Loading...
Please wait, while we are loading the content...
Similar Documents
Gestão de descarregamentos: Um estudo empírico de várias abordagens baseadas em regressão
| Content Provider | Semantic Scholar |
|---|---|
| Author | Neves, Susana Manuel, Carlos Oliveira, Milheiro De Soares, Pinto Guedes, Nuno |
| Copyright Year | 2015 |
| Abstract | Hydroelectric plants have come to occupy a position of particular importance in energy production in Portugal and other countries. Due to the growth in number and relevance, dams have raised several issues related to their construction and operation, one of them being the management of releases. Releases are indispensable for the correct operation of a dam and their good management provides a number of benefits, not only for the company responsible but also for the local communities and the environment. However, in the rainy season and with high tributary flows, this task becomes particularly difficult, and the existence of auxiliary tools for releases management is benefic. In order to address this problem, several approaches were developed to predict the future state of important hydrological variables in the release decisions. To create these models, several Data Mining techniques were applied on data from a pair of reservoirs in cascade. Specifically, hourly and sporadic data were used, from two reservoirs located consecutively, and the models were developed using time series analysis, quantile regression and prediction of rare extreme values, allowing the forecast of the tributary flow over a 10 hour prediction horizon. These last two approaches are relevant because, since the focus is on floods seasons and not in normal situations, it is important to study proper methods for extreme and rare cases. It should also be noted that, according to the search elaborated, these two methods have not been implemented on the release prediction yet, being therefore more challenging and innovative approaches than the others. A new approach was also developed, taking advantage of the previous study but with some changes in the concept, causing this to be a simulation approach. Although the desired performance was not achieved in the flood cases, several interesting conclusions can be drawn and even acceptable results are presented in the general case. The last mentioned approach made possible to obtain better results in the flood cases, comparing with the previous ones, and better predictions than the ones in the naive model. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://repositorio-aberto.up.pt/bitstream/10216/79361/2/35492.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
