Methodology for Small-Signal Stability Emergency Control in Low-Inertia Power Systems Using Phasor Measurements and Machine Learning Algorithms: A Data-Driven Approach

In the process of decarbonizing electricity generation, renewable energy sources are actively being integrated into traditional power systems. As a result, the inertia of the energy system is reduced, and the speed of transition processes is accelerated. This can lead to instability under small disturbances. This necessitates changing traditional approaches to implementing algorithms for emergency control automation. The paper proposes a methodology to solve the problem of small-signal stability analysis in low-inertia energy systems. The task of the small-signal stability analysis problem is reduced to multi-class classification problems. The proposed methodology can be divided into two main parts: selecting the most informative input features and classifying control actions. The IEEE24 mathematical model of the power system serves as a data source. Measurements from this model are received via phasor measurement units. Among the feature selection algorithms considered, the Random Forest algorithm proved to be the most effective. In terms of efficiency in solving the control action selection problem, the LightGBM algorithm proved dominant. Its accuracy in noise-free data was 98%. With 20 dB of data noise, the algorithm’s accuracy decreased slightly: 97%. The algorithm’s time delay was only 0.07 ms.