The production and processing of polycrystalline silicon, as a typical chemical production, involves dangerous and harmful factors such as explosions and leaks during production, storage, and transportation. If not detected and dealt with in a timely manner, it may lead to serious chemical hazards, which will have serious consequences for personnel safety and economic development. In order to improve the environmental safety of polycrystalline silicon production tank areas, research on temperature and humidity prediction technology for polycrystalline silicon production tank areas has been carried out. The XGBoost algorithm network is used for data analysis and processing to extract feature information of environmental factors in polycrystalline silicon storage areas. Based on this, a hybrid prediction model of CNN-LSTM is proposed. The feature Information mapping extracted from the CNN network is mapped to the LSTM network, which solves the problem that the traditional model is not accurate enough for the prediction of environmental factors in the tank farm, and realizes the accurate monitoring of environmental safety factors in the tank farm.