Due to the growth of renewable energy sources and their inter-connection with the grid, applications of power electronics devices have expanded quickly which leads to Power Quality Disturbances (PQDs). It is vital to detect, classify and mitigate these PQDs as early as possible since they have the potential to result in large losses in the power system. This paper describes a feature extraction technique based on Stockwell’s Transform (ST) for the PQDs identification. Power Quality Disturbances of nine types are generated in MATLAB as per their mathematical model and IEEE-519 standards for this study. To produce three distinct scenarios, random noise levels of 20, 30 and 40 dB are added to the PQDs. Frequency domain features of the PQDs signals are extracted using ST which are further used for the classification. Random forest classifier is used for the classification of PQDs. The performance of the proposed method for PQDs with three different noise levels has been analysed using performance metrics such as accuracy, hamming loss, precision, recall and confusion matrix. Results demonstrates that the proposed method is efficient to detect and classify the PQDs under different noise condition.