A Data-Driven Load Shedding Method Based on Generative Adversarial Networks Considering Missing Data and Unlearned Fault

Event-based load shedding (ELS) is a critical emergency countermeasure against transient rotor angle instability in power systems. ELS based on deep learning has recently achieved promising results. However, in real-world power systems, model input data may be incomplete, and faults may occur that are not in the training database. In this situation, the accuracy of the model decreases. To address these problems, this paper presents a data-driven LS model based on transfer learning. First, the missing data challenge is solved by using a generative adversarial network (GAN), an algorithm for unsupervised deep learning based on two competing neural networks. Then, a convolutional neural network (CNN) model is trained for optimal LS. Finally, when the model encounters unlearned faults in the online application, transfer learning is used to update the model with a small database. The proposed model's test on the IEEE 39 bus system shows its effective performance.