Abstract: The Thermal deformation of large gantry five-axis machine tools is one of the important factors affecting machining accuracy. The influence law of environmental temperature changes on the thermal deformation of machine tools is explored. To improve the prediction accuracy of the comprehensive thermal error of large gantry CNC machine tools in the environment, a thermal error prediction model with convolutional grey long short-term memory neural network (CNN-Grey-LSTM) is designed. Taking a certain large gantry machine tool as the research object, the thermal drift error of the tool tip point caused by environmental temperature changes is analyzed by combining finite element simulation and experiments. The thermal error models are established respectively by CNN-Grey-LSTM, CNN-LSTM and GNNMCI(1,N) and compared. The results show that compared with common neural networks, the CNN-Grey-LSTM model can better adapt to complex and changeable data characteristics and time series prediction problems, and has better prediction accuracy and robustness.