Modelling of tropical greenhouse temperature by auto regressive and neural network models

The variation of inside air temperature in a tropical greenhouse was investigated. The study included an auto regressive (AR) model with an external input (ARX), an auto regressive moving average model with an external input (ARMAX) and a neural network auto regressive model with an external input (NNARX). External and internal climatic data recorded over a year were used to build and validate models for simulating environmental conditions inside the greenhouse. The variables measured to estimate the greenhouse internal climate included external temperature, solar radiation, relative humidity and cloud cover. It was observed that models performed better when series tuning was done for fixed temperatures rather than fixed during time. Although ARX outperformed ARMAX, NNARX-predicted results were in close agreement with the measurements.