Coal and gas outbursts are dynamic disasters prone to occur during coal mining in China. The stress is one of the significant factors leading to coal and gas outbursts. However, the stress has not been used as an effective parameter for predicting coal and gas outbursts due to the difficulty of obtaining accurate in-situ stress on a large scale. In the hope of solving this problem, a method for the digital twin of the entire space-time stress field of the mine and outburst risk prediction was proposed. This method utilised geological materials to create a three-dimensional point cloud database of the mine, and realized the establishment of a proportional mine digital model through the combined application of FLCA3D and Rhino. The assimilation of the digital model and mine entity was implemented by developing an in-situ stress correction method, a rock mechanics parameter assignment method and an assimilation effect verification method. Based on this approach, the entire space-time stress field can be obtained through digital twin simulation of the mining operation. Finally, the outburst risk was initially divided by using the original stress field and rock failure condition, and then conclusively predicted by using the entire space-time stress field and the failure risk coefficient. The method had been applied in Sangshuping Coal Mine of Hancheng Mining Area. The four sets of surface movement data from the 4312 workings observed in the field and monitored by simulation were all in the range of 1.0-1.3 m, and three of them showed similar trends, thus proving that the digital model was effectively assimilated. Out of the 120 outburst accidents that were recorded, 95% of the outburst locations were distributed in the medium-high outburst risk zones predicted by this method. This method is of great significance in guiding the safe mining.