Thesis ANÁLISIS DE POLÍTICA DE MANTENIMIENTO PREDICTIVO PARA TRANSFORMADORES DE PODER.
Date
2020-11
Authors
CELEDÓN BRICEÑO, CAMILA IGNACIA
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
El an´alisis de vida de los activos de una empresa resulta importante en el campo de investigaci´on de la gesti´on de activos de ingenier´ıa. En la actualidad, los transformadores de poder son fundamentales en la industria el´ectrica, ya que de estos depende el desarrollo del pa´ıs a nivel econ´omico y social. Adem´as, en caso de existir fallas en el suministro el´ectrico la empresa se debe hacer cargo retribuyendo aquella energ´ıa no suministrada al usuario.
Debido a esto, es fundamental asegurar el funcionamiento de los elementos de transmisi´on de energ´ıa, siendo la clave mejorar la estimaci´on de salud de los transformadores de poder que permitan generar una pol´ıtica de mantenimiento adecuada que aumente su confiabilidad.
Para esto, mediante el an´alisis de datos entregados por una empresa el´ectrica, se busca determinar los par´ametros internos y/o externos de mayor relevancia, que servir´an como input al desarrollar una pol´ıtica de intervenci´on ´optima para transformadores de poder, a trav´es de un modelo Weibull PHM, para calcular y graficar la confiabilidad condicional, vida u´til remanente y curva de decisi´on, mejorando la estimaci´on de salud de los equipos.
Mediante estimadores de m´axima verosimilitud, se obtienen los par´ametros que describen el compor- tamiento estoc´astico de la funci´on de confiabilidad condicional, siendo estos β, η y γ′s, para las covariables temperatura interna (Tint), variaci´on de temperatura (∆T ) y grado de carga (%Carga). La ponderaci´on
y la relevancia de las covariables medidas permite determinar un modelo Weibull PHM, a trav´es de un an´alisis Log-Likelihood Ratio Test para las diferentes combinaciones. Se llevan a cabo dos casos de estudios, el primero es un modelo con solo una covariable, donde aquella con mayor influencia es el grado de carga, y el segundo con m´as de una covariable, donde la combinaci´on de las tres covariables son las que afectan de mayor manera, de esta forma se procede a analizar la confiabilidad condicional y RUL.
Al realizar las curvas de confiabilidad condicional, se tiene una vida u´til remanente inicial de hasta 130.622[h] para el caso 1 y 40.969[h] para el caso 2, ambos para un estado 1 de buen funcionamiento, que se ven disminuidos a la hora de trabajar la sensibilidad de los l´ımites de funcionamiento. Al realizar la curva de decisi´on, junto con los valores de testeo, se determina que en ambos casos analizados los equipos deben ser intervenidos inmediatamente.
Lo transversal del estudio permite tomar los datos desde su origen y obtener resultados que permiten
tomar una ´optima decisi´on de intervenci´on. Los resultados indican que la etapa de tratamiento y an´alisis de datos resulta la m´as importante para los posteriores c´alculos, adem´as, la definici´on de estados influyen significativamente en las curvas calculadas debido a la naturaleza de los datos. Finalmente, se concluye que el monitoreo de condiciones se hace vital a la hora implementar una pol´ıtica de mantenimiento ´optima.
La gran cantidad de informaci´on disponible debido a distintos ensayos y los ´ambitos en que influye los transformadores de potencia abre varias ´areas de futuras investigaciones, como expandir resultados de ensayos a otros puntos temporales o los ahorros o costos extras en otras ´areas de las empresas el´ectricas, m´etodos de procesamiento de datos aplicados a una pol´ıtica de mantenimiento ´optima, tales como machine learning, big data, an´alisis de inventario de repuestos o del activo mismo con respecto a su pol´ıtica de mantenimiento, entre otros posibles trabajos
Lifetime analysis of a company’s assets is an important research field of engineering asset management. Today, power transformers are essential in electrical industries, since the country’s development on a social- economic level depends on it. Also, in case there is a failure in the electrical supply, the company must pay for that energy not supplied to the user. Due to this, it is essential to ensure the operation of the power transmission elements, the key being to improve the health estimate of the power transformers that will allow generating an adequate maintenance policy that increases their reliability. For this, through the analysis of data provided by an electric company, it is sought to determine the most relevant internal or external parameters, that will serve as input to develop an optimal intervention policy for power transformers is carried out, through the development of a Weibull PHM model to calculate and graph the conditional reliability and remaining useful life, improving the health estimation of the equipment. Using maximum likelihood estimators, the parameters that describe the stochastic behavior of the condi- tional reliability function are presented, being these β, η and γ′s, for the internal covariates (Tint), tempera- ture variation (∆T ) and the degree of charge (%Charge). The weighting and the relevance of the measured covariates that allow determining a Weibull PHM model, through an analysis textit Log-likelihood relation- ship test for the different combinations. Two case studies are carried out, the first is a model with only one covariate, where the one with the greatest influence is the degree of load, and the second with more than one covariate, where the combination of the three covariates are the ones to see from In a greater way, in this way we proceed to analyze the conditional reliability and RUL. When performing the conditional reliability curves, there is an initial remaining useful life of up to 130, 622[h] for case 1 and 40, 969[h] for case 2, both for a state 1 of good operation, which is decrease when working on the sensitivity of the operating limits. When making the decision curve, together with the test values, it determined that in both cases analyzed, the equipment must be intervened immediately. The cross-sectional nature of the study makes it possible to take the data from its source and obtain results that allow an optimal intervention decision to be made. The results indicate that the data treatment and analysis stage is the most important for subsequent calculations, also the definition of states significantly influences the calculated curves due to the nature of the data. Finally, it concluded that the monitoring of conditions is vital when implementing an optimal maintenance policy. A large amount of information available due to different tests and the fields in which power transformers influence opens several areas for future research, such as expanding test results to other time points or the savings or extra costs in other areas of the electricity companies, data processing methods applied to an optimal maintenance policy, such as machine learning, big data, spare parts inventory analysis or the asset itself for its maintenance policy, among other possible jobs.
Lifetime analysis of a company’s assets is an important research field of engineering asset management. Today, power transformers are essential in electrical industries, since the country’s development on a social- economic level depends on it. Also, in case there is a failure in the electrical supply, the company must pay for that energy not supplied to the user. Due to this, it is essential to ensure the operation of the power transmission elements, the key being to improve the health estimate of the power transformers that will allow generating an adequate maintenance policy that increases their reliability. For this, through the analysis of data provided by an electric company, it is sought to determine the most relevant internal or external parameters, that will serve as input to develop an optimal intervention policy for power transformers is carried out, through the development of a Weibull PHM model to calculate and graph the conditional reliability and remaining useful life, improving the health estimation of the equipment. Using maximum likelihood estimators, the parameters that describe the stochastic behavior of the condi- tional reliability function are presented, being these β, η and γ′s, for the internal covariates (Tint), tempera- ture variation (∆T ) and the degree of charge (%Charge). The weighting and the relevance of the measured covariates that allow determining a Weibull PHM model, through an analysis textit Log-likelihood relation- ship test for the different combinations. Two case studies are carried out, the first is a model with only one covariate, where the one with the greatest influence is the degree of load, and the second with more than one covariate, where the combination of the three covariates are the ones to see from In a greater way, in this way we proceed to analyze the conditional reliability and RUL. When performing the conditional reliability curves, there is an initial remaining useful life of up to 130, 622[h] for case 1 and 40, 969[h] for case 2, both for a state 1 of good operation, which is decrease when working on the sensitivity of the operating limits. When making the decision curve, together with the test values, it determined that in both cases analyzed, the equipment must be intervened immediately. The cross-sectional nature of the study makes it possible to take the data from its source and obtain results that allow an optimal intervention decision to be made. The results indicate that the data treatment and analysis stage is the most important for subsequent calculations, also the definition of states significantly influences the calculated curves due to the nature of the data. Finally, it concluded that the monitoring of conditions is vital when implementing an optimal maintenance policy. A large amount of information available due to different tests and the fields in which power transformers influence opens several areas for future research, such as expanding test results to other time points or the savings or extra costs in other areas of the electricity companies, data processing methods applied to an optimal maintenance policy, such as machine learning, big data, spare parts inventory analysis or the asset itself for its maintenance policy, among other possible jobs.
Description
Keywords
MANTENIMIENTO PREDICTIVO , PHM , TRASFORMADORES
