In order to change the greenhouse environmental management from manual experience management to automatic management, the greenhouse with higher tomato yield in the same region was selected as the research object, and the temperature, humidity, roll film opening and closing status in the greenhouse and outdoor meteorological data were monitored in real time, the relationship between outdoor meteorological data, indoor temperature and humidity, and roll film opening and closing of the greenhouse was established. With outdoor meteorological data as input, indoor temperature and humidity and roll film opening and closing degree as output, the Cat Boost algorithm was used to build temperature and humidity prediction models and roll film decisionmaking models respectively, and the prediction results were compared and analyzed with those of Random Forest (RF)  and Extreme Gradient Boosting (XG Boost) algorithms. Finally, the film rolling control system was designed and built based on the decision model, and tested on typical sunny, cloudy and rainy days. The results showed that the root mean square error of the predicted value and the measured value of the indoor environment (indoor temperature and indoor relative humidity) of the solar greenhouse based on Cat Boost was 0.91℃ and 4.73%, respectively, which had better simulation effect and high accuracy, and the accuracy of the film rolling decision model reached 92.1%. The roll film control system based on Cat Boosts roll film decision model can accurately predict the change of roll film state and conduct accurate control, which has strong practical value and promotion significance.