEURAL NETWORKS
Palabras clave:
Armónicos, calidad de la energía eléctrica, elevaciones de tensión (swells), fluctuaciones de tensión (flicker), huecos de tensión, monitorización, redes neuronales, transformada Wavelet discreta, transitorios. (es)This paper deals about an application of Discrete Wavelet Transform (DWT) and Neural Network in detection and identification of power quality events. Some patterns based on DWT are used in order to identify low frequency events like flicker and harmonics, and high frequency events like impulsive transient and sags. The Wavelet Function Daubichies4 is used as a base function because of its frequency response and time information localization properties. A scheme based on neural networks (perceptron multilayer) taking event patterns as inputs is used as event classifier. The results are satisfactory (80 and 90 percent of success for the most events) considering that some events present resemblances in their patterns. This strategy was integrated on a MatLab ® Graphical User Interface and tested by using synthetic signals which were simulated and collected in a disturbance database.
This paper deals about an application of Discrete Wavelet Transform (DWT) and Neural Network in detection and identification of power quality events. Some patterns based on DWT are used in order to identify low frequency events like flicker and harmonics, and high frequency events like impulsive transient and sags. The Wavelet Function Daubichies4 is used as a base function because of its frequency response and time information localization properties. A scheme based on neural networks (perceptron multilayer) taking event patterns as inputs is used as event classifier. The results are satisfactory (80 and 90 percent of success for the most events) considering that some events present resemblances in their patterns. This strategy was integrated on a MatLab ® Graphical User Interface and tested by using synthetic signals which were simulated and collected in a disturbance database.