Greenhouses are closed environments that allow growing plants out of season. Hence, indoor conditions of greenhouses are critically important and adjuste to support plant growth. Controlling the indoor environment is essential to maintain an ideal microclimate, which directly affects plant health and, consequently, their yields. By optimizing environmental conditions inside the greenhouse, it is possible to increase yields while reducing energy consumption, taking into account information from both indoor and outdoor environments, as internal parameters are influenced by the external environment. Therefore, the main objective of this study is to create a predictive model of key variables, including indoor air temperature and relative humidity, in a greenhouse equipped with an integrated thermal storage system located in southern Algeria (in Ghardaïa). The greenhouse's microclimatic data were gathered daily for two months during the winter period. A total of 2833 input samples were collected and analyzed based on the Levenberg-Marquardt training algorithm model. This model uses meteorological variables as inputs and evaluates them with Artificial Neural Network techniques. The back-propagation neural network training was divided into three sets for testing (15%), validation (15%) and training (70%). The results of applying neural network technology proved highly satisfactory in predicting indoor temperature and relative humidity, with correlation coefficients estimated at 0.984 and 0.975 respectively, enabling successful management of the indoor environment for optimal yield.