Thesis IMPLEMENTACIÓN DE CONTROLADOR INTELIGENTE PARA EL CONTROL AUTOMÁTICO DE LA GENERACIÓN Y SU APLICACIÓN AL SISTEMA SING-SADI
Loading...
Date
2018
Authors
ACUÑA GONZÁLEZ, FELIPE IGNACIO
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Actualmente en Chile ya se ha implementado el uso de control automático de generación (AGC por sussiglas en inglés) para hacer regulación secundaria de frecuencia. Pero hasta hace no mucho este proceso sehacía de forma manual, lo que derivaba en no cumplimientos de las exigencias de la NTSyCS en cuanto a losrangos deseables de frecuencia. La automatización de este proceso permitió mayores velocidades de respuesta,ya que no sería necesaria la intervención humana, lo cual desencadena en mayor calidad de suministrode energía hacia los clientes.La mayoría de los modelos de AGC contemplan controladores PID, ya que se ha demostrado a lo largo dela historia que cumplen el objetivo de controlar el sistema sin estudios exagerados sobre sus parámetros. Sinembargo, estos controladores están sujetos a los puntos de operación en los que se encuentran, por lo quesería necesario cambiar los parámetros dependiendo de la situación.Durante los últimos años, gracias a la mejora de procesamiento de los computadores, además de un granaumento en el interés sobre estructuras de control no convencionales, han aparecido nuevas herramientas ycontroladores basados en inteligencia artificial.Precisamente este trabajo pretende demostrar que los métodos de inteligencia artificial pueden ser aplicadosal control de generación automática, y determinar si se puede obtener un mejor desempeño al serutilizados en el modelo de un sistema real como lo es el sistema interconectado. Para ello se contempla primeroun desarrollo de un marco teórico explicado en el capítulo 2, que permite entender el funcionamientodel AGC, la lógica difusa (fuzzy logic), las redes neuronales y sus métodos de aprendizaje, para finalmentedescribir el funcionamiento de las estructuras ANFIS. Seguidamente, en el capítulo 3, se tiene una brevedescripción de las implementaciones del AGC alrededor del mundo, además de usos de la inteligencia artificialdentro del área eléctrica. Posteriormente, en el capítulo 4, se parte implementando un modelo mínimofuncional en Simulink, para luego pasar al diseño en un caso IEEE de 14 barras en Digsilent que contempleoptimizaciones, ajustes pertinentes, y modificaciones finas, que finalmente puedan ser replicados sin mayoresproblemas en el modelo SING-SADI, que claramente es el que ocupará más poder de procesamiento.Finalmente en los capítulos 5 y 6 se tienen los resultados del sistema real, tanto para el SING, como para elsistema interconectado.Obtenidos los controladores correctos y ajustados en el caso de SING-SADI, se hicieron comparacionespertinentes con una configuración teórica del controlador PI a ser usado, donde finalmente se llegó a la conclusiónque los controladores inteligentes funcionan de buena forma, logran controlar el sistema, pero hayque tener varias consideraciones presentes a la hora de utilizar cada uno de ellos. Durante este ensayo sepudo notar que solo uno de los controladores -ANFIS- mejoró la situación actual, debido a sus grandes condicionesde adaptabilidad, que le permiten reformular sus rangos operativos de buena manera.
Nowadays, AGC has already been implemented to do secondary frequency control. However, not too longago, this process was made manually, so this used to mean that the frequency was not meeting the requirementsregarding to the desirable operative ranges. Automatization in this matter allowed better responsespeeds, because human intervention wouldn’t be necessary anymore. This also resulted in better powersupply quality towards the client.Most AGCmodels use PID controllers, because they have shownall over its existence that they are successfulcontrolling systems without any exaggerated study beforehand over its parameters. However, this goodbehavior is restricted to an operation range, so it would be preferable to change its parameter dependingon the situation. During the last few years, thanks to greater computing power and the constantly growinginterest in nonstandard control schemes, new tools and AI controller have shown up.Actually, this research pretends to show that AI methods can be applied to AGC, and determine if a betterperformance can be achieved when being used in a real model like SING-SADI connection. For that, thiswork pretends to explain everything to someone not familiar with this kind of logic, starting in chapter 2where it explains how AGC works, fuzzy logic, neural networks and its learning techniques, and ANFIS logicand how it works (available at the appendix). After this, in chapter 3, there’s a brief description about AGC inUSA, and then, short explanations about how AI is being used in Power System Industry. Later, in chapter 4,this research starts showing implementation using a minimalistic model in Simulink so it can be used laterin 14-bus IEEE Digsilent’s model, so it can also apply optimizations and custom tunning to it, that can bealso used in SING-SADI model, that is going to be the one with the most processing power required. Finallyin chapters 5 and 6, real system results are shown, considering both cases, for SING alone, and SING-SADIinterconnection.Once correct controllers are obtained and tunned for the SING-SADI case, comparissons with theoricaltunning for PI controller aremade, where it concluded that smart controllers work in proper way, achieve thegoal of controlling the system, but severals issues must be adressed before using any of them. During this testit can be noticed that only one of the controllers -ANFIS- improved the current situation, thanks to its greatadapting qualities, that let it change their operating ranges in proper way.
Nowadays, AGC has already been implemented to do secondary frequency control. However, not too longago, this process was made manually, so this used to mean that the frequency was not meeting the requirementsregarding to the desirable operative ranges. Automatization in this matter allowed better responsespeeds, because human intervention wouldn’t be necessary anymore. This also resulted in better powersupply quality towards the client.Most AGCmodels use PID controllers, because they have shownall over its existence that they are successfulcontrolling systems without any exaggerated study beforehand over its parameters. However, this goodbehavior is restricted to an operation range, so it would be preferable to change its parameter dependingon the situation. During the last few years, thanks to greater computing power and the constantly growinginterest in nonstandard control schemes, new tools and AI controller have shown up.Actually, this research pretends to show that AI methods can be applied to AGC, and determine if a betterperformance can be achieved when being used in a real model like SING-SADI connection. For that, thiswork pretends to explain everything to someone not familiar with this kind of logic, starting in chapter 2where it explains how AGC works, fuzzy logic, neural networks and its learning techniques, and ANFIS logicand how it works (available at the appendix). After this, in chapter 3, there’s a brief description about AGC inUSA, and then, short explanations about how AI is being used in Power System Industry. Later, in chapter 4,this research starts showing implementation using a minimalistic model in Simulink so it can be used laterin 14-bus IEEE Digsilent’s model, so it can also apply optimizations and custom tunning to it, that can bealso used in SING-SADI model, that is going to be the one with the most processing power required. Finallyin chapters 5 and 6, real system results are shown, considering both cases, for SING alone, and SING-SADIinterconnection.Once correct controllers are obtained and tunned for the SING-SADI case, comparissons with theoricaltunning for PI controller aremade, where it concluded that smart controllers work in proper way, achieve thegoal of controlling the system, but severals issues must be adressed before using any of them. During this testit can be noticed that only one of the controllers -ANFIS- improved the current situation, thanks to its greatadapting qualities, that let it change their operating ranges in proper way.
Description
Catalogado desde la version PDF de la tesis.
Keywords
AGC , INTELIGENCIA ARTIFICIAL , SISTEMA SING-SADI