Due to the center deviation among wellhead,rotary table and crown block,the deep groove ball bearing of rotary control head（ RCH） is affected by large lateral load. In the meantime,deep groove ball bearing can produce a large amount of heat because of the effect of friction torque,which is easy to cause the failure of bearing on account of high temperature and reduce production safety performance on site of RCH. In order to optimize and timely replace the lubrication cooling medium,furthermore,to improve the service life of the deep groove ball bearing and the overall performance of RCH,this paper pays attention to the temperature change of deep groove ball bearing. Based on the laboratory test datas,grey neural network was applied to establish mathematic model for predicting the temperature of deep groove ball bearing,and compared with BP neural network. The results show that grey neural network has the advantages of high prediction precision,good stability and less sample data,which has important application value on the temperature prediction of deep groove ball bearing and the design of RCH’s cooling lubrication system.