Fast probability power flow calculation of distribution networks considering dynamic correlation and high-dimensional uncertainty

Abstract With the access of large-scale renewable energy, probabilistic power flow has become a key method to analyze the increasing uncertainty of the distribution networks. However, the calculation of probabilistic power flow requires the solution of a large number of high-dimensional nonlinear equations. To solve the earlier problems, we propose a fast calculation method of probability power flow, which takes dynamic correlation and high-dimensional uncertainty into consideration. Firstly the scenario is generated by a machine learning method, and then the dimensionality of each sample is reduced by combining the dimension reduction method with the intrinsic dimension. Finally, by using the statistical machine learning theory, the stochastic response surface agent model considering the relation between input and output is used to calculate the probability power flow, where the selection of experimental points reduces the size of the sample set. Simulation results show that the proposed agent model method based on statistical machine learning is more efficient, faster, and more accurate than the traditional probabilistic model method in complex scenarios. Therefore the validity and practicability of the method proposed in this chapter are proved.