Artificial neural networks for the prediction of power flows applied to the Uruguayan transmission system.
Guardado en:
| Autores: | , , |
|---|---|
| Formato: | artículo original |
| Fecha de Publicación: | 2021 |
| Descripción: | In the present work, the use of artificial neural networks is proposed to solve the load flow problem. The study of the load flow of the electrical network constitutes a fundamental tool for the operation and planning of an electrical system. The load flow problem is solved by a system of non-linear equations. For this resolution, numerical methods have traditionally been used, mainly the Newton-Raphson method and its variants. These numerical methods applied to large electrical systems are very expensive in terms of computational cost. Solving a considerable number of load flows using these methods involves incurring in execution times that are prohibitive in studies of the electrical network. This problem becomes critical in contingency case studies, even using the simple N-1 contingency criterion. The construction of neural networks that approximate the resolution of load flows allows to significantly reducing the execution time of the aforementioned studies. In this work, the design of a neural network architecture for the approximation of load flows is proposed. Using the designed architecture, a load flow approximation model is implemented. The validation of the tool is carried out using the Uruguayan transmission network. The approximation obtained for this case study is evaluated by applying the MAPE metric and a value of 2.6% is obtained, which constitutes a very promising result. |
| País: | RepositorioTEC |
| Institución: | Instituto Tecnológico de Costa Rica |
| Repositorio: | RepositorioTEC |
| Lenguaje: | Español |
| OAI Identifier: | oai:repositoriotec.tec.ac.cr:2238/13634 |
| Acceso en línea: | https://revistas.tec.ac.cr/index.php/tec_marcha/article/view/6040 https://hdl.handle.net/2238/13634 |
| Palabra clave: | Power flow neural networks Flujos de carga redes neuronales |