Building a model of temperature and humidity deficit (HD) is important for the predictive control of greenhouses. In this study, we developed temperature and HD models by using the system identification method with a small number of parameters for the quick setup of a control environment in a small greenhouse of about 100 m². The temperature and HD models were created from linear auto-regressive exogenous (ARX) and output error (OE) models. The modeling data were measured in a greenhouse where green peppers were grown unheated in the fall of 2018 and the spring of 2019. The greenhouse air temperature is output from the outside air temperature and the interior solar radiation. The greenhouse HD is output from the outside HD and the interior solar radiation. Data of 5 days are sufficient for building the model. Comparing the output temperature (and HD) with the measured values, we found that the goodness of fit was highly accurate and could be maintained at 70% (80%) for 1 week in the fall and 70% (75%) for up to 2 weeks in the spring. Because the device needs few sensors and the model can be used within a week of installation, the temperature and HD models can be used in small greenhouses without an extensive sensor network.