A Data-Driven Forced Oscillation Detection Using Random Forest

This study focuses on crucial challenges in forced oscillations (FOs) detection in power system. Since FOs require specialised mitigation schemes in contrast to electro-mechanical oscillations, it is crucial to detect the same prior to implementation of any control techniques. This paper presents a timeseries classification approach to detect FOs in power systems using Random Forest. The dataset is generated on the IEEE 4- generator 11-bus system, which is utilised to create a 3D array that is exploited to train and evaluate the machine learning model. The robustness of the technique is assessed via confusion matrix, F1 score, and receiver operating characteristic (ROC). The performance of proposed methodology is compared with other machine learning methods such as deep belief networks and neural network pattern recognition.