A Federated Deep Learning Approach for Privacy-Preserving Real-Time Transient Stability Predictions in Power Systems

Maintaining the privacy of power system data is essential for protecting sensitive information and ensuring the operation security of critical infrastructure. Therefore, the adoption of centralized deep learning (DL) transient stability assessment (TSA) frameworks can introduce risks to electric utilities. This is because these frameworks make utility data susceptible to cyber threats and communication issues when transmitting data to a central server for training a single TSA model. Additionally, the centralized approach demands significant computational resources, which may not always be readily available. In light of these challenges, this paper introduces a federated DL-based TSA framework designed to identify the operating states of the power system. Instead of local utilities transmitting their data to a central server for centralized model training, they independently train their own TSA models using their respective datasets. Subsequently, the parameters of each local TSA model are sent to a central server for model aggregation, and the resulting model is shared back with the local clients. This approach not only preserves the integrity of local utility data, making it resilient against cyber threats but also reduces the computational demands for local TSA model training. The proposed approach is tested on four local clients each having the IEEE 39-bus test system.