Tendencias recientes en el pronóstico de velocidad de viento para generación eólica
Cargando...
Autores
Director
Tipo de contenido
Editor
Document language:
Español
Fecha
Título de la revista
ISSN de la revista
Título del volumen
Documentos PDF
Resumen
Este documento tiene como objetivo presentar un marco unificado para discutir, resumir y organizar los principales avances en pronóstico de velocidad de viento para generación eólica utilizando un método auditable, ordenado y reproducible. Los principales hallazgos fueron: La mayor parte de los trabajos provienen de China y Estados Unidos, las series de tiempo usadas poseen una longitud de menos de un año, comúnmente el pronóstico es realizado en un rango de 1 hora a 48 horas hacia adelante. Muchos estudios usan solamente modelos autoregresivos (Lineares y no lineares) o en muchos casos una sola variable explicatoria. Usualmente la variable pronosticada es la velocidad de viento u la potencia generada. La revisión muestra una tendencia en la que los autores están experimentando con modelos híbridos para obtener las ventajas de cada método utilizado, también, una tendencia a utilizar métodos clásicos como redes neuronales, máquinas de vectores de soporte y modelos autorregresivos
Abstract: This document aims to provide a unified frame for discussing, summarizing and organizing the main advances in wind power forecasting using an auditable, orderly and reproducible method. Our main findings are the following: most of works forecasting time series from China and United States; time series data usually cover information with a length lower than a year of data. Commonly, the forecast is done for 1 to 48 hours ahead. Many studies using only autorregresive models (linear or no linear) or, in many cases, one explanatory variable. Usually, the variables forecasted are speed and power. The review shows a tendency in which the authors are experimenting with hybrid models to obtain the advantages of each method used, also, a trend to use classical methods such as neural networks, Support Vector Machines and autoregressive models.
Abstract: This document aims to provide a unified frame for discussing, summarizing and organizing the main advances in wind power forecasting using an auditable, orderly and reproducible method. Our main findings are the following: most of works forecasting time series from China and United States; time series data usually cover information with a length lower than a year of data. Commonly, the forecast is done for 1 to 48 hours ahead. Many studies using only autorregresive models (linear or no linear) or, in many cases, one explanatory variable. Usually, the variables forecasted are speed and power. The review shows a tendency in which the authors are experimenting with hybrid models to obtain the advantages of each method used, also, a trend to use classical methods such as neural networks, Support Vector Machines and autoregressive models.