Neural models for an intelligent greenhouse - A CMAC global greenhouse model

The effectiveness of greenhouse control can be raised by the application of model based intelligent control. The large complexity of the thermal system in the greenhouse calls for a decomposed model. The key element of such model decomposition is the global greenhouse model creating predictions for all important parts of the greenhouse. A Cerebellar Model Articulation Controller (CMAC) model was built and trained to learn the thermal dynamics in the greenhouse and to provide forecasts for the future state of the house for 20 minutes ahead. The CMAC model was tested with 2 actuator configuration settings of the greenhouse, representing more than 57% of the operating time of the house.