Solar radiation prediction using wavelet decomposition

• Author(s): Coelho, J.P. ; Cunha, José Boaventura ; Oliveira, Paulo
• Date: 2008
• Persistent ID: http://hdl.handle.net/10198/2746
• Origin: Biblioteca Digital do IPB
• Subject(s): Time-series prediction; Neural networks; Wavelet decomposition; AR models
• Document Type: Conference Object
• Language: English

Description

Nowadays, a substantial part of the agricultural production takes place in greenhouses, which enable to tune the crop growing by modifying, artificially, the environmental conditions and the plant’s nutrition. The main goal is to optimise the balance between the production economic return and the operation costs of the climate actuators. Severe environment and market restrictions jointly with an increasing tendency of the fuel price motivate the development of more “intelligent” energy regulators. In order to formulate the best options for a production plan, this type of artificial supervisors must be able to formulate close predictions on a large set of variables. Considering, for instance, the air temperature control inside a greenhouse, the system must be able to close predict the evolution of the solar radiation since this is the exogenous variable which most influences the thermal load during the day. In this paper, an artificial neural network, in conjunction with a wavelet decomposition strategy, is used for forecasting, an hour ahead, the instantaneous solar radiation energy density sampled at one minute interval. The results obtained from this work encourage further exploitation of this kind of signal processing technique