Synchronized Measurements based Online Transient Stability Assessment using Gaussian Process Regression
In this paper an online post-fault Transient Stability Assessment (TSA) method using synchronized measurements or PMU measurements and Gaussian Process Regression (GPR) is presented. A post-fault multi-machine system is converted into two machine groups (Critical and Non critical) then into a suitable OMIB system using Single Machine Equivalent (SIME) method. With the help of thus obtained OMIB Pa-δ trajectory, a normalized Transient Stability Margin (TSM) is proposed offline. By using pre and during fault synchronized measurements as input, different GPR models are trained offline to predict the normalized stability margin. Keeping RMSE as a measure, a best suitable model is chosen for prediction. After a fault, the synchronized measurements are used as input to this trained model to predict the stability margin online. If the predicted margin is negative, then the post-fault system said to be unstable. If the predicted margin is positive, then the system is stable. The proposed assessment method is tested using New England 39 bus test system. The results are compared with offline simulations. High prediction accuracy rates are observed for GPR models, making them suitable for online TSA.