adaptive and predictive controllers were designed as well as hybrid models. An
adaptive-feedforward controller was developed to control temperature by a ventilation system. A generalized predictive control using a CARIMA model of air
temperature was used. A hybrid system including both a continuous and discrete set
dynamics was used also in designing hierarchical control architecture of the
greenhouse climate control problem. Also, a combination of feedforward and
adaptive controllers were designed to control the temperature inside greenhouses
(Berenguel et al. 2003). In this case, at each sampling time the adaptive controller
estimates the parameters of a linear model using recursive least squares identification, adapts parameters and calculates the signal of a PI controller, and supervises
the control system behavior. In a new direction an extended Kalman filter (EKF) for
adaptation of the parameters of a complex physical greenhouse model was used
(Speetjens et al. 2009) the first time. Researchers found better model predictions for
the online parameter estimation using the EKF than those model predictions with
fixed parameters. Also, Hameed and Sorensen (2010) performed a comparison of
the performance of the extended Kalman filter and the unscented Kalman filter
(UKF) so as to improve the air temperature and humidity predictions of a greenhouse climate model. A nonlinear physical model for temperature and humidity
ratio inside a greenhouse was linearized in order to develop the EKF. The UKF used
instead of linearization unscented transformations. Better performance was
observed in case of UKF taking into account the estimation error and signal-tonoise ratio. The adaptive controller reduced the operating cost of the greenhouse
system. Luan et al. (2012) developed a stochastic unscented Kalman filter to estimate the state variables of a greenhouse dynamic model in case of missing measured data. A two-states model including air temperature and humidity was used to
evaluate the performance of the UKF. Better accuracy was reached by the UKF
without considering missing measurements than the one with measurement loss.
14.2.7 Feedback and Feedforward Control
Setiawan et al. (2000) performed a comparison of pseudo-derivative-feedback
(PDF) controller against proportional-integral (PI) controller using a dynamic
thermal model of a greenhouse. Results showed that PDF control has a better
capacity for load handling than a PI controller. PDF control was better in comparison to PI for system with and without time-delay. Using a fifth-order ARX
model and also state-space models for nutrients supply to a greenhouse crop, both
a feedback and feedforward control actions were implemented (Gieling et al.
2000). Water uptake was considered as a disturbance and several growing systems
were tested. Results showed that the designed controller indeed compensates
changes in water uptake by the crop as a response to variations in global radiation.
An optimization methodology for irrigation and nutrients supply based on measurements of the greenhouse climate was developed (Sigrimis et al. 2001). A
feedback loop and a feedforward control loop are coupled in order to minimize the
14 Control Strategies of Greenhouse Climate
409
adaptive-feedforward controller was developed to control temperature by a ventilation system. A generalized predictive control using a CARIMA model of air
temperature was used. A hybrid system including both a continuous and discrete set
dynamics was used also in designing hierarchical control architecture of the
greenhouse climate control problem. Also, a combination of feedforward and
adaptive controllers were designed to control the temperature inside greenhouses
(Berenguel et al. 2003). In this case, at each sampling time the adaptive controller
estimates the parameters of a linear model using recursive least squares identification, adapts parameters and calculates the signal of a PI controller, and supervises
the control system behavior. In a new direction an extended Kalman filter (EKF) for
adaptation of the parameters of a complex physical greenhouse model was used
(Speetjens et al. 2009) the first time. Researchers found better model predictions for
the online parameter estimation using the EKF than those model predictions with
fixed parameters. Also, Hameed and Sorensen (2010) performed a comparison of
the performance of the extended Kalman filter and the unscented Kalman filter
(UKF) so as to improve the air temperature and humidity predictions of a greenhouse climate model. A nonlinear physical model for temperature and humidity
ratio inside a greenhouse was linearized in order to develop the EKF. The UKF used
instead of linearization unscented transformations. Better performance was
observed in case of UKF taking into account the estimation error and signal-tonoise ratio. The adaptive controller reduced the operating cost of the greenhouse
system. Luan et al. (2012) developed a stochastic unscented Kalman filter to estimate the state variables of a greenhouse dynamic model in case of missing measured data. A two-states model including air temperature and humidity was used to
evaluate the performance of the UKF. Better accuracy was reached by the UKF
without considering missing measurements than the one with measurement loss.
14.2.7 Feedback and Feedforward Control
Setiawan et al. (2000) performed a comparison of pseudo-derivative-feedback
(PDF) controller against proportional-integral (PI) controller using a dynamic
thermal model of a greenhouse. Results showed that PDF control has a better
capacity for load handling than a PI controller. PDF control was better in comparison to PI for system with and without time-delay. Using a fifth-order ARX
model and also state-space models for nutrients supply to a greenhouse crop, both
a feedback and feedforward control actions were implemented (Gieling et al.
2000). Water uptake was considered as a disturbance and several growing systems
were tested. Results showed that the designed controller indeed compensates
changes in water uptake by the crop as a response to variations in global radiation.
An optimization methodology for irrigation and nutrients supply based on measurements of the greenhouse climate was developed (Sigrimis et al. 2001). A
feedback loop and a feedforward control loop are coupled in order to minimize the
14 Control Strategies of Greenhouse Climate
409