Desarrollo de un modelo de predicción del precio horario de la energía eléctrica para el mercado diario mediante redes neuronales
[EN] This Master's thesis is based on the development of a forecasting model of the hourly price of the electric energy for the diary market by using neural networks. The project is focused in a power marketer that has the aim of introduce itself in future electric markets, and for that it...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2017 |
| País: | España |
| Institución: | Universitat Politècnica de València (UPV) |
| Repositorio: | RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia |
| Idioma: | español |
| OAI Identifier: | oai:riunet.upv.es:10251/89183 |
| Acceso en línea: | https://riunet.upv.es/handle/10251/89183 |
| Access Level: | acceso abierto |
| Palabra clave: | models prediction price networks neural Modelos predictivos Precio horario Redes neuronales INGENIERIA ELECTRICA Máster Universitario en Ingeniería Industrial-Màster Universitari en Enginyeria Industrial |
| Sumario: | [EN] This Master's thesis is based on the development of a forecasting model of the hourly price of the electric energy for the diary market by using neural networks. The project is focused in a power marketer that has the aim of introduce itself in future electric markets, and for that it needs a tool that allows it to study the prices in advance. With this purpose, it will be developed an algorithm intended to forecast the price of electric energy price with one day in advance. The model will be based on the use of neural networks and it will be fixed and validated through a case study with real data. Firstly, it will be shown the historical evolution of previous models and a series of relevant concepts to introduce the reader to the contents of the thesis. Once the different temporal horizons that can be studied have been explained, the factors that affect the final price will be analyzed. The outcome will be conditioned by the fact of choosing them correctly. The deeper is the knowledge in power market, the more complete will be the neural network, what will lead to better results. Then, the neural network architecture will be analyzed, and it will also be explained how the price is obtained from that, and how the parameters of the model are adjusted along the training, with the aim of reducing the error. Hereunder, the process followed to implement the neural network in Matlab, as well as the method used to collect the necessary data. Thus, different studies will be carried out, trying with diverse possibilities that will lead to the chosen model due to its smaller error. All the results will be argued, and potential improvements for future projects will be added. |
|---|