Driven by the carbon peak carbon neutral target, the new energy vehicle industry has ushered in a period of rapid development, and the number of electric vehicles continues to climb. The randomness and uncertainty of the spatial and temporal distribution of large-scale electric vehicles connected to the power grid for charging bring severe challenges to the power system operation. In this paper, a charging load inference and governance method based on Bayesian network is proposed for the problem of the influence of disordered charging load of electric vehicle charging pile on the stability of power grid operation. The method establishes the probabilistic dependence between the charging load and the grid operation parameters by constructing a Bayesian network model, taking into account the influencing factors such as EV charging mode, user behavioral characteristics and power battery characteristics. The study adopts Monte Carlo simulation method to generate EV charging load data, uses BIC scoring function to optimize the network structure, and determines the network parameters through maximum likelihood estimation method. The experimental results show that the data inference method based on Bayesian network achieves an average precision of 93.3% and a recall of 94.5% in charging pile state prediction, which is 7.49% and 9.63% higher than the traditional method, respectively. In the IEEE 33-node distribution system simulation, the peak system line loss occurs in the 18:00-23:00 time period when EV penetration increases from 0% to 100%. The study shows that the Bayesian network governance method can effectively reduce the adverse effects of uncontrolled charging of EVs on the voltage deviation and network loss of the distribution network, and provide a scientific basis for the optimization of power grid operation.
