WGAN-GP Generating Wind Power Scenarios and Using BP Neural Network for Classification Evaluation

Due to the influence of factors such as weather and geography, renewable energy generation has a high level of uncertainty, which poses challenges to the stability and optimization of the power grid operation. Therefore, accurately describing this uncertainty becomes a crucial problem that needs to be addressed urgently. This article proposes a method for wind power scenario generation based on the Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP). This method is model-free and data-driven. Compared to traditional probabilistic modeling approaches, WGAN-GP has higher operability, higher efficiency, and better results in generating wind power scenarios. In addition, we designed a Backpropagation Neural Network (BPNN) model for wind power scenario classification. We trained the BPNN model using real wind power data from the NREL dataset. Finally, we compared the generated wind power scenarios with the real data to validate the accuracy of the WGAN-GP model in generating wind power scenarios.