Inflow forecasting models based on artificial intelligence

Author

Aquino, Ronaldo R B ; Lira, Milde M S ; Marinho, Manoel H N ; Tavares, Isabela A. ; Cordeiro, Luiz F A

Author_Institution

Fed. Univ. of Pernambuco, Recife, Brazil

fYear

2010

fDate

18-23 July 2010

Firstpage

1

Lastpage

6

Abstract

This paper shows inflow forecasting models in the Sobradinho hydroelectric plant which are based on artificial intelligence tools: ANN and fuzzy logic. In the first models two ANNs were chosen to forecast the monthly inflow in the period of one year ahead; in the second, an ANN and an ANFIS (Adaptive Neuro-Fuzzy Inference System) were used to accomplish the forecasting in the period of one and two months ahead; in the third, an hybrid system in which an ANN provides the annual forecasting in the period of one year ahead and the ANFIS disaggregates it monthly; finally, other hybrid system similar to the previous one was developed, but instead of an ANFIS to provide the disaggregation, the fragmentation method was used to disaggregate the annual forecasting into monthly forecasting. The inflow data used were collected from the ONS (National Power System Operator) in the period from 1931 to 2004. The performance of the models was assessed on the inflow data in the period from 2005 to 2008. The models results show to be very powerful.

Keywords

fuzzy logic; fuzzy systems; hydroelectric power stations; inference mechanisms; load forecasting; power engineering computing; ANFIS; Sobradinho hydroelectric plant; adaptive neuro-fuzzy inference system; artificial intelligence; fuzzy logic; inflow forecasting models; Adaptation model; Analytical models; Artificial neural networks; Data models; Forecasting; Predictive models; Training;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks (IJCNN), The 2010 International Joint Conference on

Conference_Location

Barcelona

ISSN

1098-7576

Print_ISBN

978-1-4244-6916-1

Type

conf

DOI

10.1109/IJCNN.2010.5596943

Filename

5596943