In view of a prevention of mining accidents as well as an improvement of the safety supervision of coal mining enterprises, a three party stochastic evolutionary game model has thus been constructed among the government, enterprises, and miners, with a research focus on the establishment of reward and punishment mechanisms in coal mining safety supervision. An analysis is conducted on the stability conditions of game subjects by referring stochastic differential equations, followed by a dynamic simulation and analysis of the three party evolutionary game model by using Matlab, with a comparison with the situation without random interference. Research results reveal that under random interference, the selection strategies of game participants show a significant fluctuation trend. The greater the degree of random interference, the faster the stability speed of game participants will be. There is a relatively small influence of reward and punishment mechanisms on corporate strategy selection, while miners are more sensitive to reward and punishment mechanisms, with rewards exerting a more motivating effect on miners than punishments.