The grey theory is introduced to establish the prediction model of hazardous chemical accidents as it is suitable for uncertain system with short, medium-term sample data and little information. The GM (1,1) model of hazardous chemical accidents, unbiased grey model of hazardous chemical accidents and improved GM (1,1) model of hazardous chemical accidents are established based on the number of the hazardous chemical accidents during the period 2006–2016 in China. In addition, the number of hazardous chemical accidents during 2017–2018 in China is predicted using the three models. Predictive accuracy is proposed to compare predictive accuracy of the three models. The fitting accuracy grade of the three models are all “Pass” using the posterior variance test, which shows the three models can be used to predict the number of hazardous chemical accidents. The results show that the predictive accuracy of unbiased grey model is highest than the other two models. Therefore, the unbiased grey model of hazardous chemical accidents is relatively optimal for prediction the number of hazardous chemical accidents, which provides a new method for the prediction of hazardous chemical accidents.