Neurocomputing 43 (2002) 51–75 www.elsevier.com/locate/neucom Neural network models in greenhouse air temperature prediction P.M. Ferreira a ; ∗ , E.A. Faria b , A.E. Ruano a; c a Faculdade de Ciˆ encias e Tecnologia, Universidade do Algarve, Campus de Gambelas, 8000-Faro, Portugal b Faculdade de Engenharia de Recursos Naturais, Universidade do Algarve, Campus de Gambelas, 8000-Faro, Portugal c Instituto de Sistemas e Rob otica, Universidade de Coimbra, 3030 Coimbra, Portugal Abstract The adequacy of radial basis function neural networks to model the inside air temperature of a hydroponic greenhouse as a function of the outside air temperature and solar radiation, and the inside relative humidity, is addressed. As the model is intended to be incorporated in an environmental control strategy both o-line and on-line methods could be of use to accomplish this task. In this paper known hybrid o-line training methods and on-line learning algorithms are analyzed. An o-line method and its application to on-line learning is proposed. It exploits the linear–non-linear structure found in radial basis function neural networks. c  2002 Elsevier Science B.V. All rights reserved. Keywords: Radial basis functions; Neural networks; Greenhouse environmental control; Modelling 1. Introduction Feed-forward layered neural networks (NNs) have extensively been applied in many elds of engineering in order to perform some type of non-linear process- ing on data generated by a wide variety of systems. In the elds of modelling and identication of non-linear systems, this growing interest is due to several reasons. Some of the most general ones are that no prior knowledge about the structure of the dynamical system is needed, that multiple-input multiple-output ∗ Corresponding author. E-mail addresses: pfrazao@ualg.pt (P.M. Ferreira), efaria@ualg.pt (E.A. Faria), aruano@ualg.pt (A.E. Ruano). 0925-2312/02/$-see front matter c  2002 Elsevier Science B.V. All rights reserved. PII:S0925-2312(01)00620-8