Study on winter climatic characteristics and temperature prediction model for solar greenhouses in cold regions

Solar greenhouses are essential for sustainable year-round crop production in cold regions; however, their thermal performance is significantly influenced by climatic uncertainties. Therefore, robust temperature prediction models are necessary to ensure optimal energy management. This study examines the winter climatic characteristics and develops temperature prediction models for solar greenhouses in cold regions. Hourly nighttime temperature prediction models for sunny, cloudy, and overcast conditions were developed using multiple linear regression and random forest regression, based on historical weather forecast data and field test measurements. The model calculates the future temperature of the greenhouse by inputting initial data on water and air temperature at a specific time, and by using future meteorological data covering outdoor temperature and humidity as well as surface wind speed. The results indicate that the multiple linear regression model exhibits reliable performance, with R² values of 0.71 for sunny days, 0.75 for cloudy days, and 0.82 for overcast days. In contrast, the random forest regression model demonstrates superior accuracy in more complex weather conditions, achieving R² values of 0.78 for sunny days and 0.81 for cloudy days. Key climatic factors, including outdoor temperature, relative humidity, and wind speed, exhibit distinct correlations with indoor temperature that vary depending on weather types, thereby influencing the selection of appropriate models. The most suitable prediction model can be selected based on the current weather conditions. The findings present a data-driven framework to optimize heat release strategies in solar water heating systems and improve the accuracy of greenhouse climate predictions.