In essence, electric digital twinning uses artificial intelligence technology to model complex electric power systems, and is the development and supplement of electric power modeling technology. This paper intends to predict and analyze the steady-state risks of complex power systems based on the power digital twin. Firstly, power flow calculation and optimization are carried out for complex large power grid systems. Based on sparse matrix storage and node coding optimization, the power flow calculation speed is improved and the memory usage is reduced. The accuracy and timeliness of the continuous power flow calculation when obtaining the node power and voltage are improved by using the unit processing tangent prediction vector and the internal machine of the prediction vector to determine the prediction direction. Secondly, according to the optimization results of the power flow calculation, the multi-objective optimization problem of power system simulation is solved by using the advantages of neural network modeling, such as self-learning, self-adaptation, fault tolerance, and parallelism. Finally, the power flow calculation optimization and neural network analysis are applied to the integrated stability program of the United States Western Combined Power Grid (WSCC) power system’s nine-node model; this is in order to simulate the regional power grid for simulation analysis. Different risks in the power system under steady–state conditions are predicted and analyzed, the voltage drop in the transient voltage is reduced under multiple working conditions, and the relative power angle is improved, improving the overall stability of the power system.