A neural network dynamic model for temperature and relative humidity control under greenhouse

This paper deals with the development of dynamic models for the estimations of internal temperature and relative humidity of a greenhouse. Multilayers perceptron with 12 hidden neurons with a hyperbolic tangent as an activation function and which has been trained with Levenberg Marquardt (LM) algorithm. The data used to compute the simulation model were acquired in an experimental greenhouse using a sampling time interval of 10 seconds. The greenhouse is automated with several sensors and actuators that were connected to an acquisition and control system based on a personal computer. A comparison of measured and simulated data for both temperature and relative humidity under greenhouse showed that the elaborated models were able to identify and forecast inside greenhouse conditions reasonably well.