Abstract: In distribution network the random uncertainty of photovoltaic (abbr. PV) easily leads to the voltage out-of-limit, however at the same time PV possesses active voltage supporting capacity to take a certain self-elimination effect on the affection of voltage, therefore the voltage security of the new-type distribution system containing high proportion of PV becomes more and more complicated. For this reason, a new type of digital twin voltage security analysis and early warning method for new type of distribution network was proposed. Firstly, the realization mechanism and functional scheme for voltage security warning in twin distribution network twins were expounded and based on operating data an Markov model of random behavior of sources and loads in distribution network was established, and the voltage out-of-limit risk evaluation indices under uncertain scenario was proposed. Secondly, the Gibbs algorithm was utilized to perform Monte Carlo random sampling for sources and loads in distribution network and according to current operating state of the distribution network and by use of ultra-real-time computing capacity of the digital twins, the afterward random scenes of the distribution network was quickly simulated and the voltage risk index was computed to evaluate the voltage security. Finally, by means of calculating example the effectiveness and reasonableness of the proposed analysis method is verified.