CGAN-Based Load Scenario Generation under Typhoon Weather

Abstract

The violent fluctuation of power load level under typhoon weather threatens the power balance of power grid. Therefore, load scenario generation under typhoon weather conditions has attracted increasing attention from power supply companies. A load scenario generation algorithm based on conditional generative adversarial network (CGAN) model for typhoon weather is proposed. Firstly, considering the fact that the typhoon samples have the characteristics of scattered landing locations, different duration periods and different grades, a load sample classification and label setting method for typhoon weather is proposed. Then, a sample expansion strategy based on conditional probability is proposed to expand the sample set to solve the problem of scarce load samples under typhoon weather. Finally, in order to further improve the actual effectiveness of the sample set, based on the idea of migration training, the load samples under normal weather are firstly used to train the CGAN, and then the typhoon sample sets are applied to train CGAN. After the model training is completed, the corresponding load scenarios can be quickly generated by inputting random noise and typhoon labels. The effectiveness and advancement of the proposed model and algorithm are verified by data set from a practical power system.

Keywords

• typhoon weather
• load scenario generation
• data expansion
• label setting
• artificial intelligence