RT21 | Near Real-Time Transient Stability Analysis using Deep Learning approach

Digital twin is one of the most promising approaches enabling data-driven technologies for realizing near real-time solutions of complex systems like power systems. One of the digital twin approaches is to replace the dynamics or physics of a complex system with a machine learning model using real-time sensor data from the network. A machine learning model of the dynamics or physics based on the data from sensors can be used to perform stability analysis reducing computational burden significantly. Furthermore, machine learning model can be interfaced in parallel with an existing complex dynamical system providing real-time interaction with closed-loop feedback and hence the digital twin concept. In this presentation, we present deep learning approaches to perform a transient stability analysis of a power grid. Phasor measurement units are used in the modern power grid and it enables monitoring the dynamics online through direct variables such as voltage and angle phasor. Hence, PMU data can be considered as the power grid