It is very important to find short circuit faults of power transmission lines (PTL) quickly and efficiently. Most methods in the literature use classification algorithms for fault detection, but their use in real-time applications increases fault detection time. The reason for this that while the fault detection process is performed with the classification algorithm, the features of incoming data must be extracted continuously by using a window function. In this study, principal component analysis (PCA) or independent component analysis (ICA) algorithms that are suitable for real-time fault detection are proposed to decrease the fault detection time. Besides, time-domain statistical properties of the PTL signals computed over a period of time are proposed to increase classification speed and accuracy. The results show that PCA and ICA algorithms can detect all faults in real-time data streams, and the classification results are 100% for 10 faults with the proposed features.