Electricity price forecasting using neurofuzzy networks

Abstract:

A forecasting model for the price of electricity in Colombia using neurofuzzy networks is proposed. Two network structures including the price series in the first and the price series plus the reserve water levels in the latter are used. The results are compared with two neural networks structures and a Generalized Autoregressive Conditional Heteroscedasticity model (GARCH). Historical data were supplied by the Company XM of the ISA Group; data for 120 days were used as for training the network and the following 31 days were used for testing the pbkp_redictive capabilities of the model. The GARCH model shows better adjustment within the training period for the prices series as input, but the neurofuzzy networks have better forecasting performance for one and for two input variables.

Año de publicación:

2011

Keywords:

• Electricity price
• artificial neural networks
• Time series models
• Neurofuzzy networks

Fuente:

scopus