1.NARI Group Corporation (State Grid Electric Power Research Institute), Nanjing 211106,China;2.NARI Technology Co., Ltd., Nanjing 211106, China

In response to the current difficulty in balancing accuracy and speed in probabilistic power flow, as well as the lack of effective means to handle source and load data with arbitrary probability distributions, a probabilistic power flow method based on the arbitrary probability distribution modelling strategy and improved polynomial chaos expansion is proposed. Firstly, the system inputs are fitted to the probability distribution in the parameterized probability distribution type library. The optimal distribution is selected based on the Akaike information criterion, and the likelihood estimates of the optimal distribution and non-parametric kernel density estimation are compared to determine the final probability distribution. Secondly, to improve the accuracy of the generalized polynomial chaos expansions based on least angular regression, the pseudo spectral method and moment matching method are used to obtain a set of candidate points, and the combination probability window is used to filter them and obtain the optimal candidate points. Then, Latin hypercube sampling is performed on the original probability space to obtain supplementary configuration points, which are combined with the optimal candidate points to obtain the final configuration points. The proposed method has been validated in IEEE 30-bus and IEEE 118-bus cases, and its accuracy is significantly improved compared with the uncertainty quantification computing framework UQLab recommendation algorithm under similar time consumption.

This work is supported by National Key R&D Program of China (No. 2022YFB2404200).