Handling missing values in greenhouse microclimate dataset using PCA-SARIMAX model

The purpose of this paper is to present a novel approach to handle all-sensors losses of the internal greenhouse environmental data due to the power cut throughout the greenhouse. The proposed method is based on the principal components analysis (PCA) and the seasonal autoregressive integrated moving average model with exogenous variables (SARIMAX). The exogenous variables are derived from the external meteorological dataset provided by the weather station of the city where the greenhouse is located. The role of the PCA method is to analyze the correlation between exogenous and the available endogenous variables and then reduce the dimensions of the exogenous dataset. After selecting the best choice of the training set for the SARIMAX model, the obtained results show that the proposed approach represent a promising solution for completing the bulk missing data in internal greenhouse environmental dataset.