Federated Learning and Comparative Analysis of Machine Learning Models for Small Signal Stability Analysis

Power systems, consisting of interconnected components with dynamic properties, necessitate stable operations to ensure reliability. Small signal stability assessment (SSSA) plays a vital role in detecting and correcting small disturbances in power systems to maintain power system stability during unforeseen contingencies. While traditional model-based methods are accurate. However, these methods are computationally intensive. Machine learning (ML) techniques have emerged as efficient alternatives, offering faster and sometimes more accurate stability analysis. This paper evaluates various ML models, including neural networks and ensemble methods, for evaluation of SSSA using datasets containing eigenvalues and operating points. Additionally, it proposes a novel methodology for ensuring SSSA assessment using federated learning, addressing privacy concerns associated with data sharing in power systems research. By harnessing ML algorithms and real-world data, this study contributes to advancing SSSA methodologies, highlighting the potential of ML in enhancing power system stability assessment. The test system undertaken for this study is IEEE-14 bus system.