Generation of Large-Scale Convergent Power Flow Samples Through A Data-Driven Approach

The developments of renewable energy and electricity market have significantly increased the complexity and stochastics of power system’s operation, and the increasing possibility of unpredictable power flow scenarios’ occurrence can lead to the instability of electric system while coping strategies are not prepared. Through the analysis of diverse power flow samples, the possible status of the system can be simulated and related control method can be scheduled. In order to deal with the diversity and convergence problems occurred in traditional method, this paper proposed a fully data-driven method for power flow samples generation by using a two-stage strategy, where GAN-based model is used for initial samples generation and linear SVM is used to fine-tune some key variables for convergence. The application of proposed method in New England 39-bus system shows its advantages in generating convergent power flow samples.