We put forward a forecasting method of time-dependent correlation-based conditional probabilistic power flow for increasing the accuracy of power flow in distributed power network. Firstly

the correlation of fluctuations and forecast errors of distributed generation’s（DG） power was analyzed. Secondly

a multiple-variable generalized autoregressive conditional heteroscedasticity model based on dynamic correlation coefficient（DCC

M

VGARCH） was proposed to directly analyze the time-varying correlation between participants such as DG power fluctuations and its forecast error

wind power and photovoltaic（PV） power generation

etc. Finally

considering the power fluctuation as the conditional correlation factor

the joint conditional probability distribu

tion of WT and PV was elaborated based on the time-varying correlation coefficient. By using the Monte-Carlo simulation method

the conditional power flow calculation was then implemented. A modified IEEE 37-bus power system with a certain penetration of WT and PV was employed to validate the proposed method. The simulation was performed every 5 minutes and 15 minutes

respectively. The results show that the conditional probabilistic power flow with time-varying correlation can adjust the distribution of probabilistic power flow according to the magnitude of volatility

which makes the distribution of power flow more reasonable

and improves the credibility and accuracy of probabilistic power flow forecasting results.