Metalized Polymer-Film Capacitors Health Estimation for Three-Phase DC to AC Converters with Artificial Intelligences

Abstract

Metalized polymer-film capacitors have attained a unique role amongst numerous types of capacitors due to their self-healing ability. However, owing to the increasing usage of converters for transmissions in networks, it is necessary to enhance stability to certify the safety of system operations. Thus, a monitoring process that enables predictive maintenance is essential for evaluating the health status to ensure the stability of electrical systems. This study presents a condition monitoring strategy that utilizes frequency signal analysis to identify capacitor health in a three-phase AC-DC converter. Capacitor current is analyzed using the discrete wavelet transform and normalized by indexes, which is used as input of learning algorithms. A number of indexes, such as root-mean-squared value, variance, average, and median, are behaved as the inputs of artificial intelligent models to examine the film capacitor's factors. The estimated parameters verify that applying the discrete wavelet transform conjoined with indexes for capacitor current provides an excellent accuracy of about 99.85%. This study provides comprehensive insights into film capacitor monitoring using advanced techniques and is expected to be informative for monitoring film capacitors in practical applications.