Análisis y caracterización de fallos en redes de media tensión

Abstract

La presente tesis doctoral tiene como objetivo principal la estimación de los parámetros de confiabilidad mediante el uso de bases de datos de averías en redes eléctricas de media tensión. Históricamente, las redes de transporte han captado una mayor atención a la hora de analizar la confiabilidad del sistema, aspecto crucial en la calidad del servicio a los clientes. En este sentido, existe un consenso global en lo que respecta a la carencia de información sobre los parámetros de confiabilidad en las redes de distribución. El principal motivo de esta situación es la falta de bases de datos sólidas que registren de manera adecuada las averías que ocurren en estas redes y los diversos componentes que las conforman, dificultando la estimación de los parámetros de confiabilidad. Además, el hecho de que estos parámetros sean dependientes de las características particulares de cada red afecta a la precisión de los análisis a realizar. Actualmente, con el avance de los sistemas de información y el mejor almacenamiento de datos, los métodos probabilísticos están ganando relevancia en la evaluación de la confiabilidad de sistemas eléctricos. Estos métodos resultan ser adecuados para modelar las variaciones difícilmente predecibles de los parámetros de confiabilidad mediante el uso de distribuciones de probabilidad. Por lo tanto, el desarrollo de metodologías que proporcionen valores estimados de estos parámetros resulta de gran interés. Para evaluar la continuidad y disponibilidad del suministro eléctrico, tanto las distribuidoras como las entidades regulatorias utilizan distintos índices de calidad, tales como el SAIFI, SAIDI, NIEPI y TIEPI. En este sentido, las distribuidoras tienen la obligación de cumplir con ciertos valores de estos índices, pudiendo ser penalizadas en caso contrario. Es por ello por lo que, además de evaluar la confiabilidad del sistema, las estimaciones de los parámetros de confiabilidad permiten mejorar las estimaciones de los índices de calidad. Los parámetros de confiabilidad más comúnmente utilizados y que cuantifican estos conceptos son las tasas de fallos y el tiempo de reposición del suministro. Para estimar con mayor precisión estos parámetros de confiabilidad, es fundamental considerar las particularidades de la zona y las características de la red. Por este motivo, para el cumplimiento de los objetivos marcados en la presente tesis doctoral se han usado dos bases de datos. La primera de ellas incluye información sobre las características principales de los elementos que forman la red, mientras que el otro registro contiene datos sobre los incidentes que han ocurrido en la red. Con respecto a la estimación de las tasas de fallos, se propone una metodología que permite su estimación para distintos componentes de la red. Esta metodología se basa en un análisis de las averías acaecidas en la red, así como de los elementos que la componen. Se han considerado diferentes grupos representativos de elementos, en función de su funcionalidad y nivel de tensión dentro de la red de distribución. Para el caso del tiempo de reposición del suministro, se ha desarrollado otra metodología basada en un algoritmo capaz de ajustar los datos muestrales provenientes de los registros con diferentes funciones de densidad de probabilidad para las diferentes agrupaciones de componentes realizadas. Además, un análisis exhaustivo de la base de datos utilizada en esta tesis ha permitido considerar una gama más amplia de parámetros más allá de las tasas de fallo y tiempos de reposición. Finalmente, las metodologías desarrolladas han sido implementadas en una metodología completa que, además de estimar los parámetros de confiabilidad, permite evaluar los índices de calidad de una red eléctrica mediante el uso de un método consensuado, siendo en este caso la simulación Montecarlo. Carecería de sentido establecer procedimientos para estimar los parámetros de confiabilidad que no sea capaz de integrarse en modelados ya empleados en el cálculo de la fiabilidad. En este caso el. Todos los estudios incluidos y procedimientos establecidos están respaldados por artículos científicos que han sido publicados o están en proceso de redacción para ser presentados en revistas de alto impacto dentro del campo correspondiente.

The main objective of this doctoral thesis is to estimate reliability parameters using real fault data from medium voltage electrical distribution networks. Historically, transmission networks have garnered greater attention when analyzing system reliability, a crucial aspect in delivering quality service to customers. In this regard, there is a global consensus regarding the lack of information on reliability parameters in distribution networks. The primary reason for this situation is the absence of robust databases that adequately record breakdowns occurring in these networks and their various components, making it challenging to estimate reliability parameters. Furthermore, the fact that these parameters depend on the specific characteristics of each network affects the accuracy of the analyses to be conducted. For a more precise calculation of reliability parameters, it's crucial to consider the specificities of the area and the network's characteristics. Therefore, to achieve the goals outlined in this doctoral thesis, two real databases provided by a distribution company have been utilized. The first one encompasses information about the primary characteristics of the network elements, while the other dataset contains information on the incidents that occurred in the network between 2001 and 2013. With the advancement of information systems and improved data storage, probabilistic methods are gaining significance in assessing the reliability of electrical systems. These methods prove suitable for modeling the hardly predictable variations in reliability parameters by employing probability distributions. Hence, the development of methodologies that provide estimated values for these parameters becomes of great interest. Reliability and availability stand as key concepts essential for ensuring a The main objective of this doctoral thesis is to estimate reliability parameters using real fault data from medium voltage electrical distribution networks. Historically, transmission networks have garnered greater attention when analyzing system reliability, a crucial aspect in delivering quality service to customers. In this regard, there is a global consensus regarding the lack of information on reliability parameters in distribution networks. The primary reason for this situation is the absence of robust databases that adequately record breakdowns occurring in these networks and their various components, making it challenging to estimate reliability parameters. Furthermore, the fact that these parameters depend on the specific characteristics of each network affects the accuracy of the analyses to be conducted. For a more precise calculation of reliability parameters, it's crucial to consider the specificities of the area and the network's characteristics. Therefore, to achieve the goals outlined in this doctoral thesis, two real databases provided by a distribution company have been utilized. The first one encompasses information about the primary characteristics of the network elements, while the other dataset contains information on the incidents that occurred in the network between 2001 and 2013. With the advancement of information systems and improved data storage, probabilistic methods are gaining significance in assessing the reliability of electrical systems. These methods prove suitable for modeling the hardly predictable variations in reliability parameters by employing probability distributions. Hence, the development of methodologies that provide estimated values for these parameters becomes of great interest. Reliability and availability stand as key concepts essential for ensuring a consistent supply, distinguished as one of the fundamental pillars of electrical service quality. The most commonly used reliability parameters that quantify these concepts are failure rates and supply restoration time. However, a thorough analysis of the database utilized in this thesis has allowed for consideration of a broader range of these parameters. Regarding the estimation of failure rates, a methodology is proposed to estimate these rates for different components of the network. This methodology is based on an analysis of breakdowns occurring in the network and its constituent elements. Various representative groups of elements have been considered based on their functionality and voltage level within the distribution network. Through this process, the mean, minimum, and maximum values of the failure rates for each grouping can be obtained for the study period. Despite having been considered constant in several studies, failure rates might exhibit variation influenced by numerous factors. Analyzing this potential trend provides valuable information for decision-making and effective reliability management of the system. Therefore, the obtained failure rates have undergone the Laplace test to verify their trend within the studied period. The results indicate that, in most cases, failure rates should not be assumed constant over time. For the case of supply restoration time, an algorithm has been developed to fit sample data from records with different probability density functions. The selection of the function that best fits the data is determined by applying various goodness-offit tests. Initially, this procedure was used to estimate the probability functions of electrical supply restoration times in medium-voltage underground cables. Once validated, the procedure was applied to the remaining considered element groupings. Subsequently, this same methodology has been utilized to estimate other reliability parameters, such as localization time, network reconfiguration, start of repair, repair duration, return to normalcy, and time between failures. To assess the continuity and availability of electrical supply, both distribution companies and regulatory entities use various key performance indicator such as SAIFI, SAIDI, NIEPI, and TIEPI. In this context, distribution companies are obligated to meet certain values for these indicators and could face penalties if they fall short. Therefore, besides evaluating system reliability, estimates of reliability parameters enable improvements in estimating these indicators. Based on the aforementioned utility, leveraging the breakdown data stored by the distribution company, a methodology is introduced. It's based on Monte Carlo simulations to assess the economic risk associated with failing to meet specific indicators. This methodology was applied to a real network comprising 13 medium-voltage lines, consisting of over 100 km of underground cables and 20 km of overhead lines, serving 18,000 customers with a total contracted power close to 110 MW. Apart from obtaining probability functions for the number and duration of interruptions, a fitted function was also achieved for the SAIFI and SAIDI indicators. Initially, these indicators were used to compare the performance of the lines, determining which had lower performance, enabling the establishment of new network configuration strategies. Subsequently, these functions were used to assess the economic risk arising from potential penalties incurred by the distribution companies. The evaluation revealed that in 37.43% of cases, the distributor would receive incentives for the network's good performance, while in 33.44% of cases, the distributor would face penalties for poor network performance. The results showcased throughout this doctoral thesis have led to the publication of two scientific articles in a high-impact journal, as will be detailed in the conclusions of this document.