Identification of Transfer Functions for Control of Greenhouse Air Temperature

Abstract

The purpose of this paper is to describe a method for determining transfer functions for the air temperature in greenhouses. Such models are required for using modern control strategies such as the generalized predictive controller. The input variables are heat from the heating system and from solar radiation. The structure of the transfer functions is identified from preliminary estimates of the impulse response functions obtained by using ridge regression. In order to provide reliable and accurate parameter estimates an experiment was designed and conducted in January and February 1991. Data were sampled at 2 min intervals over two periods. During the experiment the heat supply was controlled by a pseudo random binary signal (PRBS) in order to avoid correlation between the heat supply and other variables, and in order to ensure that the dynamic characteristics of the greenhouse were present in the data. For both periods, a transfer function model was determined, and good agreement between the models for the two periods was observed. The methods used show that for the greenhouse considered, a third order model with two time constants for the response from the heating system is adequate. A short time constant of about 5 min and a longer time constant of about 25 min were found.