Prediction of System Voltage Recovery due to Single Phase Induction Motor Stall Using Machine Learning Techniques

Abstract

The stalling of single-phase induction motor loads following a fault may lead to delayed voltage recovery. This phenomenon is known as fault induced delayed voltage recovery (FIDVR). FIDVR can cause significant issues in power system and in severe cases it may result in power system blackouts. This paper proposes a machine learning-based method for predicting FIDVR duration. The detailed WECC composite load model consist of air conditioner load model, thermal protection model and proposed randomized load disconnection models have been used for producing required data for FIDVR analysis. Several power system features utilized as input for training the linear regression algorithm. By implementing proposed method, the duration of FIDVR can be assessed a few cycles after the fault. Hence, more time will be available for following emergency controls such as load shedding.