Identification of greenhouse temperature system using time series based on the NARX model

An agricultural greenhouse is a temperature-controlled habitat that has the benefit of maintaining a crop-specific climate or producing them independently of the seasons and protects against insects, illnesses, and inclement weather. Regarding the non-linearity of the system, it is necessary to take this into consideration, especially throughout the identification phase. This paper describes a novel technique for the identification phase. In this regard, This paper discusses a new strategy for identifying and predicting the internal temperature of an agricultural greenhouse located in Meknes, Morocco, using a neural network based on a nonlinear autoregressive with external input time series model. In the present study, two algorithms have been used in training. Although the Levenberg-Marquardt has shown better performance and the best accuracy. The comparison between the estimated and simulated data indicates that the NARX (Non-linear Auto-Regressive with external inputs) model is qualified to predict and identify the indoor temperature of the greenhouse with an accurate coefficient of determination equal to 0.998.