A data‐driven probabilistic power flow method based on convolutional neural networks
Dawei Wang, Kedi Zheng, Qixin Chen, Xuan Zhang, Gang Luo
Abstract
With the growing penetration of renewable energy in the smart grid, the uncertainties of power system performance are increasing. Probabilistic power flow (PPF) introduces statistical methods into the power flow calculation process, enabling fast and efficient power system operation state assessment and prediction uncertainty. The need of PPF for real-time power flow calculation makes data-driven methods attractive, because they have certain comparative advantages over traditional power flow solution methods, especially when dealing with big-data and highly nonlinear situations. In this paper, a fast power flow calculation algorithm based on convolutional neural network (CNN) is proposed. Meanwhile, the Latin hypercube sampling method is applied to process the dataset for the CNN-based method, which further improves the efficiency of the data-driven PPF by reducing the size of the test set while keeping the PPF accurate. The efficacy of the method is demonstrated with several numerical cases on IEEE 69-bus system.