Abstract: A Fused Attention Time Convolution Network (FATCN) based control method was proposed to promote the control accuracy of moisture content in the output tobacco from an HDT pneumatic tobacco dryer. Real time data processing, feature derivation and enhancement are completed using multivariate Gaussian distribution detection and a noise reduction autoencoder. Using the FATCN model, the monitored data are evaluated real-timely and the adjustment time and amounts of the control parameters are accurately identified to achieve automatic control of the tobacco drying process. The control effects of FATCN model, PID+manual adjustment and convolutional neural network model were comparatively tested on two cigarette brands in Chengdu Cigarette Factory. The results showed that compared with the other two control methods, the average deviation of moisture content in the dried tobacco was reduced by 0.105 and 0.086 with the reduction degree of 60% and 55%, respectively; the amount of over-dried cut tobacco at the start stage of drying was reduced by 15.8 and 15.1 kg with the reduction degree of 12.6% and 12.1%, respectively, under the FATCH control method. This method provides support for promoting consistency and accuracy of moisture content in dried tobacco.